Prediction of Material Removal Rate in Bonnet Polishing Using Bayesian Optimization Deep Neural Network

The material removal mechanism in robotic bonnet polishing is complex and influenced by multiple factors, necessitating an appropriate method to establish a material removal model. This study employs a Bayesian Optimized Deep Neural Network (BO-DNN) to explore the intricate relationship between polishing parameters and Material Removal Rate (MRR) using multiple sets of polishing spot experimental data as the dataset. The Tree-structured Parzen Estimator (TPE) method was employed to identify the optimal model configuration, enhancing the model's convergence speed and accuracy. Particle Swarm Optimization (PSO) is utilized to assist in inverse verification. The results demonstrate that the BO-DNN model excels in predicting polishing outcomes, with a Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) of 0.0293 and 0.0241, respectively, for the validation set, and a Pearson Correlation Coefficient (PCC) of 99.42%. For the inverse set, the RMSE and MAE are 0.0410 and 0.0382, respectively, with a PCC of 99.33%. These results demonstrate a strong correlation between the predicted and actual values, suggesting the model's potential for accurate predictions and supporting its reliability. This study offers theoretical insights and practical references for advancing robotic bonnet polishing technology.