发动机故障诊断的MONTE CARLO通解算法

给出了发动机故障诊断的 Monte Carlo通解算法,该算法可以有效地解决发动机故障诊断算法中由于故障方程存在多重共线性所引起的误诊、漏诊与多解问题。首先给出求解故障方程的 Monte Carlo算法,该算法能够保证得到满足故障方程的全部合理解;能够简单地用于各种故障相关性准则 (例如各种残差向量范数准则 )以及亚定故障方程的求解问题;并且算法简单易行,无需为每个特定情况专门编制计算程序。其次提出了基本解与通解的概念,并且给出了利用主成分分析与利用主因子模型求基本解的两种有效算法。利用基本解与通解算法可以将由于多重共线性引起的复杂故障诊断结果表示为简单明了的形式,有利于对诊断结果作出正确决策。用计算机模拟方法对算法的有效性进行了分析,对于 JT9D发动机气路方面的 2 4个实际故障样本,所给出的算法的确诊率为 86%～ 92 %

MONTE CARLO GENERAL SOLUTION ALGORITHM FOR ENGINE FAULT DIAGNOSIS

A Monte Carlo General Solution Algorithm is presented. This method is able to solve the problems such as mistaken and multiple solutions in the engine fault diagnosis algorithm based on fault equation, which result from the multi|collinearity of the fault equation. The first part of the paper provides a Monte Carlo algorithm for resolving engine fault equation, which consists of selecting every possible primary factor subset, resolving the sub|equation by using a random simulation method, isolating the reasonable solutions by using a fault correlation criterion, and selecting the independent solutions by using a similarity criterion. This method can ensure to obtain all reasonable solutions meeting the fault equation, can be applied for various fault correlation criteria (such as various residual vector norm criteria) and underdetermined fault equation, and the algorithm is very simple and facile, not necessary to compile specific program for every particular condition. The second part presents the concepts of basic and general solutions, and two methods to obtain them are given by using the principal component analysis and the principal factor model. The complicated diagnostic results caused by multi|collinearity can be transformed into a simpler and clearer form by using the basic and general solutions, which would facilitate to make correct diagnostic decision. A diagnostic effectiveness analysis of the algorithms is presented by using computer simulation, and a test of JT9D practical fault samples is made. The rate of correct diagnosis is up to 86%~92% for the 24 JT9D gas path fault samples.