Process-planning models for welding using bayesian network

A process-planning model has been developed for welding of T-joints. The model has been used for computing control variables on a T-joint with varying root gap. This research investigates how different sources of knowledge can be produced, acquired, integrated and formalised in process-planning models. The knowledge sources investigated are empirical knowledge from welding experiments, analytical knowledge from literature and knowledge obtained from interviews with welding operators and experts. A Bayesian network is selected as the tool for making the process-planning model. It is selected because it is a graphical model and it uses probabilities, which enables use of welding operator and expert knowledge. Welding made using the designed process-planning model shows good results on a T-joint with varying root gap. A benchmark with artificial neural networks for a direct process-planning model shows an advantage of the model based on a Bayesian network.