71 / 2025-03-30 20:34:49

Tool Wear Prediction Using an Ensemble Hybrid Model Based on LSTM and Transformer

tool wear prediction,complex and noisy data,LSTM and Transformer,ensemble learning,PHM2010

Tool wear prediction is crucial in manufacturing, as tool degradation affects production efficiency, product quality, and operational costs. Traditional methods, relying on empirical or physics-based analyses, often fail to address complex non-linearities and varying operational conditions. A hybrid model using Long Short-Term Memory (LSTM) networks and Transformer Encoder is proposed for tool wear prediction. LSTM captures temporal dependencies, while Transformer enhances global feature extraction through multi-head attention. Additionally, an ensemble learning approach is integrated to reduce noise and improve robustness. The ensemble LSTM-Transformer encoder (ELSTM-TE) model is validated using the PHM2010 dataset, showing significant improvements in prediction accuracy over traditional and other deep learning models.  The results demonstrate the model’s ability to handle complex, noisy data, offering a reliable solution for tool wear prediction.