基于ANFIS非线性观测器的连续小波变换故障检测

Method of Fault Detection Based on ANFIS Nonlinear Observer Using Continu ous Wavelet Transform

提出一种把信号分析和数学模型相结合的故障检测方法,通过小波变换对信号的去噪和自适应神经模糊推理系统(ANFIS)具有的离线学习功能,来获取系统输入输出的非线性动力学特性,进而实时计算出残差,并对残差序列进行MexicanHat小波变换,通过对极值点的逻辑判决,准确检测出故障的发生·该方法具有灵敏度高、克服噪声能力强的特点·仿真结果验证了这一方法的有效性·

A novel method of fault detection was proposed based on ANFIS nonlinear observer using continuous wavelet transform. This method is a combination of signal anal ysis and mathematical model. By the denoising function of wavelet transform and the offline learning function of ANFIS, the input and output nonlinear dynamic c haracteristic of system was obtained. The series of output prediction error, gen erated from the real output and ANFIS observer estimated output, was calculated in real time and transformed by Mexican Hat wavelet as a residual error to execu te logical judge of its extrema and detect occurrence of defaults exactly. This method has good sharpness of response to faults and robustness to noise. Key words:?ANFIS; nonlinear observer; wavelet transform; fault detection; residual error; l ogical judge