Tool Wear Prediction with External Signals Based on Lightweight Deep Learning Model

The discontinuity in the milling process makes the tool wear rapidly, which will affect the workpieces' precision and quality. But it is difficult to predict tool wear status with traditional signal processing methods due to inevitable interfering signals. Recently, deep learning algorithms have been widely used because its superiority in solving nonlinear problems. This paper proposes a lightweight convolutional neural network model, which is utilized to enable automatic learning of hierarchical tool wear representation in a short time. Two kinds of external signals and an internal signal are used for experiments. The results show that the classification accuracy of the model based on sound and vibration signals is 98.28% and 97.07%, the prediction from both models have high coefficient with the actual tool wear. Comparisons with related works highlight the effectiveness of our prediction method.