designing of adaptive model free controller based on output error and feedback linearization
TRANSCRIPT
Hindawi Publishing CorporationAbstract and Applied AnalysisVolume 2013 Article ID 461907 13 pageshttpdxdoiorg1011552013461907
Research ArticleDesigning of Adaptive Model-Free Controller Based onOutput Error and Feedback Linearization
S Pezeshki1 M A Badamchizadeh1 S Ghaemi1 and M A Poor2
1 Control Engineering Department Faculty of Electrical and Computer Engineering University of Tabriz Tabriz 5166614776 Iran2Department of Mechanical and Industrial Engineering Concordia University Montreal QC Canada
Correspondence should be addressed to M A Badamchizadeh mbadamchitabrizuacir
Received 20 September 2012 Accepted 30 December 2012
Academic Editor Shukai Duan
Copyright copy 2013 S Pezeshki et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
This paper introduces a new approach to design Model-Free Adaptive Controller (MFAC) using adaptive fuzzy procedure as afeedback linearization based on output error The basic idea is to transfer the control signal to an appropriate surface and thendepending on the output error of system the control signal changes around this surface Some examples are provided as well toillustrate the efficiency of the proposed approach The obtained simulation results have shown good performances of the proposedcontroller
1 Introduction
Classical control method is based onmathematical equationsof the system however this method suffers from some draw-backs For example in thismethod the performance of systemcan be affected by unmodeled dynamics of system andor bylarge delays Model-Free Adaptive Control (MFAC) as a partof modern control theory shows superiorities compared tomodel-based methods MFAC is an adaptive control methodand needs no information about system model It only uses119868119874 data for controller design and therefore it is a nonlinearcontroller designed without the need to mathematical modelof controlled system [1ndash3]
In 1994 Han and Wang introduced model-free topic andproved the stability of MFAC controllers In 1993-1994 HouZhongsheng represented applications of these controllersWith using pseudo-Jacobian matrix nonlinear systems arereplaced to near the rail line of controlled system and thenthe 119868119874 data of controlled system are used to design MFACcontroller [4]
In 1995 Jagannathan suggested a fuzzy stable controllerfor a limited class of nonlinear system in the form of 119909(119896 +1) = 119891(119909(119896)) + 119906(119896) where 119891(119909) is the unknown nonlinearfunction In 1996 this controller was used for general classof nonlinear system in the form of 119909(119896 + 1) = 119891(119909(119896)) +
119892(119909(119896))119906(119896) [5]
Another idea is introduced in 2000 [6] which uses aneural network as an adaptive controller for stabilization ofsystemTheir proposed algorithmdoes not need any complextuning and can be applied to any controllable multiinputsystems
Compared with other adaptive control schemes theMFAC approach has several advantages which make thismethod suitable for control applications First MFAC justdepends on the real-timemeasurement data of the controlledplant Second MFAC does not require any other testingsignals and any training process Third MFAC is simple andeasily implementable with small computational burden andhas strong robustness FourthMFACapproach does not needa specific controller for each specific process Finally theMFAC has been successfully implemented in many practicalapplications for example chemical industry linear motorcontrol injectionmodeling process PH value control and soon [4]
The main contribution of this work is to introduce anew MFAC approach based on output error and feedbacklinearization The proposed model has rapid monotonictracking error convergence robust stability and good distur-bance rejection
The rest of the paper is organized as follows in Section 2adaptive fuzzy control based on feedback linearization is
2 Abstract and Applied Analysis
described The proposed model-free adaptive controller andits Lyapunov stability are represented in Sections 3 and4 respectively in Section 5 simulation results are given tohighlight advantages of MFAC method Finally conclusion isstated in Section 6
2 Feedback Linearization by Means ofAdaptive Fuzzy Controller
In feedback linearization method control law is determinedin such a way that nonlinear terms eliminate the systemdynamic and replace it with appropriate reference input asseen below (for simplifying 119909(119905) = 119909)
(119905) = 119891 (119909) + 119892 (119909) 119906 (119905) 997888rarr 119906 (119905)
= 119892minus1
(119909) (minus119891 (119909) + 119907 (119905))
(1)
Conceptually Adaptive Feedback Linearization (AFL)method is the same as nonadaptive counterpart but AFLuses adaptation law for updating controller parameter toenhance system performance Notice that both methodsshould have complete description of system dynamic butour proposed method does not need any information ofsystem dynamic Due to the fact that it is difficult to obtaina mathematical model of the system and typically there existsome numerical approximations the object of this paper isto control systems just with 119868119874 data in order to improvesystem performance In order to reach this goal it is requiredto estimate unknown nonlinear function of system dynamic119891(119909) and 119892(119909) with 119891(119909) and 119892(119909) or estimate controllaw 119906(119905) with (119909) Estimation algorithm in this approachutilizes basic fuzzy membership function As an instance forestimating 119891(119909) we have
119891 (119909) 997888rarr 119891 (119909) =sum119901
119894=0
119888119894
120583119894
sum119901
119894=0
120583119894
(2)
where 119901 is the number of fuzzy roles 120583119894
(119909) is membershipfunction for 119894th fuzzy role also 119888
119894
is and the result of 119894thfuzzy role for system that is generally considered as linearcombination of some series of continuous functions such as119911(119909)
119911119896
(119909) isin 119877 119896 = 1 2 119898 minus 1
119888119894
= 1198861198940
+ 119886119894119894
1199111
(119909) + sdot sdot sdot + 119886119894119898minus1
119911119898minus1
(119909)
(3)
119911 = [1 1199111
(119909) 1199112
(119909) sdot sdot sdot 119911119898minus1
(119909)]119879
119860119879
=[[
[
11988610
11988611
sdot sdot sdot 1198861119898minus1
1198861199010
1198861199011
sdot sdot sdot 119886119901119898minus1
]]
]
119888 = 119911119879
119860
(4)
Using (3) (4) 119891(119909) can be rewritten as
120585 =
[1205831
1205832
sdot sdot sdot 120583119901
]
sum119901
119894=0
120583119894
119891 (119909) = 119911119879
sdot 119860 sdot 120577 + 120576 (119909) 997904rArr 119891 (119909) = 119911119879
sdot 119860 sdot 120577
(5)
where 120576(119909) is the approximation error of fuzzy approach119860(119905)is updated by the adaptation law below
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
sdot es (6)
119876 is a constant coefficient that usually is selected small es isthe difference between output 119909 and reference output 119909ref
es = 119909 minus 119909ref (7)
It should be pointed that AFC guarantees that 119860(119905)
approaches to its desired value 119860lowast(119905)Notice that the initial values of 119860(119905) can be chosen as
119860(0) = 0 or any selected value by information that isacquired from system dynamic or even can be determinedfrom another available control approach for instance thecolumns of 119860(0) can be taken as 119896
119891
(119896119891
is obtained fromstate feedback method 119906 = minus119896
119891
119909) or for example if we wantto estimate 119891(119909) function the columns of 119860(0) are better tocontain coefficients of system states of 119891(119909) [7] Fuzzy rulesare expressed as follows
If 119883 is 119865119894
then 119888119894
= 119891119894
(119911) (8)
119891119894
(119911) is related to 119894th row of 119911119879119860 matrix So the unknownnonlinear function 119891(119909) is estimated as 119891(119909) by fuzzyapproach
Adaptive-fuzzy control method has been divided to twoparts direct and indirect approachs this paper uses indirectmethod
21 Indirect Adaptive Fuzzy Controller (IAFC) First considerthe affine system equations given below
(119905) = 119891 (119909) + 119892 (119909) 119906 (119905)
(119905) = (120572 (119905) + 120572 (119909)) + (120573 (119905) + 120573 (119909)) lowast 119906 (119905)
(9)
120572(119909) 120573(119909) are known parts of system dynamic (which areobtained from experimental and numerical methods) and120572(119905) 120573(119905) are system equations that must be identifiedConsidering the above discussions 120572(119909) and 120573(119909) can beobtained as
120572 (119909) = 119911119879
120572
sdot 119860lowast
120572
sdot 120585120572
+ 119889120572
(119909)
120573 (119909) = 119911119879
120573
sdot 119860lowast
120573
sdot 120585120573
+ 119889120573
(119909)
(10)
119889120572
(119909) 119889120573
(119909) are approximated errors between real systemand fuzzy system with upper bound values |119889
120572
(119909)| le 119863120572
(119909)|119889120573
(119909)| le 119863120573
(119909) (119863120572
(119909) 119863120573
(119909) are known values) Soestimation of system parameters can be rewritten as follows
(119909) = 119911119879
120572
sdot 119860120572
sdot 120585120572
120573 (119909) = 119911119879
120573
sdot 119860120573
sdot 120585120573
(11)
Abstract and Applied Analysis 3
119860120572
119860120573
are matrices that will be updated adaptively with thefollowing updating laws
120572
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120572
sdot es
120573
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120573
sdot es sdot 119906 (119905)(12)
and finally the control law is described as follows
119906 (119905) =1
120573 (119905) + 120573 (119909)
(minus [120572 (119905) + (119909)] + 119907 (119905)) (13)
22 Direct Adaptive Fuzzy Controller (DAFC) Control law119906(119905) in feedback linearization method can be obtained asfollows
119906 (119905) =1
120573 (119909)(minus120572 (119909) + 119907 (119905)) (14)
The goal is to estimate 119906(119905) with consideration of basic fuzzyfunction as seen below
(119905) = 119911119879
119906
sdot 119860lowast
119906
sdot 120585119906
+ 119889119906
(119909) (15)
where (119905) is the estimated signal and 119889119906
(119909) is the errorbetween the described 119906 by fuzzymethod and desired (appro-priate) 119906lowast this error is restricted by upper known bound|119889119906
(119909)| le 119863119906
(119909) since practically it is difficult to define119863119906
(119909) so its value is specified and adjusted in designingprocess repetition this adjustment procedure continues untilthe performance of controller indicates that 119863
119906
(119909) comesclose to proper value Finally updating law 119860
119906
(119905) is obtainedas follows
119906
(119905) = minus119876minus1
119906
sdot 119911119879
119906
sdot 120585119906
sdot es (16)
The initial value of 119860119906
(119905)may be considered as 119860119906
(0) = 0
Remark 1 Thismethod is suitable for minimum phase plantsand also has the following limitation for 120573(119909)
minusinfin lt 1205731
lt 120573 (119909) lt 1205732
lt infin (17)
3 Proposed Approach
In the previous section an adaptive fuzzy controller usingfuzzy basis functions is described for stabilizing the con-trolled system In our introducedmethod the same approachhas been employed but our method does not need to utilizefuzzy basis functions for estimating system parametersInstead three proposed rules based on output errors areconsidered to control the plant
Consider a system with general form
(119905) = 119891 (119909 (119905)) + 119892 (119909 (119905)) 119906 (119905) (18)
At first nonlinear functions 119891 and 119892 must be estimated insuch a way that controller does not rely on system dynamicIt should be noted that in this paper the system is consideredcontinues time but the output is sampled and these sampled
points are used in proposed controller The estimations ofsystem dynamic can be defined as
119891 (119896) = 119879
119891
(119896)
119892 (119896) = 119879
119892
(119896)
(19)
And the updating values of 119892
119891
are given by
119891
(119896 + 1) = 119891
(119896) + 120572119903119879
(119896 + 1)
119892
(119896 + 1) = 119892
(119896) + 120573119906119888
(119896) 119903119879
(119896 + 1) 119868 = 1
119892
(119896) 119868 = 0
(20)
where 120572119891
120572119892
are constant coefficients with respect to 119891
119892
It should also be noted that usually these parameters donot need to be changed for different system dynamics Thevalue of 119868 will be defined in (27) 119903(119905) is the so-called ldquofilteredtracking errorrdquo which is the sum of errors as defined below
119903 (119896) = 119890119899
(119896) + 1205821
119890119899minus1
(119896) + sdot sdot sdot + 120582119899minus1
1198902
(119896) + 120582119899
1198901
(119896)
1198901
(119896) = 119910 (119896) minus 119910119889
(119896)
1198902
(119896) = 119910 (119896) minus 119910119889
(119896 + 1)
119890119899
(119896) = 119910 (119896) minus 119910119889
(119896 + 119899 minus 1)
(21)
where 120582119894
119894 = 1 119899 are coefficients of the predictive errorAnd119910
119889
is the desired output For eliminating nonlinear termsof system dynamic feedback linearization method has beenexploited for this purpose 119906(119896) is chosen as follows
119906 (119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (22)
where 119907(119896) is defined by
119907 (119896) = 119896119907
119903 (119896) + 119910119889
(119896 + 119899) minus
119901minus2
sum
119894=0
120582119894+1
119890119899minus119894
(119896) (23)
In (23) 119896119907
is the tracking error coefficient By choosingappropriate 119896
119907
119903(119896) is converged into zero and at last itcauses 119907(119896) to be restricted In other words if the bigger 119896
119907
is opted better performance of output and quick tendencyto the desired value are achieved while increasing the controlsignal 119906 it should be noted that if 119896
119907
is chosen very large itcan make the system unstable
As it is seen from (22) the problem of singularity occurswhen 119892 is near zero and the control signal can becomeunbounded To solve this problem separate controller com-prising two parts is proposed in the following form
119906 (119896)
=
119906119888
(119896) + 5 (119906119903
(119896) minus 119906119888
(119896)) exp (120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 1
119906119903
(119896) minus 5 (119906119903
(119896) minus 119906119888
(119896)) exp (minus120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 0
(24)
4 Abstract and Applied Analysis
In the above equation 119904 is a constant value that its amountimpacts on control signal 119906 119906
119888
119906119903
are defined as
119906119888
(119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (25)
119906119903
(119896) = minus1205831003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 sign(119903 (119896))
119892 (26)
where 120583 is robust control signal coefficient that the decreasingor increasing of this coefficient affects quantity of final controlsignal 119906 and also 119892 is a constant value that its small amountdecreases the effect of 119906
119903
on final control signal 119906 that is 119906119903
will become more effective factor for determining 119906 and 119868 isgiven by the equation below
119868 = 11003817100381710038171003817119892 (119896)
1003817100381710038171003817 gt 1198921003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 lt 119904
0 otherwise(27)
Here 119892 is lower bound of 119892 Proper selection of 119892 prevents 119892to be singular
Remark 2 120574 120583 and 119904 are designing parameters with thefollowing conditions [5] 120574 lt ln(2119904) 120583 gt 0 119904 gt 0
Remark 3 If 119906119888
is greater than 1120574 it is assumed to be equalto 1120574
The major principles of proposed MFA controller areconstructed based on the three following experimental ruleswhich make controller produce appropriate control signal
119906 (119896) =
119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 le
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus120578119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
(28)
where 120576 is threshold value for changing final control signal119906 in the above rules These laws utilize appropriate last levelcontrol signal (if error becomes less than specified value thatis called 120576) for using at next level Indeed if error at time 119896 and119896 minus 1 becomes less than 120576 and also error at time 119896 is greaterthan the previous time the controller applies negative valueof previous control signal for current time
Regarding adaptive fuzzy controllers [5 7] that esti-mate unknown nonlinear functions of the system by fuzzyalgorithm and when new reference input is applied to thesystem this estimation becomes better But the sensitivity ofproposed controller on estimation of nonlinear functions isless than that of AFCmethodThis is because output error hassignificant role in control effort rather than significant role ofestimation in AFC
4 Lyapunov Stability for the Proposed Model-Free Adaptive Controller
Let the Lyapunov function candidate 119881 be given by [8]
119881 = 119903119879
(119896) 119903 (119896) +1
120572tr (119879119891
(119896) 119891
(119896))
+1
120573tr (119879119892
(119896) 119892
(119896))
(29)
(a) For the first region 119868 = 1At first we should obtain the difference of Lyapunov
function in 119896 and 119896 + 1 level So general Lyapunov functionΔ119881 = Δ119881
1
+ Δ1198812
+ Δ1198813
can be acquired as follows
Δ1198811
= 119903119879
(119896 + 1) 119903 (119896 + 1) minus 119903119879
(119896) 119903 (119896)
Δ1198812
=1
120572tr (119879119891
(119896 + 1) 119891
(119896 + 1) minus 119879
119891
(119896) 119891
(119896))
Δ1198813
=1
120573tr (119879119892
(119896 + 1) 119892
(119896 + 1) minus 119879
119892
(119896) 119892
(119896))
(30)
If we prove thatΔ119881 le 0 it will demonstrate that 119892
119891
and 119903are bounded and they are able to stabilize system Δ119881
1
Δ1198812
and Δ119881
3
are obtained as follows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ (119879
119891
)119879
119879
119891
+ (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2(119879
119892
119906119888
)119879
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
Δ1198813
= minus (2 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120573) (119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(31)
With combination of the above three Lyapunov functions(32) is established
Δ119881 = minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896) minus (1 minus 120572) (119879
119891
)119879
119879
119891
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Abstract and Applied Analysis 5
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2120572119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus (1 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
+ 2120573(119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(32)
By simplifying (32) and using the below definition
120578 = 120572 + 120573
10038171003817100381710038171199061198881003817100381710038171003817
2
119868 = 1
120572 119868 = 0(33)
Inequality (34) is acquired as follows
Δ119881 le minus119903119879
(119896) (119868 minus (1 + 120578) 119896119879
119907
119896119907
) 119903 (119896)
+ 2120578(119896119907
119903 (119896))119879
(119892119906119889
+ 1205761015840
)
+ (1 + 120578) (119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
+120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
(34)
with assuming that 119896119907max is the maximum value of 119896
119907
and byresimplifying (34) inequality (35) is stated as given below
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2 + 21198862119896119907max 119903 (119896) (119892119906119889 + 1205761015840)+ 1198861
(119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(35)
where 1198861
and 1198862
are described as
1198861
= 1 + 120578 +120578
1 minus 120578 119886
2
= 120578 +120578
1 minus 120578 (36)
By considering Lemma A4 in Appendix A
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817=10038171003817100381710038171003817119892119906119889
+ 1205761015840
10038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896)
(37)
And with substituting inequality (37) in (35)
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2+ 21198862
119896119907max 119903 (119896) (1198880 + 1198881 119903 (119896))
+ 1198861
(1198880
+ 1198881
119903 (119896))119879
(1198880
+ 1198881
119903 (119896))
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(38)
With simplifying the above inequality the below equation isobtained
Δ119881 le minus (119868 minus 1198863
) 119903 (119896)2
+ 21198864
119903 (119896) + 1198865
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(39)
where 1198863
1198864
and 1198865
are defined as
1198863
= 1198861
1198962
119907max + 211988621198881119896119907max + 119886111988821
1198864
= 1198862
1198880
119896119907max + 119886111988801198881
1198865
= 1198861
1198882
0
(40)
By assuming that 120578 lt 1 the below statement becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0
(41)
As regards 120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 119886
1
1198862
1198863
1198864
1198865
gt 0 totally the first term in (39) minus(119868 minus1198863
)119903(119896)2
+ 21198864
119903(119896) + 1198865
becomes nonpositive if it obeysthe below condition
119903 (119896) gt 1205751199031
(42)
in which 1205751199031
is acquired from the below inequality
100381710038171003817100381712057511990311003817100381710038171003817 gt
1198865
+ radic1198862
5
+ 1198866
(1 minus 1198864
)
(1 minus 1198864
) (43)
Thus both terms of inequality (39) have nonpositive valuesand finally Δ119881 le 0
(b) For the second region 119868 = 0
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
2 Abstract and Applied Analysis
described The proposed model-free adaptive controller andits Lyapunov stability are represented in Sections 3 and4 respectively in Section 5 simulation results are given tohighlight advantages of MFAC method Finally conclusion isstated in Section 6
2 Feedback Linearization by Means ofAdaptive Fuzzy Controller
In feedback linearization method control law is determinedin such a way that nonlinear terms eliminate the systemdynamic and replace it with appropriate reference input asseen below (for simplifying 119909(119905) = 119909)
(119905) = 119891 (119909) + 119892 (119909) 119906 (119905) 997888rarr 119906 (119905)
= 119892minus1
(119909) (minus119891 (119909) + 119907 (119905))
(1)
Conceptually Adaptive Feedback Linearization (AFL)method is the same as nonadaptive counterpart but AFLuses adaptation law for updating controller parameter toenhance system performance Notice that both methodsshould have complete description of system dynamic butour proposed method does not need any information ofsystem dynamic Due to the fact that it is difficult to obtaina mathematical model of the system and typically there existsome numerical approximations the object of this paper isto control systems just with 119868119874 data in order to improvesystem performance In order to reach this goal it is requiredto estimate unknown nonlinear function of system dynamic119891(119909) and 119892(119909) with 119891(119909) and 119892(119909) or estimate controllaw 119906(119905) with (119909) Estimation algorithm in this approachutilizes basic fuzzy membership function As an instance forestimating 119891(119909) we have
119891 (119909) 997888rarr 119891 (119909) =sum119901
119894=0
119888119894
120583119894
sum119901
119894=0
120583119894
(2)
where 119901 is the number of fuzzy roles 120583119894
(119909) is membershipfunction for 119894th fuzzy role also 119888
119894
is and the result of 119894thfuzzy role for system that is generally considered as linearcombination of some series of continuous functions such as119911(119909)
119911119896
(119909) isin 119877 119896 = 1 2 119898 minus 1
119888119894
= 1198861198940
+ 119886119894119894
1199111
(119909) + sdot sdot sdot + 119886119894119898minus1
119911119898minus1
(119909)
(3)
119911 = [1 1199111
(119909) 1199112
(119909) sdot sdot sdot 119911119898minus1
(119909)]119879
119860119879
=[[
[
11988610
11988611
sdot sdot sdot 1198861119898minus1
1198861199010
1198861199011
sdot sdot sdot 119886119901119898minus1
]]
]
119888 = 119911119879
119860
(4)
Using (3) (4) 119891(119909) can be rewritten as
120585 =
[1205831
1205832
sdot sdot sdot 120583119901
]
sum119901
119894=0
120583119894
119891 (119909) = 119911119879
sdot 119860 sdot 120577 + 120576 (119909) 997904rArr 119891 (119909) = 119911119879
sdot 119860 sdot 120577
(5)
where 120576(119909) is the approximation error of fuzzy approach119860(119905)is updated by the adaptation law below
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
sdot es (6)
119876 is a constant coefficient that usually is selected small es isthe difference between output 119909 and reference output 119909ref
es = 119909 minus 119909ref (7)
It should be pointed that AFC guarantees that 119860(119905)
approaches to its desired value 119860lowast(119905)Notice that the initial values of 119860(119905) can be chosen as
119860(0) = 0 or any selected value by information that isacquired from system dynamic or even can be determinedfrom another available control approach for instance thecolumns of 119860(0) can be taken as 119896
119891
(119896119891
is obtained fromstate feedback method 119906 = minus119896
119891
119909) or for example if we wantto estimate 119891(119909) function the columns of 119860(0) are better tocontain coefficients of system states of 119891(119909) [7] Fuzzy rulesare expressed as follows
If 119883 is 119865119894
then 119888119894
= 119891119894
(119911) (8)
119891119894
(119911) is related to 119894th row of 119911119879119860 matrix So the unknownnonlinear function 119891(119909) is estimated as 119891(119909) by fuzzyapproach
Adaptive-fuzzy control method has been divided to twoparts direct and indirect approachs this paper uses indirectmethod
21 Indirect Adaptive Fuzzy Controller (IAFC) First considerthe affine system equations given below
(119905) = 119891 (119909) + 119892 (119909) 119906 (119905)
(119905) = (120572 (119905) + 120572 (119909)) + (120573 (119905) + 120573 (119909)) lowast 119906 (119905)
(9)
120572(119909) 120573(119909) are known parts of system dynamic (which areobtained from experimental and numerical methods) and120572(119905) 120573(119905) are system equations that must be identifiedConsidering the above discussions 120572(119909) and 120573(119909) can beobtained as
120572 (119909) = 119911119879
120572
sdot 119860lowast
120572
sdot 120585120572
+ 119889120572
(119909)
120573 (119909) = 119911119879
120573
sdot 119860lowast
120573
sdot 120585120573
+ 119889120573
(119909)
(10)
119889120572
(119909) 119889120573
(119909) are approximated errors between real systemand fuzzy system with upper bound values |119889
120572
(119909)| le 119863120572
(119909)|119889120573
(119909)| le 119863120573
(119909) (119863120572
(119909) 119863120573
(119909) are known values) Soestimation of system parameters can be rewritten as follows
(119909) = 119911119879
120572
sdot 119860120572
sdot 120585120572
120573 (119909) = 119911119879
120573
sdot 119860120573
sdot 120585120573
(11)
Abstract and Applied Analysis 3
119860120572
119860120573
are matrices that will be updated adaptively with thefollowing updating laws
120572
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120572
sdot es
120573
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120573
sdot es sdot 119906 (119905)(12)
and finally the control law is described as follows
119906 (119905) =1
120573 (119905) + 120573 (119909)
(minus [120572 (119905) + (119909)] + 119907 (119905)) (13)
22 Direct Adaptive Fuzzy Controller (DAFC) Control law119906(119905) in feedback linearization method can be obtained asfollows
119906 (119905) =1
120573 (119909)(minus120572 (119909) + 119907 (119905)) (14)
The goal is to estimate 119906(119905) with consideration of basic fuzzyfunction as seen below
(119905) = 119911119879
119906
sdot 119860lowast
119906
sdot 120585119906
+ 119889119906
(119909) (15)
where (119905) is the estimated signal and 119889119906
(119909) is the errorbetween the described 119906 by fuzzymethod and desired (appro-priate) 119906lowast this error is restricted by upper known bound|119889119906
(119909)| le 119863119906
(119909) since practically it is difficult to define119863119906
(119909) so its value is specified and adjusted in designingprocess repetition this adjustment procedure continues untilthe performance of controller indicates that 119863
119906
(119909) comesclose to proper value Finally updating law 119860
119906
(119905) is obtainedas follows
119906
(119905) = minus119876minus1
119906
sdot 119911119879
119906
sdot 120585119906
sdot es (16)
The initial value of 119860119906
(119905)may be considered as 119860119906
(0) = 0
Remark 1 Thismethod is suitable for minimum phase plantsand also has the following limitation for 120573(119909)
minusinfin lt 1205731
lt 120573 (119909) lt 1205732
lt infin (17)
3 Proposed Approach
In the previous section an adaptive fuzzy controller usingfuzzy basis functions is described for stabilizing the con-trolled system In our introducedmethod the same approachhas been employed but our method does not need to utilizefuzzy basis functions for estimating system parametersInstead three proposed rules based on output errors areconsidered to control the plant
Consider a system with general form
(119905) = 119891 (119909 (119905)) + 119892 (119909 (119905)) 119906 (119905) (18)
At first nonlinear functions 119891 and 119892 must be estimated insuch a way that controller does not rely on system dynamicIt should be noted that in this paper the system is consideredcontinues time but the output is sampled and these sampled
points are used in proposed controller The estimations ofsystem dynamic can be defined as
119891 (119896) = 119879
119891
(119896)
119892 (119896) = 119879
119892
(119896)
(19)
And the updating values of 119892
119891
are given by
119891
(119896 + 1) = 119891
(119896) + 120572119903119879
(119896 + 1)
119892
(119896 + 1) = 119892
(119896) + 120573119906119888
(119896) 119903119879
(119896 + 1) 119868 = 1
119892
(119896) 119868 = 0
(20)
where 120572119891
120572119892
are constant coefficients with respect to 119891
119892
It should also be noted that usually these parameters donot need to be changed for different system dynamics Thevalue of 119868 will be defined in (27) 119903(119905) is the so-called ldquofilteredtracking errorrdquo which is the sum of errors as defined below
119903 (119896) = 119890119899
(119896) + 1205821
119890119899minus1
(119896) + sdot sdot sdot + 120582119899minus1
1198902
(119896) + 120582119899
1198901
(119896)
1198901
(119896) = 119910 (119896) minus 119910119889
(119896)
1198902
(119896) = 119910 (119896) minus 119910119889
(119896 + 1)
119890119899
(119896) = 119910 (119896) minus 119910119889
(119896 + 119899 minus 1)
(21)
where 120582119894
119894 = 1 119899 are coefficients of the predictive errorAnd119910
119889
is the desired output For eliminating nonlinear termsof system dynamic feedback linearization method has beenexploited for this purpose 119906(119896) is chosen as follows
119906 (119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (22)
where 119907(119896) is defined by
119907 (119896) = 119896119907
119903 (119896) + 119910119889
(119896 + 119899) minus
119901minus2
sum
119894=0
120582119894+1
119890119899minus119894
(119896) (23)
In (23) 119896119907
is the tracking error coefficient By choosingappropriate 119896
119907
119903(119896) is converged into zero and at last itcauses 119907(119896) to be restricted In other words if the bigger 119896
119907
is opted better performance of output and quick tendencyto the desired value are achieved while increasing the controlsignal 119906 it should be noted that if 119896
119907
is chosen very large itcan make the system unstable
As it is seen from (22) the problem of singularity occurswhen 119892 is near zero and the control signal can becomeunbounded To solve this problem separate controller com-prising two parts is proposed in the following form
119906 (119896)
=
119906119888
(119896) + 5 (119906119903
(119896) minus 119906119888
(119896)) exp (120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 1
119906119903
(119896) minus 5 (119906119903
(119896) minus 119906119888
(119896)) exp (minus120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 0
(24)
4 Abstract and Applied Analysis
In the above equation 119904 is a constant value that its amountimpacts on control signal 119906 119906
119888
119906119903
are defined as
119906119888
(119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (25)
119906119903
(119896) = minus1205831003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 sign(119903 (119896))
119892 (26)
where 120583 is robust control signal coefficient that the decreasingor increasing of this coefficient affects quantity of final controlsignal 119906 and also 119892 is a constant value that its small amountdecreases the effect of 119906
119903
on final control signal 119906 that is 119906119903
will become more effective factor for determining 119906 and 119868 isgiven by the equation below
119868 = 11003817100381710038171003817119892 (119896)
1003817100381710038171003817 gt 1198921003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 lt 119904
0 otherwise(27)
Here 119892 is lower bound of 119892 Proper selection of 119892 prevents 119892to be singular
Remark 2 120574 120583 and 119904 are designing parameters with thefollowing conditions [5] 120574 lt ln(2119904) 120583 gt 0 119904 gt 0
Remark 3 If 119906119888
is greater than 1120574 it is assumed to be equalto 1120574
The major principles of proposed MFA controller areconstructed based on the three following experimental ruleswhich make controller produce appropriate control signal
119906 (119896) =
119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 le
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus120578119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
(28)
where 120576 is threshold value for changing final control signal119906 in the above rules These laws utilize appropriate last levelcontrol signal (if error becomes less than specified value thatis called 120576) for using at next level Indeed if error at time 119896 and119896 minus 1 becomes less than 120576 and also error at time 119896 is greaterthan the previous time the controller applies negative valueof previous control signal for current time
Regarding adaptive fuzzy controllers [5 7] that esti-mate unknown nonlinear functions of the system by fuzzyalgorithm and when new reference input is applied to thesystem this estimation becomes better But the sensitivity ofproposed controller on estimation of nonlinear functions isless than that of AFCmethodThis is because output error hassignificant role in control effort rather than significant role ofestimation in AFC
4 Lyapunov Stability for the Proposed Model-Free Adaptive Controller
Let the Lyapunov function candidate 119881 be given by [8]
119881 = 119903119879
(119896) 119903 (119896) +1
120572tr (119879119891
(119896) 119891
(119896))
+1
120573tr (119879119892
(119896) 119892
(119896))
(29)
(a) For the first region 119868 = 1At first we should obtain the difference of Lyapunov
function in 119896 and 119896 + 1 level So general Lyapunov functionΔ119881 = Δ119881
1
+ Δ1198812
+ Δ1198813
can be acquired as follows
Δ1198811
= 119903119879
(119896 + 1) 119903 (119896 + 1) minus 119903119879
(119896) 119903 (119896)
Δ1198812
=1
120572tr (119879119891
(119896 + 1) 119891
(119896 + 1) minus 119879
119891
(119896) 119891
(119896))
Δ1198813
=1
120573tr (119879119892
(119896 + 1) 119892
(119896 + 1) minus 119879
119892
(119896) 119892
(119896))
(30)
If we prove thatΔ119881 le 0 it will demonstrate that 119892
119891
and 119903are bounded and they are able to stabilize system Δ119881
1
Δ1198812
and Δ119881
3
are obtained as follows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ (119879
119891
)119879
119879
119891
+ (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2(119879
119892
119906119888
)119879
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
Δ1198813
= minus (2 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120573) (119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(31)
With combination of the above three Lyapunov functions(32) is established
Δ119881 = minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896) minus (1 minus 120572) (119879
119891
)119879
119879
119891
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Abstract and Applied Analysis 5
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2120572119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus (1 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
+ 2120573(119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(32)
By simplifying (32) and using the below definition
120578 = 120572 + 120573
10038171003817100381710038171199061198881003817100381710038171003817
2
119868 = 1
120572 119868 = 0(33)
Inequality (34) is acquired as follows
Δ119881 le minus119903119879
(119896) (119868 minus (1 + 120578) 119896119879
119907
119896119907
) 119903 (119896)
+ 2120578(119896119907
119903 (119896))119879
(119892119906119889
+ 1205761015840
)
+ (1 + 120578) (119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
+120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
(34)
with assuming that 119896119907max is the maximum value of 119896
119907
and byresimplifying (34) inequality (35) is stated as given below
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2 + 21198862119896119907max 119903 (119896) (119892119906119889 + 1205761015840)+ 1198861
(119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(35)
where 1198861
and 1198862
are described as
1198861
= 1 + 120578 +120578
1 minus 120578 119886
2
= 120578 +120578
1 minus 120578 (36)
By considering Lemma A4 in Appendix A
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817=10038171003817100381710038171003817119892119906119889
+ 1205761015840
10038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896)
(37)
And with substituting inequality (37) in (35)
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2+ 21198862
119896119907max 119903 (119896) (1198880 + 1198881 119903 (119896))
+ 1198861
(1198880
+ 1198881
119903 (119896))119879
(1198880
+ 1198881
119903 (119896))
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(38)
With simplifying the above inequality the below equation isobtained
Δ119881 le minus (119868 minus 1198863
) 119903 (119896)2
+ 21198864
119903 (119896) + 1198865
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(39)
where 1198863
1198864
and 1198865
are defined as
1198863
= 1198861
1198962
119907max + 211988621198881119896119907max + 119886111988821
1198864
= 1198862
1198880
119896119907max + 119886111988801198881
1198865
= 1198861
1198882
0
(40)
By assuming that 120578 lt 1 the below statement becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0
(41)
As regards 120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 119886
1
1198862
1198863
1198864
1198865
gt 0 totally the first term in (39) minus(119868 minus1198863
)119903(119896)2
+ 21198864
119903(119896) + 1198865
becomes nonpositive if it obeysthe below condition
119903 (119896) gt 1205751199031
(42)
in which 1205751199031
is acquired from the below inequality
100381710038171003817100381712057511990311003817100381710038171003817 gt
1198865
+ radic1198862
5
+ 1198866
(1 minus 1198864
)
(1 minus 1198864
) (43)
Thus both terms of inequality (39) have nonpositive valuesand finally Δ119881 le 0
(b) For the second region 119868 = 0
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 3
119860120572
119860120573
are matrices that will be updated adaptively with thefollowing updating laws
120572
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120572
sdot es
120573
(119905) = minus119876minus1
sdot 119911 sdot 120585119879
120573
sdot es sdot 119906 (119905)(12)
and finally the control law is described as follows
119906 (119905) =1
120573 (119905) + 120573 (119909)
(minus [120572 (119905) + (119909)] + 119907 (119905)) (13)
22 Direct Adaptive Fuzzy Controller (DAFC) Control law119906(119905) in feedback linearization method can be obtained asfollows
119906 (119905) =1
120573 (119909)(minus120572 (119909) + 119907 (119905)) (14)
The goal is to estimate 119906(119905) with consideration of basic fuzzyfunction as seen below
(119905) = 119911119879
119906
sdot 119860lowast
119906
sdot 120585119906
+ 119889119906
(119909) (15)
where (119905) is the estimated signal and 119889119906
(119909) is the errorbetween the described 119906 by fuzzymethod and desired (appro-priate) 119906lowast this error is restricted by upper known bound|119889119906
(119909)| le 119863119906
(119909) since practically it is difficult to define119863119906
(119909) so its value is specified and adjusted in designingprocess repetition this adjustment procedure continues untilthe performance of controller indicates that 119863
119906
(119909) comesclose to proper value Finally updating law 119860
119906
(119905) is obtainedas follows
119906
(119905) = minus119876minus1
119906
sdot 119911119879
119906
sdot 120585119906
sdot es (16)
The initial value of 119860119906
(119905)may be considered as 119860119906
(0) = 0
Remark 1 Thismethod is suitable for minimum phase plantsand also has the following limitation for 120573(119909)
minusinfin lt 1205731
lt 120573 (119909) lt 1205732
lt infin (17)
3 Proposed Approach
In the previous section an adaptive fuzzy controller usingfuzzy basis functions is described for stabilizing the con-trolled system In our introducedmethod the same approachhas been employed but our method does not need to utilizefuzzy basis functions for estimating system parametersInstead three proposed rules based on output errors areconsidered to control the plant
Consider a system with general form
(119905) = 119891 (119909 (119905)) + 119892 (119909 (119905)) 119906 (119905) (18)
At first nonlinear functions 119891 and 119892 must be estimated insuch a way that controller does not rely on system dynamicIt should be noted that in this paper the system is consideredcontinues time but the output is sampled and these sampled
points are used in proposed controller The estimations ofsystem dynamic can be defined as
119891 (119896) = 119879
119891
(119896)
119892 (119896) = 119879
119892
(119896)
(19)
And the updating values of 119892
119891
are given by
119891
(119896 + 1) = 119891
(119896) + 120572119903119879
(119896 + 1)
119892
(119896 + 1) = 119892
(119896) + 120573119906119888
(119896) 119903119879
(119896 + 1) 119868 = 1
119892
(119896) 119868 = 0
(20)
where 120572119891
120572119892
are constant coefficients with respect to 119891
119892
It should also be noted that usually these parameters donot need to be changed for different system dynamics Thevalue of 119868 will be defined in (27) 119903(119905) is the so-called ldquofilteredtracking errorrdquo which is the sum of errors as defined below
119903 (119896) = 119890119899
(119896) + 1205821
119890119899minus1
(119896) + sdot sdot sdot + 120582119899minus1
1198902
(119896) + 120582119899
1198901
(119896)
1198901
(119896) = 119910 (119896) minus 119910119889
(119896)
1198902
(119896) = 119910 (119896) minus 119910119889
(119896 + 1)
119890119899
(119896) = 119910 (119896) minus 119910119889
(119896 + 119899 minus 1)
(21)
where 120582119894
119894 = 1 119899 are coefficients of the predictive errorAnd119910
119889
is the desired output For eliminating nonlinear termsof system dynamic feedback linearization method has beenexploited for this purpose 119906(119896) is chosen as follows
119906 (119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (22)
where 119907(119896) is defined by
119907 (119896) = 119896119907
119903 (119896) + 119910119889
(119896 + 119899) minus
119901minus2
sum
119894=0
120582119894+1
119890119899minus119894
(119896) (23)
In (23) 119896119907
is the tracking error coefficient By choosingappropriate 119896
119907
119903(119896) is converged into zero and at last itcauses 119907(119896) to be restricted In other words if the bigger 119896
119907
is opted better performance of output and quick tendencyto the desired value are achieved while increasing the controlsignal 119906 it should be noted that if 119896
119907
is chosen very large itcan make the system unstable
As it is seen from (22) the problem of singularity occurswhen 119892 is near zero and the control signal can becomeunbounded To solve this problem separate controller com-prising two parts is proposed in the following form
119906 (119896)
=
119906119888
(119896) + 5 (119906119903
(119896) minus 119906119888
(119896)) exp (120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 1
119906119903
(119896) minus 5 (119906119903
(119896) minus 119906119888
(119896)) exp (minus120574 (1003817100381710038171003817119906119888 (119896)1003817100381710038171003817 minus 119904))
119868 = 0
(24)
4 Abstract and Applied Analysis
In the above equation 119904 is a constant value that its amountimpacts on control signal 119906 119906
119888
119906119903
are defined as
119906119888
(119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (25)
119906119903
(119896) = minus1205831003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 sign(119903 (119896))
119892 (26)
where 120583 is robust control signal coefficient that the decreasingor increasing of this coefficient affects quantity of final controlsignal 119906 and also 119892 is a constant value that its small amountdecreases the effect of 119906
119903
on final control signal 119906 that is 119906119903
will become more effective factor for determining 119906 and 119868 isgiven by the equation below
119868 = 11003817100381710038171003817119892 (119896)
1003817100381710038171003817 gt 1198921003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 lt 119904
0 otherwise(27)
Here 119892 is lower bound of 119892 Proper selection of 119892 prevents 119892to be singular
Remark 2 120574 120583 and 119904 are designing parameters with thefollowing conditions [5] 120574 lt ln(2119904) 120583 gt 0 119904 gt 0
Remark 3 If 119906119888
is greater than 1120574 it is assumed to be equalto 1120574
The major principles of proposed MFA controller areconstructed based on the three following experimental ruleswhich make controller produce appropriate control signal
119906 (119896) =
119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 le
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus120578119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
(28)
where 120576 is threshold value for changing final control signal119906 in the above rules These laws utilize appropriate last levelcontrol signal (if error becomes less than specified value thatis called 120576) for using at next level Indeed if error at time 119896 and119896 minus 1 becomes less than 120576 and also error at time 119896 is greaterthan the previous time the controller applies negative valueof previous control signal for current time
Regarding adaptive fuzzy controllers [5 7] that esti-mate unknown nonlinear functions of the system by fuzzyalgorithm and when new reference input is applied to thesystem this estimation becomes better But the sensitivity ofproposed controller on estimation of nonlinear functions isless than that of AFCmethodThis is because output error hassignificant role in control effort rather than significant role ofestimation in AFC
4 Lyapunov Stability for the Proposed Model-Free Adaptive Controller
Let the Lyapunov function candidate 119881 be given by [8]
119881 = 119903119879
(119896) 119903 (119896) +1
120572tr (119879119891
(119896) 119891
(119896))
+1
120573tr (119879119892
(119896) 119892
(119896))
(29)
(a) For the first region 119868 = 1At first we should obtain the difference of Lyapunov
function in 119896 and 119896 + 1 level So general Lyapunov functionΔ119881 = Δ119881
1
+ Δ1198812
+ Δ1198813
can be acquired as follows
Δ1198811
= 119903119879
(119896 + 1) 119903 (119896 + 1) minus 119903119879
(119896) 119903 (119896)
Δ1198812
=1
120572tr (119879119891
(119896 + 1) 119891
(119896 + 1) minus 119879
119891
(119896) 119891
(119896))
Δ1198813
=1
120573tr (119879119892
(119896 + 1) 119892
(119896 + 1) minus 119879
119892
(119896) 119892
(119896))
(30)
If we prove thatΔ119881 le 0 it will demonstrate that 119892
119891
and 119903are bounded and they are able to stabilize system Δ119881
1
Δ1198812
and Δ119881
3
are obtained as follows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ (119879
119891
)119879
119879
119891
+ (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2(119879
119892
119906119888
)119879
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
Δ1198813
= minus (2 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120573) (119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(31)
With combination of the above three Lyapunov functions(32) is established
Δ119881 = minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896) minus (1 minus 120572) (119879
119891
)119879
119879
119891
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Abstract and Applied Analysis 5
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2120572119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus (1 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
+ 2120573(119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(32)
By simplifying (32) and using the below definition
120578 = 120572 + 120573
10038171003817100381710038171199061198881003817100381710038171003817
2
119868 = 1
120572 119868 = 0(33)
Inequality (34) is acquired as follows
Δ119881 le minus119903119879
(119896) (119868 minus (1 + 120578) 119896119879
119907
119896119907
) 119903 (119896)
+ 2120578(119896119907
119903 (119896))119879
(119892119906119889
+ 1205761015840
)
+ (1 + 120578) (119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
+120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
(34)
with assuming that 119896119907max is the maximum value of 119896
119907
and byresimplifying (34) inequality (35) is stated as given below
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2 + 21198862119896119907max 119903 (119896) (119892119906119889 + 1205761015840)+ 1198861
(119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(35)
where 1198861
and 1198862
are described as
1198861
= 1 + 120578 +120578
1 minus 120578 119886
2
= 120578 +120578
1 minus 120578 (36)
By considering Lemma A4 in Appendix A
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817=10038171003817100381710038171003817119892119906119889
+ 1205761015840
10038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896)
(37)
And with substituting inequality (37) in (35)
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2+ 21198862
119896119907max 119903 (119896) (1198880 + 1198881 119903 (119896))
+ 1198861
(1198880
+ 1198881
119903 (119896))119879
(1198880
+ 1198881
119903 (119896))
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(38)
With simplifying the above inequality the below equation isobtained
Δ119881 le minus (119868 minus 1198863
) 119903 (119896)2
+ 21198864
119903 (119896) + 1198865
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(39)
where 1198863
1198864
and 1198865
are defined as
1198863
= 1198861
1198962
119907max + 211988621198881119896119907max + 119886111988821
1198864
= 1198862
1198880
119896119907max + 119886111988801198881
1198865
= 1198861
1198882
0
(40)
By assuming that 120578 lt 1 the below statement becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0
(41)
As regards 120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 119886
1
1198862
1198863
1198864
1198865
gt 0 totally the first term in (39) minus(119868 minus1198863
)119903(119896)2
+ 21198864
119903(119896) + 1198865
becomes nonpositive if it obeysthe below condition
119903 (119896) gt 1205751199031
(42)
in which 1205751199031
is acquired from the below inequality
100381710038171003817100381712057511990311003817100381710038171003817 gt
1198865
+ radic1198862
5
+ 1198866
(1 minus 1198864
)
(1 minus 1198864
) (43)
Thus both terms of inequality (39) have nonpositive valuesand finally Δ119881 le 0
(b) For the second region 119868 = 0
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
4 Abstract and Applied Analysis
In the above equation 119904 is a constant value that its amountimpacts on control signal 119906 119906
119888
119906119903
are defined as
119906119888
(119896) = 119892minus1
(119896) (minus119891 (119896) + 119907 (119896)) (25)
119906119903
(119896) = minus1205831003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 sign(119903 (119896))
119892 (26)
where 120583 is robust control signal coefficient that the decreasingor increasing of this coefficient affects quantity of final controlsignal 119906 and also 119892 is a constant value that its small amountdecreases the effect of 119906
119903
on final control signal 119906 that is 119906119903
will become more effective factor for determining 119906 and 119868 isgiven by the equation below
119868 = 11003817100381710038171003817119892 (119896)
1003817100381710038171003817 gt 1198921003817100381710038171003817119906119888 (119896)
1003817100381710038171003817 lt 119904
0 otherwise(27)
Here 119892 is lower bound of 119892 Proper selection of 119892 prevents 119892to be singular
Remark 2 120574 120583 and 119904 are designing parameters with thefollowing conditions [5] 120574 lt ln(2119904) 120583 gt 0 119904 gt 0
Remark 3 If 119906119888
is greater than 1120574 it is assumed to be equalto 1120574
The major principles of proposed MFA controller areconstructed based on the three following experimental ruleswhich make controller produce appropriate control signal
119906 (119896) =
119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 le
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816 lt 120576
10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
minus120578119906 (119896 minus 1) if 10038161003816100381610038161198901 (119896)1003816100381610038161003816 gt
10038161003816100381610038161198901 (119896 minus 1)1003816100381610038161003816
(28)
where 120576 is threshold value for changing final control signal119906 in the above rules These laws utilize appropriate last levelcontrol signal (if error becomes less than specified value thatis called 120576) for using at next level Indeed if error at time 119896 and119896 minus 1 becomes less than 120576 and also error at time 119896 is greaterthan the previous time the controller applies negative valueof previous control signal for current time
Regarding adaptive fuzzy controllers [5 7] that esti-mate unknown nonlinear functions of the system by fuzzyalgorithm and when new reference input is applied to thesystem this estimation becomes better But the sensitivity ofproposed controller on estimation of nonlinear functions isless than that of AFCmethodThis is because output error hassignificant role in control effort rather than significant role ofestimation in AFC
4 Lyapunov Stability for the Proposed Model-Free Adaptive Controller
Let the Lyapunov function candidate 119881 be given by [8]
119881 = 119903119879
(119896) 119903 (119896) +1
120572tr (119879119891
(119896) 119891
(119896))
+1
120573tr (119879119892
(119896) 119892
(119896))
(29)
(a) For the first region 119868 = 1At first we should obtain the difference of Lyapunov
function in 119896 and 119896 + 1 level So general Lyapunov functionΔ119881 = Δ119881
1
+ Δ1198812
+ Δ1198813
can be acquired as follows
Δ1198811
= 119903119879
(119896 + 1) 119903 (119896 + 1) minus 119903119879
(119896) 119903 (119896)
Δ1198812
=1
120572tr (119879119891
(119896 + 1) 119891
(119896 + 1) minus 119879
119891
(119896) 119891
(119896))
Δ1198813
=1
120573tr (119879119892
(119896 + 1) 119892
(119896 + 1) minus 119879
119892
(119896) 119892
(119896))
(30)
If we prove thatΔ119881 le 0 it will demonstrate that 119892
119891
and 119903are bounded and they are able to stabilize system Δ119881
1
Δ1198812
and Δ119881
3
are obtained as follows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ (119879
119891
)119879
119879
119891
+ (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2(119879
119892
119906119888
)119879
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
Δ1198813
= minus (2 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120573) (119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(31)
With combination of the above three Lyapunov functions(32) is established
Δ119881 = minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896) minus (1 minus 120572) (119879
119891
)119879
119879
119891
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Abstract and Applied Analysis 5
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2120572119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus (1 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
+ 2120573(119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(32)
By simplifying (32) and using the below definition
120578 = 120572 + 120573
10038171003817100381710038171199061198881003817100381710038171003817
2
119868 = 1
120572 119868 = 0(33)
Inequality (34) is acquired as follows
Δ119881 le minus119903119879
(119896) (119868 minus (1 + 120578) 119896119879
119907
119896119907
) 119903 (119896)
+ 2120578(119896119907
119903 (119896))119879
(119892119906119889
+ 1205761015840
)
+ (1 + 120578) (119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
+120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
(34)
with assuming that 119896119907max is the maximum value of 119896
119907
and byresimplifying (34) inequality (35) is stated as given below
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2 + 21198862119896119907max 119903 (119896) (119892119906119889 + 1205761015840)+ 1198861
(119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(35)
where 1198861
and 1198862
are described as
1198861
= 1 + 120578 +120578
1 minus 120578 119886
2
= 120578 +120578
1 minus 120578 (36)
By considering Lemma A4 in Appendix A
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817=10038171003817100381710038171003817119892119906119889
+ 1205761015840
10038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896)
(37)
And with substituting inequality (37) in (35)
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2+ 21198862
119896119907max 119903 (119896) (1198880 + 1198881 119903 (119896))
+ 1198861
(1198880
+ 1198881
119903 (119896))119879
(1198880
+ 1198881
119903 (119896))
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(38)
With simplifying the above inequality the below equation isobtained
Δ119881 le minus (119868 minus 1198863
) 119903 (119896)2
+ 21198864
119903 (119896) + 1198865
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(39)
where 1198863
1198864
and 1198865
are defined as
1198863
= 1198861
1198962
119907max + 211988621198881119896119907max + 119886111988821
1198864
= 1198862
1198880
119896119907max + 119886111988801198881
1198865
= 1198861
1198882
0
(40)
By assuming that 120578 lt 1 the below statement becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0
(41)
As regards 120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 119886
1
1198862
1198863
1198864
1198865
gt 0 totally the first term in (39) minus(119868 minus1198863
)119903(119896)2
+ 21198864
119903(119896) + 1198865
becomes nonpositive if it obeysthe below condition
119903 (119896) gt 1205751199031
(42)
in which 1205751199031
is acquired from the below inequality
100381710038171003817100381712057511990311003817100381710038171003817 gt
1198865
+ radic1198862
5
+ 1198866
(1 minus 1198864
)
(1 minus 1198864
) (43)
Thus both terms of inequality (39) have nonpositive valuesand finally Δ119881 le 0
(b) For the second region 119868 = 0
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 5
+ 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
+ 2120572119891
(119896119907
119903 (119896) + 119879
119892
119906119888
+ 119892119906119889
+ 1205761015840
)
minus (1 minus 120573) (119879
119892
119906119888
)119879
(119879
119892
119906119888
)
+ 120573(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)119879
times (119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
+ 2120573(119879
119892
119906119888
)119879
(119896119907
119903 (119896) + 119879
119891
+ 119892119906119889
+ 1205761015840
)
(32)
By simplifying (32) and using the below definition
120578 = 120572 + 120573
10038171003817100381710038171199061198881003817100381710038171003817
2
119868 = 1
120572 119868 = 0(33)
Inequality (34) is acquired as follows
Δ119881 le minus119903119879
(119896) (119868 minus (1 + 120578) 119896119879
119907
119896119907
) 119903 (119896)
+ 2120578(119896119907
119903 (119896))119879
(119892119906119889
+ 1205761015840
)
+ (1 + 120578) (119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
+120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
(34)
with assuming that 119896119907max is the maximum value of 119896
119907
and byresimplifying (34) inequality (35) is stated as given below
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2 + 21198862119896119907max 119903 (119896) (119892119906119889 + 1205761015840)+ 1198861
(119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(35)
where 1198861
and 1198862
are described as
1198861
= 1 + 120578 +120578
1 minus 120578 119886
2
= 120578 +120578
1 minus 120578 (36)
By considering Lemma A4 in Appendix A
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817=10038171003817100381710038171003817119892119906119889
+ 1205761015840
10038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896)
(37)
And with substituting inequality (37) in (35)
Δ119881 le minus (119868 minus 1198861
1198962
119907max) 119903 (119896)2+ 21198862
119896119907max 119903 (119896) (1198880 + 1198881 119903 (119896))
+ 1198861
(1198880
+ 1198881
119903 (119896))119879
(1198880
+ 1198881
119903 (119896))
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(38)
With simplifying the above inequality the below equation isobtained
Δ119881 le minus (119868 minus 1198863
) 119903 (119896)2
+ 21198864
119903 (119896) + 1198865
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(39)
where 1198863
1198864
and 1198865
are defined as
1198863
= 1198861
1198962
119907max + 211988621198881119896119907max + 119886111988821
1198864
= 1198862
1198880
119896119907max + 119886111988801198881
1198865
= 1198861
1198882
0
(40)
By assuming that 120578 lt 1 the below statement becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
+ 119879
119892
119906119888
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0
(41)
As regards 120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 119886
1
1198862
1198863
1198864
1198865
gt 0 totally the first term in (39) minus(119868 minus1198863
)119903(119896)2
+ 21198864
119903(119896) + 1198865
becomes nonpositive if it obeysthe below condition
119903 (119896) gt 1205751199031
(42)
in which 1205751199031
is acquired from the below inequality
100381710038171003817100381712057511990311003817100381710038171003817 gt
1198865
+ radic1198862
5
+ 1198866
(1 minus 1198864
)
(1 minus 1198864
) (43)
Thus both terms of inequality (39) have nonpositive valuesand finally Δ119881 le 0
(b) For the second region 119868 = 0
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
6 Abstract and Applied Analysis
In this region the proposed Lyapunov functions Δ1198811
Δ1198812
and Δ1198813
are obtained like (30) and they are stated asfollows
Δ1198811
= minus119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896))119879
(119879
119891
+ 119892119906119889
+ 1205761015840
) + (119879
119891
)119879
119879
119891
+ (119892119906119889
)119879
119892119906119889
+ 1205761015840119879
1205761015840
+ 2119891
(119892119906119889
+ 1205761015840
) + 2(119892119906119889
)119879
1205761015840
Δ1198812
= minus (2 minus 120572) (119879
119891
)119879
119879
119891
+ 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus 2 (1 minus 120572) 119891
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
Δ1198813
= 0
(44)
By combination of the above three equations (and definitionof 120578) the statement is demonstrated below
Δ119881 = minus 119903119879
(119896) (119868 minus 119896119879
119907
119896119907
) 119903 (119896)
+ 2(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119892119906119889
+ 1205761015840
)
+1
1 minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
(45)
with the consideration of Lemma A41003817100381710038171003817119892119906 minus 119892119906119862
1003817100381710038171003817 le 1198890 + 1198891 119903 (119896) (46)
By substituting inequality (46) in (45)
Δ119881 le minus (119868 minus 1198870
) 119903 (119896)2
+ 21198871
119903 (119896) + 1198872
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2 (47)
in which 1198870
1198871
and 1198872
are as follows
1198870
= 1198962
119907max + 21198891 (1198891 + 119896119907max) + (1198891 + 119896119907max)21 minus 120572
1198871
= 2 (1198890
+ 1205761015840
) 1198891
+4 minus 2120572
1 minus 120572(1198890
+ 1205761015840
) (1198891
+ 119896119907max)
1198872
=3 minus 2120572
1 minus 120572(1198890
+ 1205761015840
)2
(48)
By the assumption that 120578 lt 1 the statement below becomesnegative
minus (1 minus 120578)
10038171003817100381710038171003817100381710038171003817
119879
119891
minus120578
1 minus 120578(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)
10038171003817100381710038171003817100381710038171003817
2
lt 0 (49)
As regards120572 gt 0 120573 gt 0 120578 becomes positive too andfinally 1198870
1198871
1198872
gt 0The first term in (39) minus(119868minus1198870
)119903(119896)2
+21198871
119903(119896)+
1198872
becomes nonpositive if it obeys the below condition
119903 (119896) gt 1205751199032
(50)
in which 1205751199032
is obtained from the relation below
100381710038171003817100381712057511990321003817100381710038171003817 gt
1198871
+ radic1198872
1
+ 1198872
(1 minus 1198870
)
(1 minus 1198870
) (51)
As a result both terms of inequality (47) have nonpositivevalues which means Δ119881 le 0 And finally if ||119903(119896)|| gt
max(1205751199031
1205751199032
) Lyapunov function becomes negative
5 Implementation of MFA Controller
Example 4 Consider MIMO system with 2 input-2 outputsystem state equations which are given below [8]
1
(119905) = 1199092
(119905) + (1 + 1199092
(119905)2
) 1199061
(119905) + 1198891
(119905)
2
(119905) = minus 1199091
(119905) minus (01 minus exp (minus1199091
(119905)2
minus 1199092
(119905)2
)) 1199092
(119905)
+ (1 + 1199091
(119905)2
) 1199062
(119905) + 1198892
(119905)
1199101
(119905) = 1199091
(119905) + 1198991
(119905) 1199102
(119905) = 1199092
(119905) + 1198992
(119905)
119899 (119905) = measurement noise
1198891
(119905) 1198892
(119905) = disturbance(52)
Select control parameters as follows
120583 = 10 120572 = 1 120573 = 1 119904 = 2 120574 = 05
119892 = 8 119896119907
= [10 0 0 10] 1205821
= 001 120576 = 01
120578 = 0 119891
= [1 1] 119892
= [1 0 0 1]
(53)
Simulation results of MIMO system without disturbances1198891
(119905) 1198892
(119905) = 0 are demonstrated in Figure 1 In Figure 2external disturbances are applied to the system with power1198891
(119905) = 04 1198892
(119905) = 04 and finally the system with existenceof white Gaussian measurement noise (power 00001) andexternal disturbance is shown in Figure 3 also desired inputsare assumed 119910
1119889
= 11199102119889
= cos(2120587119905)
The aim of using this MIMO system which is broughtby [8] (along with two control signals) is surveying couplingeffect in multiinput multioutput systems As it is understoodfrom the result of Figure 1 caused by MFA controller and theresult of AFC in [8] it can be said that system outputs ofMFA controller do not have any overshoot in its responsebut in AFC this problem is observed As it is seen fromFigure 2 MFA controller rejects disturbance without anynoticeable effort in control signals and in the outputs Alsoit can be deduced from Figure 3 that the measurement noisehas intense effect on system outputs and control signals
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 7
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
Time (s)
System output 1Desired output 1
Out
put1
(a)
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
Time (s)
System output 2Desired output 2
Out
put2
(b)
0 05 1 15 2 25 3 35 4 45 5minus10010
20
30
40
50
60
70
Time (s)
Cont
rol p
ower
1
(c)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20020
40
60
80
(d)
Figure 1 Simulation results for MIMO system without disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs1 and 2
(approximately became twice) but it should be noted that theproposed controller is able to track desired output and yetremains stable As it is demonstrated in Figures 1 2 and 3the disadvantage of this controller is oscillating control signalProfits of MFAC in addition of good (also rapid) trackingand system stabilizing are decreasing settling time of outputresponse
Example 5 The second system under study is Maglev trainwhich is shown in Figure 4 Using the notations given inFigure 4 the vertical dynamics is described by [9]
System state equations are described as follows
1198981198892
119911 (119905)
1198891199052= minus119865 (119894 119911 119905) + 119891
119889
+ 119898119892
= minus1205830
1198732
119886119898
4[119894 (119905)
119911 (119905)]
2
+ 119891119889
+ 119898119892
119889119894 (119905)
119889119905=119894 (119905)
119911 (119905)
119889119911 (119905)
119889119905
minus2
1205830
1198732119886119898
119911 (119905) (119877119898
119894 (119905) minus 119906 (119905))
(54)
Here 119911 is the distance between rail and train 119894 is the current ofwindings 119898 is the train mass 119891
119889
is the force of disturbance1205830
is the permeability coefficient which equals 4120587 lowast 10minus7 119873the number of turns in winding 119886
119898
is the section surface ofwindings and 119877
119898
is the wire resistance in windings
Consider the state vector as below thus the aboveequations convert to new state equations
1
(119905) = 1199092
(119905)
2
(119905) = minus1205830
1198732
119886119898
4119898[1199093
(119905)
1199091
(119905)]
2
+1
119898119908 + 119892
3
(119905) = minus2119877119898
1205830
1198732119886119898
1199093
(119905) 1199091
(119905) +1199092
(119905) 1199093
(119905)
1199091
(119905)
+21199091
(119905) 119906 (119905)
1205830
1198732119886119898
119910 (119905) = 1199092
(119905) + 119899 (119905)
119899 (119905) = measurement noise119908 (119905) = external disturbance
(55)
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
8 Abstract and Applied Analysis
Time (s)
Out
put 1
System output 1Desired output 1
0 05 1 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
(c)
0
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
05 1 15 2 25 3 35 4 45 5minus80minus60minus40minus20
(d)
Figure 2 Simulation results for MIMO system with disturbance (a) (b) system outputs 1 and 2 (c) (d) control signals 119906 relative to inputs 1and 2
Maglev trains work in one of two ways both methods arebased on the same concept but involve different approaches
(1) Electromagnetic Suspension is based on magneticattraction it is very complex and somewhat unstable
(2) Electrodynamic Suspension is based on the repulsionofmagnetsThemagnetic levitation force balances theweight of the car at a stable position Controlling EMSis more difficult than EDS train because normality ofthis dynamic state is unstable
Magnetic trains have two important issues levitationand propulsion the target of controller is first goes uptrain to desired level along with guarantee stabilizing againstsome uncertainty such as wind and changing train mass inboltroads And second adjust train speed at the workingfrequency of magnets In this paper just levitation part(important section of train in controlling) is discussed Thetrain in this example goes from 10mm to 16mm level By
considering these assumption values for train and controlleras follows
119898 = 15 kg 119873 = 280 turns 119886119898
= 1024m2
119877119898
= 11Ω 120583 = 32 120572 = 1 120573 = 1
119904 = 140 120574 = 01 119892 = 9 119896119907
= 1 1205821
= 0
120576 = 0001 120578 = 1 119891
= minus 4 119892
= 1
(56)
The result of simulating without considering disturbancehas been displayed in Figure 5 in Figure 6 input step asexternal disturbance with amplitude 119889(119905) = 1 is appliedto the system for Figure 7 sinusoidal measurement noisewith amplitude 1mm in presence of disturbance is con-sidered Also white Gaussian noise in existence of externaldisturbance is applied to the system with power 8 times 10minus9(it is considered small because 119868119874 amplitude is small)which is shown in Figure 8 The purpose of presentingthis system is studying on sort of complicated practicalsystem dynamics such as Maglev in which they have some
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 9
Time (s)
Out
put 1
System output 1Desired output 1
0 051 15 2 25 3 35 4 45 50
02
04
06
08
1
12
(a)
Time (s)
Out
put 2
System output 2Desired output 2
0 05 1 15 2 25 3 35 4 45 5minus15minus1minus05005
1
15
(b)
0102030405060
Time (s)
Cont
rol p
ower
1
0 05 1 15 2 25 3 35 4 45 5minus30minus20minus10
(c)
0
20
40
60
80
Time (s)
Cont
rol p
ower
2
0 05 1 15 2 25 3 35 4 45 5
minus20minus40minus60minus80(d)
Figure 3 Simulation results for MIMO system with the measurement noise and disturbance (a) (b) system outputs 1 and 2 (c) (d) controlsignals 119906 relative to inputs 1 and 2
Figure 4 Vertical slice of magnetic levitation and rail in a magnetic train [9]
nonlinear terms like existence of first state in denominator(that affect directly the destabilizing system) and so forthAs it is demonstrated in Figure 5(a) the system outputthat is distance between rail and train has no overshootafter about 25 second it goes from 10 to 16mm smoothlySuch as the last example it is acquired from Figure 6 thatdisturbance did not have any effect on the output control
effort and settling time In Figure 7 it is illustrated thatsinusoidal measurement noise with no changes appears intooutput but no noticeable alternation in control signal hasbeen seen And finally in Figure 8 an impalpable effectof white Gaussian measurement noise on output and con-trol signal can be shown in this example some repeatedbenefits which are seen in Maglev system response are
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
10 Abstract and Applied AnalysisD
istan
ce (m
)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 5 Simulation result for Maglev without disturbance (inwhich it goes from 10 to 16mm) (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 6 Simulation result for Maglev with disturbance (a)Distance between train and rail (b) control signal
expressed such as appropriate system stabilizing good andrapid tracking disturbance rejection and small settlingtime
6 Conclusion
Model-free adaptive controller is an adaptive controller thatjust uses system outputs and does not require another systemstates (so does not need observer) and with this as model-free controller it performs good and also has some meritssuch as good stability appropriate tracking robust againstuncertainty disturbance rejection and good decoupling andthe biggest advantage of MFA controller is no requiringto system dynamic and even does not need to any priorexperiment about system
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 7 Simulation result for Maglev with sinuous measurementnoise at 15Hz and disturbance (a) Distance between train and rail(b) control signal
Dist
ance
(m)
Time (s)0 05 1 15 2 25 3
001001200140016
(a)
5
15
Time (s)
Cont
rol p
ower
0 05 1 15 2 25 3minus15minus5(v
olt)
(b)
Figure 8 Simulation result for Maglev train with white Gaussianmeasurement noise along with disturbance (a) Distance betweentrain and rail (b) control signal
Appendices
A Some RequirementsLemma A1 For each time 119896 119909(119896) is bounded by
119909 (119896) le 1198890
+ 1198891
119903 (119896) (A1)
which is 119903(119896) called ldquotracking errorrdquo and 1198890
1198891
are computablepositive constants
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 11
Proof [see [8 Lemma 32]] Let 119909(119896) isin 119880 in which 119880 is acompact subset of 119877119899 Assume that ℎ(119909(119896)) isin 119862infin119880 that isℎ(119909(119896)) is a smooth function 119880 rarr 119877 so that the Taylorseries expansion of ℎ(119909(119896)) exists Then using the bound on119909(119896) express the function ℎ(119909(119896)) on a compact set as linearform of parameters Thus the upper bound for ℎ(119909(119896)) canbe obtained as
ℎ (119909 (119896)) le 1198880
+ 1198881
119903 (119896) (A2)
where 1198880
and 1198881
are constant matrices [8]
Remark A2 120576119892
120576119891
are estimation errors of 119892 119891 respectivelyand 119889 is external disturbance These errors are bounded andhave these definite bounds
10038171003817100381710038171003817120576119891
10038171003817100381710038171003817le 120576119873119891
10038171003817100381710038171003817120576119892
10038171003817100381710038171003817le 120576119873119892
119889 le 119889119872
(A3)
Remark A3 According to (27) if 119868 = 1 then 119906119888
le 119904 so 119906119888
is bounded and if 119868 = 0 then 119906119888
gt 119904 but by consideringthe mentioned condition at Remark 3 in Section 5 in whichif 119906119888
is greater than 1120574 it is assumed to be equal 1120574 Thus itcan be deduced that in both regions and in all of times 119906
119888
isbounded and its bounds are shown as follows
10038171003817100381710038171199061198881003817100381710038171003817 le
119904 119868 = 1
1
120574119868 = 0
(A4)Lemma A4 With regards to that the proposed controlalgorithm is divided into two regions 119868 = 0 and 119868 = 1 Wewill prove the following relationship which is established in eachregion
10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le 1198880
+ 1198881
119903 (119896) 119868 = 1
1003817100381710038171003817119892119906 minus 1198921199061198621003817100381710038171003817 le 1198890 + 1198891 119903 (119896) 119868 = 0
(A5)
Proof (a) First region 119868 = 1 or 119906119888
le 119904 and 119892(119909) gt 119892In consideration of Lemma A1 it is demonstrated that
119892(119909) le 11988801
+ 11988811
119903(119896) where 11988801
11988811
are constant matrices
10038171003817100381710038171199061198881003817100381710038171003817 le 119904 997888rarr
10038171003817100381710038171199061199031003817100381710038171003817 le
120583
119892
10038171003817100381710038171199061198881003817100381710038171003817 997888rarr bounded 1003817100381710038171003817119906119903
1003817100381710038171003817 (A6)
By using the result of relation (A6) the following inequalityis obtained
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A7)
As regards 119906119888
le 119904 rarr 120574(119906119888
minus 119904) le 0
1003817100381710038171003817119906 minus 1199061198881003817100381710038171003817 le
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(10038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198882
(A8)
By considering (A7)1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817
1003817100381710038171003817(119906 minus 119906119862)1003817100381710038171003817
le 1198882
(11988801
+ 11988811
119903 (119896)) = 11988802
+ 11988812
119903 (119896)
(A9)
Also by utilizing Remark 3 (A10) is stated as follows10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817le10038171003817100381710038171003817120576119891
10038171003817100381710038171003817+10038171003817100381710038171003817120576119892
10038171003817100381710038171003817
10038171003817100381710038171199061198881003817100381710038171003817 + 119889
le 120576119873119891
+ 120576119873119892
+ 119889119872
= 120576119873
(A10)
By using (A9) and (A10) the following inequality is acquired10038171003817100381710038171003817119892 (119906 minus 119906
119862
) + 120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le1003817100381710038171003817119892 (119906 minus 119906119862)
1003817100381710038171003817 +10038171003817100381710038171003817120576119891
+ 120576119892
119906119888
+ 11988910038171003817100381710038171003817
le 11988802
+ 11988812
119903 (119896) + 120576119873
le 1198880
+ 1198881
119903 (119896)
(A11)
Thus the assumption is proved in the first region(b) Second region 119868 = 0 or 119906
119888
ge 119904 and 119892(119909) lt 119892 andthe mentioned condition for the proposed MFA controller inRemark 3 in Section 5 is stated as follows
if 119906119888
gt1
120574997904rArr 119906119888
=1
120574 (A12)
In consideration of Lemma A1 it is demonstrated that119892(119909) le 119888
01
+ 11988811
119903(119896) where 11988801
11988811
are constant matricesBy using the result of relation (A6) and the condition that isgiven above the following inequality is obtained
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2119890minus120574(119906119888minus119904)1003817100381710038171003817100381710038171003817 (A13)
As regards 119906119888
gt 119904 rarr minus120574(119906119888
minus 119904) le 0
119906 le10038171003817100381710038171199061199031003817100381710038171003817 +
1003817100381710038171003817100381710038171003817
119906119903
minus 119906119888
2
1003817100381710038171003817100381710038171003817le1
2(310038171003817100381710038171199061199031003817100381710038171003817 +
10038171003817100381710038171199061198881003817100381710038171003817) le 1198892
(A14)
By considering (B5)1003817100381710038171003817119892119906
1003817100381710038171003817 le10038171003817100381710038171198921003817100381710038171003817 119906 le 1198892 (11988801 + 11988811 119903 (119896))
= 11988902
+ 11988912
119903 (119896)
10038171003817100381710038171198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921003817100381710038171003817
10038171003817100381710038171003817100381710038171003817
1
120574
10038171003817100381710038171003817100381710038171003817
le 1198893
(A15)
By using (A10) and (A15) the following inequality isobtained
1003817100381710038171003817119892119906 minus 1198921199061198881003817100381710038171003817 le
10038171003817100381710038171198921199061003817100381710038171003817 +
10038171003817100381710038171198921199061198881003817100381710038171003817
le 11988901
+ 11988911
119903 (119896) + 1198893
le 1198890
+ 1198891
119903 (119896)
(A16)
Finally the assumption is proved in the second region Thusinequality (A5) is correct in both regions
B Obtain 119903(119896)
In consideration of error definition 119903(119896) in (21)
119903 (119896 + 1) = 119890119899
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
I+ 1205821
119890119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
II+ sdot sdot sdot + 120582
119899minus1
1198901
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
III
(B1)
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
12 Abstract and Applied Analysis
(I) 119890119899
(119896 + 1) = 119909119899
(119896 + 1) minus 119910119889
(119896 + 119899) = 119891(119909) + 119892(119909)119906(119896) +
119889(119896) minus 119910119889
(119896 + 119899)
(II) 1205821
119890119899minus1
(119896 + 1) = 1205821
[119909119899minus1
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
119909119899(119896) minus 119910119889
(119896 + 119899 minus 1)] =
1205821
119890119899
(119896)
(III) 120582119901minus1
1198901
(119896 + 1) = 120582119901minus1
[1199091
(119896 + 1)⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
1199092(119896) minus 119910119889
(119896 minus 1)] =
120582119901minus1
1198902
(119896)
Thus by substituting (I) (II) and (III) in the above equation
119903 (119896 + 1) = 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896) minus 119910119889
(119896 + 119875)
+ 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
(B2)
Equation (25) can be rewritten as follows
119907 (119896) = 119892 (119896) 119906119888
(119896) + 119891 (119896) (B3)
By using (23) (25) and (B2)
119903 (119896 + 1) = 119907 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) 119906 (119896) + 119889 (119896)
minus 119910119889
(119896 + 119875) + 1205821
119890119875
(119896) + sdot sdot sdot + 120582119875minus1
1198902
(119896)
= 119896119907
119903 (119896) minus 119907 (119896) + 119891 (119909) + 119892 (119909) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
(B4)
Also by considering the above equation the tracking errors119891
and 119892
are acquired as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + (119892119906 minus 119892119906119888
)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906119888
minus 119892119906119888
)
+ 119889 (119896) + 119892119906119889
(B5)
where 119906119889
= 119906minus119906119888
By the following definitions about the119892(119909)119891 119891(119909) and 119892 (B5) is rewritten as follows
119891 (119909) = 119906119879
119891
+ 120576119891
(119909) 997904rArr 119891 = 119879
119891
119892 (119909) = 119906119879
119892
+ 120576119892
(119909) 997904rArr 119892 = 119879
119892
119891
= 119906119891
minus 119891
119892
= 119906119892
minus 119892
119903 (119896 + 1) = 119896119907
119903 (119896) + (119906119879
119891
minus 119879
119891
) + (119906119879
119892
119906119888
minus 119879
119892
119906119888
)
+ 120576119891
+ 120576119892
119906 + 119889 (119896) + 119892119906119889
= 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 120576119891
+ 120576119892
119906 + 119892119906119889
(B6)
with substitution of 1205761015840 = 120576119891
+ 120576119892
119906119888
+ 119889 in (B6) final relationof tracking error is demonstrated as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + 119879
119891
+ 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
(B7)
Remark B1 In consideration of (20) in this region thefollowing equalities are given
119891
= 119906119891
minus 119891
997904rArr 119891
(119896 + 1) = (1 minus 120572) 119891
(119896)
minus 120572(119896119907
119903 (119896) + 119879
119892
119906119888
+ 1205761015840
+ 119892119906119889
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = (1 minus 120573) 119879
119892
(119896) 119906119888
minus 120573(119896119907
119903 (119896) + 119879
119891
+ 1205761015840
+ 119892119906119889
)119879
(B8)
By using equation (B4) in the second region tracking errors119891
119892
are obtained as follows
119903 (119896 + 1) = 119896119907
119903 (119896) + (119891 (119909) minus 119891) + (119892 (119909) 119906 minus 119892119906119888
) + 119889 (119896)
= 119896119907
119903 (119896) + (119891 (119909) minus 119891) + 119892119906119889
+ 119889 (119896)
(B9)
Remark B2 By utilizing (20) in this region 119891
and 119892
areacquired as follows
119891
= 119906119891
minus 119891
119891
(119896 + 1) = (1 minus 120572) 119891
(119896) minus 120572(119896119907
119903 (119896) + 119892119906119889
+ 1205761015840
)119879
119892
= 119906119892
minus 119892
997904rArr 119892
(119896 + 1) = 119892
(119896)
(B10)
References[1] M Yan C Xue and X Xie ldquoA novel model free adaptive
controller with tracking differentiatorrdquo in Proceedings of theIEEE International Conference onMechatronics and Automation(ICMA rsquo09) pp 4191ndash4196 Changchun China August 2009
[2] A Xu Y Zheng Y Song and M Liu ldquoAn improved modelfree adaptive control algorithmrdquo in Proceedings of the 5th Inter-national Conference on Natural Computation (ICNC rsquo09) pp70ndash74 Tianjian China August 2009
[3] Q Gao D Ren C Dong and Z Jia ldquoThe Study of Model-FreeAdaptive Controller based on dSPACErdquo in Proceedings of the2nd International Symposiumon Intelligent InformationTechnol-ogy Application pp 608ndash611 Shanghai China December 2008
[4] H Zhongsheng and J Shangtai ldquoA novel data-driven controlapproach for a class of discrete-time nonlinear systemsrdquo IEEETransactions on Control Systems Technology vol 19 no 6 pp1549ndash1558 2011
[5] S Jagannathan ldquoAdaptive fuzzy logic control of feedback lin-earizable discrete-time nonlinear systemsrdquo in Proceedings of theIEEE International Symposium on Intelligent Control DearbornSeptember 1996
[6] S G Cheng ldquoModel-free adaptive process controlrdquo UnitedStates Patent 6055524 Washington United States Patent andTrademark Office of theUnited States Department of Com-merce 2000
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Abstract and Applied Analysis 13
[7] ROrdonez J Zumberge J T Spooner andKM Passino ldquoAda-ptive fuzzy control experiments and comparative analysesrdquoIEEE Transactions on Fuzzy Systems vol 5 no 2 pp 167ndash1881997
[8] S Jagannathan ldquoAdaptive fuzzy logic control of feedbacklinearizable discrete-time dynamical systems under persistenceof excitationrdquo Automatica vol 34 no 11 pp 1295ndash1310 1998
[9] P K Sinha and A N Pechev ldquoNonlinear Hinfin controllersfor electromagnetic suspension systemsrdquo IEEE Transactions onAutomatic Control vol 49 no 4 pp 563ndash568 2004
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Game Theory
Journal ofApplied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex Systems
Journal of
ISRN Operations Research
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Abstract and Applied Analysis
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Industrial MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
OptimizationJournal of
ISRN Computational Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Complex AnalysisJournal of
ISRN Combinatorics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Geometry
ISRN Applied Mathematics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
thinspAdvancesthinspin
DecisionSciences
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2013
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Algebra
ISRN Mathematical Physics
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2013
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2013