3D flexible roll forming with machine learning

Project description


Flexible roll forming is a variant of the conventional roll forming process and allows forming long structural components from high strength materials with complex shape variation along the length of the component. The process is a suitable alternative to the conventional stamping operation and reduces cost while improving flexibility and efficiency in low volume production of the structural components for EVs (electric vehicles). However, the commercialisation of the process faces challenges in the form of shape defects such as end flare and twist. Several studies have shown that the input material, geometric and process parameters have a major influence on the level of shape defects and for few conventional forming processes, various regression models were successfully developed to establish the link between input parameters and the defect severity. Based on this relationship, the necessary corrective action can be implemented at the start for in-line defect compensation. The existing models and conventional defect compensation tactics are not suitable for flexible roll forming, as the component formed in this process has variable cross sections, which creates varying levels of plastic deformation along the length or forming direction.  Although some studies have shown approaches for end flare compensation in flexible roll forming, they are mostly trial-and-error based and often lack the flexibility to accommodate localised changes along the forming line. Therefore, there is currently a need to a build a reliable data model for flexible roll forming, which can predict the expected shape deviation to suggest the amount of flexible corrective action required for the tool.


The 3D roll forming centre established at Deakin University is a state-of-the-art technology for flexible roll forming of high strength materials into components with variation in height and width. The machine offers a stable forming approach by clamping the sheet while forming the flanges, and the tool movements are flexibly controlled by robotic arms with multiple degrees of freedom (DoFs). The recent improvement has further enhanced the capabilities by allowing a variable rotation of the tool to adjust the local bend angle along the forming direction. This opens the possibility of in-line shape compensation with variable corrective action, which is necessary for the flexible roll forming process.

Project objectives

  1. The development of solutions for on-line part quality monitoring: A new strategy should be developed for measuring the part shape in-real time using a calibrated sensor.
  2. A flexible part shape control that enables the targeted correction of forming defects: Using the new and improved capabilities of the machine, an optimum control strategy will be developed for localised part shape improvement.
  3. Algorithms that enable an in-line shape compensation based on the monitored shape: Regression models and advanced genetic algorithms need to be employed for building a predictive model.
  4. Validation of the smart technology for in-line shape control by the manufacture of prototype components: Manufacture and testing of industry components with variation in both width and depth.

Expected outcomes

  • New methodologies: Solutions for part quality monitoring in the FRF process.
  • New insights: Fundamental understanding of shape defects when FRF complex shape and their relation to material parameters and variation.
  • New Technologies: Solutions for flexible part shape compensation to overcome springback and end flare + A smart technology for in-line shape control.

Student requirements

  1. Qualifications: H1 grade average or equivalent, OR, Masters Degree (Research).
  2. Recommended skills: Numerical modelling (Abaqus, Marc), Advanced data analysis and machine learning skills, Experience of working with manufacturing systems.
  3. Language skills: Band B (https://www.deakin.edu.au/international-students/entry-requirements/english)