Parameter optimisation in constitutive equations for hot forging

Various methods for parameter optimisation in constitutive equations applied to the hot deformation of a popular α–β titanium alloy have been examined. The use of direct search and gradient methods are shown to be effective, even with a limited dataset, and reliable confidence limits can be computed in each case. However, a hybrid approach, whereby genetic algorithms are used to find an initial parameter starting point, and then a direct search (simplex) method is applied to obtain a global minimum, is particularly promising. For comparison, an artificial neural network approach, which does not require the use of any constitutive equations, has also been implemented.