This invention provides a method, apparatus,
computer equipment, and readable storage medium for predicting slope
welding stress. It establishes a
database containing slope factors,
welding process factors, and asymmetric stress factors through thermo-mechanical coupled numerical
simulation and conducts experimental
verification. After data preprocessing, an LSTM-ResNet-
Diffusion hybrid model is constructed and trained. The model predicts asymmetric stress factors, and SHAP analysis clarifies the correlation patterns and critical thresholds of these factors. This invention offers high prediction accuracy and can precisely reveal the correlation mechanism between various influencing factors and asymmetric stress, providing a scientific basis for optimizing slope
pipe welding processes.