adaptive controller for siemens pcs7 v7.0/v7 · pdf filepcs7 v7.0/v7.1 : handling operator...

i.p.a.s.-systeme

June 2010 of 28 Ingen ieu rbüro fü r P rozeßau tomat ions - und So f twaresys teme • Owner : D ip l . - Ing . Bernhard Wurm An der Landwehr 6 • D -60437 F rank fu r t • Phone +49 69 95 04 18 - 0 • Fax +49 69 95 04 18 – 19

E-Mai l : se rv ice@ipas -sys teme.de h t tp : www. ipas -sys teme.de

ADCO V7.1

ADaptive COntroller for Siemens PCS7 V7.0/V7.1

1. REVISIONHISTORY ...................................................................... 2 2. INTRODUCTION .......................................................................... 3 3. CONTROLLER CONFIGURATION/PROGRAMMING ................................. 5 4. CONTROLLER TUNING ................................................................. 17 5. SAVING TUNING PARAMETERS ..................................................... 19 6. CONTINUOUS ADAPTATION .......................................................... 19 7. CASCADE CONTROL LOOP ............................................................ 20 8. MULTIRANGE-CONTROLLER ......................................................... 21 9. TIPS AND TRICKS ...................................................................... 23 10. CONTROLLER INSTALLATION ....................................................... 26 11. GENERAL DIRECTIONS FOR USE ................................................... 27 12. FACEPLATES ............................................................................ 28

i.p.a.s.-systeme



1. Revisionhistory Version Date Description 5.3 22. May 2002 Adding of BATCH-Input/Output 5.4 6.0

29. Aug 2002 04. Dec 2003

Adding of BYPASS Input Adjusting the concept of Operation and Interaction of ADCO to the Siemens Standard Controller CTRL_PID (FB61) for PCS7 v6.0

6.1 04. Jul 2005 Adding of fled-forward control of disturbance,

31. Jul 2006 Controltolerance, extended Sensitivity Adjusting the concept of Operation and Interaction of ADCO to the Siemens Standard Controller CTRL_PID (FB61) for PCS7 v6.1

7.1 21. Dec 2009 Adjusting the concept of Operation and Interaction of ADCO to the Siemens

Standard Controller CTRL_PID (FB61) for

29. Jun 2010 PCS7 v7.0/v7.1 Handling Operator authorisation for manipulated variable with interconnected AUTO-Parameter

i.p.a.s.-systeme

June 2010 of 28

2. Introduction In most cases the tuning (optimization) of PID-controllers is based on so-called “trial and error” methods. This requires special experience and also takes a lot of time especially when trying to control sluggish processes (e.g. temperature processes). Above that the control quality does not correspond to the optimum and still leaves quite some room for improvements. The tuning procedure gets even more difficult if there are non-linear or time-variant processes to be controlled. The adaptive controller ADCO provides solutions to all these problems. It automatically adapts itself to changing process characteristics but it can also be operated as a controller with constant parameters. In this instance the adaptation is turned off after the initial optimization step and the controller then serves as a better alternative to a regular PID-controller. If necessary the adaptive mode can be turned back on any time during the operation of the controller. Besides standard lag processes ADCO is especially suited to control processes with integrating characteristics and also processes with significant deadtimes. It is common knowledge that regular controllers have problems with these types of processes. As opposed to PID-controllers ADCO provides an equally optimal control behaviour in setpoint control as well as disturbance control (non-measurable signals acting on the process variable) tasks.

ProcessController

ProcessIdentification

Validation

ControllerOptimization

Process Variable

Setpoint

ManipulatedVariable

Figure 2.1: Block structure of the adaptive control loop

I ngen ieu rbüro fü r P rozeßau tomat ions - und So f twaresys teme • Owner : D ip l . - Ing . Bernhard Wurm An der Landwehr 6 • D -60437 F rank fu r t • Phone +49 69 95 04 18 - 0 • Fax +49 69 95 04 18 – 19


i.p.a.s.-systeme



ADCO basically consists of two main parts: * The process model estimation is based on a method which is known as

DSF (Discrete Square Root Filtering) or SRIF (Square Root Information Filter). This procedure calculates a parameter model of the process to be controlled by evaluating the process signals (manipulated variable / controller output, process variable) according to the method mentioned above.

* The controller optimization is based on an estimated process model

which is validated through a supervisory function. The algorithm delivers an optimal state controller. Besides the actual control error a few more states allowing a prediction about upcoming process variable values are fed into the calculation of the manipulated variable. Since the state controller evaluates more information about the process behaviour than any PID-controller it provides a superior control quality even when acting on simple “linear” processes. After a setpoint change or a disturbance of the process variable all state deviations are reduced to 0. The control behaviour depends on one tuning parameter (controller sensitivity) which can adapt values between -100 and 150. The default value for this parameter is 25 and does not have to be changed in most applications. Increasing the sensitivity basically means increasing the activity of the controller, i.e. the controller is acting stronger onto the process using up more energy.

Outstanding advantages compared to regular controllers: * essentially faster control parameter tuning * better control quality controlling “easy to handle” processes * significantly better control behaviour controlling processes with

integrating and/or deadtime characteristics * optimal tuning for setpoint and disturbance control * adaptation to changing process characteristics * basically no overshoot

i.p.a.s.-systeme



3. Controller Configuration/Programming Like standard function blocks within PCS7 (esp. CFC – Continuous Function Chart) the adaptive state controller is to be configured by connecting block inputs and outputs to variables, constants or to inputs/outputs of other function blocks. Following the inputs and outputs of the function block are explained in more detail: Inputs: PV_IN: Process variable (e.g. temperature, pressure, level etc.) of

the control loop in physical units. QC_PV_IN: Quality-Code. Indicates the following status informations:

“Uncertain device”, „Uncertain process, “Bad process”, “Bad device”, „Testmode“. Will be indicated in the faceplate.

NM_LMNLR: Low range of the manipulated variable (controller output). NM_LMNHR: High range of the manipulated variable (controller output). LMN_LLM: Low limit of the manipulated variable within the controller

output range. LMN_HLM: High limit of the manipulated variable within the controller

output range. DISV: By the use of this input parameter, a fled-forward control

of disturbance can be configured. The value that will be accepted depends on the range of the Process Variable (NM_PVLR, NM_PVHR). If this range for example goes from -100 … +100, DISV can reach values between -200 and +200. This means, that the range of measurement NM_PVHR-NM_PVLR will be projected in both directions around the 0-point. This value will be added / subtracted to/from the controller output (LMN).

TTIME: The transition time (for an exact definition see chapter 3)

must be defined during the configuration or later during runtime (in single loop displays) just before the controller is to be optimized. The dimension of this entry is [min]. The transition time determines the internal scan rate of the

i.p.a.s.-systeme



controller (internal scan rate = transition time/60). The main scan time of the function block has to be adjusted in a way that the internal scan rate according to the formula above can be realized. The reason for the need to specify a separate (internal) scan rate is that it does not make sense to acquire data for a sluggish process (like a temperature process) with a frequency in the higher [Hz] range. In this case the differences of the process signals (between two scans) would not deliver any information about the process behaviour. The differences are then based on disturbances with a share of almost 100%. If a value of 0 is being entered into this field then the internal scan rate is set equal to the scan rate of the function block.

DTIME: The model based state controller is especially suited to

control deadtime processes. The process deadtime (for an exact definition see chapter 3) however is not calculated or estimated by the control algorithm but it has to be entered (in [min]) by the user. During the calculation of the manipulated variable this entry is evaluated. In other words the calculation of the manipulated variable is not based on the actual process variable but on a process variable in the future which is predicted by means of the estimated process model and the specified deadtime. The deadtime can be changed on-line to accomodate changing process characteristics.

SENS: The sensitivity of the controller basically is the only “tuning

parameter” (0 ... 101) to be adjusted by the user. This field has a default value of 50 which in most cases does not have to be modified. Increasing this value also means increasing the activity of the adaptive controller. Decreasing the value of course means decreasing the activity.

DIRECT: A lot of industrial controllers require the specification of the

so-called controller action as part of the tuning procedure. This entry usually determines whether a decreasing process variable (below the setpoint) should be controlled by an increasing (1 or TRUE: direct) or a decreasing (0 or FALSE: reverse) manipulated variable. In this algorithm the specification of the controller action is used to validate the process model estimated by the identification routine. “direct” means that the process must have a positive gain factor, “reverse” of course means the opposite. The estimated process model will only be conveyed to the controller optimization if - besides other checks - the

i.p.a.s.-systeme



estimated process gain factor corresponds to the gain factor derived from the specification of the controller action. For processes with an integrating characteristic this entry is not relevant since a gain factor is not defined for this type of process.

NO_VAL: Before an estimated process model is handed over to the

controller optimization procedure it is validated by applying different checks. Only if all checks show a positive result the process model is released and can be used to base a set of control parameters on. By means of this selection field the process model check can be turned off. This should only be done when - because of very noisy signals - a valid process model can not be found. However this should happen very rarely (0 or FALSE: model validation; 1 or TRUE: no model validation).

LMN_DEL: The value of this input limits the change velocity of the

manipulated variable (output change - i.a. % - per [min]). This limitation can be applied to valves where the opening and closing speed have to be limited (due to process related reasons) to a maximum value. This entry is not relevant in output track and manual mode. A value of 0 means “no limitation”.

LMN_INI: The initial (after the first system start) manipulated

variable value of a newly configured controller is determined in this field. In most cases this value remains at 0 (default value). A deviating value should be entered when setting up split-range control loops. In this case the value is usually specified so that both valves involved are started up in a de-energized state.

LMN_TRK, LMN_SEL: If this mode (LMN_SEL; 0 or FALSE: no track mode; 1 or

TRUE: track mode) is enabled the controller output (manipulated variable) is overwritten by a predefined value (LMN_TRK).

SP_TRK_ON: To ensure a bumpless transfer from manual to automatic

mode the setpoint can be defined to track the process variable value (in manual mode only). After switching to automatic the control error (setpoint - process variable) is 0 which means that no step change is generated at the controller output, i.,e. the manipulated variable (0 or FALSE: no setpoint track; 1 or TRUE: setpoint track).

i.p.a.s.-systeme



NO_BUM: In the inner controller of a cascade loop it is not possible to apply the setpoint track mode ensuring a bumpless transfer to automatic. By means of this input it is nevertheless possible to do the job. When enabled the manipulated variable is temporarily processed through a low pass filter (0 or FALSE: no low pass filtering; 1 or TRUE: low pass filtering).

LMN_STB, LMN_STBON: The control algorithm offers a mode (LMN_STBON; 0 or

FALSE: no standby mode; 1 or TRUE: standby mode) where the adaptive controller can be operated parallel to an already existing and active control concept (e.g. a PID-controller). By means of the input LMN_STB the manipulated variable of the active controller is usually transferred into the adaptive state controller. Based on that signal and also on the process variable a process model can be estimated and subsequently a controller can be optimized. In this mode the manipulated variable of the adaptive controller should not modify the corresponding valve position (this has to be ensured by an overall function block layout). The manipulated variable should just be recorded. By comparing the manipulated variable of the adaptive controller with the output of the active (acting on the valve) controller it should be possible to make a statement about the control quality of the adaptive algorithm. In this way the adaptive controller can be tested without taking any risk of upsetting the process.

AD_OVR: If this input is activated (0 -> 1) the controller is forced

into the non-adaptive mode. Forcing the controller into this mode is appropriate e.g. when the process variable value can no longer be acquired because of a sensor failure. Otherwise (when continuing to run the controller in the adaptive mode) a disrupted relationship (manipulated variable <-> process variable) would possibly be projected into the estimated process model. If the reason (e.g sensor failure) for the switch over (override) is no longer existing the controller does not automatically go back to the adaptive mode. If necessary this has to be done by a manual user interaction.

i.p.a.s.-systeme



AU_OVR: If this input is activated (0 -> 1) the controller is forced into manual mode. Forcing the controller into this mode may be applicable e.g. during certain emergency shutdown strategies. If the reason for the switch over (override) does no longer exist, the controller does not directly go back to the automatic mode. If necessary this has to be done by a manual user interaction.

IL_VAL, ILCK: If this mode (ILCK; 0 or FALSE: no interlock; 1 or TRUE:

interlock) is activated the controller goes into the interlock state, i.e. the adaptation is deactivated, the controller changes to manual and the controller output adapts a predefined value (IL_VAL).

OSHT: By means of this input it is possible to define whether it

should be possible to operate the controller immediately after the switch over to the interlock state (one-shot) or whether the controller should be forced to this state as long as the interlock input is set (0 or FALSE: regular mode; 1 or TRUE: one-shot mode).

STRT_MAN: If this input is set the controller (after a system start)

always comes up in manual mode. Otherwise the controller starts up in the mode the controller was in during system shutdown.

RNG_ADA: f this input is set the controller increases the process

variable range by 50% if the process variable approaches its current high limit. Likewise it decreases the range by 50% if the process variable approaches its current low limit value.

SUB_ZER: If this input is set and RNG_ADA is set then the low limit of

the process variable range can (during a range modification step) adapt a value below 0. Otherwise this is not possible.

ADA_DIS: If this input is set then the user should not be able to

modify the adaptation status (adaptation ON or OFF) in the corresponding controller faceplate. This input is currently not evaluated in the attached faceplate.

M_SUP_AH: If this input is set then the high/high alarm checking of the

Process Value is disabled. (0: active; 1: disabled).

i.p.a.s.-systeme



M_SUP_WH: If this input is set then the high alarm checking of the Process Value is disabled. (0: active; 1: disabled).

M_SUP_WL: If this input is set then the low alarm checking is disabled.

(0: active; 1: disabled). M_SUP_AL: If this input is set then the low alarm checking is disabled.

(0: active; 1: disabled). M_SUP_DH: If this input is set, the alarm checking of the high deviation

(SP - PV_IN) is disabled. (0: active; 1: disabled). M_SUP_DL: If this input is set, the alarm checking of the low deviation

(SP - PV_IN) is disabled. (0: active; 1: disabled). LMNOP_ON: Allows the operator to enter a value for the Manipulated

Value via the Detail Faceplate. ADCO has to be in MANUAL mode.

LIOP_MAN_SEL: Value = „1“: Switching of AUTO/MAN-mode by interconnections in CFC Value = „0”: Switching of AUTO/MAN-mode by manual operation (operator) in WinCC-Runtime

AUTOP_EN: Value = „1“: ADCO can be switched to AUTO-mode via the Detail Faceplate Value = „0”: Switching to AUTO-mode via the Detail Faceplate is disabled

MANOP_EN: Value = „1“: ADCO can be switched to MAN-mode via the

Detail Faceplate Value = „0”: Switching to MAN-mode via the Detail Faceplate is disabled

SP_OP_ON: Value = „1“: Enables the input of a Setpoint-value by the

operator via the Detail Faceplate LIOP_INT_SEL: Value = „1“: Switching of Setpoint-mode (INT/EXT) by

interconnections in CFC Value = „0”: Switching of Setpoint-mode (INT/EXT) by manual operation (operator) in WinCC-Runtime

i.p.a.s.-systeme



SPEXON_L: If LIOP_INT_SEL = „1“, the Setpoint-mode will be set by this connectable Input - Value = “1”: external Setpoint-mode - Value = “0”: internal Setpoint-mode

SPEXT_EN: Value = “1“: Allows the operator to switch ADCO to

External Setpoint-mode via the Detail faceplate Value = “0”: Switching ADCO to External Setpoint-mode via the Detail faceplate is disabled.

SPINT_EN: Value = “1“: Allows the operator to switch ADCO to

Internal Setpoint-mode via the Detail faceplate Value = “0”: Switching ADCO to Internal Setpoint-mode via the Detail faceplate is disabled.

CTOL: Control Tolerance in [EngUnits]. The controller „tolerates“ Process values which differ from the Setpoint value (in both directions + and -) by the value “Controller tolerance”, which has been defined in the Detail faceplate. The controller tries to control the process to the Setpoint value after the Control Tolerance has been left. Otherwise, the controller ensures that the process runs in fixed proportions, i.e. the controller ensures that the Process value keeps robust inside of the Control tolerance. The value = “0.0” disables the “Tolerance control”.

CTABS: Absolute Control Tolerance

If a Control Tolerance is specified AND CTABS is active (TRUE), the controller does not alter its output if the Process value is inside the Controller tolerance (around the Setpoint value). If CTABS = FALSE, the controller tries to run the process in fixed proportions inside of the Control tolerance (see CTOL).

BA_EN: BATCH ENABLE (like CTRL_PID) CSF: CONTROL SYSTEM FAULT (like CTRL_PID) OCCUPIED: OCCUPIED BY BATCH (like CTRL_PID) BA_NA: BATCH NAME (like CTRL_PID) BYPASS: Will be usefull for bypassing the indicator „B“ in the UDO in

WinCC (have a look at Standard-FB Valvecontrol)

i.p.a.s.-systeme



MO_PVLR: Low range of the bargraph of Setpoint-/Process value and Alarm limits in the Detail faceplate.

MO_PVHR: High range of the bargraph of Setpoint-/Process value and Alarm limits in the Detail faceplate.

AUX_PR05 bis AUX_PR10: Message accompany texts of the ALARM_8P FB (SFB35).

Free configurable.

MSG_LOCK: If MSG_LOCK = TRUE, the output of all Messages and Alarms in WinCC, coming from this specific controler-FB, will be disabled. Configured Alarm limits will be monitored furthermore and can be used in the programs on AS-level (see QHH_ALM bis QDL_ALM), but there is no transfer to the message lists in WinCC.

MSG_EVID: This (WORD)input is internally connected to the EV_ID input of the ALARM_8P block.

QC_LMN_I: Input value Quality-Code. Will be copied to the output

value QC_LMN. In-/Outputs: NM_PVLR: Low range of the process variable in physical units. NM_PVHR: High range of the process variable in physical units. SAMPLE_T: Scan time of the function block in [sec]. SP_OP: Setpoint of the control loop in physical units. SP_EXT: An external setpoint (in physical units) can be connected to

this input. This feature is necessary to set up e.g. cascade control loops.

SPEXTSEL_OP: If LIOP_INT_SEL = 0, the setpoint mode of the controller

will be set via the Detail faceplate. Value = “1”: External Setpoint mode. Value = “0”: Internal Setpoint mode. The binary inputs SPEXT_EN / SPINT_EN will be evaluated additionally!

SP_LLM: Low limit for setpoints (in physical units).

i.p.a.s.-systeme



SP_HLM: High limit for setpoints (in physical units). SP_DEL: This value limits the change velocity of the setpoint (in

physical units / [min]). LMN_IN: Manipulated variable (controller output) in manual mode. AUTO: This input defines the controller mode (0 or FALSE:

manual; 1 or TRUE: automatic). ADAP: This input defines the adaptive mode of the controller (0 or

FALSE: adaptation off; 1 or TRUE: adaptation on). RESET: With this input it can be determined whether the adaptive

controller is to be reset. Reset means that the controller looses all previously gathered information about process characteristics which in turn means that it has to be optimized again (0 or FALSE: no reset; 1 or TRUE: reset).

HH_LIM: High/high alarm limit (within the range of the process

variable). H_LIM: High alarm limit (in physical units within the process

variable range). L_LIM: Low alarm limit (in physical units within the process

variable range). LL_LIM: Low/low alarm limit (within the range of the process

variable). DH_LIM: High alarm limit for the deviation (SP – PV_IN). DL_LIM: Low alarm limit for the deviation (SP – PV_IN). BA_ID: BATCH-ID (like CTRL_PID) STEP_NO: BATCH STEP NUMBER (like CTRL_PID) Outputs: LMN: Manipulated variable. QC_LMN: Quality-Code. Indicates the following status informations:

“Uncertain device”, „Uncertain process, “Bad process”,

i.p.a.s.-systeme



“Bad device”, „Testmode“. Will be indicated in the faceplate.

SP: Setpoint of the controller. IDENT: If the adaptation is turned on and if the controller detects a

“sufficient” dynamic movement of the process variable then the process model estimation procedure within the adaptive algorithm is activated. An active estimation routine is indicated by setting this block output.

VAL_M: The estimated process model is checked before it is passed

to the control parameter optimization procedure. This check contains several validation steps. Only if all validation steps show a positive result the process model is released to the optimization procedure. A positive result is indicated by setting this block output.

ORIG_M: This output indicates that a first valid process model has

already been found and that therefore the controller can be switched to automatic.

QAUTO: This output represents the controller mode (0: manual; 1:

automatic). QAUTOP: Indicates, if the operator is legitimated to switch the

controller to AUTO mode. QMANOP: Indicates, if the operator is legitimated to switch the

controller to MANL mode. QADAP: This output indicates whether the controller works in the

adaptive mode (0: adaptation off; 1: adaptation on). QLMN_SEL: With this output it is possible to relay the track-mode

(LMN_SEL) to other function blocks. QLMNOP: Indicates, if the operator is legitimated to change the

controller’s Output value when the controller is in MANL mode.

Q_SP_OP: Indicates, if the operator is legitimated to change the

controller’s Setpoint value. QSP_EXT: With this output it is possible to relay the external-

setpoint-mode (SP_EXT_ON) to other function blocks.

i.p.a.s.-systeme



QSPEXTEN: Indicates, if the operator is legitimated to switch the controller into External Setpoint mode.

QSPINTEN: Indicates, if the operator is legitimated to switch the

controller into Internal Setpoint mode. QILCK: With this output it is possible to relay the interlock-state

(ILCK) to other function blocks. ER: Current control error, i.e. the deviation of the Process

value from the Setpoint value (SP – PV_IN). TAGNAME: Tagname of the controller (max. 14 characters; e.g. „FIC

1234.56“). F_SCAN: During the first scan after the configuration/programming

(F_SCAN = 0) of a new controller certain instance variables of the adaptive controller have to be initialized. During the initialization the value F_SCAN is set to TRUE or 1 which means that during the following scans the initialization routine is skipped.

DIMENS: Physical unit of the process variable (max. 8 characters;

e.g. „m³/h“). VERSIO: Version number of the adaptive controller (max. 14

characters; e.g. „Rev. 2.0“). QHH_ALM: This output indicates, whether the high/high alarm limit is

violated. This output will only be set if the corresponding alarm is enabled (M_SUP_AH := FALSE).

QH_ALM: This output indicates, whether the high alarm limit is

violated. This output will only be set if the corresponding alarm is enabled (M_SUP_AW := FALSE).

QL_ALM: This output indicates, whether the low alarm limit is

violated. This output will only be set if the corresponding alarm is enabled (M_SUP_WL := FALSE).

QLL_ALM: This output indicates, whether the low/low alarm limit is

violated. This output will only be set if the corresponding alarm is enabled (M_SUP_AL := FALSE).

QDH_ALM: This output indicates, whether the high deviation alarm

limit is violated. This output will only be set if the corresponding alarm is enabled (M_SUP_DH := FALSE).

i.p.a.s.-systeme



QDL_ALM: This output indicates, whether the low deviation alarm limit

is violated. This output will only be set if the corresponding alarm is enabled (M_SUP_DL := FALSE).

QMSG_ERR: The ERROR output of the internal ALARM_8P block is

copied to this output. QMSG_SUP: This output will be set if messaging system of PCS7 is

disabled. MSG_STAT: The STATUS output of the internal ALARM_8P block is

copied to this (WORD)output. MSG_ACK: The ACK_STATE output of the internal ALARM_8P block is

copied to this (WORD)output. DEL_LOW: Absolute low deviation (DEL_LOW := SP + DL_LIM). DEL_HIGH: Absolute high deviation (DEL_HIGH := SP + DH_LIM).

i.p.a.s.-systeme

June 2010 of 28

4. Controller Tuning If a new controller has been configured or an existing one has been reset the control algorithm does not have any information about process characteristics. Therefore the controller optimization which is based on an estimated and validated process model can not be performed. In this situation the algorithm prevents the controller from being switched to automatic. Through manual stimulation (changing the manipulated variable) knowledge about the process behaviour has to be relayed to the identification routine. First of all the adaptive control algorithm needs some basic information about the process dynamics (transition time) and possibly about process deadtimes. The transition time (see figure 4.1) is defined for lag as well as for integrating processes. Concerning lag processes the transition time is the time necessary for the process to reach a new steady state after a step change of the manipulated variable (controller output). Dealing with integrating processes the transition time is the time the process needs - starting out at a steady state - to change its process variable by n/2 % as a response to a step change of the controller output of also n % (e.g. 20 % step change of the manipulated variable -> 10% change of the process variable). It is sufficient to enter the transition time as an approximate value in [min]. The control algorithm is so robust that the entered



Startup ProcedureDead-

time andtransition time

known ?

START

Adaptation on ?

Adaptation off ?

Enter transition and

deadtime

Turn onadaptation

Turn offadaptation

Adjustadaptation mode

(on/off)

Switch to automaticand adjust the

setpoint

Change manipulatedvariable

Wait until processvariable reaches new

stationary value

Read deadtime andtransition time outof transfer function

END

N

J

N

J

N

J

0

0 ,2

0 ,4

0 ,6

0 ,8

1

1 ,2

0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5Time in [min]

Deadtime Transition TimeLag Process

ControllerOutputStep

0

0 ,2

0 ,4

0 ,6

0 ,8

1

1 ,2

0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5

Integrating ProcessDeadtime

ControllerOutputStep

Transition Time

Time in [min]

Change manipulatedvariable

Wait until processvariable reaches new

stationary value

Figure 4.1: Startup procedure for the adaptive controller

i.p.a.s.-systeme



value can be five times higher or five times lower than the real transition time without impairing the resulting control quality. The deadtime [min] should have a higher degree of accuracy. If these times are unknown they have to be determined by applying a step change to the manipulated variable (with the adaptation turned off). The necessary numbers can be classified by taking a look at the resulting process variable graph. During the following learning phase (adaptation turned on!!) a classical transfer function (answer to a step change of the manipulated variable) can be recorded. Furthermore it is also possible to adjust the controller output several times during the learning phase. So it is conceivable that the process variable is manually controlled and led to its setpoint. As soon as the algorithm detects its first valid process model the controller can be switched to automatic, i.e. the internal interlock to force the controller to manual mode is no longer effective. With the majority of processes it is not necessary to operate the controller in a continuously adaptive mode. The control algorithm can then work with a constant control parameter set (after turning off the adaptation).

i.p.a.s.-systeme

June 2010 of 28

5. Saving Tuning Parameters After the tuning parameters of the controller have been optimized they should be saved to hard disk so that after download of the PCS7 program the tuning parameters are still available and the optimization procedure has not to be repeated. This can be reached by copying the corresponding instance data blocks from the “on-line-window” to the “off-line-window” and by consequently saving the whole project to disk. If a PCS7 control program is compiled the instance data blocks are generated again, i.e. the information in the old instance data blocks is lost and the controllers would have to be re-tuned. Therefore all instance data blocks related to the adaptive controller ADCO have to copied to other data blocks before a complete compilation run is started and re-copied to the original blocks after the compilation. This procedure is not necessary if only changes are compiled.

6. Continuous Adaptation If processes show a distinct non-linear or time-variant behaviour it may be necessary to operate the controller in a continuously adaptive mode. Through this the estimated process model is being adapted to changing process characteristics which again causes the control parameters to track the optimal controller settings.

ADCOProcess Variable

Signal Failure

PV_IN

AD_OVRLMN_SEL

Manipulated VariableLMN

LMN_TRK

Figure 6.1: Block diagram for continuous adaptation



i.p.a.s.-systeme

June 2010 of 28

Since the process identification evaluates all process signals and projects these signals into a model, attention should be payed to the fact that signal failures can severely disrupt the information gathered in the process model so far. In the block diagram of figure 6.1 a signal failure turns off the adaptation and also switches the controller to the output track mode. Since the manipulated variable is fed back into the track variable the controller output keeps its last (before the signal failure) valid value.

7. Cascade Control Loop The adaptive state controller can also be applied to cascade control loops. The function block structure in figure 7.1 shows one possible layout.

SP_EXT

ADCO2

LMN PV_IN

ADCO1 PV_IN

LMN_SEL QILCKQAUTO QLMN_SEL

Process-variable 2

AD_OVR

Interlock

Setpoint

QSP EXT

>=1

manualTrack Mode

Process- variable 1

LMN_TRK

LMN Manipulated variable

External Setpoint

Figure 7.1: Function block structure in a cascade loop If the inner controller (ADCO2) is switched to manual mode, to track mode, to interlock mode or to local setpoint then the adaptation of the outer (primary) controller is turned off and the track mode of this controller is activated. The controller output (manipulated variable) of ADCO1 (setpoint for ADCO2) now tracks the process variable of the inner controller. This means that ADCO2 can be switched back to its regular mode without causing a bump. Furthermore the functionality of the complete cascade is re-established by simply adjusting the desired mode of the inner controller.



i.p.a.s.-systeme



If necessary the adaptation feature of ADCO1 (outer controller) has to be re-enabled through a separate user interaction. In a cascade control loop set up with two adaptive state controllers it should never happen that both controllers are operated in the continuously adaptive mode at the same time. This could cause mutual disturbances which most likely diminish the control quality of the cascade loop.

8. Multirange-Controller Besides the regular adaptive controller the software also contains the so-called Multirange-Controller. A special feature of this controller is that it can be subdivided in up to 8 different zones and that these zones can individually be optimized. The switch over between zones can be initiated by the user or any other process event. If the process shows a distinct non-linear behaviour then the process variable range can be split up into e.g. 8 sections. Through this the process can be “linearized” section by section. Since individually optimized control parameter sets are assigned to the different “linearized” zones an essential improvement of the control quality can be achieved. Another conceivable application is the control of batch processes where process characteristics are predictably changing during a production lot. Here the zone specific control parameters can automatically be activated depending on the progress of the corresponding batch. Transition time, deadtime and controller sensitivity are defined only once per Multirange-Controller. They are equally valid for all controller zones. The startup (tuning) procedure for the Multirange-Controller basically coincides with the procedure described in chapter 3. The only difference is that the procedure - except for the definition of the transition time and deadtime - has to be performed for each configured (up to 8) range.

i.p.a.s.-systeme



Inputs: RANGE: The definition of the controller range (1 ... 8) determines which

control parameter set is to be loaded during the operation of the controller.

In-/Outputs: RES_RN: The algorithm of the Multirange-Controller contains 2 ways

to reset or initialize the controller. Either the currently selected zone (RES_RN) or all controller ranges (RESET) can be initialized.

FILL: By means of this block input the tuning parameters of the

actual range are copied to all remaining ranges which are not tuned yet. This block input is defined as a so-called “IN-OUT”-input, i.e. it can be modified like a regular block input as well as through the internal algorithm. This however means that the output of another upstream block must not be connected to this input type.

Outputs: ORI_M1 ... ORI_M8: These outputs indicate whether the corresponding

range (1 ... 8) has already been optimized. QRANGE: This output contains the number (1 ... 8) of the currently

selected and active range.

i.p.a.s.-systeme

June 2010 of 28

9. Tips and Tricks - Basically the learning phase to set up a process model and to optimize

a state controller based on this model can be started at any time. During the first learning phase (i.e. after a new controller is placed into the PCS7-program or after an existing controller has been initialized) the process model estimation should be started in a “nearly static” operating point and should end in a different but also “nearly static” operating point. The reason herefore is that during the transition from a static to a dynamic phase and also during the transition from a dynamic to a static phase the “best process information” can be transferred to the process model (see figure 9.1). A consequent fine tuning optimization (based on an already existing process model) can also be started in the course of a dynamic transition without impairing the resulting control quality.

Figure 9.1: Transition phases with essential information



i.p.a.s.-systeme



- If a controller is operated in the continuously adapting mode then it

makes sense to limit the change rate of the manipulated variable (LMN_DEL). Assumed that no limitation is introduced the manipulated variable can get into a oscillating state if a wrong process model – despite all checks – is conveyed to the optimization procedure (this should happen very rarely, but it can not be guaranteed that it never happens). A high-frequent oscillation can lead to a static process variable which in turn means that no process information can be extracted from the process variable, i.e. the model can no longer be improved and the control algorithm is locked.

The value for a limited change rate of the manipulated variable depends on the dynamic behaviour of the process and on the requirements on the control quality of the loop. A generally valid value can not be indicated.

- If a process contains a significant deadtime characteristic then the

manually entered deadtime value (DTIME) should always be lower as or equal to the real process deadtime. If the indicated deadtime is too low the control quality diminishes very slowly. However if it is too high the quality of the control loop is strongly affected. The reason for that behaviour is still to be examined.

- To establish an optimal control quality an exactly reproduceable and

constant scan time is necessary. This means that the adaptive controller has to be triggered by a cyclic OB (Organization Block). Triggering the controller through OB1 (i.e. in the so-called free cycle or PLC-mode) would impair the control performance. Moreover the basic scan or cycle time of the controller must not be lower than the scan time for external inputs (esp. the process variable). If this is not been followed then the control algorithm (i.e. the process model estimation) is dealing with identical numbers in subsequent scan steps even if the process is in a dynamic transition. This disturbs the process model and reduces the control quality.

- If process characteristics show a defined difference in certain

transitions (e.g. a temperature process where heating up takes more time than cooling down) the faster transition should be the basis for a process model estimation and a subsequent controller optimization. In the example above (provided that cooling down shows faster dynamics) the system should first be heated up with the adaptation turned off. Then the adaptation should be turned on and the system should be cooled down. The resulting controller (with constant tuning parameters) is capable of handling both the heating up and cooling down phase.

i.p.a.s.-systeme



- If the controller is acting too strong on the process, i.e. it produces an oscillating controller output (manipulated variable) and thereby approaches stability limits, the following actions should be taken (in that order): • Reduce the sensitivity factor (SENS) step by step – if necessary

down to -100. • Limit the change velocity (LMN_DEL) of the manipulated variable

(controller output). The value to be entered depends on the process dynamic. As a first guess the value can be adjusted so that a 100% change of the controller output is possible within one oscillation period. E.g. if the oscillation period of the instable or nearly instable control loop is 0.5 min then LMN_DEL can be set to 200.

- If the controller is acting too sluggish on the process then the following

actions can be taken (in that order): • Increase the change velocity (LMN_DEL) of the manipulated

variable (controller output) or set it to 0 to disable the limitation completely.

• Increase the sensitivity factor (SENS) step by step – if necessary up to 150.

i.p.a.s.-systeme



10. Controller Installation 10.1 General Procedure Before the installation please close active tasks. Insert the installation disk in the drive. Start the program "ADCO v7.1 - ADaptiveCOntroller for PCS7.exe”. The Setup-program will then copy all the necessary files into subdirectories of the <Simatic-path>, which are specified by Siemens. 10.2 Installation into STEP7-Environment The Library will be added to the STEP7-Libraries. The Function Blocks should only be edited with STEP7-tools (not with WINDOWS-Tools as some other files will be connected as well). The Library will be automatically included in the CFC Editor, its name is “PCS_7_ADCO_Library_V71”. For further details please see the documentation on ADCO for PCS7. 10.3 Installation into WinCC-Environment The typicals will be copied into the path

<Simatic-Path>\WinCC\options\Pdl\FaceplateDesigner_V6 If you create a new PCS7 project AFTER the installation of ADCO, the faceplates will be copied by PCS7 automatically to the right destination folder. If you install ADCO and you will use it with an existing PCS7 project, you must proceed in the following way : In the OS Projecteditor, check the radio button „Komplettprojektierung“ (see Screenshot). While startup of the OS-Projecteditor, the faceplates will be copied to the current project.

i.p.a.s.-systeme

June 2010 of 28

Figure 10.1: OS-Projecteditor 10.4 De-Installation of the Software For De-Installation of the Software, open Windows-Systemcontrol and start „Applications“. Here you will find „ADCO v7.1 - ADaptiveCOntroller for PCS7“. Pressing the button <Delete> will uninstall ADCO from you system completely.

11. General Directions for Use 11.1 General Directions on AS-Function Blocks The AS-Function Blocks of the PCS7-Libraries are compatible with S7-400, only (not with M7). The Function Blocks will have to be imported from the Library into the respective project in order to use them together with computer programs or graphical projection tools. CFC Application: Simply install the respective Function Block in the diagram by Drag and Drop procedure.



i.p.a.s.-systeme



The Function Block will be automatically imported from the Function Block Container into the Chart Container. For further details please see the paragraph “Importing Function Blocks” in the CFC-Directory. You can start the online-help by selecting the ADCO FB’s and pressing <F1>. The Function Blocks will be provided without any Sourcecode. In addition, the AS-Function Blocks carry the attribute “Know How Protect”. However, default-values and attributes can still be modified by use of the progamming language tools, especially for those Function Blocks with FB/FC numbers. For further details please see the documentation on ADCO for PCS7. 11.2 General Directions on OS-Function Blocks Surface and message texts of the Typicals are in English and German language. It depends on the language configuration in PCS7, which texts might be displayed. For further details please see the documentation on ADCO for PCS7.

12. Faceplates The Faceplates for ADCO and ADMR consist of some more parts. The operation of the faceplates is described in operations manual of CTRL_PID. Please have a look at the documentation of Siemens CTRL_PID.

adaptive controller for siemens pcs7 v7.0/v7 · pdf filepcs7 v7.0/v7.1 : handling operator...

Documents