On the stretchability prediction of sheet metals using an artificial neural network based on experimental data

Abstract

The utilization of artificial intelligence has aided researchers in many scientific and engineering fields to derive plausible predictions without expending costs on experiments and modeling. In the field of metal forming, due to several parameters affecting formability, analytical and experimental have difficulties in addressing all effects. In the present study, we aim to present an artificial neural network to incorporate microstructure effects on the stretchability of plain carbon steel sheet metals. In this regard, spheroidized carbon steels with AISI 1008, 1012, and 1045 grades are considered with different levels of spheroidization and distributions of cementite. The experimental works use a semi-spherical punch test to evaluate forming limit diagrams (FLDs). The experimental dataset is utilized to train the designed artificial neural network (ANN). The results indicated that with more uniform distribution of the spheroidized cementite, the stretchability of the sheet metals increases. Moreover, an increase in carbon content in the fully spheroidized sheets (>90%) causes a decrease in formability. The trained ANN demonstrates satisfactory performance in predicting forming limits considering variation in the cementite patterns.

KEYWORDS:

• Forming limit diagram (FLD)
• neural network
• spheroidization rating
• carbon content
• plain carbon steel

Disclosure statement

No potential conflict of interest was reported by the author(s).