Machine Learning Assisted Prediction of the Hardness Distribution for the Cast-Forged Az80 Magnesium Alloy

This paper presents a framework for prediction of the hardness of the cast-forged AZ80 magnesium alloy. AZ80 material is cast at four different cooling rates and then die-forged into an asymmetric I-crossbeam geometry at three forging temperatures. The two-dimensional color-coded hardness contour maps of 12 cast-forging conditions are constructed based on the Rockwell 30T superficial hardness measurements across a cross-section of the cast-forged components. Artificial neural network models are trained on the experimental data to predict the hardness value based on a given combination of cooling rate, forging temperature, and the x and y coordinates of the desired location on the component. This data-driven model is then used to predict the hardness values for all the measurement points on an unseen combination of cooling rate and forging temperature, producing the predicted hardness contour maps.  Our statistical analysis demonstrated that the model accurately captures the input-output nonlinear relationship attributed to the effect of processing parameters on the evolution of several different microstructural features. Different model structures and sampling schemes are investigated. The prediction of the hardness distribution is performed for one-I-beam-out and one-condition-out scenarios, resulting in the percent of the average hardness errors of 1.44% ± 0.82 and 2.05% ± 1.25, respectively. The predicted contour maps accurately predict the hardness distribution of unseen cast-forged conditions and resemble the actual contour maps in distinguishing the hardness changes that come from different microstructure evolutions.