系统工程与电子技术
繫統工程與電子技術
계통공정여전자기술
SYSTEMS ENGINEERING AND ELECTRONICS
2014年
12期
2566-2570
,共5页
测试性验证%Bayes理论%动态增长模型%小子样%变总体
測試性驗證%Bayes理論%動態增長模型%小子樣%變總體
측시성험증%Bayes이론%동태증장모형%소자양%변총체
testability verification%Bayes theory%dynamic growth model%small sample%varying population
现有测试性验证试验方案一般都需要较大的故障样本量，利用装备研制阶段的测试性数据则可能有效解决这一问题。但研制阶段的测试性数据和现场试验数据一样属于“小子样”且具有“变总体”特点，为此提出一种基于Bayes理论的测试性验证方法。首先根据装备研制各阶段积累的测试性试验信息构建装备测试性参数故障检测／隔离率（fault detection rate／fault isolation rate，FDR／FIR）的动态增长模型，用以揭示装备测试性水平的动态变化规律，并对其测试性指标进行预测。然后根据最大熵原理计算系统FDR／FIR的先验分布。最后结合少量现场试验数据，根据Bayes最大验后风险准则制定装备测试性验证方案，对装备测试性指标进行验证。通过实例对比分析表明，该方法有效融合了研制阶段数据与现场试验数据，能在小样本的情况下提高验证结论的置信度，降低评估风险。
現有測試性驗證試驗方案一般都需要較大的故障樣本量，利用裝備研製階段的測試性數據則可能有效解決這一問題。但研製階段的測試性數據和現場試驗數據一樣屬于“小子樣”且具有“變總體”特點，為此提齣一種基于Bayes理論的測試性驗證方法。首先根據裝備研製各階段積纍的測試性試驗信息構建裝備測試性參數故障檢測／隔離率（fault detection rate／fault isolation rate，FDR／FIR）的動態增長模型，用以揭示裝備測試性水平的動態變化規律，併對其測試性指標進行預測。然後根據最大熵原理計算繫統FDR／FIR的先驗分佈。最後結閤少量現場試驗數據，根據Bayes最大驗後風險準則製定裝備測試性驗證方案，對裝備測試性指標進行驗證。通過實例對比分析錶明，該方法有效融閤瞭研製階段數據與現場試驗數據，能在小樣本的情況下提高驗證結論的置信度，降低評估風險。
현유측시성험증시험방안일반도수요교대적고장양본량，이용장비연제계단적측시성수거칙가능유효해결저일문제。단연제계단적측시성수거화현장시험수거일양속우“소자양”차구유“변총체”특점，위차제출일충기우Bayes이론적측시성험증방법。수선근거장비연제각계단적루적측시성시험신식구건장비측시성삼수고장검측／격리솔（fault detection rate／fault isolation rate，FDR／FIR）적동태증장모형，용이게시장비측시성수평적동태변화규률，병대기측시성지표진행예측。연후근거최대적원리계산계통FDR／FIR적선험분포。최후결합소량현장시험수거，근거Bayes최대험후풍험준칙제정장비측시성험증방안，대장비측시성지표진행험증。통과실례대비분석표명，해방법유효융합료연제계단수거여현장시험수거，능재소양본적정황하제고험증결론적치신도，강저평고풍험。
The fault sample size which is required in existing testability demonstration test schemes may be reduced by using data in the development phase.However,the growth test data are“small samples”with“varying population”.A new testability verification method based on the Bayes theory is proposed.Firstly,the proposed method establishes a dynamic growth model of the test parameters based on multiple phases’samples,which is used to describe the changing rule of equipment’s testability and predict the fault detection rate (FDR)and the fault isolation rate (FIR).Then,the prior distribution of the system’s FDR/FIR is calculated based on the maximum entropy principle.Finally,a new testability determination scheme is defined to verify FDR/FIR,ac-cording to Bayes maximum posterior risk rules with a small size of field trial data.The practical comparison shows that this method,in which the growth test data and field trial data can be fused effectively,can reach an evaluation conclusion with a high confidence level under small sample circumstance,and reduce the risk of evaluation.
