Hybrid modelling of aluminium–magnesium alloys during thermomechanical processing in terms of physically-based, neuro-fuzzy and finite element models

Abstract

For modern metals industries using thermomechanical processing, off-line modelling and on-line control based on physical knowledge are highly desirable in order to improve the quality of existing materials, the time and cost efficiency, and to develop new materials. Neural network and neuro-fuzzy models are the most popular tools, but they do not embed physical knowledge. On the other hand, current physically-based models are too complex for industrial application and are less efficient than neural networks. A combination of neuro-fuzzy and physically-based models has therefore been developed, which is termed a “hybrid model”. The hybrid model has been applied to predict flow stress and microstructural evolution during thermomechanical processing. Comparison with experimental data shows generally good agreement for Al–1% Mg alloy deformed under thermomechanical processing conditions. The hybrid model was then embedded into a finite element model and the simulated results show a very similar distribution to those calculated using empirical models.