Development of neural network models for the prediction of solidification mode, weld bead geometry and sensitisation in austenitic stainless steels
Quantitative models describing the effect of weld composition on the solidification mode, ferrite content and process parameters on the weld bead geometry are necessary in order to design composition of the welding consumable to ensure primary ferritic solidification mode, proper ferrite content and to ensure right choice of process parameters to achieve good bead geometry. A quantitative model on sensitisation behaviour of austenitic stainless steels is also necessary to optimise the composition of the austenitic stainless steel and to limit the strain on the material in order to enhance the resistance to sensitisation. The present paper discuss the development of quantitative models using artificial neural networks to correlate weld metal composition with solidification mode, process parameter with weld bead geometry and time for sensitisation with composition, strain in the material before welding and the temperature of exposure in austenitic stainless steels.
Keywords: artificial neural networks, quantitative modelling, austenitic stainless steel welds, solidification mode, weld bead geometry, sensitisation behaviour, weld composition, process parameters, strain, nuclear reactors, nuclear energy, nuclear power