THE ROLE OF MODELING

AND SIMULATION
CHAPTER 2

MD SHOAIB MAHMUD
LECTURER, DEPARTMENT OF IPE
MILITARY INSTITUTE OF SCIENCE AND TECHNOLOGY
(MIST)
SIMULATION DEFINITION

Definition
 Simulation is now regarded as a third and indispensable avenue of
scientific pursuit, alongside theory and experimentation. It enables
us to construct a virtual replica of a real‐world system or process—
capturing its essential dynamics without the cost or risk of real‐
world trials .
SIGNIFICANCE
Significance
 Insight Gathering: Exploring unknown phenomena when little data exist.
 Model & Experiment Validation: Checking analytic models or experimental
findings against a detailed virtual testbed.
 Training: Allowing operators—pilots, surgeons, soldiers—to rehearse complex
tasks safely.
 Statistical Analysis: Running controlled “experiments” to estimate means,
variances, and distributions.
 Computer Animation: Driving realistic motion in graphics and virtual
environments.
 Real-Time Control: Embedding a live simulation within a physical process (e.g.,
adaptive drilling systems).
 Prediction & What-If Analysis: Forecasting weather or testing new system
designs under hypothetical conditions .
IMPACT IN EVERYDAY LIFE

Impact in Everyday Life

Weather Forecasting: Ensemble simulations yield probabilistic
outlooks for the next several days.
Flight Simulators: Trainee pilots experience emergency scenarios
without endangering lives.
Medical Training: Surgeons practice complex procedures in a risk-
free virtual operating room.
SPECTRUM OF USE

Spectrum of Use
 Problem Solving (Low Uncertainty): A mature, validated model
is treated as a trusted decision tool (e.g., computing a projectile’s
trajectory).
 Insight Gathering (High Uncertainty): An evolving, sometimes
disposable model is used to explore and learn (e.g., tuning early
pandemic spread models) .
PROBLEM SOLVING WITH SIMULATION: CHARACTERISTICS

Characteristics
Validated Context: In problem-solving scenarios, the simulation is
assumed valid for its intended context. Users trust outputs without
delving into internal logic—so long as the underlying assumptions (e.g.,
a stationary firing platform in a ballistics model) hold true, results are
treated as meaningful.
Low Epistemic Uncertainty: The model’s structure and equations
are considered sufficiently accurate; only inherent randomness remains
(aleatory uncertainty), which is managed via probabilistic processes.
User Focus on “What If”: Users frame their inquiries in concrete,
decision-oriented terms, relying on the simulation to answer specific
performance and feasibility questions.
PROBLEM SOLVING WITH SIMULATION: KEY QUESTIONS
ADDRESSED
Key Questions Addressed
 What? “What firing velocity will damage the weapon muzzle?”
 How? “How will a projectile with these characteristics travel through
the air?”
 Why? “Why did the munition deviate from its intended course?”
 Can? “Can this munition travel more than 10 km under given
conditions?”
 Does? “Does the simulation output match analytical ballistics
formulas?”
 Should? “Should we select alloy X for this application?” .
PROBLEM SOLVING WITH SIMULATION: EXAMPLES
Examples
Ballistics Computations
One of the earliest simulation applications (1940s). By encoding projectile
dynamics, air resistance, and propellant behavior, modelers can rapidly explore
firing scenarios without costly live tests. Questions such as “What velocity ensures
impact at 5 km?” are answered in minutes.
Coin Tossing
Simulating 1,000 tosses of a “fair” coin to verify the expected binomial distribution,
and to probe why extreme outcomes (e.g., 900 heads/100 tails) are exceedingly
unlikely. This exercise also illustrates the efficiency trade-off: analytic combinatorics
could answer these questions in seconds, whereas a simulation run may take hours .
Weather Prediction & Other Domains
Tomorrow’s forecast via ensemble modeling.
INSIGHT GATHERING: CHARACTERISTICS

 High Epistemic Uncertainty: The model structure or key

parameters are not fully known, so the simulation is treated as
an adaptable, even disposable, exploratory tool rather than a fixed
decision‑support system .
 Designer‑Driven: Insight gathering is the domain of the simulation
designer—a scientist or researcher using the simulation itself to
learn about the phenomenon or the model’s behavior.
INSIGHT GATHERING: KEY QUESTIONS

 What has the greatest influence on outcome X?

 How will X and Y interact under varying conditions?
 Is there a way to make X happen (or never happen)?
 Why did unexpected behavior Z occur in the simulation?
 What new behaviors might emerge when inputs or structure
change? .
DESIGNER’S ROLE & PROCESS

 Set Up Validation Experiments: Using external data or theory (e.g.,

lab tests, expert opinion), the designer configures the simulation to
reproduce known results. Discrepancies guide model refinement.
 Iterative Refinement: Adjust structure or parameters, rerun
simulations, and compare outputs against benchmarks.
 Risk Management: If uncertainty cannot be resolved or the cost is
prohibitive, the designer may accept residual uncertainty and
document its implications.
PARAMETERS & CONTROLS

 Context Parameters (e.g., distributions for air‑current strength

when simulating coin tosses) are varied systematically to assess
model sensitivity.
 Control Parameters (e.g., algorithm selection, time‑step size) are
tuned to reveal how numerical methods influence insight.
 Adaptation Loop: Each round of insight leads to model
adjustments—either tweaking parameters or replacing submodels
entirely—to better reflect the true phenomenon.
 ILLUSTRATIVE EXAMPLES

Coin Tossing
 Epistemic Uncertainty: Is the coin fair, or is it a “smart” (biased) coin?
 The designer simulates 1,000 flips under different bias assumptions, then
compares simulated distributions to experimental data, revealing whether
the model captures true coin behavior.
Yucca Mountain Corrosion Study
 Validation Questions: “What influences container lifespan most?” and “Why
does failure occur under certain moisture/temperature thresholds?”
 By configuring the simulator with extreme rainfall or temperature profiles,
the designer tests whether predicted corrosion rates align with laboratory
findings, then refines material‑interaction submodels accordingly.
TYPES OF UNCERTAINTY IN SIMULATION

 Aleatory Uncertainty (Random Variation): Phenomena treated as inherently

random because we lack a better way to characterize them—e.g., air currents,
surface irregularities, or flick variations in coin tossing. Such effects are either
simulated via probabilistic (stochastic) processes or deliberately ignored when
their impact is negligible.
 Epistemic Uncertainty (Structure Uncertainty): Gaps in our knowledge of the
system’s true behavior or model structure—e.g., whether a given coin is fair,
biased, or “smart.” When handed an unfamiliar coin, one must either gather
data (transforming epistemic to aleatory uncertainty) or make educated
guesses to simulate its flips.
IMPACT OF UNCERTAINTY ON SIMULATION

 Any significant epistemic uncertainty signals that the simulation is still in an

insight‑gathering phase. Designers must either:
1. Reduce uncertainty (through experiments, theoretical analysis, or further
simulation‑based exploration), adapting the model as insight accumulates; or
2. Accept residual risk, if resolving the uncertainty is impractical or if remaining
gaps do not materially affect decision needs.
 Once epistemic uncertainty is sufficiently resolved—i.e., the model is
structurally sound—the simulation graduates to a problem‑solving role,
managing only aleatory randomness.
REAL-WORLD CONSEQUENCES

 A February 2006 NSF Blue Ribbon Panel highlighted that

quantifying uncertainty in complex simulations remains a
“daunting challenge” for engineering science. Epidemiological
models of diseases like smallpox or bird flu exemplify this: just four
out of hundreds of critical variables can drive extreme sensitivity in
predictions, leading to widely conflicting public‑health strategies
with billions of dollars and millions of lives at stake.
LIFE CYCLE: BIRTH OF A SIMULATION

 A simulation is “born” when a designer decides a virtual model of a phenomenon

would be useful.
 “Meaningful” can mean:
Complete: No further modifications planned; the designer believes the model is
valid from the outset.
Crude: Sufficient to begin gathering insight, with significant refinement expected.
 Paths at Birth:
If truly complete, the simulation may go Birth → Insight Gathering → Problem
Solving with no repeats.
More commonly, the designer’s understanding evolves, leading to multiple cycles of
insight gathering and problem solving—or, in rare cases, the simulation is scrapped if
understanding proves inadequate.
LIFE CYCLE: KEY STAGES

 Insight Gathering (Designer‑Driven): Under high epistemic

uncertainty, the designer adapts model structure or parameters.
Iterative revalidation and calibration ensure the simulation
increasingly reflects reality.
 Problem Solving (User‑Driven): Once validated, the simulation is
deployed by users as a trusted decision‑support tool. Users focus on
“What? How? Does?” questions without modifying internal logic.
LIFE CYCLE: TRIGGERS FOR ROLE SHIFTS

 Problem Solving → Insight Gathering:

1. Designer gains a new understanding of the phenomenon.

2. User or designer discovers the simulation is invalid for certain aspects.
3. Desire to add new or remove existing modeled phenomena.
4. Need to apply the simulation in a different context than originally
designed.
 Insight Gathering → Problem Solving:
 Occurs when the simulation is finally deemed valid for its intended use
and is calibrated using known or hypothesized data .
LIFE CYCLE: LONGEVITY & CYCLES

 Many simulations endure for decades, cherished by research groups

as ongoing projects.
 Some enter a stable problem‑solving role and remain unchanged for
years.
 Others regularly cycle between insight gathering and problem solving
as new requirements or discoveries arise.
 Even “one‑time” simulations often re‑enter the lifecycle, eventually
stabilizing into long‑lived decision‑support tools.
 PRACTICAL APPLICATIONS: PROBLEM SOLVING
 Problem‑Solving Simulations
 Ballistics: Engineers ask: What firing velocity is required to damage the weapon
muzzle? How does a munition with given characteristics travel through the air? Does
the simulation output match mathematical ballistics formulas?
 Commercial Nuclear Waste Storage (Yucca Mountain): Decision‑makers evaluate:
What rainfall levels would leach dangerous isotopes? Why do Alloy 22 containers
resist corrosion under site conditions? Should the U.S. store waste at Yucca Mountain?
 Coin Tossing: Users explore: What are the most likely heads/tails sequences in 1,000
flips? How probable is an extreme outcome (e.g., 900 heads, 100 tails)? Does the
simulated distribution follow the binomial curve?
 Weather Prediction & Disease Spread: Meteorologists run ensemble forecasts to
predict tomorrow’s weather. Public‑health officials calibrate smallpox “R” values
against historical outbreaks to inform vaccination strategies
PRACTICAL APPLICATIONS: INSIGHT GATHERING

 Insight‑Gathering Simulations
 Coin Tossing: Designers probe: Why are flips independent? Is there any bias if a
“smart” coin is introduced? Iteratively refine the pseudo‑random generator model
to capture true flip dynamics
 Yucca Mountain Corrosion Study: Researchers ask: What factor most influences
container lifespan (e.g., alloy vs. environment)? How do moisture and temperature
interact to drive failure? Why does corrosion accelerate past certain thresholds?
 Disease‑Spread Models: Epidemiologists use simulation to reduce structural
uncertainty in key parameters (contact rates, transmission probabilities), exploring
sensitivity to variables that have driven conflicting public‑health policies
CONCLUSION & KEY TAKEAWAYS: DUAL ROLES

 Dual Roles of Simulation

 Simulations serve two primary roles—problem solving (trusted, user‑driven) and
insight gathering (exploratory, designer‑driven). Which role dominates depends
on:
 Who is using the simulation (user vs. designer)?
 How much epistemic uncertainty remains about the phenomenon?
 How confident are we that the simulation accurately represents reality?
 The more we reduce epistemic uncertainty, the more we can deploy the
simulation directly for problem solving without further designer intervention. We
can often infer a simulation’s current role simply by the type of questions it is
used to answer.