latest forecasting methods applied to lean production systems
DESCRIPTION
TRANSCRIPT
Latest forecasting methods applied to lean
production systems
José Fragozo
Dept. Of Industrial Engineering, Universidad Del Norte
Barranquilla, COLOMBIA
Abstract- Nowadays technology increases in an
exponential rate that is only overcome by our
ambition of keep growing, every single day
appears a new product, that is designed taking in
account the life cycle and obsolescence time, this
happens specially with the information technology
companies (ITC) that in our days are the most
powerful giants of world`s industry in the other
hand we are using too much energy than the
planet is capable to provide in an undefined period
of time for that reason alternative energy is
earning value, this paper focuses on two new
forecasting methods, the first one help the
forecasting for parts on a technology supply chain,
and the second one help us to forecast the wind
speed in the energy industry in order to increase
the efficient of wind energy technology.
I. INTRODUCTION
Having the knowledge of the limited classic literature
that is available and useful for this kind of forecasting
it was necessary to entrepreneur in this new field
using the statistics tools that are around us, we have
heard that the universe has an equilibrium equation,
natural events can be modeled in mathematics
models, in the same way market behavior can be
modeled too, using statistical tools like Bayes
theorem and Weibull probability distribution there are
new advances in this areas we are going to review
two papers, the first one “Bayesian forecasting of
parts demand” published by Elsevier B.V where
applies Bayes theorem in the forecast of demand
of technology parts and the second one
“Development of wind speed forecasting Model
Based on the Weibull Probability Distribution”
published by Ruigang Wang, Wenyi Li and B.
Bagen where use Weibull probability distribution
models to forecast the wind speed.
II. BAYESIAN FORECASTING OF PARTS
DEMAND
Manufacturing of high technology products like
computer is an exacting business were the supply
chain as we as industrial engineers know must be
synchronize optimizing the information and materials
flow, no all the times this job is easy, in this particular
case the demand of computer parts is a very complex,
because no matter if computers is an exacting
business, computers parts business is very complex
because it depends of the life cycle of the part and the
obsolescence of the part in this case Sun
Microsystems Inc is a vendor of computer products
that it is fettered to the supply chain
In order to have an idea of the type of market
behavior that we they are dealing with it appear in the
next graphics.
Fig. 1. Demands
Where the solid lines present de demands, the
horizontal axe represent the financial planning
periods of roughly one month’s duration, we can see
how demand`s behavior depends of the life cycle and
obsolescence, short life cycles means that the
individuals parts frequently do not have sufficient
observed demand values to support reliable
extrapolation, Bayesian model take in account
predictive conditionals, life cycle curves, uncorrelated
errors, auto correlated errors, scale factors, priors
parameters, distributions of parts demand, estimation,
the Bayesian model requires an investment of
$10.000 USD due that is an heuristic algorithm of 482
forecast each one has 4000 iterations in the Gibbs
Sampler so are required 16 computers processors, this
model describes the behavior of the demand better
than classic forecast methods, its limited for this kind
of demands that depends of life cycles and
obsolescence, it has a investment but the forecasting
is a vital tool in the planning so for giants vendors
like in this case this new forecasting method is a very
good option.
III. DEVELOPMENT OF WIND SPEED
FORECASTING MODEL BASED ON
THE WEIBULL PROBABILITY
DISTRIBUTION
In a unsustainable world like our world where oil
provides us with the major percent of our energy
demands, worlds population is around 6.000.000 and
in 2050 it is forecasted that will be around 9.000.000,
we consume more energy than the energy that the
planet is able to provide, so in this point of time is
essential to look alternative ways to produce energy,
since many years ago those alternative methods exist
but are far away to be compared with the oil energy,
is too less efficient and is more expensive, so oil
energy still being the best option, taking in account
that oil is a non renewable resource, alternative
methods needs to be improved, in this paper develop
a forecasting method to forecast the wind´s speed,
wind energy is a variable energy source that, the
power output of a wind turbine generator (WTG) unit
fluctuates with the wind speed variations, existing
forecasting methods presents significant errors in the
forecast what make no reliable to analyze power
networks impact, so in this paper present an improved
probability method based on Weibull distribution,
with two parameters Weibull fit with the actual wind
speed perfectly. Although there are only two
parameters on Weibull distribution, the wind model is
very sensitive two those parameters, so if the
parameters are designed with the proper accuracy the
wind speed forecasting model can represent the actual
speed variation.
Fig. 2. Probabilities density function
This paper improves existing Weibull method
combining the mean wind speed and standard
deviation method with the maximum likehood
method, and wind speed is modeled as a random
variable with a Weibull distribution.
It also compare time series methods like AR(p) and
MA(q) with the new method that is proposed, the
accuracy of this method is significant better than the
other ones, obviously every forecast includes a
natural errors, the wind variation change the behavior
depending of annual period, season period and diurnal
period, we can see the comparison of the methods in
each period in the next table.
Chart 1. Comparison of methods
Improved method has smaller errors in each period,
what means that is describing and forecasting the
winds speed variation better than the other methods,
accurate wind speed forecasting are necessaries in the
network energy planning on a wind energy station so
this new methods are very useful in the industry.
IV. CONCLUSIONS
Forecasting methods need to be developed in the
same rate and time that the new markets behavior is
appearing, classic methods are good backgrounds
when a forecasting its necessary but are not useful in
a lot of cases where the behavior of the data is
particular of the case like in this paper cases,
knowledge is a continuos in the universe, several
times the change resistance difficult the
implementation of new methods that are better in the
majority of the cases, in this two reviews the new
forecasting methods will help in the company
evolution, will save a lot of money, will increase the
utilities, will optimize alternative energies, will help
to save the world etc.
V. ACKNOWLEDGMENTS
This paper was supported by “Universidad Del
Norte”, Ing. Daniel Romero and Ing. Carlos Paternina
that provides us with the knowledge in classic
forecasting methods and always emphasized us to
investigate in the new methods.
VI. REFERENCES
[1] Phillip M. Yelland, Bayesian forecasting of parts demand,
International Journal of [1] Forecasting, Volume 26, Issue 2, Special
Issue: Bayesian Forecasting in Economics, April-June 2010, Pages 374-
396, ISSN 0169-2070, DOI: 10.1016/j.ijforecast.2009.11.001.
[2] Ruigang Wang; Wenyi Li; Bagen, B.; , "Development of Wind Speed
Forecasting Model Based on the Weibull Probability Distribution,"
Computer Distributed Control and Intelligent Environmental Monitoring
(CDCIEM), 2011 International Conference on , vol., no., pp.2062-2065,
19-20 Feb. 2011
doi: 10.1109/CDCIEM.2011.333