Altair HyperStudy Capabilities
Design of Experiments
Design of Experiments (DOE) methods in HyperStudy include:
- Full Factorial
- Fractional Factorial
- Box-Behnken
- Plackett-Burman
- Central Composite Design
- Latin HyperCube
- Hammersley
- User defined and direct input of external run-matrix.
The study matrix can consist of continuous, discrete or character strings that can be either controlled or uncontrolled.
Approximations
The approximation module allows creation of different approximations for different responses. Available approximation methods are Least Squares Regression, Moving Least Squares and HyperKriging. Response surfaces created using the approximations module can be used for performing optimization and stochastic studies.
Multi-Disciplinary, Reliability, and Robustness Optimization
HyperStudy offers multidisciplinary study capabilities as well as reliability and robustness optimization. HyperStudy's comprehensive optimization algorithms include:
- adaptive response surface method
- sequential quadratic programming
- method of feasible directions
- genetic algorithm
- sequential optimization and reliability analyses.
Optimization studies can be performed using either exact simulation or approximation model. In addition, HyperStudy provides an API to incorporate external optimization algorithms into a study.
Stochastic Studies
The stochastic study capability in HyperStudy allows engineers to assess reliability and robustness of designs and provide qualitative guidance to improve and optimize based on these assessments. HyperStudy includes Simple Random, Latin Hypercube and Hammersley sampling methods along with statistical distribution functions such as normal, uniform, triangular, Weibull and exponential. Stochastic studies can be performed using either exact simulation or the approximation model.
Post-Processing and Data Mining
HyperStudy helps engineers to gain a deeper understanding of a design through extensive post-processing and data-mining capabilities. This significantly simplifies the task of studying, sorting and analyzing results. Study results can be post-processed as statistical data, correlation matrices, scatter plots, interaction effect plots, histograms, and snake view plots. In addition, HyperStudy provides a series of data-mining tools such as principal components analysis and clustering analysis.
Evaluation and Rating
A large database for signal analyses and comparison functions allows engineers to perform correlations. These correlations can then be evaluated and rated based on user defined criteria.