Importance measure analysis of design variables and uncertain parameters in multidisciplinary systems

Abstract

The traditional reliability-based multidisciplinary design optimization consumes large computation resources due to a large number of design variables and uncertain parameters in multidisciplinary systems. Therefore, the screening of important design variables and uncertain parameters contributes to reducing the difficulty of optimization. However, the current methods ignore the interaction between design variables and uncertain parameters. Via a probabilistic and nonprobabilistic reliability model, the work in this paper establishes the concept of a comprehensive reliability-based importance measure that considers both design variables and uncertain parameters. An efficient method based on the Bayesian theorem is applied to calculate the above-mentioned indices of the proposed importance measure, which can estimate importance measure accurately adopting the same samples with reliability analysis. Therefore, the probabilistic method has faster convergence speed than traditional methods. The nonprobabilitic analysis method can promote the computation efficiency but ensure accuracy when the system performance functions are monotonic. Furthermore, the developed approach contributes to decreasing the time cost of reliability-based optimization in multidisciplinary systems. Additionally, the developed approach can also be applicable to single-disciplinary systems. Moreover, two practical engineering examples and the multidisciplinary design of a hypersonic wing are employed to demonstrate the validity and applicability of the proposed method.