A review of prediction and control methods on welding-induced distortions in large metallic structures

Welding enables cost-effective fabrication of large structures in shipbuilding, aerospace and automotive industries, but repeated thermal cycles during multi-stage assembly inevitably cause distortions that often exceed design tolerances, degrading structural performance and requiring costly corrections. Despite significant advances in local weld modelling, predicting distortion in large, complex assemblies remains challenging due to interactions among material properties, structural geometry, heat input, welding sequence and restraint conditions, particularly for lightweight, thin-walled high-strength steels and aluminium alloys. This review synthesises research from 2000 to 2025 to assess distortion prediction and mitigation strategies for large metallic structures, emphasising industrial applicability. Thermo-elasto-plastic finite element methods provide high accuracy but at significant computational cost and inherent strain-based approaches offer efficient large-scale simulations but rely on calibration and simplifying assumptions, while local/global methods balance accuracy and computational efficiency. Data-driven surrogate models are emerging for rapid prediction and optimisation. No single approach is universally transferable across structural configurations, with scalability generally achieved at the expense of calibration requirements and application-specific assumptions. Distortion mitigation is most effective at the design stage through welding sequence optimisation and structural planning. In-process techniques such as thermal tensioning and controlled restraint further reduce distortions, while post-welding corrections remain less efficient. The literature is strongly concentrated on conventional structural steels, with only approximately 10% and 1% of the reviewed references addressing large aluminium and high-strength steel structures, respectively. Overall, major limitations concern scalability, standardisation and data availability, highlighting the need for integrated physics-based and data-driven frameworks for efficient large-scale distortion control.