Computational simulation models support a rapid design process. Given model approximation and operating conditions uncertainty, designers must have confidence that the designs obtained using simulations will perform as expected. The traditional approach to address this need consists of model validation efforts conducted predominantly prior to the optimization process. We argue that model validation is too daunting of a task to be conducted with meaningful success for design optimization problems associated with high-dimensional space and parameter spaces. In contrast, we propose a methodology for maximizing confidence in designs generated during the simulation-based optimization process. Specifically, we adopt a trust-region-like sequential optimization process and utilize a Bayesian hypothesis testing technique to quantify model confidence, which we maximize by calibrating the simulation model within local domains if and when necessary. This ensures that the design iterates generated during the sequential optimization process are associated with maximized confidence in the utilized simulation model. The proposed methodology is illustrated using a cantilever beam design subject to vibration.
Maximizing Design Confidence in Sequential Simulation-Based Optimization
Contributed by the Design Automation Committee of ASME for publication in the JOURNAL OF MECHANICAL DESIGN. Manuscript received June 24, 2012; final manuscript received March 12, 2013; published online June 10, 2013. Assoc. Editor: Timothy W. Simpson.
- Views Icon Views
- Share Icon Share
- Cite Icon Cite
- Search Site
Li, J., Mourelatos, Z. P., Kokkolaras, M., Papalambros, P. Y., and Gorsich, D. J. (June 10, 2013). "Maximizing Design Confidence in Sequential Simulation-Based Optimization." ASME. J. Mech. Des. August 2013; 135(8): 081004. https://doi.org/10.1115/1.4024470
Download citation file:
- Ris (Zotero)
- Reference Manager