Hybrid model based on recursive feature elimination with cross validation and Tradaboost for workpiece surface topography prediction of five-axis flank milling

Surface topography prediction is a crucial part of error compensation, which is important to improve the machining quality. Because the surface topography is often affected by numerous comprehensive machine tool factors, predicting the surface topography more accurately and efficiently has become a challenge faced by the modern manufacturing industry. To solve the problem of surface topography prediction in the five-axis flanking milling field, this paper proposed a hybrid data-driven model based on recursive feature elimination with cross validation (RFECV) and Tradaboost. The dependence of machine tool factors in the milling operation and the requirement of data quantity in the model training were reduced through this model. A case study of an "S" test piece was carried out to verify the effectiveness of the model where the point contour error (PCE) was used to evaluate the surface topography. For the same "S" test piece, one contour line was selected as source contour, and all points were measured. Other contour lines were selected as target contour, on which only a few points were measured. The parameters which have vital influence to the prediction result were selected by RFECV. The measurement points and selected parameters in source and target contour were used to train the model. Then, the PCEs of target contour were predicted using the trained model. The results of comparative experiments with other methods showed that the proposed model has a lower MSE, which validated the effectiveness of the model.