Data-based model building and process optimization with artificial neural network for injection molding.
Program Director UROP
- +49 241 80-90695
- Send Email
- Project Offer-Number:
- UROP International
- Mechanical Engineering
- Organisation unit:
- Institute for Plastics Processing (IKV) in Industry and Craft at the RWTH Aachen University
- Language Skills:
- Fluency in English
Excellence Cluster: Internet of Production
In the field of injection molding, the optimization of machine parameters is a crucial part of the set-up process in order to ensure efficient and stable manufacturing. Due to its high non-linearity, the injection molding process is frequently optimized utilizing expert knowledge, e. g. from experienced technicians or engineers. However, this strategy guarantees neither an optimized process nor reproducibility, as it depends solely on the technician's experience. A possible solution to an unbiased set-up is the usage of data-based methods that are fitted to the specific process by gathered data samples. The already has vast experience with artificial neural networks and evolutionary algorithms in this field.
One highly important aspect to data-based modelling is the necessity of an adequate amount of data samples to train the models. As data acquisition is costly, in terms of time and material, the objective is to reduce the necessary amount of data to train the artificial neural networks. One possible approach to this is so-called transfer learning. During your time at the IKV, you will assist in conceptual as well as practical research to achieve this goal. Some of your tasks may be: - Literature research with latest publications, describing the state-of-the-art of artificial neural networks and transfer learning (generally and in the field of plastics processing) - Injection molding simulations - Construction of new molds for manufacturing - Data acquisition with injection molding machines on a fully connected production/shop floor - Measurements on produced parts for quality evaluation - Implementing new models of artificial neural networks in a Python-microservice