Simulation Models for Manufacturing Systems

Transcription

1 MFE4008 Manufacturing Systems Modelling and Control Models for Manufacturing Systems Dr Ing. Conrad Pace 1 Manufacturing System Models Models as any other model aim to achieve a platform for analysis and evaluation of real systems. models tend to be ideal when dealing with Complex systems Dynamic, time-variant behaviour of systems Random behaviour/characteristics of systems Complex control strategies that depend on numerous factors The need to map data gathered from the real system into the model 2 1

2 Manufacturing System Models is a procedure for experimentation with a model Proper experimental design is required What are the performance parameters we are after? What are the conditions/ decision variables we want to analyse to determine the effect on the performance parameters? What are the constraints/ characteristics that bound our simulated system? Results are typically statistical data Average values Distributions Variances 3 Components in a Manufacturing System Model Languages A special purpose computer language Includes constructs for building models, performing experimental analysis, and recording results General Purpose Languages used for simulation (C/C++, VISUAL BASIC) Specifically developed languages (SIMAN/ SIMULA) 4 2

3 Components in a Manufacturing System Model Package / System Integrates a simulation language with other supporting software tools Graphical model editors Statistical analysis tools Animation tools Report and graph generators Example ARENA, EXTEND, SIMPLE++ 5 Model Examples 6 3

4 Model Examples 7 Model Examples 8 4

5 Model Examples 9 Model Examples 10 5

6 System Model Entities Resources Queues Events Performance Measures 11 System Model Entities Process System Model Entity Entity Examples Parts Batches Orders Tags Represent the elements that move around through the system model Changes status while transitioning through the model They influence and are influenced by other entities as well as the system state 12 6

7 System Model Entities Entities are the dynamic elements of the model Entity Creation ENTERING MODEL Process Model Entity Creation Examples : Part arrives on the line Batch is released for production Order for Production EXITING MODEL Entity Disposal Entity Disposal Examples : Finished Product leaves line Batch is shipped 13 System Model Process Model System Model Entities Attributes of Entities Time entity enters system Time entity starts being processed at a station Part Code Due Date, etc

8 System Model Process Model System Model Entities System Model as an entity Attributes (Global Attributes) : Time Parts in System Parts waiting in Queue Mean throughput time, etc.. 15 System Model Resources Process Model Resource Example Processing Station Material Handling System System Model Inspection Unit Handle entities by servicing the entities 16 8

9 System Model Resources Server Unit of a Resource Resource Entities waiting to be serviced by Resource Entities compete to be serviced by the resource. Resources consist of one or more servers multiple servicing of entities 17 System Model Resources Resource Entities waiting to be serviced by Resource Example of Servers Operators at a processing station Material Carrying Devices in a Material Handling System 18 9

10 System Model Resources SEIZE Resource Entities waiting to be serviced by Resource SERVICE Release A server (resource unit) can service a single unit at any instance in time. Entity seizes a server to be serviced Entity is serviced by server Entity releases server following service completion 19 System Model Resource Resources Attributes of a Resource Current State of servers busy idle blocked starved Broken/Failed etc

11 System Model Queues Entities that cannot proceed to the next operation Attributes of Queues Capacity Available Handling of entities in queue FIFO LIFO etc.. Example of Queues Buffer Storage ahead of a station Orders to be processed Waiting Servicing for machines 21 System Model Events Time Instance State Change Event A Event B Event C Time Entity Arrives (Created) Entity Enters Queue Entity Seizes Resource and starts service 22 11

12 System Model Events Typical Events Arrival of a Part in the system Commencement of servicing at a resource server Finalisation of servicing at a resource server Departure of a Part from the system Event Attributes Time of Event occurrence Event Identifier (type of event) Future events to be scheduled by event 23 System Model Events Event Scheduling Event List/ Calendar T A Event A T B Event B T C Event C. T A Event A T B Event B T C Event C Entity Arrives (Created) Deterministic (ex. 1.2 mins) Entity Enters Queue Time Between Events Time Entity Seizes Resource and starts service Random (ex. Unif. Dist. Between 1.0 and 1.4 mins) 24 12

13 System Model Performance Measures Parameters used to interpret behaviour of simulation model Time Persistent Measurable at any simulation time Observed Measurable at event times 25 System Model Performance Measures Time Persistent Queue Lengths State of Resource Work in Process No of Parts Produced Observed Part Processing Time Time Spent in Queue Time Resource is idle Statistical Accumulators 26 13

14 System Model General Statistical Accumulators Used Number of parts produced Total waiting time in queue Number of parts passing through queue Longest time spent in queue Longest time spent in system 27 System Model General Statistical Accumulators Used Area under the queue-length curve Area under the server-busy function Makespan Percentage Failures Congestion 28 14

15 System Model Other Parameters Clock Initialisation / Termination Mode 29 Event An Example Processing of Blank Parts at a Drilling Station Part Arrives If Station is Idle Part Starts Servicing Immediately If Station is Busy Part Enters a First-In-First- Out Queue 30 15

16 Event Drill Example Entities Resource Events Parts Drill Press (Single server) 1. Part Arrives 2. Part Starts Service 3. Part Completes Service 31 Event Drill Example Event Based Model Event 1. Part Arrives 2. Part Starts Service 3. Part Completes Service State Change No. of Parts in System Increases No. of Parts in Queue Increases No. of Parts in Queue Decreases Drilling Machine becomes Busy No. of Parts in System Decreases Drilling Machine becomes Idle 32 16

17 Event Drill Example Event Based Model Event Interaction : Using an Event Graph Inter arrival time (if station queue not empty) Part Arrives (if machine idle) Part Starts Service Service time Part Completes Service Event Trigger : Machine Inter-arrival Service Time Becomes time Idle Event Action : Place Part waiting part Releases in FIFO in Resource queue queue Event Action Scheduling : Part Resource : Part Seizes Arrives becomes resource, Idle Event Scheduling Resource : Part Starts becomes Service busy Event Scheduling : Part (conditional) (Conditional) Completes Service 33 Event Drill Example Event Based Model Models the system through the Definition of Events, the subsequent state changes and the Event Interactions 34 17

18 Event Drill Example Process Based Model Alternative Viewpoint Model the system as an Ordered Sequence of Events followed by an Entity Process Conditions can Modify the sequence of Events followed by the Entity Facilitates representation of Manufacturing Processes NOTE : The Model is Still Driven by EVENTS 35 Event Drill Example Process Based Model Process Steps Part Arrives Wait Till Part s Turn to be serviced Wait for Servicing Time Event : Part Arrives Part Enters Queue Event : Start Service Part Is Serviced Event : Complete Service Part Leaves System 36 18

19 Event Drill Example Running the 2 Events Part Arrival Part Departure 2 Resource States IDLE BUSY Part K Part Arrives Event : Part Arrival (K) Part Enters Queue Event : Part Departure (K-1) Or Part Arrival (K) Part Is Serviced Event : Part Departure (K) Part Leaves System 37 Event Drill Example Running the Event Routine Part Arrival Start Add Part to queue Yes Is Drill Busy? No Seize Drill and make Drill State = Busy Schedule Part Departure Event End 38 19

20 Event Drill Example Running the Event Routine Part Departure Start Dispose of Finished Part Make Drill Press = Idle Yes Is Queue Empty? No 1 st Part in Queue seizes Drill Queue length reduced Schedule Part Departure Event End 39 Event Drill Example Running the Data Part Number Arrival Time Inter-arrival Time Service Time

21 Event Drill Example Running the Initialisation - Time 0.0 Resource Status Number in queue System State I 0 Times of arrival Time Event List Arr Event Drill Example Running the Event : Part 1 Arrival - Time Resource Status Number in queue System State BI 0 Times of arrival Time Event List Arr Dep

22 Event Drill Example Running the Event : Part 2 Arrival - Time Resource Status Number in queue System State B Times of arrival Time Event List Arr 1.73 Dep Arr Event Drill Example Running the Event : Part 1 Departure - Time Resource Status Number in queue System State 2.90 Time 0 B 1.73 Event List Dep Arr. Dep Times of arrival 44 22

23 Event Drill Example Running the Event : Part 3 Arrival - Time Resource Status Number in queue System State 3.08 Time B Event List Arr Dep.4.66 Times of arrival 45 Event Drill Example Running the Event : Part 4 Arrival - Time 3.79 System State 1.73 Resource B Status Number in queue 01 2 Times of arrival Time Event List Arr Dep

24 Event Drill Example Running the Event : Part 5 Arrival - Time 4.41 System State Time Resource B Status Event List Arr Number in Dep.4.66 Arr queue Times of arrival Event Drill Example Running the Event : Part 2 Departure - Time 4.66 System State 1.73 Resource B Status Number in queue 4.41 Times of arrival Time Event List Dep Arr

25 Event Drill Example Running the Event : Part 3 Departure - Time Resource Status Number in queue System State B Times of arrival Time Event List Dep Arr Event Drill Example Partial Trace of the Process Model Interpretation Time Process Step Description Part 1 arrives 1.2 Part 1 enters buffer 1.3 Part 1 starts service First part arrives First part enters buffer First part begins processing Part 2 arrives 2.2 Part 2 enters buffer Second part arrives and queues for service Second part enters buffer Part 1 completes service 2.3 Part 2 starts service First part is finished Second part begins processing Part 3 arrives 3.2 Part 3 enters buffer Third part arrives and queues for service Third part enters buffer Part 4 arrives 4.2 Part 4 enters buffer Fourth part arrives and queues for service Fourth part enters buffer Part 5 arrives 5.2 Part 5 enters buffer Fifth part arrives and queues for service Fifth part enters buffer Part 2 completes service 3.3 Part 3 starts service Second part is finished Third part begins processing 50 25

26 Event Drill Example Running the Time Persistent Statistics No. of parts in Queue 51 Event Drill Example Running the Time Persistent Statistics Busy State of Drilling Press 52 26

27 Part Number Randomness in the Drill Example Note that source of Data can possibly come from recorded behaviour of a real system Is the period of recording a true representation of normal operation? Will input values be always as illustrated in the table? Arrival Time Inter-arrival Time Service Time Service Time = Random Function Inter-Arrival Time = Random Function 53 Randomness in the Drill Example Part Number Arrival Time Inter-arrival Time Service Time Inter-Arrival Time = Exponential Distribution Function (mean = 5 min.) Prob. Density Function

28 Randomness in the Drill Example Part Number Service Time = Triangular Distribution Function (min = 1, mode = 3, max = 6 min.) Arrival Time Inter-arrival Time Service Time Prob. Density Function Randomness in the The Necessity of Random Parameter Values Where considerable uncertainty is present in system parameter data values Probability Distribution originates from a statistical analysis of parameter behaviour Various distributions are used to represent various behaviours of system components Important to ensure that the selected distribution fits closely on the statistical data

29 Randomness in the The Necessity of Random Paremater Values Statistical interpretation of Real System Data Random Inputs System Model Random Outputs Statistical Analysis of Results 57 Replications Statistical Analysis of Results Numerous experimental data can be generated with Probability Functions for the system model inputs Run Multiple Simulated Experiments Replications Will behave similarly following the same probability functions Will not necessarily give the same results 58 29

30 Randomness in the Statistical Analysis of Results In certain cases, a simulated test may give extreme results (e.g. unlikely long processing times, etc..) Far from the expected operating conditions Normal to run multiple replications Provide more stability in results interpretations (e.g. through confidence interval evaluation) 59 Comparing Alternatives Running Experiments under different Conditions Parameter Values (e.g. Inter-Arrival Times) Control Logic (e.g. changing order in which parts are handled in a queue) We may even want to introduce changes in the simulation model Introducing or Eliminating Buffers Introducing the effect Machine Downtime, etc

31 Part A Inter-Arrival = 10mins. Part B Production Line Example Line Characteristics 3 stations 2 limited capacity buffers 1 infinite capacity buffer 2 types of parts 10 units 10 units Inter-Arrival = Expo (15mins) Service Time A : Unif (3,7min) B : 6min Service Time Trian (min. 2, mode 5.88, max 11mins) Service Time Expo (mean 5.5min) 61 Production Line Example Line Characteristics Station 3 Failure MTTF = 40hrs Part A Inter-Arrival = 10mins. MTTR = 2hrs 10 units 10 units Part B Inter-Arrival = Expo (15mins) Service Time A : Unif (3,7min) B : 6min Service Time Trian (min. 2, mode 5.88, max 11mins) Service Time Expo (mean 5.5min) 62 31

32 Production Line Example Objective transient behaviour of buffer queue lengths following a failure of station 3 Portion of time a station is blocked Distribution of unit time in the system Part A Part B Production Line Example Results Running the for 5 replications duration = 40hr week Station 1 Station 2 Station 3 Arrival rate (unit/mins) Service rate (unit/mins) Utilisation Percentage blocked Mean parts at station Mean time at station (min) Mean parts queued Mean time in queue (min) 1/6 1/5.4 86% 13% /6 1/ % 6% /6 1/5.5 85% N/A

33 Production Line Example Results Buffer 3 queue length 10.0 Buffer Occurrence of Station 3 failure Day 65 Production Line Example Results Buffer 2 queue length 10.0 Buffer Occurrence of Station 3 failure Day 66 33

34 Production Line Example Results Buffer 1 queue length Buffer Occurrence of Station 3 failure Day 67 Production Line Example Results Distribution of Unit Time % Percentage (%) % 6.2% 10.1% 18.5% 12.4% 0 < > 10 Part Time in system (hrs) 68 34

35 Part A Inter-Arrival = 10mins. Part B Inter-Arrival = Expo (15mins) Production Line Example Evaluating impact of increasing buffer sizes Running the with buffers 2 and 3 having a capacity of 20 units units units Production Line Example Results (2) Buffer 3 queue length 20.0 Buffer Occurrence of Station 3 failure Day 70 35

36 Production Line Example Results (2) Buffer 2 queue length 20.0 Buffer Occurrence of Station 3 failure Day 71 Production Line Example Results (2) Buffer 1 queue length 26.0 Buffer Occurrence of Station 3 failure Day 72 36

37 Production Line Example Results (2) Other Results % blocking time for station 1 Reduced from 13% to 5% % blocking time for station 2 Reduced from 6% to 2% Average Throughput per week Increased from 380 to 399 (with an absolute max theoretical throughput of 408 parts) 73 Production Line Example Results (2) Distribution of Unit Time Percentage (%) % 20% 18.5% 14.6% 8.5% 0.0% < > 10 Part Time in system (hrs) 74 37