一种改进的粒子滤波算法应用于故障诊断

Improved Algorithm of Particle Filter Applied to Fault Diagnosis

为了解决粒子滤波技术中粒子退化的问题,出现了重采样算法。传统的重采样算法如系统重采样,分层重采样等普遍运算时间较长,耗费去很多机时,有时难以满足对实时性要求较高系统的故障诊断。在对粒子滤波技术进行分析的基础上,提出了一种新的重采样算法—"斐波那契查找重采样"。并利用基于残差生成的系统故障诊断方法,将改进的重采样算法应用于传感器故障诊断中。通过对电磁流量传感器的信号处理系统为实验对象进行仿真分析,可以看出,该算法预测系统状态的精度与其它算法基本一致,能有效的进行故障诊断,并且实时性较好,运算时间较快。

To solve the degeneracy problem with particle filters technology, Resampling algorithms appeared. The common characteristics with traditional resampling algorithms, such as system resampling, residual resampling, were all high complexity, and more machine-time was cost. So it could hardly satisfy the real-time requirement of the online diagnostic system. Particle filters were analyzed and a new resampling algorithm called ＂Fibonacci Search resampling＂ was proposed. And using the method of system fault diagnosis based-on residual generation, new resampling algorithm was applied to the sensor faults diagnosis. The simulation of signal processing system of an electromagnetic flow sensor results show that the proposed algorithm has the same precision, and it is effective for fault diagnosis ,better real-time and faster running time.