Thermal Deformation Analysis and Outlier Prediction for Assembly Accuracy Estimation in Adjustment Piping Fabrication

The adjustment pipes are components used to connect fixed pipes where step have occurred in ship blocks. This study proposes a method for predicting the deformation of adjustment pipes that occurs during the fabrication. Thermo-elasto-plastic finite element analysis (FEA) was employed to numerically simulate deformations occurring under various welding conditions. By integrating both experimental and analytical data, a model was developed to quickly and accurately predict welding deformation and compliance with tolerance limits using machine learning algorithms. To manage the complexity of high-dimensional variables, Principal Component Analysis (PCA) was utilized, and predictive models were constructed using KNN, Decision Tree, and XGBoost algorithms. The prediction results indicated that XGBoost algorithm achieved the highest accuracy in predicting the amount of deformation, and Decision Tree algorithm excelled in predicting compliance with tolerance limits. The results of this study are expected to contribute to the design and quality control of adjustment pipes by providing a methodology that provides reliable predictions under various welding conditions.