Process Improvement with Discrete Event Simulation Luis Garcia Guzman, PhD Asst Research Scientist and Adjunct Professor Industrial and Operations Engineering The University of Michigan
Biografía 2 Educación: IIS-90 (ITESM-Campus Estado de México) MSE and PhD (U of M) --- Industrial & Operations Engineering ExperienciaLaboral: Investigador y Profesor– Ingeniería Industrial, Universidad de Michigan Ingeniero en Logística, Ingeniero de Producto y de Calidad~ Duroplast (Naucalpan), AMP Industries (Michigan), Daimler Chrysler (Michigan) y GM (Michigan). Docencia: Probabilidad y Estadística Ingeniería Estadística Diseño de Experimentos Control de Calidad Simulación de EventosDiscretos Seis Sigma – Cursos de Green Belt y Black Belt
Topics Overview of Simulation Models Steps in a Simulation Study Process Simulation Examples 3
What is Simulation Modeling? A model is an imitation of a system (or process) in real-world over time. A system is a collection of interrelated elements (or processes) which function cooperatively to achieve a stated objective. There is a measurement of performance Model of a system (or process) should reflect and mimic the behavior of the system (or process) Understanding the model implies at least some understanding of the real system 4
System and Model System Environment System Endogenous Exogenous Model State Entity, Attribute Activity, Event Model Scope System Boundary
Started/finished aisle, enter cashier queue, exit queue
Entity Attribute Activity State of a system Event Endogenous/exogenous (activity, event)
Types of Simulation Models Dynamic versus Static Stochastic versus Deterministic Discrete versus Continuous Since models mimic real-world systems, these definition apply to systems as well.
Why Simulate? Typical Decision Support Problems: Evaluate alternative configurations of a system capacity, utilization, bottlenecks, scrap, etc. Identify the desirable/feasible configuration(s) of the system for a specified objective (optimization) Identify a robust strategy to achieve a specified objective for a system Go – No Go decisions for project management Evaluate the value and the risk of an asset 8
Ways to Study a System System Why model? - describe - explain - predict - demonstrate Experiment With the Actual system Experimentwith a model of the system Physicalmodel Mathematicalmodel Analyticalsolution Simulation 9
Experiment with Actual System Advantages Don’t have to spend time/resource to model the system No loss of accuracy Disadvantages May interfere with current operation, or is cost inhibitive May be difficult to repeat, e.g. war game Not possible if there is no real system yet 10
Analytical Methods Advantages Low requirement on modeling efforts Provide great insights on relationships among variables Answer is exact (not necessarily accurate) Disadvantages May need lots of variables or distributions Closed form solution may not exist or is difficult to derive 11
Advantages of Simulation Models Most complex systems cannot be accurately described by the alternatives (e.g., analytical math models) Allows estimating the performance of an existing system under some projected set of operating conditions without disrupting ongoing operations without committing resources for acquisition of new hardware Promotes the understanding of how the system works Test hypotheses about how or why phenomena occur Obtain insight about the interaction of variables Obtain insight about the importance of variables to performance Bottleneck analysis Control over experimental conditions Allows great flexibility for ‘what-if’ analysis Enables comparison of alternative system designs 12
Disadvantages of Simulation Simulation models can be expensive and time consuming to develop Lots of upfront work, e.g. input modeling, computer coding Requires special training, open to interpretations Simulation results may be difficult to interpret Each run produces only estimates of a model’s true characteristics for a particular set of input parameters Computer model may be wrong, e.g. programming bugs The large volume of numbers or the persuasive impact of realistic animation often creates a tendency to place greater confidence in the results than is justified Possibility of misinterpretation of random results 13
Simulation is not appropriate when… The problem can be solved using common sense The problem can be solved analytically It is easier to perform direct experiments The costs exceed possible savings Resources are not available Time is not available No data, not even estimates, are available Not enough time to verify and validate Managers have unreasonable expectations The system behavior is too complex or cannot be defined
II. Steps in a Simulation Study Problem formulation Setting of objectives and overall project plan Model conceptualization Data collection Model translation Verified? Validated? Experimental design Production runs and analysis More runs? Documentation and reporting implementation
Steps in a Simulation Study What is the problem? Formulation Is simulation appropriate? Define alternative systems Project planning Define Project Goal & Plan How? Data Collection Model Conceptualization An Art Start simple Then expand Model Translation No Verified? Is code OK? Yes Represents the system well? No No Validated? Yes
Steps in A Simulation Study What runs to make to answer question efficiently? Experimental Design Production Runs & Analysis Estimate the performance measures Yes Yes More Runs? No Customer acceptance Documentation & Reporting Program and Progress scope of this class Implementation
Process Simulation – Queuing ModelsDescribed by Customer Population Queue Channels and Phases Customer Arrival Process Service Process Queue Discipline 18
1. Customer Service Populations Infinite Cars Passing Toll Booth Supermarket, Bank, Restaurant Customers Telephone Calls at Service Center Finite Geriatric Patients under nursing care TV Networks Students in course 19
2. Queue Channels and Phases Servers Single Server (Single Channel) Multiple Server (Multiple Channel) Phases Single Phase (Single Service) Multiple Phase (Multiple Sequential Services) 20
3. Customer Arrival Processes Constant Example: Scheduled Outpatient Care Variable Arrivals (random variable) Independence (between customers) Single Customer Example: Emergency Room Care Batches of Customers 21
4. Service Process Constant Service Rate Automated Assembly Line Automated Car Wash Streaming Video Distance Learning Variable Service Rate (Random) Gasoline Station Shopping Center 22
5. Queuing Discipline First Come, First Served Priority Customers Shortest Processing Time Reservations First Limited Needs Other 23
Simulation and Six Sigma Six sigma is a data-driven methodology for improving quality in many aspects of a company’s products and services Phases of six sigma methodology typically are: Define, Measure, Analyze, Improve and Control (DMAIC) for existing processes or Define, Measure, Analyze, Design, Verify (DMADV) for new processes or major changes or re-designs (Design for Six Sigma) Simulation is one of the available tools in a Six-Sigma initiative. Particularly within the Analyze and Improve of the DMAIC project or Analyze and Design of a DMADV project or Design and Optimize of a IDDOV project
Simulation and Six Sigma Benefits of simulation in the context of six sigma: Considers process variances, uncertainties and interdependencies Easy to include and study alternative solutions Models can be developed without disruptions to existing processes Takes subjectivity and emotion out of decision making (data-driven=six sigma) Animation tool helps illustrate and convince others on the best solutions Reusable models can encourage continuous improvement
III. Process Simulation Examples Process Simulation Examples OEM Paint Shop Operations OEM Work In Process Inventory (WIP) reduction Supply Chain Optimization
1. North American OEM Paint Shop 27 Problem Description: The paint shop assembly line at an OEM plant is complex and can be improved. Currently, 80% of the painted vehicle bodies are declared a success. Project goal: To increase the number of successfully painted vehicle bodies by: Decreasing system down time, Optimizing color sorting, and/or improving paint robot success rates.
NA OEM Paint Shop 29 Process improvement opportunities: System down time - paint machine color cartridge replacement process Machine operating speed, machine age, and total machine operating time. wait time between locations. Approach: First, a model of the actual system was constructed. Then the model was verified and validated. Alternative configurations developed and tested to find best solution Results Recommend layout solution, increased the yield from 80% to 90% Reduced downtime costs by $2,700 per day
2. OEM WIP Reduction 30 Problem Description: Excessive WIP in the Assembly Area Project Goal: to decrease excess WIP in the workshop. Process Improvement Opportunities: large lot sizes long set-up times long lead times Ineffective production scheduling Breakdowns of machines Non-value-adding activities of Operators
OEM WIP Reduction 31 Approach: First, a model of the actual system was constructed. The reasons for excess WIP in the workshop were analyzed and identified. Then the model was verified and validated. After that, the problem solving approach was developed. By testing the results of changes on variables, the minimum stock level was reached.
OEM WIP Reduction 32 Recommendations: The proposal for decreasing WIP were divided into two groups: Scheduling: creating lot sizing methods material pulling to the system (the number of pieces going into the systems should be equal to the required number of output) lead time monitoring and lead time reduction through waste elimination machine-operator assignments done according to priority of jobs increasing the number of multi-process material handling operators Technological: reduction of set up times methodical improvements automation of machines where possible layout optimization the preventative and productive maintenance
OEM WIP Reduction 33 Results: There was a 48% reduction on the average WIP in the assembly floor As a result of the improvements in WIP the cost of material was reduced by the same amount. There was a 14% improvement by implementing only the scheduling rules.
3. Supply Chain Optimization 34 Problem Description: Excessive lead time for the distribution of confectionary products in India Project Goal: to cut the lead time from factories to retail depots. Determine the optimal amount of trucks to be utilized to minimize lead time at a reasonable cost. Approach: First, a model of the existing supply chain. Then the model was verified and validated. After that, alternative supply chain model was built and simulated to compare with initial model.
Supply Chain Optimization 35 This model is based around a central warehouse used for storage and as a distribution point for some routes. Existing Supply Chain
Proposed Supply Chain 37 Products are shipped directly from the factories to the individual depots much of the burden is shifted to the factories increase in the number of trucks required to meet demand. higher cost, however, cost savings occur due to the lack of maintenance of a larger distribution center and reduction in lead time.
Supply Chain Optimization 39 Results: The proposed supply chain cuts costs by 50% Lead time would be reduced by almost six times.
Logistics – Energy Services Company 40 Problem: High level of maintenance costs at local maintenance centers (26 locations around the world) Long delays in completing maintenance jobs Goal of simulation: Study the effects of maintaining a single global maintenance center where experts can perform the job more quickly and cost effectively.
Initial Results 42 Results: The proposed model could cut maintenance costs by 20% Increase the service level (e.g. probability of having available tools at the oil rigs from 70% to 85%) Lead time could be reduced by almost 30%.
Call Center Evaluation 43 Comparison of 2 different layouts: Current layout Planned improvement to a cell fashion layout Results: Reduction of number of lost calls Reduction of average holding time Reduction of maximum hold time
44 Summary A simulation model is an imitation of a system (or process) in real-world over time. Simulation can be a useful tool in decision making Allows great flexibility for ‘what-if’ analysis Enables comparison of alternative system designs Simulation models are “run” rather than solved Assumptions of model should be validated based on model characteristics and behavior Simulation applications are vast particularly in manufacturing and transactional processes