A Hybrid CBiGRUPE Model for Accurate Grinding Wheel Wear Prediction

In grinding machining, monitoring grinding wheel wear is essential for ensuring process quality wear and reducing production costs. This paper presents a hybrid CBiGRUPE model to predict grinding wheel wear, which integrates the advantages of convolutional neural networks (CNNs), bidirectional gated recurrent unit (BiGRU), and the Performer encoder. Time-domain features are extracted from the spindle motor current signals of a surface grinding machine. The structure and hyperparameters of CBiGRUPE are optimized using Bayesian optimization. Experimental validation of the model demonstrates superior performance, with mean absolute error (<i>MAE</i>), root mean square error (<i>RMSE</i>), and coefficient of determination (<i>R</i>²) values of 3.041, 3.927, and 0.920, respectively. Compared to models like CNN, BiGRU, and Transformer, the CBiGRUPE model offers more accurate and stable wear predictions. This paper also discusses the advantages and limitations of various models for estimating grinding wheel wear, emphasizing the effectiveness of the proposed approach. This study establishes a foundation for compensating wheel wear and accurately determining the optimal dressing time.