In-Process Tool Condition Forecasting of Drilling CFRP/Ti Stacks Based on ResNet and LSTM Network

Abstract

Tool condition forecasting (TCF) is a key technology for continuous drilling of CFRP/Ti stacks, as the tool wear is always rapid and severe, which may further induce unexpected drilling quality issues. However, for drilling CFRP/Ti stacks, the cutting spindle power and vibration signals change are complex, influenced by many factors due to the different materials properties. The TCF for drilling CFRP/Ti stacks remains challenging, as the sensitive features are difficult to extract, which decide the accuracy and robustness. Aiming to monitor and forecast tool wear of drilling CFRP/Ti stacks, an in-process TCF method based on residual neural network (ResNet) and long short-term memory (LSTM) network has been proposed in this paper. Using the cutting spindle power and vibration signals preprocessed by the proposed method, the LSTM network with the ResNet-based model integrated can forecast tool-wear values of the next drilling holes. A case study demonstrated the effectiveness of TCF, where the results using raw measured signals and preprocessed datasets are tested for comparison. The mean absolute error (MAE) using raw signals is 45.01 μm, which is 2.20 times bigger than that using preprocess signals. With the proposed method, the data preprocessing for drilling CFRP/Ti stacks can improve the tool-wear forecasting accuracy to MAE 20.43μm level, which meets the demand for online TCF.