Neural network modelling of the mechanical properties of nickel base superalloys.

Modelling techniques are being developed with the aim of reducing the cost and time associated with the development of new alloys for critical aerospace components. In this paper the yield and tensile strengths of commercial polycrystalline wrought Ni-base superalloys are modelled using an artificial neural network technique. Neural networks of this type are capable of realising a great variety of non-linear relationships of considerable complexity. They are 'trained' using existing experimental data which is presented to the network in the form of input-output pairs, thus the optimum relationship is found between the tensile properties and those parameters which are considered to be of importance. Through a series of tests it was found that with appropriate training a neural network can reliably reproduce metallurgical experience and knowledge. These results demonstrate that neural network models could be successfully used in the development of new alloys, reducing the amount of experimental work required and thus the time taken for a new alloy to be introduced into service.