The application of artificial intelligence to predict extrusion processes has shown that simulation results can be predicted accurately, thereby reducing time-consuming simulations. However, a current challenge is to efficiently transfer developed models to new processes and data sets without having to build up an extensive database each time.

This thesis will investigate how transfer learning can be used for extrusion processes. The aim is to further develop existing models so that they can be transferred to new processes and products while still achieving a high degree of prediction accuracy. In particular, this should improve applicability for companies that regularly introduce new products and want to build on existing knowledge.

The work packages and the individual adaptation of the task will be determined in a personal interview. The applicant's personal interests and previous experience will be taken into account.

Research method

Experimental, theoretical, numerical