Robustness Evaluation
The robustness of a model, thus the reliable function mode within permitted tolerances, is a key topic in the virtual product development.
Method overview
Matrix of quadratic correlations
- Stochastic variables with correlation and distribution information
- Types of distribution (normal, log-normal, uniform, truncated normal, Weibull, discrete)
- Latin Hypercube Sampling, Monte Carlo Sampling
- Statistical postprocessing (analysis of variance, correlation analysis, principal components, coefficients of determination)
Features
Histogram of a single response with fitted probability distribution
optiSLang, a set of possible design realizations is generated by means of suitable sampling methods and these design realizations are computed. As result, the important input scatter can be identified and the variation of responses can be quantified. Robustness evaluations secure thereby operativeness, safety and reliability for relatively frequent events (up to 2-sigma level of a normal distribution).
Focus
Combination of optimization and robustness evaluation
Often, the robustness is an essential part of the design optimization task. From our experience of introducing optimization and robustness evaluation methodology in virtual product development processes, it is absolutely necessary to understand both disciplines to be able to formulate combined problems. Therefore, it is recommended to start with a consecutive approach of using sensitivity analysis, robustness evaluation and deterministic optimization to achieve iterated knowledge for a robust and optimal design.
Postprocessing
The result monitoring in optiSLang is based on interactive windows which allow the user to get the desired information simply and quickly.
Distributors
optiSLang is sold worldwide. A list of our distributors can be found here.
Your Contact
