Modeling Worldviews

• Event Scheduling – Event Graphs

State Machines

• Activity Scanning – Event Graphs

and Petri Nets

• Process Interaction – Event Graphs

and Block Diagrams

Resident Entities

Variables describe the state.

Focus on cycles. (busy-idle-busy...)

Transient Entities

Variables describe state and location.

Focus on flow paths.

System Entities

Block LanguagesProcess “World View”• Focus on Transient Entity Flow – “JOB DRIVEN”.

• Preprogrammed blocks of code.• Block execution is triggered by a

Transient Entity entering a block. • Execution Controlled by Job Moves.• Common Languages

Arena (Siman), AutoMod, GPSS, SLAM (AweSim)

Block LanguagesCommon Languages

SLAM SIMAN GPSS (fundamental Blocks)

Generate - Transactions enter modelQueue - Transactions wait for facilitySeize - Transactions take resourceDepart - Transactions leave queueAdvance - Transactions are delayedRelease - Transactions release resourceTerminate - Transactions leave the model

Block LanguagesCommon Languages

SLAM SIMAN (fundamental blocks)GPSS

Create - Entities enter modelQueue - Entities wait for facilitySeize - Entities take resourceDepart - Transactions leave queueDelay - Entities are delayedRelease - Entities release resource (Entities may leave using “Dispose” )

Single Queue

QUEUE

SEIZE

DEPART

ADVANCE

RELEASE

GENERATE

TERMINATE

1

1

1

1

GPSSSIMANCREATE

DELAY

RELEASE

1

QUEUE

SEIZE1

1

Main Event Scheduling Algorithm for Process Interaction

Cancel Events

Initialize

Advance Clock to Time of Next Event

Execute Event

Remove Event From Listand move “tokens” until blocked or delayed

State

Event List

Change State

Schedule Events

Summary StatisticsStop YesNo

Loading Unloading

Processing

Time

Labor Modeling “Machine Interference”

tool1

Oper.

tool2

Loading

Processing

Unloading

QUEUE

SEIZE

GENERATE

TERMINATE

1

1

SEIZE2

DEPART

Load

RELEASE

1

1

Cycle

RELEASE2

SEIZE1

RELEASE1

Unload

Semiautomatic Machine

Worker is Resource 1Machine is Resource 2

Single Queue with FailuresSIMAN

CREATE

DELAY

RELEASE

1

QUEUE

SEIZE1

1

CREATE

DELAY

RELEASE

1

QUEUE

SEIZE1

1

(Job Processing) (Failure Processing)

TTR

TBF

Proc. time

IA time

HIGH PRIORITY

JOB

Siman Blocks View (File/Template/Attach/Blocks)

This is what the Arena Processes viewgenerates..

Eventually you will need to understand this to use Arena

In a real problem...

Bank Queue with Jockeying

MACHINEMACHINE MACHINE MACHINE

TRANSFER SEGMENT

TOOL BANK TOOLBANK

TOOLBANK

TOOLBANK

TRANSFERSEGMENT 1

TRANSFERSEGMENT 2

TRANSFERSEGMENT 3

TRANSFERSEGMENT 4

TRANSFERSEGMENT 5

DIESEL ENGINE MANUFACTURING LINE MODEL(SIMAN IV)

WALKING CRANE

ONLINE BUFFER

OFFLINE BUFFER

• ; *****************************• ; * Mechanical Failures *• ; ***************************• CREATE;• ASSIGN: J = 0;• DUPLICATE:NumSegs-1; !create entity for each segment• ASSIGN: J = J + 1: !increment counter• SegNum = J; !set segment number• MechLoop DELAY: ED(MTBF(SegNum)); !delay for mean time btwn failures• QUEUE, MechFail1_Q; !wait for machine• SEIZE: Mach(SegNum); !seize machine• ASSIGN: MStat(SegNum) = 6: !segment status to mech fail• State = 0;• BRANCH,2: IF,TR_Policy.EQ.2.OR.TR_Policy.EQ.3,PM:• ALWAYS,M_Rep;• M_Rep ASSIGN: DTime = ED(MTR(SegNum)); !assign mean time to repair• TALLY: MechFail_1 + SegNum - 1, DTime;• DELAY: DTime; !delay for mean time to repair• BRANCH,1: IF,TR_Policy.EQ.2.OR.TR_Policy.EQ.3,PM_GetR:• ELSE,MF_Done;• MF_Done ASSIGN: MStat(SegNum) = 2; !segment status to starved• RELEASE: Mach(SegNum): !release machine• NEXT(MechLoop);

Block LanguagesProcess World View:+Transient Entity Focus Helps Animation+Easy to Use for Simple Applications-Limited to Pre-programmed blocks-Difficult to Model Complex Logic-Difficult to Model Resource Failure or (say, on

amount of time worked.)-Hard to Learn Well-Simulation Slows (or stops) as the System

Becomes Congested ! ! !

Block Languages

Why the simulation slows (or stops) as the system becomes congested.A record of each active transient entity

must be maintained during a run.e.g.: Process Flow in a semi-conductor

fabrication facility. A resident entity model will run hundreds

of times faster.

Block Languages

Resident Entity models are required to have practical simulations of any system with many active transient entities.e.g.

• Communications Networks• Large Scale Manufacturing Systems

and• Continuous Flow Processes

Screen Capture of ARENA Output: Svc Time = 1.85

Ref: Kim Branch and Jay Gillespie

Maximum of 100 Entities Exceeded! (Svc Time = 18.5)

What is the effect on Bias of ignoring runs that did not terminate correctly?

Resource/Job Driven Simulations

0

50

100

150

200

250

300

350

400

450

0 0.5 1 1.5

traffic intensity

tim

e p

er1

00

K j

ob

s

transient entity

resident entity

RunTime Ratios

24.6 25.5 27.9 29.4

100

60

45

30

0

20

40

60

80

100

120

0.10 0.20 0.90 0.95 0.99 1.11 1.25 1.43

Traffic Intensity

Ru

n T

ime

Ra

tio

(A

ren

a/S

igm

a)

Speed comparisons with Automod ASAP

M/M/1 System

100.0

150.0

200.0

250.0

300.0

350.0

400.0

40% 50% 60% 70% 80% 90% 100% 110%

CP

U T

ime

(S

eco

nd

s/1

,000

,00

0 U

nit

s)

RD DESJD NetworkJD DESASAP

6.0

8.0

10.0

12.0

14.0

16.0

18.0

40% 50% 60% 70% 80% 90% 100% 110%

Traffic Intensity

Little Slowdown from using PUT/GET

Runtimes, no queue capacity

0

500

1000

1500

2000

2500

3000

3500

0 200 400 600 800 1000

# servers

Ru

nti

me

in s

eco

nd

s

Arena RD

Arena JD

SIGMA RD

SIGMA JD

Using Activity Scanning to Develop Process Models

“Color” tokens that represent Transient Entities - - track these tokens’ paths.Parse Activities into “SEIZE”, “DELAY” and “RELEASE” Blocks.

SEIZE

DELAY

RELEASE

Idle Resources

Queue

Process -> EG Mapping

Process World View:Automated SLAM to Event Graphs

(Woodward and Mackulak, ASU)

Resource deadlock detection

SIMAN to Event Graphs (Barton and Gonzalez, PSU)

Premature run termination

Notes: 1. Model "footprint" is "number of servers" not "size of queue" - insensitive to congestion – complexity grows with resources!2. Q statistics are O.K. but what about Delays? - ambiguity of the mystery blocks - Use Little's Law or transient entity (PUT/GET) method.

(S>0) (Q>0)

TA

GEN

1QUEUE

DEPART

TS

SEIZE

RELEASE

yes

TERM

yes

TERM

1

1

1

Conventional EG Model

SEIZE Release

{Q=Q+1} {R=R-1,Q=Q-1}

{R=R+1}

(R>0)

(Q>0)

QUEUE

Ta

Ts

NOTE: Can change Distns.dat while model is running!

//CREATE block 1 - interval.5*TRI{.5}

//DELAY block 1 - duration

2*TRI{.5}

//DELAY block 2 -duration

2+3*BET{.5;.5}

Run (S[0],S[1])

Enter

{ I=(Q[0]>10), Q[I]++ }

Start(I)

Finish(I)

I~(Q[I])

tS

tA

{Q[I]--, S[I]--}

{ S[I]++ }

I

~

(S[I])