Modelling of the hot flow behaviors for Ti-13Nb-13Zr alloy by BP-ANN model and its application

Abstract

The plastic deformation mechanisms and the constitutive model of flow behaviors at different deformation conditions in biomedical titanium alloy are an essential step to optimize the design of any forging process for implant productions. A series of isothermal compressions tests on Ti-13Nb-13Zr alloy in a wide range of true strain, temperature and strain rate were conducted on a thermomechanical simulator. The hot flow behaviors with different softening mechanisms, including dynamic recrystallization and dynamic recovery, were characterized based on true strain-stress curves. A back-propagational artificial neural network (BP-ANN) method was conducted to evaluate and predict this non-linear problem by self-training to be adaptable to the material characteristics. The flow stress of this material a wide deformation condition range can be predicted accurately by the BP-ANN model obtained in this study. The prediction ability of this BP-ANN Model was evaluated by three accuracy indexes, Absolute error, Relative error and Average absolute relative error. Sequently, the developed BP-ANN model was programed and implanted into the finite element (FE) analysis platform, Msc.Marc software. The results have sufficiently articulated that the well-trained ANN model has excellent capability to deal with the complex flow behaviors and has great application potentiality in hot deformation processes.