Title: Optimisation of BiLSTM for end mills wear predictive study

Authors: Chunlong Zou; Lin Zhou; Chen Wang; Xuxiang Lu

Addresses: College of Mechanical Engineering, Hubei University of Automotive Technology, Shiyan, Hubei, China ' College of Mechanical Engineering, Hubei University of Automotive Technology, Shiyan, Hubei, China ' College of Mechanical Engineering, Hubei University of Automotive Technology, Shiyan, Hubei, China ' College of Mechanical Engineering, Hubei University of Automotive Technology, Shiyan, Hubei, China

Abstract: An effective method of milling tool wear monitoring is important to improve product quality and extend the life of milling tools. A Genetic Algorithm (GA) optimised Bidirectional Long- and Short-Term Memory Neural Network (BiLSTM) deep learning model (GA-BiLSTM) is proposed to monitor the wear value of end mills. First, the important time-domain, frequency-domain and time-frequency features of the cutting tool wear signal are extracted by wavelet transform, and then processed by cross-validation and correlation analysis to obtain the input time-series samples for the prediction model. Then, the model parameters of the BiLSTM are optimised by GA, mainly optimising the learning rate (Learning Rate), generation (Epoch) and loss rate (Dropout). At last, the comparison experiment of different models is carried out: GA-BiLSTM is better than RNN, LSTM, CNN-BiLSTM and its evaluation index MAE and RMSE are lower than the comparison model, and the model operation time cost is in the reasonable scope. It shows that GA-BiLSTM method is effective and feasible, and improves the precision of wear prediction.

Keywords: milling tool wear; genetic algorithm; BiLSTM; wear monitoring.

DOI: 10.1504/IJWMC.2026.154161

International Journal of Wireless and Mobile Computing, 2026 Vol.30 No.4, pp.325 - 334

Received: 25 Mar 2024
Accepted: 20 Jun 2025
Published online: 15 Jun 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article