摘要
小波变换去噪中最关键的问题是最优小波基的选取,使其能够将噪声从原始信号中分离出来。针对电机故障的特点,提出了一种基于信号的最优小波基选取方法。将信号小波变换的能量阈值曲线作为小波基函数的适用性评价指标。通过训练神经网络,选取适合该信号的最优小波基,最后采用平移不变量(TI)小波阈值法实现信号去噪。在此基础上对750W化纤电机进行了测试,实验结果表明,该方法能准确找出适合特定信号的最优小波基。训练后的神经网络可直接用于其它类型电机的信号去噪处理,具有实用价值。
In the field of wavelet denoising, an essential problem is how to determine the wavelet basis that eliminates noise parts from original signal. According to the characteristic of electromotor fault, a method of optimal wavelet basis selection based on signal is put forward, which regards the energy-threshold curve as the applicability criterion of wavelet basis. The optimal wavelet basis is ascertained by training the neural network, and the noise of original signal is eliminated by the arithmetic of translation invariant wavelet threshold. The method is used in an example of 7.50W electromotor, the results show that the optima/wavelet basis can be ascertained. The trained neural network can be used to other type of electromotor signal directly.
出处
《现代机械》
2009年第3期33-35,39,共4页
Modern Machinery
关键词
小波变换
电机
最优小波基
能量阈值曲线
特征提取
wavelet transform
motor
optimal wavelet basis
energy-threshold curve
feature extraction
