Chapter 3_Simulation Principles

System

A collection of interacting components that work together and can be studied to understand their behavior.

System → Components → Interactions → Behavior

System Framework

Model

A simplified representation of a real system used to understand, analyze, or predict its behavior.

Modeling

A process that creates a representation of a system, defines relationships and assumptions, and answers "What should the system look like?" — producing a model as its output.

Simulation

A process that uses a model to imitate how a real system behaves over time, answers "What happens when the system operates?" by experimenting with the model, and produces observations/results.

Real System → Model → Simulation → Results

Simulation Framework

Simulation Workflow

• The structured sequence of steps followed when building and running a simulation study.

• Ensures simulations are systematic, reliable, and useful for decision-making.

• Acts as a roadmap from identifying a problem to interpreting simulation results.

Experimental Design

• Defines the purpose and scope of the simulation.

• Without clear _____, simulation can generate lots of data but no actionable insight.

Goal Setting

Determines what question you are trying to answer (e.g., What is the optimal staffing level in a call center?).

Scenarios and Parameters

Identifies which variables you will test, such as arrival rates, service times, and resource allocations.

Performance Measures

Decides what outputs you'll evaluate, such as average waiting time, throughput, or system utilization.

Replication Requirements

Plans how many simulation runs are needed for statistically valid results.

Model Initialization

Before running the simulation, the model must be prepared with initial conditions.

Parameter Setup

Inputs starting values, such as the initial number of patients in a hospital ward or the initial traffic flow rate.

Data Loading

Historical or experimental data may be used to initialize variables.

Warm-up Period Consideration

Some systems require discarding initial runs, like the first few minutes of a queue simulation, to avoid bias.

Simulation Execution

This is where the computational model is run under the specified conditions.

Simulation Engine

Executes events, updates states, and advances time.

Replication

Multiple runs with different random seeds ensure that stochastic (random) effects are captured.

Duration

Defines run length, such as simulating 1 day, 1 month, or until a threshold is reached.

Data Collection & Storage

Raw simulation output is captured for later analysis.

Output Tracking

Captures metrics like waiting times, utilization, throughput, or system failures.

Data Organization

Outputs are structured in tables, logs, or databases.

Traceability

Metadata about which input scenario generated each dataset is critical for comparing results later.

Output Analysis

This is where the real value of simulation emerges.

Statistical Analysis

Uses averages, variance, confidence intervals, and hypothesis tests.

Visualization

Graphs such as line charts, histograms, and boxplots highlight system performance and variability.

Comparative Insights

Compares multiple scenarios, such as 30s vs 45s vs 60s green light cycles in traffic.

Verification

• "Building the model right"

• ensures the model is implemented correctly as intended.

• Focus: programming/model implementation, logic, calculations, event handling.

• Example: checking code logic in a traffic simulation.

Validation

• "Building the right model"

• ensures the model accurately represents the real-world system.

• Focus: real-world behavior, assumptions, historical data, expert judgment, comparison with actual observations.

• Example: an epidemic model matching actual case data.

Interpretation & Decision Making

The final step is making sense of the results in the context of the original question.

Translate Numbers into Insights

Example: "Increasing nurses from 5 to 7 reduces patient wait times by 40%."

Decision-Making Support

Simulation guides policy, design, and operational choices.

Communication

Results must be explained to stakeholders (managers, engineers, policymakers) in clear terms.

Workflow Automation & Reproducibility

In large-scale studies, automated workflows make it possible to run hundreds of scenarios quickly.

Batch Runs

Software can vary parameters systematically.

Reproducibility

Using scripts, workflows can be rerun in the future with identical results.

Transparency

Ensures results can be validated by others.

Deterministic Simulation

Output is fully predictable given the inputs; same input produces the same output every run.

Stochastic Simulation

Incorporates randomness or probability distributions; same input can produce different outputs due to randomness.

Monte Carlo Simulation

• A major example of stochastic simulation that uses random sampling to estimate outcomes;

• repeating many simulations can build a probability distribution

• (e.g., random number generator, histograms).

Discrete Event Simulation (DES)

• System evolves by discrete events at points in time;

• time jumps to the next scheduled event (increments).

Continuous Simulation

System evolves smoothly over time, governed by differential equations; variables change at every time instant.

Agent-Based Simulation (ABS)

• Models the system by representing individual agents and their behaviors and interactions;

• emergent behavior arises from interactions • (e.g., agents as nodes in a network/environment).

System Dynamics Simulation (SDS)

• Represents a system using overall relationships, flows, feedback, and changes in quantities over time;

• represented by stocks, flows, and feedback arrows.

Hybrid Simulation

• Combines two or more simulation approaches to represent different aspects of a complex system, capturing more realism

• (e.g., a hospital using continuous patient health plus discrete arrival/departure times).

Choosing the Right Simulation Type

You need to choose the right simulation type to ensure the model accurately represents the system being studied, leading to valid and meaningful results for decision-making.

Why Verify and Validate?

To ensure credibility, reliability, and usability of simulation results, and to prevent costly errors in policy or engineering decisions.

Calibration

• The iterative process of adjusting a model's input parameters to minimize the difference between its outputs and observed data from the real-world system it represents.
• Ensures the model accurately reflects reality and is suitable for analysis and decision-making.
• Tends to adjust parameters to improve model accuracy (e.g., climate models tuned using past temperature data; setting more sheep than wolves in a NetLogo simulation).

Case Example: Epidemic Modeling

• Verification: ensure correct coding of infection rules.

• Validation: compare infection curve with actual outbreak data.

• Calibration: adjust transmission rate to fit reality.

• Both verification and validation are essential to ensure the relationship between data and information among simulated models is credible and unbiased.