Abstract: To address the challenge of balancing computational efficiency and accuracy in numerical simulation of welding deformation for large-scale engineering structures, this paper proposes a finite element simulation method based on a local-global mapping strategy. The method establishes a refined local weld model for temperature field analysis and applies the obtained thermal cycle curves to the global structural model through a mapping algorithm, thereby establishing a rapid prediction system for welding-induced stress and deformation. Specifically targeting the prevalent multi-layer multi-pass welding process in EMU (Electric Multiple Unit) underframe fabrication, a thermal cycle superposition equivalent algorithm is developed to establish the cumulative relationship between multi-pass heat inputs and an equivalent single-pass heat source, enabling simplified modeling of multi-pass welding processes. Experimental validation demonstrates the effectiveness of the proposed method, revealing that when using the equivalent single-pass loading derived from thermal cycle superposition, the relative error in welding deformation remains below 5%, with a significant reduction in computational time. The study confirms that this method significantly enhances computational efficiency through thermal source mapping and superposition algorithms while maintaining accuracy, providing a novel numerical tool for process optimization in large-scale complex welded structures.