关键词: 振动信号, 表面粗糙度, 铣削, 特征排序, 神经网络

Abstract: To improve the accuracy of predicting workpiece surface roughness with vibration signals during milling process,a feature selection-neural network algorithm was proposed.The features of vibration signals were extracted by using “wavelet packet decomposition + statistics”,and the feature base was constructed.The correlation analysis was used to sort the extracted features.According to the ranking results,the orthogonal experiment was utilized to optimize the number of selected features and the number of neurons in the hidden layer of the neural network.Taking processing data onto 6061 aluminium alloys under different milling factors as an example,the arithmetic mean deviation of the contour was applied to measure surface roughness.The results suggested that the prediction error of the proposed method was within a reasonable range and the average deviation was 6.57%.The predicted value was close to the actual value.

Key words: vibration signal, surface roughness, milling, feature selection, neural network