In this paper a cycle-based alternating anti-optimization approach using Bounded-But-Unknown uncertainties is studied on the basis of a practical application. The underlying optimization technique is based on use of Response surface and undergoes series of cycles before it can co
...
In this paper a cycle-based alternating anti-optimization approach using Bounded-But-Unknown uncertainties is studied on the basis of a practical application. The underlying optimization technique is based on use of Response surface and undergoes series of cycles before it can converge. The basic anti-optimization technique looks at the worst case scenario by finding the worst settings of the uncertainties for each constraint evaluation separately. This Rigorous anti-optimization technique involves two-level optimization, in which anti-optimization is nested within the main optimization, making it computationally exhaustive. In order to make the anti-optimization technique computationally efficient, in the cycle-based alternating technique, anti-optimization is carried out at the end of every cycle of the main optimization instead of at every design during the cycle. Additionally, in the present paper, a nested parallel computing strategy is developed in order to make the cycle-based alternating technique computationally efficient when a cluster of computers is available for function evaluations in parallel. This is particularly essential in case of practical problems involving expensive function evaluations, such as using Finite Element Analysis. In this paper, the cycle-based alternating technique combined with nested parallel computing is applied to the uncertainty-based shape optimization of a Shape Memory Alloy Microgripper.@en
