Experimental and numerical analysis of roll bending process of thick metal sheets

In metalworking, bending forming processes are used extensively for a wide range of applications referring to several industrial fields, such as oil&gas field, naval field or automotive field. In cylinders and truncated cones bending process, three-point roll forming of metal sheets represents a concrete option, which mainly consists of pyramid-type machines made of two lower rolls and a top roll. Each roll can be designed to work as a driven roll or as an idler and to move both vertically and horizontally during the rolling process, depending on the sheet and final product features (mechanical properties, thickness and final curvature mainly). Final products can be affected by defects, which represent a concrete problem for metalworking industries because of the further works to be done to meet customers needs; in three-point roll bending, warping defect and hourglass (or barrel) defect represent two of the most common problems. In this paper, an experimental and numerical analysis is proposed to detect defects on roll-bent thick cylinders referring to several companies experiences; the occurrence of thickness variability and circumferences variability along cylinder length and considerable differences between the internal and external length of cylinders themselves, referable to the so-called "hourglass defect" represents a common problem in this industrial field and a major cost for metalworking companies. A Finite Element (FE) 3D model of both the roll-bending machine and the metal sheet is used to simulate the mechanical behaviour of the metal sheet during the forming process, checking for the influence of the spring-back effect and the bending moments from the rolls. Both the pre-bending and the roll bending phases have been simulated. Rolls are modelled as rigid. Both rolls and sheet have been modelled using solid elements and contact elements to take into account rolls/sheet friction. Obtained results are compared with experimental results to check the reliability of the FE model and the simulation. The model itself is then implemented with a central lower roll which serves as an idler and is placed half-way between the original ones. The updated model is then used to simulate the process again and test the central idler as an hourglass-defect solution.