System Dependent Failure Probability Model Based on Monte Carlo-Neural Network

Dependent failure can significantly reduce the redundant system reliability,so it has been widely paid attention to in the engineering practice.According to the reliability theory and analysis of failure mechanism on the components,a component’s failure probability was regard to be the conditional failure probability with its stress,so that the mathematical expression for system dependent failure probability was given.Using Monte Carlo simulation,distributed type of conditional failure probability of the components was obtained;Extracting dependent failure information from given low-fold failure data and establishing the neural network model,its distributed parameters were obtained.The model can predict any multiplicity dependent failure probability,and also be applicable to other system with same components and environment and different sizes.An example is provided to illustrate the application,and its predictability is tested,which shows that the approach is accurate and feasible.