A time-dependent reliability analysis method based on neural network response surface

Uncertainty exists widely in the engineering field, and the reliability of products has time-dependent characteristics due to the accumulation of product use time and material performance degradation. Traditional time-dependent reliability solutions such as span rate and Monte Carlo are difficult to be applied in engineering practice due to their complex principles and low computational efficiency. Therefore, this paper proposes a time-dependent reliability analysis method based on neural network response surface learned form the idea of limit value. This method uses inverse reliability sampling and establishes a response surface model between the design variables and the limit value of the limit state function to convert the time-dependent reliability into time-invariant reliability. Then, the traditional method is used to solve the time-dependent reliability, which improves calculation efficiency and accuracy of time-dependent reliability analysis. Finally, two examples are used to verify the effectiveness of the method.