基于后验风险确定故障样本量的Bayes方法

针对目前测试性故障样本量的确定方法过于粗糙和试验样本量过大的问题,提出了合理运用信息熵方法对装备系统各单元的测试性先验信息进行信息融合,得到装备系统级测试性试验数据。在此基础上得到测试性指标的先验分布,并进一步通过Bayes后验风险准则确定故障样本量及试验方案。以某型电动舵机系统各模块的试验数据为例,以故障检测率为测试性指标,经过分析和计算,发现运用所提方法得到的故障样本量相比传统方法明显减少,从而减少了试验成本,同时得到的测试性指标相对误差较小,保证了可信度。

Bayes method for determining fault sample size based on posterior risk

In view of the problem that the old approach of determining the sample size is too rough and the test sample size is too large, a reasonable application of the information entropy method is proposed to obtain the test data of the equipment system by fusing the prior information of each unit of the equipment system. The testability indices of the prior distribution are obtained, and the fault sample size and the test scheme are further determined by Bayes posterior risk criterion. The test data of an electric actuator system’s module are taken as an example and the failure detection rate is adopted as the test index. Compared with the traditional method, the failure sample quantity obtained by this method is reduced obviously through analysis and calculation, so the test cost is reduced. Meanwhile, the relative error of the test index is little, which guarantees the credibility. It is proved that the method reduces the test cost and guarantees the credibility.