Artificial Neural Network for Predicting Edge Stretchability in Hole Expansion Test With Gpa-Grade Steel

Abstract

This paper mainly proposes an artificial neural network (ANN) model for predicting edge stretchability of GPa-grade steels, which is substantially difficult to predict due to the complex nonlinear relation among the numerous sheared edge qualities. We newly suggest the physically characterized parameters, such as material properties, deformed shape, and work hardening of sheared edge, to predict the various materials and punching methods, simultaneously. The proposed parameters are trained with the predamage strain which is calculated by inherent fracture strain and experimental results in terms of hole expansion ratio. To prevent the overfitting issues, cross validation method with additional datasets from a different kind of edge stretchability test such as sheared edge tensioning test are utilized. Experimental validations have been conducted with various GPa-grade steels and sheared edge conditions, which are compared with the proposed ANN model and numerical simulation. The proposed ANN model exhibits remarkable performance in the prediction of hole expansion ratio having a mean absolute error of 1.5% when compared to the previous studies such as numerical simulation and ANN model with utilizing the maximum hardness measured at the sheared edge. INDEX TERMS Artificial neural network, edge cracking, edge stretchability, GPa-grade steels, sheared edge quality.

Prediction of Hole Expansion Ratio for Advanced High-Strength Steel with Image Feature Analysis of Sheared Edge

“…Many methods for edge damage quantification have been suggested to understand the underlying physical phenomena or to provide an estimation of edge damage, such as microhardness [ 13 , 14 , 15 ], metal flow angle [ 16 ], void inspection [ 17 , 18 ], and geometry of the edge. Except for the edge geometry, the methods require microscopy inspection and cutting of the specimen, limiting its application.…”