Prediction of flow stress in isothermal compression of hydrogenated TC17 alloy using multiple prediction models

Isothermal compression tests were carried out on TC17 alloys with different hydrogen contents using a Gleeble3800D thermal simulator with deformation temperatures ranging from 760 degrees C to 860 degrees C, strain rates ranging from 1 to 0.001 s-1, and deformations of 50%. Six flow stress prediction models were used to predict the experimental data, namely, an improved model of coupled hydrogen content based on the strain-compensated Arrhenius model (MSCA), a constitutive model based on a self-consistent (CBSC) model, an artificial neural network (ANN) model, an optimizable ensembles of decision tree (OEDT) model, a support vector machine (SVM) model, and an optimizable Gaussian process regression (OGPR) model. Compare the predictions of the six models. Regression coefficients (R2), mean absolute error (MAE), and root mean square error (RMSE) were used to assess the predictability of the model. The data with less than 10% error in the prediction results were also defined as passed data, and the stability of the model was evaluated using the pass rate. The results show that the MSCA model with coupled hydrogen content has good predictive ability and model stability. The optimal prediction model for the flow stresses of TC17 alloy with different hydrogen contents is the OGPR model with the values of R2, MAE, RMSE, and pass rate of 0.995, 1.6 MPa, 6.2 MPa, and 97%, respectively.