代价参考粒子滤波在故障诊断中的应用

Cost Reference Particle Filter Application on Fault Diagnosis

针对故障诊断中经常遇到非高斯、非线性的噪声,利用代价参考粒子滤波进行噪声处理,并将该方法中固定的遗忘因子改为自动调整的遗忘因子,同时对该方法中的概率质量函数用代价函数表示改为用风险函数来表示,提高了处理突变事件的能力。仿真结果表明,该方法比传统的粒子滤波方法具有更好的滤波效果。

To deal with the non-Gaussian and nonlinear noise encountering frequently in fault diagnosis by the method of cost reference particle filter. In this method the fixed forgotten factor was replaced by automatically adjust forgotten factor, and the cost function was replaced by the risk function in the expression of probability mass function. This method improves the handling capacity of the catastrophic event. Simulation results shows that the method have the better filtering effect than the traditional particle filter method.