Physically-based and Neuro-fuzzy Hybrid Modelling of Aluminium Alloys during Thermomechanical Processing

摘要

During industrial thermomechanical processing such as forging, rolling and extrusion, local regions within the stock undergo different strain histories. These involve a continuous change in strain rate and/or temperature and change in strain path during deformation. These strain path and strain rate histories have a significant effect on the evolution of internal state variables and thus on subsequent recrystallisation behaviour. Physically based models have been developed and successfully applied to describe the effect of strain path and strain rate histories on the evolution of internal variables and subsequent recrystallisation behaviour. However, because of the complexity of the evolution of internal variables during transient deformation, there is a need to develop more powerful and more efficient tools to model the microstructural evolution and flow stress. The present paper shows some developments of so-called hybrid modelling, i.e., the combination of physically-based and neuro-fuzzy modelling. The modelling results show an encouraging agreement with experimental data.