Laser Cladding Substrate Distortion Prediction Based on Temperature Characteristics

Abstract The thermodynamic processes associated with laser cladding technology often result in an inhomogeneous distribution of thermal stress within the substrate, leading to substrate distortion. Such distortions can significantly compromise the assembly precision and functional performance of subsequent components. Consequently, the accurate prediction of substrate distortion behavior is imperative for the optimization of process parameters, enhancement of production efficiency, and broadening of the industrial applicability of laser cladding technology. In this study, a systematic analysis was conducted to characterize the temperature field distribution across various regions of the substrate's upper surface during the laser cladding process. The effects of key process parameters, including laser power, scanning speed, and powder feed rate, on the substrate's average temperature and end warping distortion were thoroughly investigated. Leveraging these insights, a novel predictive model for substrate distortion was developed using a Long Short-Term Memory (LSTM) neural network. The model incorporates the initial temperature of the deposited layer and the process data of the substrate's end distortion as input features, with the distortion data at time T + N serving as the predictive output. A comparative evaluation of the LSTM network's predictive performance, with and without the inclusion of temperature features, demonstrated that the integration of temperature data significantly reduced the model's prediction error from 3.7–0.01%, thereby substantially enhancing prediction accuracy. These findings provide a robust theoretical foundation and technical framework for the intelligent control and optimization of laser cladding processes.