Research on Laser Cladding Morphology Prediction of Shield Machine Main Bearing Based on BP Neural Network
This paper revolves around the laser cladding repair process for shield machine main bearings.In addressing the challenge of accurately characterizing the mapping of process parameters to the clad area,and aiming to predict the morphology dimensions of a single cladding pass for enhancing the overall performance of the clad region,we propose a BP neural network-based predictive model for the laser cladding process morphology of main bearing.Firstly,a 3-factor 4-level orthogonal experiment is conducted using 42CrMo bearing steel as the substrate and stellite6 as the cladding powder material,with clad height,width,and depth as the indicators.Subsequently,based on the indicators obtained from the orthogonal experiment and the process parameters,a BP neural network model is designed and trained.Finally,the predicted morphology dimensions are compared with the actual dimensions to calculate the errors.Analysis of the experimental results reveals that the model exhibits prediction errors within 2%for clad height and width,but the prediction performance for clad depth is less satisfactory.The analysis attributes this to the limitation of data leading the network model's weights to converge to a local optimum.
main bearinglaser claddingneural networkorthogonal experimentmorphology prediction
NETL
NSTL
万方数据
