An artificial neural network model was established from a commercially available system, and the model was fed with data from a series of simple trials using D and DH 36 steel plate. Two thickness of steel plate were used, 6 and 8mm. Plate topography was measured before and after welding using a simple grid system, and the measurement of distortion was based on I-numbers. A sensitivity analysis was carried out on the data generated to feed the model. This revealed a number of factors—some already established and some which were previously unknown or unrelated to welding distortion. The carbon equivalent (CEV) of the steel was a significant factor, and when broken down further was found to be strongly influenced by the carbon content of the steel plate. Some other factors which have been known to affect distortion were identified, e.g. plate thickness, heat input. In addition to the carbon content of the steel plate, the yield strength to tensile strength ratio, the steel grade, the edge preparation mode and the rolling treatment were also identified as being factors which appeared to influence the sensitivity of the plate to welding distortion. Each of the factors has been discussed along with possible mechanisms for their effect on distortion. The potential for extending the use of the model is also discussed. A comparison has been made between artificial neural networks and finite element models as predictors of distortion.