USING NUMERICAL SIMULATION TO PREDICT THE DISTORTION OF LARGE MULTI-PASS WELDED ASSEMBLIES

Welding is a complex manufacturing process involving interactions between thermal, metallurgical and mechanical phenomena, which cause distortion and affect properties and performance of materials. The fabrication process induces high thermal gradients in a localized zone around the welding joint that creates plasticity and phase transformations in hardenable steel. As a result, the welded assembly undergoes deformation and microstructure changes, and high residual stress gradients appear in the heat-affected zone. The effort to generate weld quality and assemblies that are within defined tolerances causes upstream design cost, prototyping cost, material cost, repair cost, labor cost, energy cost and tooling cost. Understanding the consequences of a welding process on an assembly prior to generating the parts can highly reduce the cost of these welded assemblies. Testing and prototyping has been proven quite effective in understanding the consequences of the welding processes but they are also very costly. Thanks to numerical simulations technology, the consequences of weld quality and weld distortion can now be computed before any material is cut and any parts are welded. Using numerical simulation in the design cycle allows engineers to get a better understanding of the fabrication process and help them generate better designs. It also reduces cost to make prototypes by reducing the amount of try-outs required. It reduces tooling cost by providing a good understanding of the behavior of the entire assembly under different clamping conditions and helps reduce the amount and type of tooling required. It helps welding engineers analyze different welding sequences and clamping configurations to increase production rate without compromising the quality of the parts. Prediction of weld distortion of huge and thick-walled assemblies, which are joined with multipass welding, is a big challenge in computer simulation.. These computer models can take up to several days or even a couple of weeks to run. In order to manage this task within a reasonable time, a local-global method was developed and is now available as a commercial tool in order to address such memory and computer intensive simulation tasks. Using an industrial part - a turbine housing from SKODA - the local-global approach is applied to prove that computer simulation methods can be effectively used within reasonable time to produce accurate results.