| Popis: |
In the aeronautical sector, because parts are mainly of large dimensions and in high performance materials, products are forged in small batches. Forming these complex parts requires energy-controlled production means, such as screw presses or, more generally, forging hammer. With these machines, several successive strokes are necessary to obtain the parts desired geometry and mechanical characteristics. However, for these small quantities, the automation of the manufacturing process is not necessarily possible or profitable and consequently, the control of the machine remains dependent on the know-how of the operators, in particular with regard to the quantity of energy to be delivered blow after blow, the temperature, the lubrication conditions, etc. The main challenge is to provide flexibility and robustness particularly adapted to small batches, thus limiting the impact of process parameters variability on the part final quality. To reach that goal, the implementation of a digital twin is proposed. The objective of the project is to develop a digital twin in the context of forming materials on an energy-controlled screw press. The scientific challenge is to obtain an accurate, predictive and reactive twin that will allow real-time control of the process as well as access to information that cannot be measured during the manufacturing process. A methodology for the creation of a predictive meta-model based on a calibrated numerical simulation and updated by machine learning is proposed. This meta-model will compose the digital twin. Our approach is validated on a case study: the uni-axial compression of a copper cylinder. Finally, the following development phases of the digital twin are presented. |
