论文部分内容阅读
针对航空零件的加工特点,采集刀具在不同磨损状态下的声发射(AE,Acoustic Emission)信号,对AE信号进行时频分析和小波变换,运用快速傅里叶变换(FFT,Fast Fourier Transform)以及db8小波5层分解,提取AE信号幅值的均方根和主能量频段的能量作为特征向量,对特征向量进行归一化处理后作为输入向量对小波神经网络进行训练.小波神经网络运用参数调整算法,在权值和阈值的修正中加入动量项.测试结果表明:AE信号对刀具磨损敏感的频率范围在10～150kHz,网络实际输出与期望结果的误差小于0.03,该方法能够对刀具不同磨损状态进行正确的识别。
Aeroacoustic emission (AE) signals of the tool under different wear conditions are collected according to the processing characteristics of the aviation parts. The AE signals are analyzed by time-frequency analysis and wavelet transform. Fast Fourier Transform (FFT) db8 wavelet 5 layer decomposition, extract the amplitude of the AE signal root mean square and energy of the main energy band as the eigenvector, the eigenvector normalized as an input vector to train the wavelet neural network using wavelet neural network parameter adjustment The results show that the sensitivity of AE signal to tool wear is in the range of 10 ~ 150kHz and the error between the actual output of the network and the expected result is less than 0.03. The method can measure the wear of tool Status of the correct identification.