发动机故障诊断主因子模型故障相关性准则的研究

】主因子模型是发动机状态诊断（基于故障方程的故障诊断）理论的一个新进展。故障相关性准则的正确选取是主因子模型的核心问题。本文给出三种建立故障相关性准则的途径，即利用线性回归分析、时序分析和系统辨识理论中的定阶准则建立故障相关性准则，以及利用数理统计中的线性模型线性假设显著性检验理论和区间预测理论建立故障相关性准则。文中对所得到的１１个准则进行了比较。研究结果表明，利用线性模型线性假设显著性检验理论建立的两个准则（变量显著性准则ＶＳＣ和子集显著性准则ＳＳＣ）要比近代回归分析、系统辨识或时序分析中通常采用的选择最优变量子集的定阶准则具有更强的故障辨识能力和鉴别线性模型多重共线性的能力。

A Study on the Fault Correlation Criterion for Jet Engine Diagnosis

The Primary Factor Model is a significant progress in the engine state diagnosis (engine fault diagnosis based on fault equation) theory.The proper choice of fault correlation criterion is a critical problem for the Primary Factor Model. Three avenues for developing fault correalation criterion are presented in the paper: to develop fault correalation criteria by using the order definition criteria in the system identification theory (the criteria for selecting optimal variable subclass in modern regression analysis), and by using the significance test theory of linearhypothesis and the interval prediction theory. Eleven fault correalation criteria are obtained and a comparision between them is made. It is shown that the VSC and SSC criteria derived from the hypothesis test theory are more effective in fault and multi-collinearity identification than the order definition criteria for selecting optimal variabe subclass.