ADVANCED DECISION MAKING TOOLS IN THE PRODUCTION
COMPANY
=GHND9LGHFNi
1
1
Ing. Zdenka Videcka, Ph.D. Facul y o Business and Managemen . B no Uni e si y o Technology. Technicka
2, 616 69 B no, Czech Republic. E-mail: idecka@ bm. u b .cz
ADVANCED DECISION MAKING TOOLS IN THE
PRODUCTION COMPANY
=GHND9LGHFNi
Abs ac :
Today’s companies collec a la ge amoun o ope a ional da a ela ing o all kinds o
ac i i ies. This da a holds and hides he expe ience o company’s his o y. The impac
o “ lood o da a” pe o m unsa e si ua ion. I managemen use hese da a and don’
disco e mu ual ela ionship o hese da a i can lead o w ong decision. I igh
decision ools doesn’ exis in he ame o companies e y o en is decision based on
he pas expe ience o manage s. I can lead o w ong decision i si ua ion on he
decision a ea is changed. P ope ly analyse o da a can ha e signi ican e ec on a
company’s pe o mance and p o i abili y. This pape desc ibe possibili y o using
ad anced decision ools in day- o-day decision manage s.
Key wo ds: Simula ion, da a mining, ERP, BPR
ADVANCED DECISION MAKING TOOLS IN THE
PRODUCTION COMPANY
=GHND9LGHFNi
1. In odu ion
In o ma ion sys ems o p oduc ion planning and con ol (PPC) we e implemen ed
al eady long ago in almos all indus ial companies. They a e using in day- o-day
p ac ise. P oduc ion Planning and Con ol Sys ems a e ocused o p o ide decision
on ope a ional and ac ical le el.
The economic e olu ion o indus y is a ec ed h ough he implemen a ion o
inno a i e in o ma ion echnology. In o ma ion sys ems o p oduc ion planning and
con ol al eady long ago implemen ed in almos all indus ial companies a e in hei
unc ionali y limi ed. ERP sys ems o e a la ge amoun o asks ha can help o
manage s making decision. Un o una ely he e is no ools o disco e mu ual
ela ionship da a – knowledge inside ERP sys ems. The e is as de elopmen o da a
mining ools in he las i e yea s. They a e used oge he wi h ERP sys ems. This is
oge he , wi h many o he s ad an ages, he eason why is inc easing he usage o
simula ions ools, which could do nea ly unlimi ed numbe o es s and expe imen s in
wide scope. Wi h all his so wa e ools doesn’ need o be only specialized bu also
hey can be wide uni e sal applicable and hei usage need no only special
knowledge and high educa ion. The main aim o his connec ion is disco e
knowledge in a la ge amoun o da a. Decision suppo sys em based on knowledge
sys ems is called business in elligen sys ems.
2. P edic i e Tools
Decision Suppo Sys ems (DSS) use di e en kinds o echniques and models which
help o igh decision. Gene al de ini ion o decision suppo sys em doesn’ exis .
“Decision Suppo Sys ems a e in e ac i e compu e sys ems ha use da a and
models o solu ion nons uc u ed p oblems o help o decision subjec s” Sco Mo on
(1971). E ec i e ool is “wha -i ” analysis. One o Decision Suppo Sys ems is
disc e e e en simula ion. The p inciple o simula ion is expe imen wi h a compu e
model – an exac image o s ochas ic disc e e sys em. The e is allowed dynamic
ep esen a ion o ma e ial o cus ome low, simula ion expe imen s a e, p ocesses,
machine using, e c. A use can moni o dynamic p ocess. I means ha he has
necessa y in o ma ion o decision and sys em e alua ion commensu a e wi h cosen
c i e ias. He ecei e answe o ques ion: “Wha is he ma e i I change (se up,
in es , …) …?”. The sys em beha iou is moni o ed h ough epo s and s a is ics.
P edic ion o sys em beha iou is e icien ool in p ojec phases, eenginee ing o
p ocesses and on-line simula ion nowadays. The e is possible ind an op imal se up
o pa ame e s o he bes solu ion (minimum o maximum o objec i e unc ion). On
one hand i is possible o ind op imal solu ion h ough simula ion on he o he hand
he e is no possible ind ou ela ionship be ween di e en pa ame e s o a scope o
he bes solu ion o decision suppo . The e is no possible o mine knowledge o
models sa ed in da a ob ained du ing expe imen s wi h compu e model.
The powe ul ool o a la ge amoun da a analysis is da a mining which help o
disco e ela ionship o da a and hidden ules. The e is no uni ied de ini ion o da a
mining. I is possible o say ha da a mining pe mi disco e hiden ules and
ela ionship be ween da a in wide da abases. The e is pa adox: i we ha e mo e da a
he e is di icul and ime consuming analyse hem wi h adi ional me hods. The e is
possible as and ully isola e aluable and use ul in o ma ion h ough analy ical
echnologies o da a mining. I is s a e when models and expe ience can be ecie ed
om da a.
Simula ion ools
His o ically, simula ions a e de eloped o -line using cus om so wa e
packages/languages wi h limi ed di ec connec ions o he ac ual da a gene a ed by
he p oduc ion sys em (D ake and Smi h 1996) [4] This adi ional simula ion
gene ally examines long- e m sys em pe o mance, mos ly o planning and design
pu poses. These models a e usually “one-sho models” because hey a e seldom
used a e he p ojec is inalized (i - hen mos ly only o checking he sys em abili y
unde new p ojec condi ions). P ima y easons o his in lexibili y a e ha he inpu
da a o he simula ion a e collec ed and analyzed ou side he simula ion model and
simula ion en i onmen and ha he simula ion sys em canno communica e
au oma ically wi h he in o ma ion sys em - En ep ise Resou ce Sys em (ERP) ha is
esponsible o collec ing, adminis a ion and dis ibu ion o s a us in o ma ion. On-
line simula ion in eg a es he in o ma ion sys em wi h he simula ion model. On-line
simula ion o p ocess scheduling, eal- ime “in elligen ” con ol, pe o mance
o ecas ing, p ocess capabili y es ima ion, eal- ime con ol sys ems emula ion, eal-
ime displays o sys em s a us, and sho e m decision making is nowadays ac i e
a ea o so wa e de elopmen . By using he mos cu en in o ma ion sys em,
accu a e p edic ions abou he sys em and u u e con ol al e na i es canno be
de eloped. The eason o using simula ion is ha simula ion can be e cap u e and
desc ibe he complex in e ac ions whe e analy ical me hods ail. Cu en simula ion
so wa e ools a e al eady able o communica e wi h da abases – bu he ull
in eg a ion is mo e o less s ill u u e o see p obably only in au omo i e indus y by
using he concep o digi al ac o y. In his way s ays he combina ion o simula ion,
in o ma ion sys em and eal- ime con ol e y p omising amewo k o op imiza ion o
dynamic cha ac e is ic in lexible p oduc ion sys ems.
Gene ally he simula ion models can be buil by gene al p og amming language as
C++. Bu his way needs “special” p og amming knowledge. People able o build
simula ion model in his way a e specialized and ha en’ un o una ely on he o he
hand o en p ac ise expe iences wi h he eal sys em. This ac leads o building o
bigge eams dealing wi h simula ion o o buying simula ion so wa e ool.
Cu en ly he e a e se e al so wa e ools a ailable. Wi h he i s al eady men ioned
g oup hese ools can be di ided in o ee basic classes:
− Gene al pu pose simula ion language – weakness o his g oup is ha he use
has o be p o essional p og amme and compe en simula ionis in one.
− Simula ion on -end – essen ially p og ams be ween use and simula ion
language.
− Simula o s – so wa e packages u ilize cons uc and e minology common o he
businesses communi y (manu ac u ing, se ices…) and o e g aphical
p esen a ion and anima ion.
Las g oup, in case o disc e e simula ions, is e y well usable o simula ion o
p oduc ion sys ems. Building he model is ela i ely easy and needs ne e y speci ic
knowledge o p og amming. Simula ion expe s can educe he ime needed o

building he simula ion model. Cu en ly is possible o buy many o hese packages
wi h ollowing common cha ac e is ic (needed o be success ul on ma ke ):
− Rela i ely good execu ion speed.
− In e aces o da a inpu /ou pu ( om ex e nal da abases).
− Easy syn ax.
− Random a iable gene a o s o o en-used dis ibu ions.
− G aphical en i ies (elemen s) ep esen ing unc ion o eal sys em objec :
 P oduc ion machines.
 Assembly machines.
 Mul iple machines.
 Con eye s.
 S o ages.
 Pa s.
 e c.
− In e ace wi h gene al used language – possibili y o add own p og ammes
− Possibili y o impo g aphical objec s ( o layou s o elemen s design)
− On-line anima ion and s a us indica ion
− Op imisa ion
− Good building help and documen a ion.
These a e he main ea u es cha ac e ising p esen so wa e packages. Bu no each
o p esen so wa e ools has all o hese. You can ind eally good ools bu speci ied
on only one ype o p oduc ion as o example on low shop p oduc ion. One o his
specialised so wa e is Simp o. This is used in au omo i e indus y and i s elemen s
could be eally good used o his kind o p oduc ion. So i is easy o simula e special
kinds o con eye s. Disad an age is bu he di icul y o in o ma ion low by models
ha should simula e o he kinds o p oduc ion as o example mass p oduc ion. In his
case you ha e o use mo e ic i e elemen s and so ha e p oblems wi h s a is ic
documen a ion by in e p e a ion.
3. The ela ionship o simula ion and da a mining
Da a mining me hods a e used on he ame o eal da a s o ed in da abases o
En e p ise Resou ce Planning (ERP) sys ems, CRM sys ems and wa enhouses.
The e a e implemen ed analy ical me hods om he classical s a is ical me hods,
neu al ne wo ks, decision ee, e c. Excep hese sys ems he special so wa e
p oduc s o da a mining a e used o da a mining p ojec s.
App op ia e solu ion o using hidden knowledge om expe imen al modelling da a is
connec ion o disc e e e en simula ion and da a mining. The ou pu o simula ion
(see Fig.1) a e pa ame e s’ se up o he op imized sys em and consequen ial mining
o knowledge hidden in expe imen al da a. I allows de ine app op ia e manage ial
sys em model .
Figu e1: Rela ionship o disc e e e en simula ion and da a mining
4. Case s udy o da a mining using in p ac ise
Case s udy is ocused o using o disc e e e en simula ion and model op imising o
manu ac u ing p ocess. The esul o op imalisa ion is ile o expe imen al da a. The
da a a e used in da a mining p ojec . The aim is knowledge ha a e neccesa y o
inc easing p oduc i i y o manu ac u ing p ocess.
Scena y is simula ed h ough compu e simula ion. The p ocess s a wi h cus ome
equi emen s, manu ac u ing include building and es ing o compu e s.
Compu e Simula ion
Sys em Op imisa ion
Da a Mining
Se up o Sys em Pa ame e s
Using o Knowledge o a Sys em
Simula ion model
The inpu o model a e cus ome s’ equi emen s. The speci ica ion is cla i ied h ough
phone communica ion. The compu e s a e assembled, es ed and dis ibu ed (see
ig.2). They a e wo skills labo s (A and B).
Figu e 2: The model o manu ac u ing sys em
The op imalisa ion is aimed o maximize o p o i . The pa ame e s o expe imen a e
numbe o A skill labo s, B skills labo s, numbe o assembly and es ing wo kplace.
The bes esul s we e eached using Adap i e The mos a is ical SA me hod. The
op imal esul is 11 758 money uni s. I was ound a e 65 expe imen al uns.
Expe imen al da a we e expo ed o Mic oso Excel. File o expe imen al da a is
sou ce ile o da a mining p ojec (see ig. 3).
Figu e3: S uc u e o p ojec in Wi ness Mine
The esul o da a mining p ojec (see ig.4) is decision suppo o maximum o p o i
and ecommenda ion o pa ame e s’ se up. I means numbe o enginee s highe
hen 2, numbe o es ing wo kplaces highe han 4 and numbe s o B skilled labo s
highe han 4. This solu ion co espond o aim o case s udy – elimina ion o labo
cos s and he o he sou ces o maximize p o i o compu e building.
Figu e 5: Knowledge disco e y and da a mining esul