Charpy impact toughness data for manual metal are and submerged are we
ld metal samples have been analysed using a neural network technique w
ithin a Bayesian framework. In this framework, the roughness can be re
presented as a general empirical function of variables that are common
ly acknowledged to be important in influencing the properties of steel
welds. The method has limitations owing to its empirical character; b
ut it is demonstrated in the present paper that it can be used in such
a way that the predicted trends make metallurgical sense. The method
has been used to examine the relative importance of the numerous varia
bles thought to control the toughness of welds.