| Summary: | In order to overcome the dilemma of insufficient effective sample number for subsurface microcrack depth in the lapping of brittle materials with fixed abrasives and achieve accurate prediction, a co-training SVR was used to construct the prediction model. The effects of different labeled training set partitioning methods on the mean square error of the test set were compared. Then the predictive performance of supervised learning PSO-SVR model was compared with that of the model. Finally, brittle materials such as microcrystalline glass and calcium fluoride, which were not included in the labeled training set, were taken as processing objects for lapping and angular polishing experiments to examine crack depth values. The examined subsurface microcrack depths of four groups were compared with the predicted values of the co-training SVR model. The results show that the co-training SVR model under the separate partitioning method has a smaller mean square error. Compared with the PSO-SVR model, the mean square error and the mean absolute percentage error of the co-training SVR model are reduced by 9% and 17%, respectively. The prediction error of the model for the four groups of verification experiments is between 1.2% and 13.8%. The above results show that the co-training SVR model can predict the subsurface microcrack depth accurately when lapping brittle materials with fixed abrasives.
|
|---|