When Should I Use Simulation?
1 like1,446 views
The document discusses the applications and benefits of various simulation methods in manufacturing, particularly focusing on discrete event simulation (DES), mathematical modeling, and Monte Carlo simulation. It outlines how these methods can be used for system design, operational procedures, and performance evaluation to enhance efficiency in production processes. Additionally, the presentation touches on the integration of Six Sigma techniques and the importance of data collection and statistical distribution in building robust simulation models.
When Should I Use Simulation?
- 1. When should I use simulation? Prof. Brian Harrington
- 2. Introductions Brittany Hagedorn, MBA, CSSBB - SIMUL8’s Healthcare Lead for North America - Experienced Six Sigma Blackbelt and Healthcare Consultant - Here to answer your questions at the end
- 3. Introductions Brian Harrington, CSSBB - 20 years in simulation at Ford Motor Company - Experienced Six Sigma Blackbelt and Simul8 Manufacturing Consultant - Director of MTN-SIM, a simulation specialist consulting firm - Our presenter for today
- 4. Agenda • • • • • • • Manufacturing issues Different types of simulation Using Math Using Excel/Monte Carlo simulation Using Discrete Event Simulation Simulation for Six Sigma Q&A
- 5. Manufacturing Dilemma • Any product development process involves extensive prototyping; • Yet, costly manufacturing production systems are typically not prototyped
- 6. Simulation in Manufacturing • System Design • Operational Procedures • Performance Evaluation
- 7. System Design • • • • • • • Plant Layout Effects of introducing new equipment Location and sizing of inventory buffers Location of inspection stations Optimal number of carriers, pallets Resource planning Protective capacity planning Biggest Bang for the Dollar! Contains Operational Procedures & Performance Metrics.
- 8. Operational Procedures • Production Scheduling - Choice of scheduling and dispatching rules • Control strategies for material handling equipment • Shift patterns and planned downtime • Impact of product variety and mix • Inventory Analysis • Preventative maintenance on equipment availability Continuous Improvement
- 9. Performance Evaluation • Throughput Analysis (capacity of the system, identification of bottlenecks); Jobs per Hour • Time-in-System Analysis • Assessment of Work-in-process (WIP) levels • Setting performance measure standards; OEE If you can measure it, you can manage it!
- 10. Agenda • • • • • • • Manufacturing issues Different types of simulation Using Math Using Excel/Monte Carlo simulation Using Discrete Event Simulation Simulation for Six Sigma Q&A
- 11. Why Simulation? • • • • • Competition drives the following: Leaner production environment Shorter product development cycles Narrower profit margins Flexible Manufacturing (1 Facility, 1 Process, Multiple Models)
- 12. Types of Simulation • Mathematical Modeling – e.g. Queuing Theory • Monte Carlo Simulation – e.g. Excel based models • Discrete Event Simulation – e.g. Using simulation software
- 14. Question Time: Which of the following Simulation techniques do you use: 1. Math, Queuing Theory 2. Excel Based, Monte Carlo 3. Discrete Event Simulation 4. None
- 15. Agenda • • • • • • • Manufacturing issues Different types of simulation Using Math Using Excel/Monte Carlo simulation Using Discrete Event Simulation Simulation for Six Sigma Q&A
- 16. A Queuing System Input Source Service Process Queue Arrival Process Service Mechanism Jockeying Queue Balking Reneging Served Customers Queue Structure
- 17. Queuing Concepts Relationships for M/M/C 1 Po = C-1 Σ n=0 (λ/µ) n! n + (λ/µ) c! c cµ ( ) cµ - λ c Lq = (λ/µ) (λ µ) Po (c – 1)! (cµ – λ) 2 λ = mean arrival rate µ= mean service rate C = number of parallel servers ρ = utilization These are messy to calculate by hand, but are very easy with appropriate software or a table.
- 18. Queuing Concepts A Comparison of Single Server Models 2 M/G/1 L = q M/D/1 L q = M/M/1 L = q λ σ 2 2 + (λ/µ) 2(1 - λ/µ) (λ/µ) 2 2(1 - λ/µ) 2 (λ/µ) (1 - λ/µ) Note that M/D/1 is ½ of M/M/1
- 19. Benefits & Common Uses Proven mathematical models of queuing behavior; the underlying framework of more comprehensive models. • Computer Networks – data buffering before loss of data transmission • Healthcare – optimizing staffing levels according to patient arrivals • Traffic & Parking lots – Traffic lights, toll booths • Service Industry – Number of servers, checkouts, lanes, ATM machines, etc.
- 20. Limitations on Queuing Models • What if: – we don’t have one of these basic models? – we have a complex system that has segments of these basic models and has other segments that do not conform to these basic models? • Then – simulate!
- 21. Excel Based Simulations • Uses Data Table functions • Each Row might be one iteration of a simulation • Each Col is a random variable generated in the simulation • RAND(), VLOOKUP(), COUNTIF(), NORMINV() • Calculation & Iteration • >>> Using VBA to bring in Probability functions
- 22. Monte Carlo Simulation • Named after the gaming tables of Monte Carlo • Also referred to as a Static Simulation Model in that it is a representation of a system at a particular point in time • In contrast, a Dynamic Simulation is a representation of a system as it evolves over time • Might be accomplished using Excel and the Random()
- 23. Monte Carlo Simulation A Simple Example Day RN Demand Units Sold Units Unsold Units Short Sale s Rev Return s Rev Unit Cost Good Will Profit $ 1 10 16 16 2 0 4.80 0.16 2.70 0.00 2.26 2 22 16 16 2 0 4.80 0.16 2.70 0.00 2.26 3 24 17 17 1 0 5.10 0.08 2.70 0.00 2.48 4 42 17 17 1 0 5.10 0.08 2.70 0.00 2.48 5 37 17 17 1 0 5.10 0.08 2.70 0.00 2.48 6 77 18 18 0 0 5.40 0.00 2.70 0.00 2.70 7 99 20 18 0 2 5.40 0.00 2.70 0.14 2.56 8 96 20 18 0 2 5.40 0.00 2.70 0.14 2.56 9 89 19 18 0 1 5.40 0.00 2.70 0.07 2.63 10 85 19 18 0 1 5.40 0.00 2.70 0.07 2.63 Avg 2.50 Where do these numbers come from?
- 24. Benefits & Common Uses Proven technique that captures random behavior (at a specific point in time); can go further than mathematical solutions. • Business risk assessment – Demand & Profit • Sizing of a market place – Consumption rate • Project schedules (best case, worst case)
- 25. Limitations & Disadvantages • Stochastic, but static! Usually the time evolution of a manufacturing system is significant! • Excel based models, soon start to use VBA, and become very complicated • Might require 1000’s of iterations; Data Tables become slow • Difficult to communicate results to management.
- 26. Agenda • • • • • • • Manufacturing issues Different types of simulation Using Math Using Excel/Monte Carlo simulation Using Discrete Event Simulation Simulation for Six Sigma Q&A
- 27. Benefits of using DES Simulation • Mathematical & Excel based models only go so far • Less difficult than mathematical methods • Adds lot of “realism” to the model. Easy to communicate to end users and decision makers • Time compression • Easy to “scale” the system and study the effects • User involvement results in a sense of “ownership” and facilitates implementation Sim Tree
- 28. Manufacturing Models • The element that the system evolves over time is important • Contain several complicated queuing systems • Internal process steps are significant to achieve the desired result • Conditional build signals (Batch, In-Sequence) • Several sources of stochastic behavior • Contain several shared resources and conditional decisions
- 30. Plant Example cont… How do you simulate an entire plant?
- 31. DES Building Blocks The 8 Core Building Blocks: Start Point, Queue, Activity, Conveyor, Resource, and End Point. Then the Logical aspect Labels & Conditional Statements.
- 32. 8 is all you Need 1. Work Item Types: Can represent parts, carriers, signals, phone calls, just about anything that requires a “Label Profile”. 2. Activities: Work Centers, machines, tasks, process steps, anything that requires a “Cycle Time”. 3. Storage Areas: Buffers, de-couplers, banks, magazines, anything that requires a finite space to occupy over time. 4. Conveyors: Moving parts from pt A to pt B; Number of parts & Speed of conveyor.
- 33. …8 is all you Need… 5. Resources: Manpower, crews, forklifts, tugs; anything that require a certain resource to be present. 6. End Pt: Keep track of statistics and free memory! 7. Labels: The attributes of a Work Item. 8. Visual Logic: The ability to create conditional statements; variables, loops, commands & functions.
- 34. Question Time… How do you use 6-Sigma techniques within your current role? 1. I don’t use 6-Sigma 2. I use 6-Sigma on specific types of projects 3. I use 6-Sigma on all my projects 4. I use an integrated toolset which includes 6-Sigma
- 35. Agenda • • • • • • • Manufacturing issues Different types of simulation Using Math Using Excel/Monte Carlo simulation Using Discrete Event Simulation Simulation for Six Sigma Q&A
- 36. Less is More using 6-Sigma DES Steps: • Objective, Assumptions, Data Collection, Build Model, Verify, Validate, Experimentation, Results DMAIC or DMADV steps: • Define, Measure, Analyze, Improve, Control • Define, Measure, Analyze, Design, Verify Very similar steps!
- 37. Y=f(x’s) Transfer Function Six Sigma focuses on Key Input Factors (x’s) to deliver your Response. All of the x’s can be measured & controlled to increase accuracy & precision of hitting your Target (Y). Trivial Many (N’s) Inputs (N’s & X’s) System/Process Vital Few (X’s) Output (Y)
- 38. The P-Diagram The P-Diagram not only helps engineers to define the Key Parameters for a robust design, but also acts as an excellent communication tool for team reviews.
- 39. Leverage Statistical Distributions! • Curve fit your data! Instead of using lengthy spreadsheets. • Black-box; entire segments of the model can be collapsed using distributions. • If using empirical datasets, drop them into a “Probability Profile Distribution”
- 40. Graph your Data! One of the most basic steps in 6-Sigma; Exploit your data! Stat-Fit for SIMUL8
- 41. Use Known Distributions The data collection phase of modeling can be the lengthiest and most time consuming. Downtime (MTBF & MTTR); such as Exponential & Erlang respectively. Cycle times often use a Fixed distribution; that is the “Design Cycle Time”.
- 42. Steady State A common data collection error is to capture all data points, and attempt to force them into one distribution. – Filter out the outliers; usually catastrophic points are outside the scope of the steady state system. 42
- 43. Concluding Thoughts • Queuing Theory & Monte Carlo Simulations can meet your specific objectives in certain applications. Yet, can become overwhelming when pulling them beyond their intent. • Most Manufacturing, Healthcare objectives go much further beyond these capabilities. Where the dynamic aspects of time are critical! • Discrete Event Simulation is a user friendly tool that is built on the foundations of queuing theory & statistical sampling.
- 44. Q&A