A Shop Floor Approach to Springback Prediction

This paper presents a new methodology called a shop floor approach to springback prediction. It begins with the development of a springback model based on die and stamped part historical data by Multiple Regression (MR) technique. This model is then used to validate a numerical springback model using Finite Element Analysis (FEA). Three different pressed parts were selected from industry for this study. They represent three different levels of springback severity namely high, medium and small. Significant factors affecting springback such as die radius, die angle, die depth, punch radius and die width were used as independent or predictor variables meanwhile shape deviation refer springback response between stamped part and die surface were used as dependent or criterion variable. Initial results show that the multiple regression models for the three parts are linear with a prediction error of less than 9%. However a combined parti and part3 regression model improves the prediction accuracy by 7% compared to values predicted by individual regression model. An on-going work is being conducted to validate the model by using commercial FEA software.