Machine learning assisted hardness prediction and experimental verification of laser cladding coatings

To quickly determine the influence of laser cladding process parameters on hardness, artificial neural network (ANN) and support vector machine (SVM) models were used to predict the hardness of cladded coatings, this approach substantially reduces the labor and raw material costs associated with conventional hardness testing. The results indicate that the prediction accuracy of the SVM model is higher than that of the ANN model . In the SVM model, the hardness prediction accuracy for 316 L alloy is superior to that for Inconel 718 alloy. It was found that laser power, scanning speed, and powder feeding rate have significant effects on hardness. An orthogonal experimental design was adopted, and six sets of data were selected for experimental verification. The RMSE value for 316 L alloy was 9.41, while that for Inconel 718 alloy was 23.3. Therefore, the experimental values are generally consistent with the model predictions, and the SVM model is more suitable for predicting the hardness of 316 L alloy. The model predictions and experimental results in this study provide engineering application value for the rapid selection of laser cladding process parameters in hardness prediction and experimental validation, as well as an effective and feasible method for problems with limited single material data and difficult prediction. • A hardness prediction model is established using the ANN and SVM models. • A feasible method is provided to address the issue of limited data in machine learning. • By combining orthogonal design, the number of validation experiments is reduced.