Weld design is an area where neural network modelling has proved invaluable in reducing development costs.

The mechanical properties of a weld (its resistance to impact, its strength, its ability to sustain deformation) depend on a vast number of variables, which include the welding process, the composition of the alloy, the heat-treatments received before and after welding, etc.

There is no physical model available that can account for all these variables and predict the mechanical properties of a weld.

On the contrary, neural network models have had enormous success and are now routinely used by the best weld consumable manufacturers.