Bio Eng.D in Production Systems from Ecole des Mines de Paris, France Ph.D in Operations Research from MIT Research: Revenue Management, Dynamic Pricing, Auctions, Procurement, Stochastic Models of Manufacturing Systems Experience in ecommerce Fulfillment, Electronics, Aeronautics, Transportation and Software
What is Simulation? Scenario Generation Input A System Model Outcome Input B System Model Outcome Input Data Collection Modeling Validation / Accuracy Assessment Inference Output Data Collection Input A Real System Outcome Input B Real System Outcome
Types of Simulation Static / Monte-Carlo (Crystal Ball) Discrete Events (Simul8) Continuous Time (application-specific) Examples: Examples: Examples: Options Contracts Business Processes Weather models Insurance Education Engineering design Demand Models Military Scientific research Entertainment Education
Simulation Module Goals Develop the practical skills necessary to design, implement and analyze discreteevent simulation systems; Practice of Modeling!!! Cover the basic theory underlying discreteevent simulation methods.
Class Date Topic Reading Assignment TUT MIT 11/2 SG 11/3 Simul8 Tutorial Introduction to Simul8 1 MIT 11/3 SG 11/4 Introduction, Simulation Process and Stochastic Modeling ClearPictures, Inc. (in this document) Proba/Stat Review 1 (in this document) 2 MIT 11/8 SG 11/9 Monte-Carlo Theory and Examples (with Crystal Ball) Common Probability Distributions for Simulation Modeling 3 MIT 11/10 SG 11/11 Discrete-Event Framework and Examples 4 MIT 11/15 SG 11/16 Design and Analysis of Simulation Experiments Proba/Stat Review 2 (in this document) MIT 11/30 SG 12/1 Individual Assignment Due (in this document)
Optional References Law, A. and W. Kelton, Simulation Modeling and Analysis, 3 rd ed., McGraw-Hill (2000). Ross, S., Simulation, 3 rd ed., Academic Press (2002). Swain, J., Power Tools for Visualization and Decision-Making, OR/MS Today, February 2001. Available online at http://www.lionhrtpub.com/orms/surveys/simulation/simulation.html
From the Trenches There have been a lot of course changes, basically because I didn't have the problem I was trying to solve and the model structure well thought out before starting. So if I can offer you one piece of advice it would be to spend as much time up front as you need thinking about exactly what you want to model. LFM Intern, 2003
The Simulation Process 1 2 3 4 5 6 7 Define the simulation goal Model the system Implement the model Debug, Validate, Sensitivity Design the experiment Run the experiment Analyze and communicate Never skip! Keep the goal in mind! Customer feedback! Choice of tool is key Never skip! Customer feedback! Run length, warm start, variance reduction Use confidence intervals!
Simulation Goal System Design Vs. System Analysis Strategic? Tactical? Key Performance Measures? Control? What about ClearPictures, Inc.?
ClearPictures: Simulation Goals Estimate the average and standard dev. of delivery leadtime through the pull section; Estimate the average and standard dev. of WIP inventory through the pull section; Determine the production bottleneck; Estimate the impact of purchasing more machines on leadtime and inventory;
System Modeling Everything should be made as simple as possible, but not simpler. Albert Einstein. The simulation goal should be the guiding light when deciding what to model Start to build your model ON PAPER! Get client/user feedback early, and maintain model + assumption sheet for communication purposes For random variables, collect data and fit distributions after modeling the system, with sensitivity analysis in mind!
System Modeling 2 Model the system 1 Define simulation goal 2a Model on paper Process Flow Diagram, Flow Chart 2b User/Client feedback Model assumption sheet! 2c Sensitivity analysis Theoretical: TOC, Queueing theory 2d Data collection & Fit Prioritize, mock data example, fitting software 3 Implement the model 4 Debug, Validate, Sensitivity 5 Design the experiment
ClearPictures Production Model Exp(12) Box / Sensor Board Assembly Sensor Firmware Test U[15,25] Inspection N[9.5,4] assy TRIAN[5,10,15] 1 ST2 ST1 2 insp. 15% 85% 360 Customer Notes and Assumptions: 1/ Service time at ST1 is U[13,24] first passage, U[10,15] rework. Rework has priority. 2/ There can be at most 1 rework cycle for each part (passes inspection second time).
Model Implementation General programming language (C++, Java ) Simulation-oriented language (MODSIM ) Simulation software with GUI (Simul8, Witness ) Excel Add-in (Crystal Ball, @Risk ) FLEXIBILITY COST REQ. SKILLS DEV. TIME RUN TIME Genera l Prog. Language Very High Low Very High High Low Simulation Language High High High Medium Low Simulation Software Medium High Medium Low Medium Exce l Add-in Low Low Low Lowest High
Validation & Debugging Slow Graphical Animation Step-by-step event list
Experiment Design Warm-up Period? Run Length? Number of Trials? How to analyze and interpret the results?
1. Simulation Process Class 1 Wrap-Up 2. Modeling 3. Choice of simulation tool