An approach to system reliability analysis with fuzzy random variables

Nondeterministic variables of certain distributions are employed to represent uncertainties, which are usually treated as the stochastic factors to reliability models. However, model parameters may not be precisely represented due to some factors in engineering practices, such as lack of sufficient data, data with fuzziness and unknown or non-constant reproduction conditions. To address these issues, fuzzy random variables are implemented and two developments are made in this paper. The first development is that the Saddlepoint Approximation (SAP)-simulation is extended to conduct reliability analysis accounting for the time-dependent degradation process and fuzzy random variables, and we attempt to give a method to select a proper saddlepoint. The second development is that two system reliability analysis methods are proposed for different scenarios of reliability modeling processes. It could be suitable for the system consisting of structural components with gradual failure, whose reliability can be obtained by the method in the improved SPA-simulation, also for system consisting of components with sudden failure, whose reliability can be acquired from site field or experiments. An illustrated example is followed to testify the proposed methods.

► A fuzzy physics-based reliability analysis method is proposed.
► A selection method for proper saddlepoint is presented.
► System reliability analysis methods for complex systems are provided.