Steps of simulation study and advantages

Steps of a simulation study

Simulation is the imitation of a real system's operation over time using a model. A study normally follows these steps:

1. Problem formulation. State the problem clearly. If the client states it, the analyst checks that it is understood; if the analyst states it, the client must agree with it.
2. Setting objectives and overall project plan. Decide the questions the simulation must answer and whether simulation is the right tool. Plan the people, cost, time and the alternative systems to study.
3. Model conceptualization. Build the abstract model of the system: its entities, attributes, activities and the logic between them. Start simple and add detail only as needed. The client should take part.
4. Data collection. Collect input data (arrival times, service times, failure rates) and fit probability distributions. This runs in parallel with model building because it takes a long time.
5. Model translation. Code the model in a general language (C, Java, Python) or a simulation language/package (GPSS, Arena, Simulink).
6. Verification. Check that the program works as the conceptual model intends (debugging, tracing, step-by-step runs). "Is the model built right?"
7. Validation. Check that the model is an accurate representation of the real system by comparing its output with real data. Calibrate and repeat until accuracy is acceptable. "Is the right model built?"
8. Experimental design. Decide the alternatives to run: run length, warm-up period, number of replications, initial conditions.
9. Production runs and analysis. Run the experiments and estimate performance measures with confidence intervals.
10. More runs? Decide from the analysis whether additional runs or design changes are needed. If yes, return to step 8.
11. Documentation and reporting. Write program documentation (for future use and modification) and a progress/final report with results and recommendations.
12. Implementation. The client uses the findings. Success depends on how well the earlier steps were done and how involved the client was.

Formulate -> Plan -> Conceptualize + Collect data -> Translate
 -> Verify -> Validate -> Design experiments -> Run + Analyze
 -> More runs? (yes: redesign) -> Document -> Implement

Advantages of simulation

• New designs, policies and layouts can be tested without disturbing the real system.
• Time can be compressed (years in minutes) or expanded (a fast process studied slowly).
• It gives insight into which variables are important and how they interact.
• It can handle complex, stochastic systems that cannot be solved analytically.
• Bottlenecks and causes of problems can be found by observing the model closely.
• It is safer and cheaper than experimenting on the real system (aircraft, power plants).
• It is useful for training and for building consensus among stakeholders, since animation makes results easy to see.
• Assumptions about how the system works can be tested before any investment is made.