Simulation Lab (EX-606) - Unit 3 Short Notes

UNIT 3: SIMULATION LAB - CORE CONCEPTS & PRACTICAL IMPLEMENTATION

1.0 Introduction & Foundations of Simulation Lab

• Purpose: To bridge theoretical simulation concepts (from course) with hands-on model building, experimentation, and analysis using specialized software.

• Scope: Focuses on discrete-event simulation (DES) as the primary paradigm for studying system dynamics over time.

• Theory vs. Practice: Theory provides mathematical models (queues, inventories); Lab provides the tool to implement, test, and analyze these models under complex, realistic conditions.

• Lab Ethics & Best Practices: Maintain model integrity, document all assumptions and changes, use version control (e.g., Model_v1.2.sim), and ensure reproducibility of results.

• Software Ecosystem: Common tools include Arena Rockwell, Simio, AnyLogic (multi-method), and FlexSim. Choice dictates specific module names but core concepts are transferable.

[!TIP] Exam Focus: Be prepared to define the primary goal of the simulation lab (implementation & validation of theory) and list 2-3 key software packages by name.

2.0 Software Environment & Tool Proficiency

• Core Paradigm: Most DES tools are flowchart-based (drag-and-drop modules representing processes). Some (AnyLogic) also support agent-based and system dynamics.

• UI Navigation: Key panels typically include:

  • Library/Toolbar: Contains modules (Create, Process, Dispose).

  • Model Window/Canvas: Where the flowchart is built.

  • Properties Panel: To define module parameters (e.g., processing time, resource name).

  • Project/Data Panel: Manages variables, resources, and external data files.

• File Operations: Always use Save As with version numbers. Understand the project file structure (often a folder containing model file, data files, reports).

[!TIP] Common Pitfall: Confusing the Properties Panel (for a selected module) with the Data Panel (for global definitions like variables). Always check you are editing the correct entity.

3.0 Model Development: Building Blocks & Constructs

3.1 Entity Creation and Management

• Entity Types: Represent items/customers/jobs flowing through the system (e.g., PartA, Customer).

• Attributes: Entity-specific data (e.g., Priority, Size, Destination). Defined in the Entity Type properties. Used for routing (Decide module) or calculations.

• Routing: Decide module for conditional routing (e.g., If Attribute.Type == "Express"). Route or Transfer modules for unconditional paths.

3.2 Resource Definition and Allocation

• Capacity & Schedules: Define number of identical units (Capacity=3). Use Shift/Schedule definitions to model breaks, shifts, and 24/7 operation.

• Failures: Define MTBF (Mean Time Between Failures) and MTTR (Mean Time To Repair) distributions in resource properties.

3.3 Process Logic and Flow Control

• Core Modules:

  • Create: Generates entities according to an inter-arrival time distribution (e.g., Exponential(5)).

  • Process: Models an operation (delay) requiring a resource and/or a processing time distribution (e.g., Triangular(2,5,8)).

  • Decide: Routes based on a condition (probability or attribute value).

  • Assign: Sets/updates variables or entity attributes.

  • Release: Frees resources seized earlier.

  • Dispose: Removes entity from the system.

• Flow Control: Use Hold/Signal or Batch/Separate for synchronization. Loops are created by routing back to a previous module.

3.4 Data Structures: Variables, Attributes, and Expressions

• Global Variables: System-wide counters or parameters (e.g., Total_Completed, Shift_Number). Declared in the Data panel.

• Entity Attributes: Per-entity data (see 3.1). Crucial for routing logic and personalized processing times.

• Expressions: Used everywhere for dynamic values. Syntax is tool-specific but often similar to Excel (e.g., Resource.Capacity, TN(Entity.Attribute)).

3.5 Time and Scheduling

• Time Distributions: Must be selected based on real data. Common ones:

  • Exponential(λ): For random, memoryless arrivals/service.

  • Uniform(a,b): For equally likely times within a range.

  • Triangular(min, mode, max): For uncertain times with a most likely value.

  • Normal(μ, σ): For times with symmetric variation (clip negative values!).

• Schedules/Calendars: Define working periods (e.g., 8:00-12:00, 13:00-17:00). Modules use these to know when they are "active."

3.6 Input/Output (I/O) and Data Integration

• Reading External Data: Use Read/File modules or import CSV/Excel files into Tables. Link table columns to module parameters (e.g., processing time from a column ProcTime).

• Writing Results: Configure Output Reports or use Write modules to log specific statistics (e.g., entity cycle time) to a file for external analysis (Excel, R).

[!TIP] Exam Critical: Know the difference between an Attribute and a Variable. Attribute = per-entity; Variable = global. Misusing them is a top modeling error.

4.0 Experimentation & Analysis Framework

4.1 Replication vs. Single-Run Analysis

• Single Run: One long simulation. Useful for animation/debugging and warm-up analysis. Cannot provide statistical confidence.

• Replication (Multiple Runs): Standard for output analysis. Run the model n times (e.g., 30) with different random number streams. Each run is independent after warm-up.

• Warm-up Period: Initial transient period where system starts empty. Must be discarded to achieve steady-state. Determine by:

  • Time Series Plot of a key metric (e.g., WIP). Look for stabilization.

  • Rule of Thumb: Discard at least 10% of total run length, but verify visually.

4.2 Input Parameter Variation (What-If Analysis)

• Manual Tuning: Change a parameter (e.g., # of servers) and re-run replications.

• Experimenter Tool: Built-in tool to automate runs across a grid of input values (e.g., servers=1,2,3; inter-arrival time=5,10). Collects output for all combinations.

4.3 Defining and Collecting Performance Metrics

• Time-Persistent vs. Time-Average: Utilization is time-persistent (value at each instant). Avg. Queue Length is a time-average (integral over time).

• Setting Statistics: In module properties, check Tally (for per-entity values) or Statistic (for time-persistent values) to be included in reports.

[!TIP] Exam Formula: Confidence Interval for Mean (across replications):

$$\bar{X} \pm t_{\alpha/2, n-1} \times \frac{S}{\sqrt{n}}$$

Where $\bar{X}$ = mean of replication means, $S$ = std dev of replication means, $n$ = # of replications, $t$ = t-distribution value.

5.0 Output Analysis & Interpretation

5.1 Understanding Generated Reports

• Standard Reports: Overview (summary), Entity (cycle time, etc.), Resource (utilization, schedule), Queue (length, wait time).

• Custom Reports: Filter by category or create custom reports aggregating specific statistics from multiple modules.

5.2 Statistical Treatment of Output Data

• Point Estimators: Mean, Median, Min, Max from the replication means (not from a single long run!).

• Confidence Intervals (CI): Mandatory for any comparative conclusion. A 95% CI not overlapping with another system's CI suggests a statistically significant difference.

• Comparing Systems (Paired-t Test): When comparing two alternatives (A vs. B), run them with identical random number streams (paired design). Use the paired-t test on the replication outputs for more statistical power.

  • Hypothesis: $$\displaystyle H_0 $$: $$\displaystyle \mu_A = \mu_B $$ (no difference) vs. $$\displaystyle H_1 $$: $$\displaystyle \mu_A \neq \mu_B $$.

  • Test Statistic: $$\displaystyle t = \frac{\bar{d}}{S_d / \sqrt{n}} $$, where $\bar{d}$ is mean of differences $$\displaystyle (A_i - B_i) $$.

5.3 Graphical Analysis

• Time Series Plot: For warm-up determination (see 4.1). Also shows system behavior over time (e.g., cyclic utilization).

• Histogram: Shows distribution shape of an output (e.g., cycle times). Check for normality (important for t-tests).

• Box Plot: Excellent for comparing multiple scenarios (e.g., cycle time for 1, 2, 3 servers). Shows median, quartiles, outliers.

• Scatter Plot: To visualize relationship between input parameter and output metric.

[!TIP] Critical Rule: Never compare results from a single run. Always use replications and report confidence intervals. This is a top exam question.

6.0 Advanced Modeling Constructs

• Submodels/Hierarchical Modeling: Encapsulate a recurring logic block (e.g., "Inspection Station") as a reusable submodel. Promotes clean design.

• Advanced Routing: Transporter for mobile resources; Conveyor for paced lines; Network nodes for path-based movement (e.g., in warehouses).

• Custom Logic/Scripting: Use embedded VBA (Arena), JavaScript (Simio), or Java (AnyLogic) for complex logic not possible with standard modules (e.g., custom scheduling, complex calculations).

• Animation & 3D Visualization: Not just for pretty pictures. Used for debugging (watch entities move) and communication with stakeholders. Can be simple 2D shapes or full 3D.

• Debugging Techniques:

  • Step Execution: Run model one event at a time.

  • Breakpoints: Pause execution when a condition is met (e.g., Entity.Attribute > 100).

  • Tracing: Log detailed event sequence to a file.

7.0 Model Verification & Validation (V&V)

[!TIP] Common Mistake: Skipping V&V in the lab report. Always dedicate a section to: 1) How you verified (debugging steps), 2) How you validated (expert review, sensitivity tests).

8.0 Lab Report Writing & Presentation

• Standard Structure:

  1. Problem Statement & Objectives: What decision is being supported?

  2. Conceptual Model: Flow diagram of the real system, key entities, resources, processes.

  3. Detailed Model Specification: Software used, list of all assumptions, input data sources & distributions (with justification!), entity/resource definitions.

  4. V&V Plan & Results: How you checked correctness and realism.

  5. Experimental Design: Scenarios tested, number of replications, warm-up period justification.

  6. Results & Analysis: Tables/figures of key output metrics with 95% CIs. Statistical comparisons (paired-t test results).

  7. Conclusions & Recommendations: Clear answer to the original problem. Limitations of the model.

• Effective Visuals: Use screenshots of the model flowchart (annotated), tables with CIs, box plots for comparison. Avoid wall-of-text.

• Management Summary: A 1-page executive summary at the front, stating the problem, key findings, and recommendation in non-technical language.

9.0 Common Lab Case Studies & Applications

• Manufacturing: Job shop (routing via attributes), assembly line (batching, conveyors), push/pull systems (Kanban).

• Service: Call center (skill-based routing, Abandon module), hospital ED (priority triage, resource sharing), bank queue (multiple tellers, drive-up).

• Logistics: Warehouse picking (resource travel time), distribution center (cross-docking).

• Validation Exercises: Build and compare simulation results of M/M/1 and M/M/c queues to their analytical formulas (e.g., $$\displaystyle L_q = \frac{\rho^2}{1-\rho} $$ for M/M/1). This tests model logic correctness.

[!TIP] Final Exam Strategy: For any case study, first identify: Entities? Resources? Key Process (with distribution)? Primary KOVs? Then map to standard modules (Create -> Process (with resource) -> Dispose).