An output reliability assessment method for initiating devices based on fiducial inference

To address the issues of limited sample size and low accuracy of parameter models in the output reliability assessment of initiating devices, a method based on fiducial inference is proposed. A fiducial inference framework is first established. To verify its effectiveness, numerical simulations are conducted to compare it with the normal statistical tolerance limits method specified in GJB376A-2019. The results show that the proposed method achieves higher estimation accuracy under most sample size conditions. In cases with fewer than 10 samples and a high coefficient of variation in output parameters, the estimation bias is reduced by approximately 20% compared to the traditional method. Moreover, under extremely small sample conditions (n = 3), the fiducial inference method remains stable, demonstrating strong applicability and robustness. It is therefore better suited to scenarios with limited data, effectively mitigating the sample size dependence of conventional methods and providing a novel approach to assessing the output reliability of initiating devices. Its performance is also compared with Bootstrap and Bayesian approximate limit methods to highlight its superiority.