Prediction of Laser Cladding Layer Cross-Sectional Morphology - Based on SSA-BP Neural Network

In the face of challenges such as difficulties in controlling the morphology of laser cladding layers and high costs in laser cladding technology, this paper utilizes the Sparrow Search Algorithm (SSA) to optimize the error Back Propagation (BP) neural network. The SSA-BP network is applied to predict laser cladding cross-sectional morphology. A comparison is made between the predicted and the experimental values, using R2, RMSE, and MAPE as evaluation metrics for the model. The results indicate that the predicted heights of the cladding layers are 0.972, 0.076, and 3.948%, respectively, while the predicted widths are 0.862, 0.099, and 2.004%, respectively. Furthermore, experiments beyond the input range of process parameters are conducted to validate the universality of the model. The results demonstrate that the predicted heights of the cladding layers are 0.952, 0.118, and 4.771%, respectively, and the predicted widths are 0.813, 0.209, and 3.688%, respectively. Moreover, we also compared the predictions from models of SSA-BP, MPA-BP, SOA-BP, and PSO-BP. In conclusion, this paper's model exhibits excellent predictive capabilities and good universality, providing reference for the prediction and control of cross-sectional morphology in laser cladding technology.