Verification and validation are essential processes in computational science and engineering. Verification asks whether the numerical equations and computer implementation have been solved correctly. It may involve checking code, comparing with analytical solutions, testing simplified cases and examining numerical convergence. Validation asks whether the mathematical model is suitable for representing the real-world system. It often involves comparing model predictions with experimental measurements or observational data. A numerical result may be verified but still fail validation if the underlying model does not accurately represent reality. Both processes are therefore needed before computational results can be trusted in practical applications.
Verification and Validation of Numerical Results
Verification checks whether a numerical method and its implementation solve the mathematical problem correctly, while validation examines whether the mathematical model adequately represents the real system.
r = b − Ax
Explanation
Variables
| Symbol | Meaning |
|---|---|
r | Residual vector. |
A | Coefficient matrix. |
x | Computed solution vector. |
b | Right-hand-side vector. |
Example or application
A heat-transfer simulation may be verified by comparing its numerical solution with a known analytical solution for a simple case. It may then be validated by comparing predictions with laboratory temperature measurements.
Definition
Verification is the process of checking the correctness of a numerical model's implementation and solution procedure. Validation is the process of assessing whether the model adequately represents the real system for its intended purpose.
Purpose
To establish methods for checking the mathematical correctness, computational implementation and real-world relevance of numerical results.
Importance
Verification and validation reduce the risk of relying on incorrect simulations or misleading predictions. They are particularly important in engineering, climate science, medicine, aerospace and other fields where numerical decisions may have significant consequences.
Learning Objectives
Define verification and validation.
Explain the difference between the two processes.
Understand the role of residuals.
Identify methods for verifying numerical implementations.
Recognize the importance of comparison with experimental evidence.
Key Concepts
Verification
Validation
Residual
Analytical benchmark
Experimental data
Model accuracy
Computational testing
Working Principle
Verification examines the mathematical and computational correctness of the numerical procedure. Validation compares the model's predictions with observations or experiments to determine whether the model is suitable for its intended application.
Process
Define the numerical model and intended purpose.
Check the mathematical formulation.
Test the implementation using known or simplified cases.
Calculate residuals or other numerical checks.
Examine convergence and consistency.
Compare model predictions with experimental or observational data.
Document the limitations and reliability of the results.
Applications
Engineering simulation
Computational fluid dynamics
Climate modelling
Structural analysis
Scientific software testing
Medical and biological modelling
Worked Example
Problem
Verify a computed solution of the linear system Ax = b using a residual.
Steps
Let A = [[2, 0], [0, 3]], x = [1, 2]ᵀ and b = [2, 6]ᵀ.
Calculate Ax = [2, 6]ᵀ.
Calculate the residual r = b − Ax.
Therefore, r = [2, 6]ᵀ − [2, 6]ᵀ = [0, 0]ᵀ.
The computed solution satisfies this system exactly in the given arithmetic.
Answer
The residual is the zero vector, indicating that the stated solution satisfies the given linear system.
Common Misconceptions
Verification and validation are not interchangeable.
A small residual does not automatically prove that a mathematical model represents reality accurately.
A model may be validated for one purpose but unsuitable for another.
Passing a few test cases does not guarantee that software is free from all errors.
Practice Questions
| Question | Answer |
|---|---|
What is verification? | Checking whether the numerical implementation and solution procedure correctly solve the specified mathematical problem. |
What is validation? | Assessing whether the mathematical model adequately represents the real system for its intended purpose. |
What does a residual measure in a linear system? | The difference between the right-hand side and the product of the coefficient matrix with the computed solution. |
Key Takeaway
Verification checks whether the computation is performed correctly, while validation checks whether the model is suitable for representing reality.
Prerequisites
SCI-MAT-014-006
SCI-MAT-014-009
SCI-MAT-014-011
SCI-MAT-004-111
Reference
Roache, P. J. Verification and Validation in Computational Science and Engineering.
Oberkampf, W. L., and Roy, C. J. Verification and Validation in Scientific Computing.
Related Topics
SCI-MAT-014-008
SCI-MAT-014-024
SCI-MAT-015-521
SCI-MAT-015-550
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.