基于PCA的装甲车辆PHM系统性能度量指标筛选

Screening of Performance Measure Indices for PHM System of Armored Vehicles Based on PCA

构建系统全面的指标体系是确保产品性能科学评价的前提与基础,而指标筛选是指标体系构建的关键环节。针对装甲车辆PHM系统性能度量指标体系构建过程中存在的原始指标集冗余性较高、针对性不强的问题,采用主成分分析法对原始指标集进行约减。本文以智能化程度、系统失效率等14个用户性能度量指标为例,首先依据主成分载荷值的大小进行指标筛选,保留了前三个主成分中载荷值较大的10个指标;然后考虑到初选指标之间可能存在信息重叠的现象,对其进行相关性分析,剔除了人员设备使用效率和实时性两个指标,用剩余的8个指标替换了原始指标集;最后分析计算了指标的信息贡献率,结果表明:用57%的指标表达了原始指标集94.38%的信息,验证了该方法的可行性和有效性。

Building a comprehensive index system is the premise and basis for the scientific evaluation of product performance, and the index screening is the key in the construction of index systems. To solve the problems of high redundancy and low pertinence of the original index set in the constructing process of product performance index system,the principal component analysis method and the correlation analysis were used to reduce the indices. This paper used 14 user performance indices for example. The principal component analysis method was used to screen them, and the 10 performance indices with the largest principal component load value were chosen. Due to the possibility of overlapping information between the selected indices, the correlation analysis was used to remove two relevant indices. The remaining eight indices were used to replace the original index set. To verify the feasibility and effectiveness of the method, the information contribution rate of the remaining indices was calculated. The results show that the remaining indices can present 94. 38% of the information the original index set cein do. It can be assumed that this method is feasible and effective.