Material removal depth prediction in robotic belt grinding of TC4 blade based on hybrid-driven model
Robot abrasive belt grinding is increasingly recognized as an effective technique for machining complex blade parts,driven by advancements in robot technology.Accurate prediction of material removal depth(MRD)is crucial for ensuring precise closed-loop operations in robot belt grinding.Traditional MRD prediction models,which rely on prior knowledge,often lack accuracy.Conversely,purely data-driven models achieve high accuracy but fall short in terms of interpretability and generalization.This study introduces a novel hybrid approach to predict MRD in robotic belt grinding,combining a mechanism model with a data-driven method.Initially,the mechanism model is developed based on abrasive particle morphology and Hertz contact theory.The predictions and residuals of this model are then fed into an XGBoost data-driven model,which predicts the residuals of the mechanism model.This creates a sequential combination of the two models.Simultaneously,the residuals predicted by the data-driven model and the values predicted by the mechanism model are combined in parallel to deliver a hybrid MRD prediction.To enhance the accuracy of the hybrid model,the sparrow search algorithm is employed to optimize its hyperparameters.Experimental tests conducted on Ti-6A1-4V specimens and compressor blades demonstrate that the proposed hybrid method significantly reduces the root mean square error by 71.66％compared to the mechanism model alone and by 64.73％compared to the data-driven model.Furthermore,the hybrid approach excels in few-shot learning scenarios and model generalization.
万方数据
