Reliability analysis for multidisciplinary systems with the mixture of epistemic and aleatory uncertainties

Abstract

SUMMARYEpistemic and aleatory uncertain variables always exist in multidisciplinary system simultaneously and can be modeled by probability and evidence theories, respectively. The propagation of uncertainty through coupled subsystem and the strong nonlinearity of the multidisciplinary system make the reliability analysis difficult and computational cost expensive. In this paper, a novel reliability analysis procedure is proposed for multidisciplinary system with epistemic and aleatory uncertain variables. First, the probability density function of the aleatory variables is assumed piecewise uniform distribution based on Bayes method, and approximate most probability point is solved by equivalent normalization method. Then, important sampling method is used to calculate failure probability and its variance and variation coefficient. The effectiveness of the procedure is demonstrated by two numerical examples.

Reliability modeling and optimization of die-casting existing epistemic uncertainty

“…The RBMDO procedure presented in this paper is based on the frame of SORA, which shows high computational efficiency [26]. Moreover, the RBMDO procedure is also based on J-C method, which possesses high computational efficiency [24]. Therefore, the computational efficiency of the RBMDO procedure is high, which can be verified by this engineering application.…”

“…Just proved in Ref. [24], the reliability based on the piece-wise uniform distribution in focal element is conservative, which is rational for engineering practice. Therefore, the RBMDO procedure presented in this paper is suited to optimize the multidisciplinary system with epistemic uncertainty.…”

“…In reference [24], the authors presented a novel reliability analysis procedure for multidisciplinary system with epistemic and aleatory uncertain variables. First, the epistemic uncertain variables are represented by ET and their probability density function is assumed piecewise uniform distribution based on Bayes method, and approximate most probability point is solved by equivalent normalization method(named as the modified J-C method).…”

“…The modified J-C method presented in Ref. [24] may be used to solve MPP for uncertainties. In this procedure, conventional MDO method such as multidisciplinary feasible method and bi-level integrated system can be used in this RBMDO method.…”

“…It has been pointed out by Oberkampf and Helton [19] and Bae et al [20] that ET has an intrinsic capability to handle both aleatory and epistemic uncertainties. Recently, some researchers have investigated reliability design based on ET and its engineering applications [21][22][23][24].…”