Topic 11 - Validation & Verification of Simulation Models
Key Concepts
Verification:
Definition: Building the model right
Focus: Comparison of conceptual model vs. computer representation
Questions:
Is the model correctly implemented in the computer?
Are input parameters and logical structure accurately represented?
Validation:
Definition: Ensuring the model accurately represents the real system
Achieved through calibration:
Iterative comparison between model and actual system behavior
Adjustments made based on discrepancies until acceptable accuracy is reached
Simulation Modeling Process
Problem Formulation
Objectives Definition
Model Construction and Experimental Design
Simulation Results
Model Validation (Operational Validation, Verification)
Programming & Implementation
Verification Techniques
Code Review: Have the code checked by someone other than the programmer
Flow Diagrams: Create diagrams to outline possible actions in model logic
Output Examination: Check model output against a variety of input settings
Parameter Tracking: Ensure input parameters aren't altered during execution
Self-Documentation: Define every variable and describe the purpose of code sections
Calibration and Validation
Iterative process comparing the model with the real system
Repeat revisions until model behavior matches expectations
Validation of Simulation Models (Naylor and Finger's Approach)
Build model with high face validity
Validate model assumptions
Compare input-output transformations with real systems
Validation of Model Assumptions
Structural assumptions: How does the system operate?
Data assumptions: Use reliable data and correct statistical analysis
Examples: Interarrival and service times for customers in a bank
Data Analysis Steps
Identify appropriate probability distribution
Estimate parameters of the distribution
Validate statistical model using goodness-of-fit tests (chi-square, Kolmogorov-Smirnov)
Example: Fifth National Bank of Jaspar
Service time assumed random sample from an underlying population
Data collection over 90 customers during peak hours
Arrival process modeled as Poisson process, rate: customers per hour
Service times normally distributed: Mean = minutes, Std Dev = minutes
Discussion Questions
Q1: What is the primary purpose of verification in simulation modeling?
Q2: How can you validate a simulation model?
Q3: Why are both verification and validation necessary in simulation modeling?