χ~2统计检验的风险分析及改进算法  

Risk analysis and improved algorithm of chi-square statistical test

作　　者：李金梁[1] 张宗麟[1] 胥勇军[1] 

机构地区：[1]空军工程大学工程学院,西安710038

出　　处：《中国惯性技术学报》2008年第1期108-112,共5页Journal of Chinese Inertial Technology

基　　金：国防科技重点预研项目(51309060302)

摘　　要：为了更好地满足航空领域对故障检测漏检风险的特殊要求,研究了x^2检验法的误检和漏检风险问题,通过具体算例和仿真分析得出x^2两类风险与故障幅度之间的关系,并指出单历元x^2检验法在小故障幅度条件下的误检和漏检风险不能很好地满足航空需求。因此,利用统计理论和序贯检验思想,提出了一种基于χ2?CUSUM统计检验的改进算法,能够兼顾数据窗前端和终端的检测可靠性,有效降低x^2检验的漏检风险,最后的联邦滤波仿真曲线和相关数据表明新算法的有效性和可行性。Aiming at the special demand of aviation domain to fault missed detection risk, the undesirable false alarms and undesirable lemissed detections risk problem of the chi-square test was searched, and material example and simulation analyzing indicate that snapshot algorithm is incapable of satisfying the aviation requirements for detection risks. Consequently, by statistics theory and sequential test idea, an improved algorithm based on χ^2- CUSUM statistics test was proposed, which is good for giving attention to both the head and terminal detection reliability of data window and decreasing missed detection risk of chi-square test. The final federated filtering simulation shows its validity and feasibility.

关 键 词：故障诊断 Χ^2检验 漏检风险 序贯检验 联邦滤波 

分 类 号：U666.1[交通运输工程—船舶及航道工程]