| 引用本文: | 邱文昊,连光耀,闫鹏程,黄考利.考虑贡献率和可信度的测试性试验优化方法[J].哈尔滨工业大学学报,2022,54(12):95.DOI:10.11918/201905219 |
| QIU Wenhao,LIAN Guangyao,YAN Pengcheng,HUANG Kaoli.Testability demonstration test optimization method considering contribution rate and credibility[J].Journal of Harbin Institute of Technology,2022,54(12):95.DOI:10.11918/201905219 |
|
| 摘要: |
| 针对现有测试性试验方法中系统级先验信息获取困难，先验分布赋权不合理以及样本量过大等问题，提出了基于分系统贡献率和先验分布可信度的测试性验证试验优化方法。首先，系统分析了测试性多源先验信息，定义了分系统贡献率，在此基础上利用信息论方法对分系统先验数据进行折算得到系统级先验数据；然后，通过相似性度量检验先验数据与实装试验数据的相容性，并提出采用逼近理想点排序-层次分析法(Technique for order preference by similarity to ideal solution-analytic hierarchy process, TOPSIS-AHP)计算先验分布可信度，进而确定混合先验分布；最后，基于分系统先验信息确定的混合先验分布，运用序贯验后加权检验(Sequential posterior odd test, SPOT)方法制定试验优化方案。实例分析表明，由基于贡献率的数据折算和基于可信度的权值计算方法得到的混合先验分布更加准确，与序贯概率比检验(Sequential probability ratio test, SPRT)方法相比，该试验方案样本量平均减少18.6%，与Bayes方法相比平均减少61.1%，而且该方法可以有效降低双方风险。考虑贡献率和可信度的SPOT试验方案在先验信息获取、先验分布权值、试验样本量、双方风险等方面均具有较好的应用效果。 |
| 关键词: 测试性 试验方案 先验分布 贡献率 可信度 序贯验后加权检验 样本量 |
| DOI：10.11918/201905219 |
| 分类号:TJ760, V219 |
| 文献标识码:A |
| 基金项目: |
|
| Testability demonstration test optimization method considering contribution rate and credibility |
|
Wenhao QIU1,2, Guangyao LIAN2, Pengcheng YAN2, Kaoli HUANG1,2
|
|
1.Shijiazhuang Campus, Army Engineering University of PLA, Shijiazhuang 050000, China;2.Unit 32181 of PLA, Xi'an 710000, China
|
| Abstract: |
| A testability test optimization method based on subsystem contribution rate and prior distribution credibility was proposed to deal with the problems such as the difficulty in obtaining system-level prior information, the unreasonable calculation of prior distribution weight, and the large sample size in existing methods. First, the testability multi-source prior information was systematically analyzed and the subsystem contribution rate was defined. On this basis, the subsystem prior data was converted to obtain the system-level prior data by the information theory. Then, the similarity measure of prior distribution was introduced to characterize the compatibility between prior distribution and experimental data. Next, the technique for order preference by similarity to ideal solution-analytic hierarchy process (TOPSIS-AHP) method was proposed to determine the credibility of prior distribution, and then the mixed prior distribution was obtained. Finally, on the basis of the mixed prior distribution determined by subsystem prior information, the sequential posterior odd test (SPOT) method was used to make the test optimization plan. Case analysis results show that the mixed prior distribution obtained from the data conversion method based on contribution rate and the weight calculation method based on credibility was more accurate. The sample size of the SPOT method decreased by 18.6% on average compared with the sequential probability ratio test (SPRT) method, and was 61.1% smaller than the Bayes method. Besides, this method could effectively reduce the risk of both sides. The SPOT method considering contribution rate and credibility has good application effects in acquisition of prior information, weight of prior distribution, number of test samples, and risk of both sides. |
| Key words: testability test planning prior distribution contribution rate credibility sequential posterior odd test sample size |
