Accelerated learning using Gaussian process models to predict static recrystallization in an Al-Mg alloy

This paper describes an investigation into the suitability of Gaussian process models for predicting the microstructure evolution arising from static recrystallization. These methods have the advantage of not requiring a prior understanding of the micromechanical processes. They are wholly empirical and use a Bayesian framework to infer the probability distribution of data, given a 'training set' comprising observed outputs for known inputs. Given the evidence from the training set, they can make a prediction and assess its certainty, taking into account the noise in the data. In addition, non-uniform deformation geometries were chosen to provide the training data, both to approximate typical manufacturing processes with complex strain paths and to investigate whether learning could be accelerated by using only a small number of test samples containing a distribution of deformation histories. The model was trained and tested on data from samples of a cold-deformed and annealed aluminium-magnesium alloy.