Managing Uncertainty in Bilevel Robust Design Optimization

点击翻译按钮获取中文摘要

This grant provides funding for the development of a collaborative optimization framework for the robust design of complex engineering systems such as automobiles, aircraft or consumer products. The collaborative optimization framework will account for and manage the uncertainties in the performance predictions generated by the computer simulation tools used for the design of these systems. Each computer simulation model is an abstraction of reality and has some uncertainty associated with its performance predictions. This uncertainty must be accounted for in the simulation based design process. An implicit method for estimating system performance uncertainties within a bilevel optimization algorithm that employs decomposition techniques to facilitate distributed computation will be developed. The methodology will account for both the uncertainty associated with design inputs and the uncertainty of performance predictions from each of the simulation tools. A mathematical proof of convergence will be developed to validate the bilevel optimization algorithm being developed in this investigation. The framework will be implemented in a distributed computing environment providing for parallel computation and concurrent design. Industry partners will test the framework and measure the computational improvements using a suite of benchmark test problems.It is anticipated that the use of the collaborative optimization framework developed in this research will lead to reduced product development times at reduced cost and risk. The collaborative optimization framework will facilitate the concurrent design of complex engineering systems in a parallel-computing environment. The benefits of parallel computation lead to reductions in product development times. The ability to manage uncertainty and risk in this parallel design environment will ensure robust performance of the resulting system. The development of the non-deterministic collaborative optimization framework will demonstrate that designers can effectively manage both the uncertainty and risk associated with the simulation based design of new products in a parallel computing environment.