Fusion welding for joining of metals is an important manufacturing process widely used in industry, and very appreciated for its usefulness. This thesis presents a strategy dealing with the problem of designing feedback control for robotised welding. The idea is to use off-line programming where computer aided robotics for weld sequences is integrated with finite element modelling for simulations and analyses of weld processes. By this approach the design, evaluations, trials and visualisation can be made “off-line”, beside or prior to continuous production. The focus is to reduce the amount of manpower and need for physical experiments. Focuses on the results are to ensure a high quality weld with limited residual stress and deformation. Different models for two types of austenitic steels and two types of weld sources has been calibrated and validated to form a basis for this strategy. Suggestions for systematic model calibration methods have been proposed including global and local optimisation methods. Experimental work has been performed to support and verify the simulation results and the usefulness of the method. The simulation based strategy has been evaluated and proven to work successfully in two different types of applications. The method is not independent of physical experiments since it is based on models that have to be calibrated, but the experiments needed are assumed to be carried out in a simple and cost efficient way. The thesis suggests how these experiments can be performed. The use of all these technologies is assumed to form an efficient tool for the welding engineer in order to obtain high weld quality in robotised welding. The research presented indicates that the methods work well in real situations and that further work for more robust industrialisation will be beneficial for the welding community.