TECHNICAL NOTE 17 SIMULATION

Review and Discussion Questions

1. Why is simulation often called a technique of last resort? Simulation is called a technique of last resort because simulation models are time consuming to build (flow charting, coding, etc.) and do not guarantee an o timal solution or indeed any solution. !herefore, it ma"es sense to in#estigate other roblem sol#ing methods such as linear rogramming or waiting line theory before embar"ing on simulation. $. What roles do statistical hy othesis testing lay in simulation? % simulation can be loo"ed u on as a test of a hy othesis. &. What determines whether a simulation model is #alid? !he only true test of a simulation is how well the real system erforms after the results of the simulation ha#e been im lemented. !he roof of the udding is in the eating. 'rior to this stage of a lication, howe#er, the simulation user can certainly e#aluate the general #alidity of the model by com aring its results with ast data or sim ly as"ing the question( )s the information ) am getting reasonable? *. +ust you use a com uter to get good information from a simulation? ,- lain. % com uter is a must for any but the most sim le simulation roblems. .ecause simulation is a sam ling rocess, it stands to reason that a large number of obser#ations is desirable, and the com uter is the only ractical way of ro#iding them. /f course, com uteri0ation is no guarantee of good information. Simulating an in#alid model on the com uter will only ro#ide a larger #olume of questionable data. 1. What methods are used to increment time in a simulation model? 2ow do they wor"? !ime incrementing methods include fi-ed time increments and #ariable time increments. With fi-ed time increments, uniform cloc" times are s ecified (minutes, hours, days, etc.) and the simulation roceeds by fi-ed inter#als from one time eriod to the ne-t. %t each oint in cloc" time, the system is scanned to determine if any e#ents ha#e occurred and time is ad#anced3 if none ha#e, time is still ad#anced by one unit. With #ariable time increments, the cloc" time is ad#anced by the amount required to initiate the ne-t e#ent. )t is interesting to note that #ariable time incrementing generally is more difficult to rogram unless one is using a s ecial simulation language such as 4'SS.

$*5

!echnical 6ote 17

8. What are the ros and cons of starting a simulation with the system em ty? With the system in equilibrium? !he ros of starting a simulation with the system em ty are that this enables e#aluation of the transient eriod in terms of time to reach steady9state and the acti#ities which are eculiar to the transient eriod. /ne con is that it ta"es a longer eriod of time to erform the simulation. % second is that the model will be biased by the set of initial #alues selected, since the time to achie#e steady9 state and the acti#ities which ta"e lace during the transient eriod will be affected by the initial #alues. Ste s must be ta"en to remo#e these initial #alues if steady9state results are needed. !he ad#antages of starting the system in equilibrium are that the run time may be greatly reduced, and that the aforementioned bias may be eliminated. !he disad#antage of starting the simulation in equilibrium is, in essence, that it assumes that the analyst has some idea of the range of out ut he is loo"ing for. !his, in a sense, constitutes beating the model and may lead to incorrect conclusions from the simulation run. 7. :istinguish between "nown mathematical distributions and em irical distributions. What information is needed to simulate using a "nown mathematical distribution? % "nown mathematical distribution is one that can be generated mathematically and is amenable to the laws of statistical robability. ,-am les of such distributions are the normal, binomial, 'oisson, 4amma, and hy ergeometric. %n em irical distribution is one that is obtained from obser#ing the robability of occurrence of henomena relating to a s ecific situation. While it may be ossible to define the moment generating function for such distributions, their a licability to other situations is li"ely to be small. !he information required to simulate using a "now distribution, of course, de ends on the "nown distribution selected. 4enerally s ea"ing, howe#er, at a minimum, the analyst must be able to estimate the mean and standard de#iation of the o ulation to be sam led since they are arameters of distributions. ;. What is the im ortance of run length in simulation? )s a run of 1<< obser#ations twice as #alid as a run of 1<? ,- lain. !he length of the simulation run generally determines how much confidence one may ha#e in the results generated by the model. Since simulation is in a large measure a sam ling rocess, it stands to reason that the larger the sam le (i.e., the longer the simulation run) the more re resentati#e the findings. /ne the other hand, the selection of #ariables and arameters for a simulation study re resent ma=or sources of otential error and sim ly running a simulation for a long eriod of time will not o#ercome inaccuracies introduced through these factors. % run of 1<< may be twice as good as a run of 1<, it may be 1< times as good, or may sim ly be no better. )t de ends on the si0e of the model, the nature of the distributions selected, the stream of random numbers used, and the ob=ecti#e of the study.

$1<

Simulation

P o!"e#s
!y e of 'roblem @ariable ?i-ed time time Simulation increment increment design Aes Aes Aes Aes Aes Aes Aes Aes Aes Aes Aes Aes

'roblem 1 $ & * 1 8 7 ; 5 1< 11 1$

:ifficulty +oderate ,asy +oderate +oderate +oderate +oderate :ifficult ,asy +oderate :ifficult :ifficult +oderate

6ew 'roblem

+odified 'roblem

>hec" figure in % endi- % Aes

Aes

Aes

1. !he following decision rules were used for this simulation. 1. !otal orders laced is equal to demand in the ast three eriods (for eriods immediately before the start of the game, ;<< cases was used. $. 1B of the total orders laced were flash fro0en e-ce t for the two wee"s before +ardi 4ras. 6ote that solutions will change with different decision rules and with different sets of random numbers. )t may be interesting to see what decision rules students used as well as they rofitability.

!o determine the amount recei#ed the following random number assignment table was used(
>hange 9 1<B <B D 1<B 'robability .&< .1< .$< Candom number <<9$5 &<975 ;<955

!wo digit random numbers are then drawn from a uniformly distributed random number table such as % endi- .. ?or the demand, the equation : E ;<< D 1<< is used, with selected from a normally distributed random digits tables such as % endi- >.

$11

!echnical 6ote 17

!he following table illustrates one set of numbers obtained for the change in orders recei#ed and the demand(
Wee" 1 $ & * 1 8 7 ; 5 1< Candom number ;< *7 15 &$ *7 11 ** ;$ ;; ;5 >hange(B) D1< < < < < 91< < D1< D1< D1< F 9<.$&;1$ <.1;1$* 91.&1;;$ 91.*1*<$ 9<.$;1;1 1.18111 9<.$;$7; 1.8**1 <.&$*8; <.&*1<8 :emand (:) 778 ;1; 88* 811 77$ 517 77$ 58* ;&$ ;&1

Wee"

?lash fro0en in#entory

/rders laced ?lash Cegular fro0en

/rders recei#ed ?lash Cegular fro0en

%#ailable (regular and flash fro0en)

:emand (;<< D 1<<F)

Sales (minimum of demand or a#ailable)

,-cess Cegular ?lash Shortages

1 $ & * 1 8 7 ; 5 1<
total

< ** 1; 1; 58 &8 < < < *; &<<

78< *< ;&8 ** 71$ *< 71$ *< 71; *< 71; *< 711 &; 711 &; 878 &8 878 &8 88$ &1 158 &$ 999999999+%C:) 4C%S999999999 775 *1 ;17 *1 ;*< ** 5$* *; ;1& *& ;5* *7 71<; &7<

;;< ;&8 ;18 ;11 ;<; 88* < 5<$ 57$ 5;5

778 ;1; 88* 811 77$ 517 77$ 58* ;&$ ;&1

778 ;1; 88* 811 77$ 88* < 5<$ ;&$ ;&1 851;

8<* < 5* 8< < < < < 5$ 15 &81

** 1; 1; 58 &8 < < < *; 51 &51

< < < < < $1& 77$ 8$ < < 1<;7

'rofits Ce#enue from sales (G1< - col. 7) Ce#enue from sal#age (G* - col. ; reg.) !otal Ce#enue >ost of regular urchases (G&< - col. * reg.) >ost of flash9fro0en urchases (G&* - col.* flash) >ost of holding flash Hfro0en (G$ - col.; flash) >ost of shortages (G$< - col. 5) !otal cost 'rofit G&*1,5<< G1,*8< G&*7,&8< G$1&,$*< G1$,1;< G75< G$1,7*< G$*;,&1< G55,<1<

$1$

Simulation

$.
!ime between arri#als (minutes) 1 $ & * 1 'robability .<; .&1 .&* .17 .<8 C6 assignment <<9<7 <;9*$ *&978 7795& 5*955

Ser#ice !ime (minutes) 1.< 1.1 $.< $.1 &.< &.1

'robability .1$ .$1 .&8 .15 .<7 .<1

C6 assignment <<911 1$9&$ &&98; 859;7 ;;95* 51955

>ustomer number 1 $ & * 1 8

C6 <; $* *1 &1 *1 1<

)nterarri#al time $ $ & $ & $

%rri#al time $ * 7 5 1$ 1*

Ser#ice begins $.< *.1 7.< 5.1 1$.< 1*.<

C6 7* &* ;8 &$ $1 87

Ser#ice time $.1 $.< $.1 1.1 1.1 $.<

Ser#ice ends *.1 8.1 5.1 11.< 1&.1 18.<

Waiting time <.< <.1 <.< <.1 <.< <.<

)dle time $.< <.< <.1 <.< 1.< <.1

%#erage waiting time E 1.<I8 E 1I8 minute, and a#erage teller idle time between customers E *.<I8 E *I8 minute. %#erage teller idle time ercent E *.<I18 E .$1 or $1B.

&. / tion a(
Simulation 1 $ C6 7 < !ime9 dist. 1 1.1 1.< C6 & 1 !ime9 dist. & 1.1 $.< C6 < & 'ay9 dist. * G$Icrew G&Icrew !ime9 C6 dist. 1 * *.< 1 *.< %#erage !otal time 7.< 7.< 7.< !otal ay G8.<< G5.<< G7.1<

/ tion b(
Simulation 1 $ C6 8 8 !ime9 dist. 1 1.1 1.1 C6 1 8 !ime9 dist. 8 1.< 8.< %#erage !otal time 8.1 7.1 7.<

$1&

!echnical 6ote 17

/ tion c( We need two more distributions and random number assignments(

!> at air ort? Aes 6o 'robability .1< .5< C6 < 195

!> refuses G1<? Aes 6o

'robability .&< .7<

C6 <9$ &95

Simulation 1 $

C6 * 7

!ime9 dist. $ 1.< 1.<

C 6 ; 7

!> at air ort? 6o 6o

C6 * 1

!ime9 dist. ; &.< $.<

C6 5 7

!> refuses G1< 6o 6o

C !ime9 6 dist. 7 < 1.< < 1.< %#erage

!otal !ime 1.< *.< *.1

Summary(
/ tion a b c %#erage tri time 7.< 7.< *.1

*.
+achine brea"down number 1 $ & * 1 )nterarri#al time (hours) 1.< <.1 1.< <.1 &.< .rea"down time (hours) 1.< 1.1 $.1 &.< 8.< Ser#ice time (hours) &.< *.< <.1 1.< &.< Ce airman 1 begin 1.< *.< *.1 8.< end *.< *.1 1.1 5.< Ce airman $ begin 1.1 end 1.1 :own time (hours) &.< *.< $.< $.1 &.<

C6 &< <$ 11 $; ;8

C6 ;1 51 <; ** ;*

%#erage down time is 1*.1I1 E $.5 hours

$1*

Simulation

1. %nswers will #ary based u on the random numbers selected.

)nterarri#al time (min). 1 $ & * 1 8 7 ; 5 1< 'robability .<1 .1< .1< .11 .11 .$< .1< .<; .<1 .<$ C6 <<9<* <191* 119$* $19&5 *<91* 1197* 719;* ;195$ 5&957 5;955

Ser#ice time (min). 1 $ & * 1 8 7 ; 5 1<

'robability .1< .11 .11 .$< .11 .1< .<; .<* .<$ .<1

C6 <<9<5 1<9$* $19&5 *<915 8<97* 719;* ;195$ 5&958 5795; 55

>ustomer number 1 $ & * 1 8 7 ; 5 1<

C6 18 ;& 11 *7 ;* <; &8 <1 $8 *$

)nterarri#al time (min.) 8 7 8 1 7 $ * $ * 1

%rri#al time (min.) 8 1& 15 $* &1 && &7 &5 *& *;

Ser#ice begins 8 1* $$ $7 &1 && &7 *1 ** *5

C6 51 51 88 17 <& $1 17 &1 85 5<

Ser#ice time (min.) ; ; 1 $ 1 $ * & 1 7

Ser#ice ends 1* $$ $7 $5 &$ &1 *1 ** *5 18

!ime in system (min.) ; 5 ; 1 1 $ * 1 8 ;

%#erage time in system E 18I1< or 1.8 minutes.

$11

!echnical 6ote 17

8. %nswers will #ary de ending u on whether minor or ma=or in=uries are calculated first, and the assignment of C6 to ma=or in=uries.
6umber of minor in=uries < 1 $ & * 1 'robability .$<< .1<< .$$< .<1< .<$1 .<<1 C6 <<<9155 $<<9855 7<<9515 5$<9585 57<955* 5519555

+a=or in=ury? Aes 6o

'robability .<1< .51<

C6 <<<9<*5 <1<9555

4ame number 1 $ & * 1 8 7 ; 5 1<

.eginning number of C.s 8 * * $ * & & * * &

C6 <** ;5; 5;8 11; 177 &<1 <$* ;7; $;1 ;8$

+a=or in=ury? Aes 6o 6o 6o 6o 6o Aes 6o 6o 6o

C6 &5$ 811 515 &1& ;88 ;1& 1;5 <$& **$ ;*;

6umber of minor in=uries 1 1 & 1 $ $ < < 1 $

,nding number of C.s * & 1 1 $ 1 $ * & 1

7.
!ime between brea"downs (minutes) * 1 8 7 ; 5 'robability .1< .&< .$1 .$< .1< .<1 C6 assignment <<9<5 1<9&5 *<98* 819;* ;195* 51955

Ser#ice !ime (minutes) * 1 8 7 ; 5

'robability .1< .*< .$< .11 .1< .<1

C6 assignment <<9<5 1<9*5 1<985 7<9;* ;195* 51955

$18

Simulation

>ondition 1
.rea"do wn number C6 !ime between brea"down s .rea"down time C6 Ser#ice time Start end Ce airma +achine :own9 n idle wait time time time

>ondition $
1 $

Start

,nd

Start ,nd

)dle )dle :owntim time H time H e re airm re airm an 1 an $

1 $ & * 1 8 7 ; 5 1< 11 1$ 1& 1* 11 18 17 1; 15 $< $1 $$ $& $* $1

;1 18 81 78 5& 55 81 7< 1; ** $ ;1 1 57 8& 1$ 1& 11 8$ $; ;* ;$ $7 $< &5

; 1 7 7 ; 5 7 7 8 8 * ; * 5 8 8 8 1 8 1 7 7 1 1 1 7 1 1 8 7

; 1& $< $7 &1 ** 11 1; 8* 7< 7* ;$ ;8 51 1<1 1<7 11& 11; 1$* 1$5 1&8 1*& 1*; 11& 11; 181 17< 171 1;1 1;;

8; $8 ;1 11 18 $8 51 87 57 7& 71 8* $8 *1 1 ;7 $< 1 15 &8 85 ;5 ;1 ;1 $ 1 1< 11 $* &8

8 1 ; 1 1 1 5 8 5 7 7 8 1 1 * ; 1 * 1 1 8 ; 7 7 * * 1 8 1 1

; 1* $< $; &1 ** 11 8< 88 71 ;$ ;5 51 1<< 1<1 1<5 117 1$$ 1$8 1&1 1&8 1*& 111

1* 15 $; && *< *5 8< 88 71 ;$ ;5 51 1<< 1<1 1<5 117 1$$ 1$8 1&1 1&8 1*$ 111 11;

; < 1 < $ * $ < < < < < < < < < < < < < < 1 < < < < < < < < 1;

< 1 < 1 < < < $ $ 1 ; 7 5 1 * $ * * $ $ < < & 1 7 * & & & 1 ;7

8 8 ; 8 1 1 5 ; 11 1$ 11 1& 1* 1< ; 1< 5 ; 7 7 8 ; 1< 1$ 11 ; ; 5 ; 8 $8&

; $< &1 ** 11 8* 7* ;$

1* 1& $; $7 *< *5 8< 1; 7& 7< ;1 ;; ;8 51 77 8* &$ 1;

; 1& 8 5 7 * $ $8 * 8 1 1 5 7 1 $ 11& 11 ; $$ & $ < $ 1 1*; 11 1 &< $ 11; 18 $ &

51 1<< 1<1 1<1 1<7 111

8 1 ; 1 1 1 5 8 5 7 7 8 1 1 * ; 1 * 1 1 8 ; 7 7 *

11; 1$* 1$5 1&8 1*&

1$$ 1$5 1&* 1*$ 111

11; 181 181 185 185 17& 17; 1;* 1;5 17& 17; 1;* 1;5 15*

11& 18<

$8 7< $7 $8 $; $1 $5 *1 &< ;1 !otal

181 17< 171 1;1 1;;

185 171 1;1 1;8 15&

1 * 1 1 < 8 < 1 $ &1 1 81 1*5 178

$17

!echnical 6ote 17

b.
Ce airman idle time :owntime :elay time >ost H downtime >ost H ser#ice !otal cost >ondition 1 1; minutes $8& minutes ;7 minutes ($8&I8<)JG*<) E G171.&& (15*I8<)JG1$ E G&;.;< G$1*.1& >ondition $ $1< minutes 178 minutes < minutes (178I8<)JG*< E G117.&& $J(15&I8<)JG1$ E G77.$< G15*.1&

;.
)nterarri#al time (min). 1< $< &< *< 'robability .*< .&1 .$< .<1 C6 <<9&5 *<97* 7195* 51955

)nterarri#al time (min). 1 1< 11 $<

'robability .*1 .&< .$< .<1

C6 <<9** *197* 7195* 51955

>ustomer number 1 $ & *

C6 55 7& &; 7$

)nterarri#al time (min.) *< $< 1< $<

%rri#al time (min.) *< 8< 7< 5<

Ser#ice begins *< 8< 7< 5<

C6 << <5 1& 51

Ser#ice time (min.) 1 1 1< 11

Ser#ice ends *1 81 ;< 1<1

!ime in system (min.) 1 1 1< 11

6o cars go to another station. !he a#erage time s ent at the station is &1I* E ;.71 minutes

$1;

Simulation

5.

(1) >riteria for ma"ing recommendation( a. a#erage line length under different chec"9out minimums b. customer comments on different minimums c. cost of em loying different numbers of chec"ers d. com etiti#e ractices and regional differences in sho er urchases

e. desired holding time in line (to e- ose customers to im ulse items) f. range of other si0es to be considered (remember, the a#erage store has nine stands, others may ha#e more or less)

($) %dditional data to setu simulation( a. arri#al atterns and rates to counters b. ser#ice atterns and rates c. seasonal #ariation in store atronage d. distribution of indi#idual customer urchases, etc. (&) 'reliminary data gathering( a. wor" sam ling or estimates by store management b. model one store for ilot analysis, etc. (*) 'roblem setu for simulation( .asically follow the general rocedure listed in the >ha ter, "ee ing in mind what ty e of out ut data you are loo"ing for. 2ere, we might suggest de#elo ing a tradeoff table that indicates( a. the a#erage line lengths for different chec"out minimums b. the ser#ice cost and im uted waiting costs for customers associated with items in a. (1) ?actors affecting a licability to all stores( a. stoc"ing and sales atterns b. cultural differences (>anada #s. K.S., stores in large cities #s. small cities, etc.) c. com etition faced by different stores d. si0e of stores themsel#es

$15

!echnical 6ote 17

1<. ?irst begin by using by simulation the rocess sequence and associated rocess times. !hen schedule as a static sho using ?>?S. %nswers will #ary based u on the random numbers selected and the sequence in which rocess sequence is determined. Sequence determination(
Starting rocess ?rame ,ngine 'ainting 'robability .1 .& .$ C6 <9* 197 ;95

?rom ?rame to ,ngine 'ainting /ut

'robability .* .* .$

C6 <9& *97 ;95

?rom ,ngine to ?rame 'ainting /ut

'robability .& .* .&

C6 <9$ &98 795

?rom 'ainting to ?rame ,ngine /ut

'robability .1 .1 .;

C6 < 1 $95

'rocessing times(
?rame H 'rocessing time (hours) 1 $ & 'robability .$ .* .* C6 <91 $91 895

,ngine H 'rocessing time (hours) 1 $ &

'robability .8 .1 .&

C6 <91 8 795

'ainting H 'rocessing time (hours) 1 $ &

'robability .& .& .*

C6 <9$ &91 895

$8<

Simulation

>ycle 6umber C6 'rocess 1 1 ,ngine 1 'aint ; /ut $ < ?rame < ,ngine * 'aint 5 /ut & 8 ,ngine < ?rame 1 'aint 8 /ut * 5 'aint 1 ,ngine 8 'aint 8 /ut 1 & ?rame 7 'aint ; /ut

C6 ; * & $ 5 1 $ & & 8 5 1 *

'rocess time & $ $ 1 & 1 $ $ $ $ & $ $

4antt Load >hart

Ti#e P ocess
Frame
2

2
5

4
3

10

12

14

16

Engine

Paint

)t a ears that the 1 cycles can be com leted within the *< hours allotted. 2owe#er, additional re lications should be conducted to confirm this answer.

$81

!echnical 6ote 17

11.
!ime between arri#als (minutes) 1 $ & * 'robability .*< .&< .11 .11 C6 <<9&5 *<985 7<9;* ;1955

Ser#ice time (minutes) 1 $ &

'robability .$< .*< .*<

C6 <<915 $<915 8<955

>ar number 1 $ & * 1 8 7 ; 5 1<

C6 &7 8< 75 $1 ;1 71 *; &5 &1 &1

!ime between arri#als 1 $ & 1 * & $ 1 1 1

%rri#al time 1 & 8 7 11 1* 18 17 1; 15

,nter system? Aes Aes Aes Aes Aes Aes Aes 6o Aes Aes

Ser#ice begins 1 * 7 1< 1& 11 1; 999 15 $1

C6 88 7* 5< 51 $5 7$ 17 999 11J 11

Ser#ice time & & & & $ & 1 999 $ 1

Ser#ice ends * 7 1< 1& 11 1; 15 999 $1 $$

J/ne random number could be s"i

ed, using 11 rather than 11.

%nswer( one car does not enter system due to two cars already being in the system.

$8$

Simulation

1$.
Sales er wee" 1 8 7 ; 5 1< 11 1$ 1& 1* 'robability .<1 .<1 .1< .1< .1< .$< .$< .1< .<1 .<1 C6 <<9<* <19<5 1<915 $<9$5 &<9&5 *<915 8<975 ;<9;5 5<95* 51955

Wee" 1 $ & * 1 8 7 ; 5 1< 11 1$ 1& 1* 11

.eginning in#entory $1 17 &* $$ 1< $7 $$ 1$ &< $* 11 &< $< &7 $8

C6 $& 15 ;$ ;& 81 << *; && <8 &$ ;$ 11 1* 88 11

Sales ; 1< 1$ 1$ 11 1 1< 5 8 5 1$ 1< 1< 11 1<

,nding in#entory 17 7 $$ 1< <J $$ 1$ & $* 11 & $< 1< $8 18

/rder? Aes

Aes

Aes

Aes Aes

Aes

J%n assum tion must be made in this situation. !he abo#e assumed a lost sales a roach. )n other words, since the item was not on hand, the customer urchased elsewhere. %nother assum tion would be a bac"order. )n that case, the ending in#entory would be <, with a bac"order of 1. +any times this is re resented by a H1. !he im act on the simulation is that the beginning in#entory for wee" 8 would be $8, not $7, and the following in#entory le#el would need to be ad=usted. !he in#entory olicy a ears to be acce table. 2owe#er, during eriod 1 there was a shortage of cars. :ifferent olicies (different reorder oints and order si0es) should be e-amined. %dditionally, longer or multi le simulation runs of each olicy should be conducted to im ro#e the reliability of the results.

$8&