Reliability Evaluation of Small-sample Binary Systems Based on Relative Entropy Inheritance Factor

In the reliability evaluation of small-sample binary system, the inheritance factor in the hybrid Beta prior Bayesian method is determined by expert experience and subjective judgment, which affects the accuracy of the prior distribution and reliability evaluation results. A quantitative calculation method for the relative entropy inheritance factor is proposed. A relative entropy-chi-square fitting inheritance factor (RE-CSIF) model is proposed by analyzing the probability distribution differences and similarity boundary conditions of prior samples and test samples, which enables the quantitative calculation of relative entropy inheritance factor. To improve the input data quality of the RE-CSIF model, a four-stage test data preprocessing framework including data cleaning, data conversion, quality evaluation and data fusion is constructed to solve the problem of test data quality assessment. The effectiveness of the proposed method is verified through Monte Carlo simulation. Compared with traditional inheritance factor calculation methods, the RE-CSIF Method improves the accuracy of reliability estimation by 17% and the coverage confidence level by 41% when the reliability levels of prior information and actual test samples are highly consistent; and when the reliability levels are relatively close, it improves the accuracy of reliability estimation 5.4% and the coverage confidence level by 7.9%. The applicability of the proposed method is verified by taking the reliability test data of a fuze system as an example.