Novel approach for data-driven modelling of multi-stage straightening and bending processes

Abstract. In multi-stage bending and straightening operations cross-stage and quantity-dependent effects crucially affect the quality of the end product. Using punch-bending units in combination with a mechatronic straightening device can improve the accuracy and repeatability of product features remarkably well. In this work a concept for an innovative hybrid model of a roll straightener in a multi-stage straightening and multi-stage bending process is proposed. This model combines data-driven elements with expert knowledge and aims to minimise residual errors of the roll straightener to reliably decrease the risk of disadvantageous cross-stage and quantity-dependent effects on a subsequent punch-bending process.