Predicting the flow stress of Inconel 617 superalloy using constitutive equation and artificial neural network approach

Predicting the flow behavior of materials is a crucial step to optimize the design of forming processes. In this research, artificial neural network (ANN) and hyperbolic sine methods were used to estimate the flow behavior of Inconel 617 superalloy, IN 617. For this purpose, several hot compression tests were performed in the temperature range of 850-1150 degrees C, strain rates of 0.001-1 s-1 and strain value of 0.6, and the stress-strain curves were plotted. To precisely evaluate and compare the accuracy of the two models, standard statistical parameters such as correlation coefficient (R), average absolute relative error (AARE), and average root mean square error (RMSE) were used. R and AARE values for ANN model training data were 0.9999 % and 0.8414 %, and for the hyperbolic sine model were 0.9931 % and 5.4744 %, respectively. The RMSE value for the training data of the ANN model was 1.5516 and for the hyperbolic sine model, it was calculated to be 18.2536. The results indicate that the ANN model's ability to predict the flow behavior of IN 617 superalloy is more than the hyperbolic sine model. Furthermore, it is worth mentioning that AARE distribution for both ANN and hyperbolic sine models indicates the high accuracy of both models in predicting the flow behavior of IN 617 superalloy at relatively high strain rates and temperatures.