A comparison of the macromodel and neural network approaches for the prediction of microstructural evolution during hot forging

Abstract

Abstract

Data on the microstructural evolution of a titanium aluminide, derived from axisymmetric compression tests, have been used to define different models for the prediction of microstructure during hot deformation. Isothermally forged turbine blades have been manufactured and sectioned to provide the necessary metallographic validation data. Numerically based computer models, using finite element techniques, Gaussian methods and recurrent neural networks, have been used to obtain representative thermomechanical histories and microstructural predictions for the blades. It has been shown that, for the limited training data available, the macromodels provide more satisfactory predictions than the purely empirical neural network approaches.