Design Theory and Methods using CAD/CAE

4.8. Summary

Efficient and accurate sensitivity analysis is essential to the success of gradient-based optimization, especially for large-scale applications that require substantial computational effort for function evaluations. We discussed three major approaches for sensitivity analysis—the brute-force finite difference, discrete, and continuum methods. We pointed out the pros and cons of individual approaches. In practical implementation, the semianalytical and continuum-discrete methods are general and efficient. In addition, considering accuracy, the continuum-discrete method is, in our opinion, the best approach for support of DSA in general and large-scale structural problems. However, it requires a more in-depth study for the reader to ...