关键词: 数字装备试验, 贝叶斯方法, 假设检验, 样本量有限

Abstract:

The fidelity verification of digital equipment is often limited by experimental funding, time, and resources, resulting in an extremely limited sample size available. How to determine the minimum sample size to reduce experimental costs while ensuring the accuracy of indicator verification has become a key issue that needs to be urgently addressed. Therefore, a fusion Bayesian dynamic sample size estimation method is proposed for digital equipment tests. This method is based on the Bayesian hierarchical framework, and prior information and the hypothesis testing principle of the frequency school are effectively utilized to construct a posterior distribution model of digital equipment test response. On this basis, preliminary sample size estimates are derived under Bayesian posterior confidence considering alternative hypotheses. Furthermore, a dynamic adjustment mechanism for the sample size is designed to analyze mid-term test data using half of the samples. Based on this, the final required sample size is dynamically updated and optimized to efficiently solve the design problem of sample size with limited resources. The proposed method is applied to the fidelity verification test of anti-ship missile and radar digital models. Compared with the sample size determination methods of the frequency school, the composite error and composite false judgment rate of the proposed method are improved by more than 18.4%, which demonstrates the effectiveness of the proposed method.

Key words: digital equipment test, Bayesian method, hypothesis test, limited sample size