System Reliability-Based Design Optimization Under Input and Model Uncertainties

摘要

For reliability-based design optimization (RBDO), sensitivity analysis capability is a major bottleneck for broader use of RBDO methods in multidisciplinary M&S applications of Army complex physical systems. To overcome this bottleneck, a sequential sampling-based dynamic Kriging (DKG) method is developed. The DKG method has been integrated with the Iowa-RBDO software system that is developed under US Army TARDEC sponsorship. For large- scale simulation models, the total number of simulations carried out for surrogate modeling could be limited due to computation resource. In such cases the inaccuracy and uncertainty of the surrogate model needs to be quantified. For this, the weighted Kriging variance method is developed using the corrected Akaike information criterion (AICc) to generate a confidence interval of the surrogate model; and the upper bound of the confidence interval is used to obtain confidence-based RBDO optimum design that satisfies the target reliability. In practical industrial application, often input data are not sufficient enough to generate true input distribution models for reliability analysis and RBDO. When only the limited input data are provided, uncertainty is induced on the input probability model and this uncertainty propagates to the reliability output which is defined as the probability of failure. Thus, the reliability output is considered to have a probability distribution in this research, which is obtained as a combination of consecutive conditional probabilities of input distribution type and parameters using Bayesian approach. Using the probability of the reliability output as constraint, a confidence-based RBDO (C-RBDO) problem is formulated. For effective C-RBDO process, the design sensitivity of the new probabilistic constraint is derived. As an alternative for DKG, a virtual support vector machine (VSVM) is developed to improve the efficiency of the sampling-based RBDO.