Multi-regression modeling for springback effect on automotive body in white stamped parts

• Published in: Materials in engineering Vol. 46; pp. 175 - 190
• Main Authors: Kamal Bashah, Nagur Aziz, Muhamad, Norhamidi, Md Deros, Baba, Zakaria, Ahmad, Ashari, Shaharum, Mobin, Achmed, Mohd Abdul Lazat, Mohd Safuan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.04.2013
• Subjects: Automotive bodies; Automotive components; Body in white; Crashworthiness; Errors; High speed tool steels; Mathematical models; Multi-regression technique; Springback; Stamped part; Statistical methods; Stamped part; Multi-regression technique; Body in white; Springback
• ISSN: 0261-3069
• DOI: 10.1016/j.matdes.2012.10.006

► Springback models were synthesized by actual springback-prone of BIW stamped parts. ► Models quantify the springback angle deviates from stamped and die simulation part. ► Models were verified via statistical technique and validated by physical technique. ► Models predicted results approximate to the actual springback measurements. ► Adequate to predict springback with high R2 values backed by measured of springback. The use of high strength steel (HSS) materials in automotive body in white (BIW) stamped parts has increased the occurrence of springback after the forming process. Although HSS exhibits superior strength, weight reduction, and crash energy, it strongly influences springback impact on the sustainable development of BIW stamped parts. In this study, an empirical springback prediction model was synthesized based on the contemporary data sets of springback-prone components of automotive BIW stamped parts. Two different BIW stamped parts from an actual industrial stamping production line were selected as pilot parts for this study. A statistical multi-regression (MR) analysis was used to model the springback prediction effect by examining the sensitivity of springback input parameters on existing die geometry. The outputs represent the total springback values of the stamped parts. A total of 240 data from samples of selected stamped parts were tabulated to synthesize the springback prediction model. The results show that the MR models for the two parts were linear with the springback estimated errors between the measured and predicted values between 0.5° and 3°, which is acceptable from an industrial viewpoint. The proposed MR models are capable of predicting the springback effect with minimal error by incorporating all possible variations that are inherent in the shop floor process.