Tool Wear Prediction Combining Global Feature Attention and Long Short-Term Memory Network

• DOI: 10.46604/peti.2024.14201
• Publication Date: 2024-11-07T01:46:43Z
• Authors: Wanzhen Wang; Sze Song Ngu; Miaomiao Xin; Xiaomei Ni; Beibei Kong; Kaiyuan Wu; Ruyue Han
• Keywords: Feature (linguistics)

This study aims to accurately predict tool flank wear in milling and simplify the complexity of feature selection. A hybrid approach is proposed eclectically integrate advantages between long short-term memory (LSTM) network global attention (GFA) module. First, matrix calculated using multi-domain extraction method. Subsequently, a parallel employed achieve fusion. The stacked LSTM extracts temporal dependencies features GFA module used adaptively complement key representing information samples. Finally, output are concatenated, prediction achieved through fully connected layer. Experiments demonstrate optimal performance predicting wear. Specifically, GFA-LSTM framework reduces mean absolute error (MAE) by 36.9%, 17.7%, 25.2% three experiments compared simple LSTM, validating effectiveness