A multi-stage, parallel implementation of probabilistic design optimization in an MDO framework

Abstract

Until recently, the computational expense of probabilistic analysis has made its application to all but very simplistic design problems impractical. Consequently, probabilistic optimization has been considered prohibitively expensive, particularly for complex multidisciplinary problems. With the increase in computing power and availability of probabilistic analysis and optimization approaches and tools, however, the combination of these technologies can facilitate effective probabilistic analysis and optimization for complex design problems. In this paper we present a multi-stage implementation of probabilistic design analysis and optimization, developed in an MDO framework that supports parallel execution of design points required for probabilistic analysis. This MDO framework is provided by the commercial software package iSIGHT. Probabilistic analysis and optimization tools incorporated are described, as well as their multi-stage, parallel implementation within this MDO framework for efficient, automated probabilistic design optimization. This implementation is then demonstrated for the probabilistic preliminary design of a commercial aircraft. Probabilistic solutions presented demonstrate the need for including uncertainty in design analysis and optimization, and the resulting trade-off necessary with respect to achievable objective values and desired design reliability. A comparison of sequential versus parallel execution for each stage in the approach is also presented; parallel execution results in up to 60% reduction in execution time for this problem.