Learning HIP dynamics with neural networks

The authors investigate backpropagation neural networks for learning the dynamics of densification during hot isostatic pressing (HIP). The micromechanical description of the dynamics is extraordinarily messy, contains over 27 hard-to-measure parameters, and required 10+ years to develop. Thus, supervised learning is quite an attractive alternative. The authors´ results indicate that it is a feasible alternative. It only took a few hours of training, with one set of data, and very little prior information about the process for a backpropagation neural network to acceptably learn HIP densification dynamics