The capability of coupled fuzzy logic and adaptive neural network in the formability prediction of steel sheets

Abstract

In this study, an adaptive neural network fuzzy inference system (ANFIS) is employed to obtain a model demonstrating a cold rolling effect on the forming limits of sheet metals. Artificial intelligence-based methods require valid datasets for training and testing designed neural networks. In this regard, comprehensive experiments are conducted to achieve different thickness reductions in cold rolling for 304L sheet metals. The effect of cold rolling on the uniaxial tensile curves is determined experimentally. In addition, metallography and tensile tests are performed to determine the stretch in grains due to cold rolling. Moreover, experimental FLDs are obtained using the hemisphere punch test. The experimental data are further utilized to train and test the ANFIS. Subsequently, the model is used to predict variations of FLD for cold rolling thickness reduction. It is shown that with extremely lower computational cost in comparison to the experimental method, ANFIS can qualitatively predict the dependency of forming limits on the cold rolling thickness reduction.