发动机故障诊断的变记忆约束滤波算法

为提高故障诊断可靠性，给出了3种利用多次测量信息减少测量误差影响的方法，即变记忆约束滤波算法、分段算术平均预平滑算法以及指数顶平滑算法。其中，变记忆约束滤波算法能够有效地避免由于故障方程多重共线性所引起的病态解。证明了多故障诊断问题中滤波算法与分段算术平均算法的等价性，证明了在衰减滤波的条件下（第一变记忆因子小于1），随测量点数的增加，变记忆滤波算法的第一变记忆因子与指数平滑的平滑系数之和趋近于1。利用计算机仿真方法对于所得结论进行了检验。

Variable-Memory Constrained Filter Algorithm for Engine Fault Diagnosis

This paper presents three engine fault diagnosis algorithms, all by using multi measuremet re-sults to reduce measurement errors and increase diagnosis reliability. These are variable-memory constrained filter algorithm(VMCF), stepped mathematical average preliminary smoothing algorithm and preliminary exponential smoothing algorithm. The VMCF is capable of effectively preventing an abnormal solution caused by the multi-colinearity of a fault equation. The paper proves the equivalence beween the VMCF and the stepped mathematical average algorithm with regard to multi-fault diagnosis. It also demonstrates that the sum of the first forgetfulness factor of the VMCF and the smoothing coeffident of the exponential smooting algorithm would approach unity with the number of measurements increased and in the case of the smoothing first forgetfulness factor of the VMCF being less than unity. Three conclu-sions are then confirmed by subjecting them to a computer simulation test.