Abstract
Numerical models of fusion welding traditionally compute temperature field for a given set of welding conditions in a forward manner. The reliability of computed temperature profile depends on the accuracy of a number of model input parameters, values of which are uncertain in nature. Here, the authors show that a genetic algorithm (GA) assisted integrated numerical model, following either convection or conduction based calculations, can identify the suitable values of the uncertain model input parameters and in turn provide reliable computed results. Powered with GA, the integrated model is used further in a reverse manner to predict multiple sets of welding conditions for a target weld geometry. The convection based calculations have been able to provide more reliable multiple welding variables in reverse calculations.