Advanced characterization methods to model nonlinear strain paths and localization for sheet forming

Abstract

Advanced material characterization methods are proposed to model nonlinear material behaviors such as the Bauschinger effect, localized necking initiation, and compressive fracture strain. The Bauschinger effect is an early re-yielding behavior after load reversal. For the Bauschinger effect calibration, a simple bending/reverse bending tester is developed and utilized to optimize material constants of the Yoshida-Uemori model. The calibrated material property for the Bauschinger effect is verified with the application of U-draw bending test. The springback prediction result is comparatively evaluated with other loading/reverse loading test properties. Localized necking is manifested as rapid thickness thinning prior to ductile fracture in the tensile deformation. The localized necking is visually expressed through the concave shape of materials. A convolution neural network (CNN) is constructed to detect the onset of localized necking by the edge lines of a specimen. The performance of the CNN model is evaluated through the class activation mapping (CAM) and the confusion matrix of test dataset. Compressive ductile fracture exhibits extremely large strain, which has been regarded as being difficult to be measured. This study proposes the compressive fracture characterization method using bilinear strain paths: pre-tension and compression. Compressive loading after pre-tensile loading can result in a premature compressive fracture. The verification of the proposed characterization method is performed by comparing experimental data and simulation results with different pre-tensile strain levels.