| MANATEE software | e-NVH Laboratory |
|---|---|
| Sub-module | LAB.SA |
| Function | Sensitivity analysis and parameter sweep |
| Validation | NUM_006, MS_SPMSM_SS_001 |
| See also | LAB.MCO Multi-objective Constrained Optimization, LAB.MDE Multidimensional Design Explorer |
Objectives
This article details LAB.SA module to optimize the design of electrical machines using sensitivity analysis and parameter sweep under MANATEE software.
These methods can be useful to identify the best design variables to be changed in order to have an influence on a specific response of the model.
Summary
This module of MANATEE software can calculate the sensitivity of a response variable with respect to
design variables and quantify the correlation factors, using different sampling stategies of the design space to be explored (e.g. Uniform Sampling or Latin Hypercube Sampling).
The analysis can also include the sensitivity to geometrical constraints, and user-defined design constraints.
This module can be coupled to any MANATEE elementary models (electrical, electromagnetic, structural, acoustics).
It can be used to calculate the sensitivity of some design variables to some objectives before launching the optimization process. Indeed, optimization algorithms are generally time consuming so it is therefore important to include only the most relevant design variables in the optimization process.
The results of the module can be conveniently visualized with the multidimensional design explorer.
- Example of parameter sweep made on an IPMSM rotor mmf harmonics in EV HEV NVH application
Graphical post-processing
The complete list of plot commands is available on MANATEE plot commands webpage. The following visualizations are more particularly related to this module:
- Correlation matrix between design variables and response variables (plot_MS_correlations)
- All response variables as a function of the simulation cases (plot_MS_all_resp_var)
- 5D design visualization tool (plot_MS_viewer)
- A specific response variable as a function of the simulation case (plot_MS_resp_var)
- Design response variables in 3D (plot_MS_resp3D)