As TWS enhances our finite element software, verification and validation activities are paramount. These activities ensure that what we develop not only performs as mandated by mathematics, but also mirrors what occurs on the shop floor. In the end, Third Wave Systems’ verification and validation processes for its finite element models are cyclical, with interrelated development and evaluation activities.
Verification: Mimicking the Equation
At its essence, TWS software is a compilation of mathematical equations that represent the physical world. Therefore, when a new feature or enhancement is suggested for development, Computational Mechanics Engineers (CMEs) and Applications Developers must first identify an equation (analytical solution) that delivers the anticipated result. This equation is then implemented into AdvantEdge FEM as a model algorithm. The Product Quality and Development teams test the algorithm, verifying that it performs as intended (matches the analytical solution), and identifying any unintended results (bugs). As the TWS process stands today, once a feature/enhancement has been proven to consistently produce the intended result without adversely affecting software performance, it is considered ready for release.
Validation: Evaluating the Equation
During verification, the assumption is made that an equation will deliver results that mimic the physical world. To see if this assumption was correct, TWS’ Applications Engineering team uses complex equipment to conduct a mirage of physical tests. Data gathered from these tests is then sent to Third Wave’s CAE and Product Quality teams for comparison against software results. The original equation is reviewed and, with input from the TWS Development team, a new equation may be identified that produces a stronger correlation between measured data and the software model’s output. Alternatively, the measurement strategy is also evaluated, and an improved method of testing may be identified from the model. The cycle begins again.
The kicker? A wild card called measurement uncertainty—in short, no measurement is exact. When a quantity is measured, the outcome depends on the measuring system, the measurement procedure, the skill of the operator, the environment, and other effects. Even if the quantity were to be measured several times, in the same way and in the same circumstances, a different measured value would likely be obtained each time, assuming that the measuring system has sufficient resolution to distinguish between the values. In other words, measurement uncertainty is a margin of error, and this margin of error has yet to be defined for measurement (and as there is not an international authority for standards, each country may define different expectations).
Despite this obstacle, TWS strives to deliver the most accurate finite element software possible. The verification/validation cycle is endless, but even though outputs and measured data will never be identical, the company is committed to validating its software to the best of its ability.
 Back
|
VERIFICATION |
VALIDATION |
Done by |
Development, Product Quality, Applications Developers |
Applications Engineering, CAE,
Development, Product Quality |
Measures |
Agreement between model
algorithm output and analytical solution |
Agreement between model output and measured data
|
In Other Words... |
If analytical solution is F=ma, model algorithm outputs F=ma |
Does F=ma give us results that align with measured data? |
Considerations
& Questions |
- Does the model algorithm match the analytical solution?
|
- Does the measured data model the salient/appropriate physics?
- Are there numerical influences affecting the measured data?
- What errors and uncertainties are in the measured data?
- What errors and uncertainties are in our model?
|
Goal |
Confirm that model algorithm
implementation was done
correctly. |
Establish correlation between modeling results and measured data. |
Outcome |
- Bugs resulting from algorithm functionality have been identified and resolved
- Model algorithm output matches analytical solution
- Confidence that implemented algorithm is well understood
|
- Model output more closely represents measured data
|
Back |
