New BMBF reserch project on Machine Learning

SCALE launches a new BMBF project on machine learning together with Fraunhofer IWU Dresden and the Professorship for Artificial Intelligence in Chemnitz.

On December 1, 2018, SCALE started a new research project in collaboration with the partners Fraunhofer IWU Dresden and the Faculty of Computer Science, Professorship of Artificial Intelligence of TU Chemnitz. The project has a term of three years and is funded by the Federal Ministry of Education and Economy (BMBF). The goal is to use machine learning to forecast process parameters and component quality in automotive body manufacturing.

 

The subject of the research project is the model development and the application of machine learning (ML) to accelerate the planning and series startup in car body manufacturing. The input and output data generated in the process development or production start-up phase should be used for the development of forecasting models for process optimization or for quality management in production. In addition to the process or component-related parameters, the manufacturing expertise should also be considered in the model development.

 

By using different methods of machine learning in the body manufacturing the following results should be achieved:

 

  • Significant dynamization of the development cycles of body assemblies
  • Illustration of indirect and direct interactions between input and output variables (eg fluctuating semi-finished product properties, machine parameters, defined assembly properties)
  • Improvement of the robustness by regulation of the quality by situational adaptation of the influencing variables
  • Using data and forecast models of the development phase for data-driven models for quality assurance in production and thus increasing productivity by reducing downtime and production waste

The scientific challenges related to machine learning include:

  • Learning from data sets with different data dimensions (development phase, production start-up, series production)
  • Merging heterogeneous data sources (laboratory experiment, FEM simulation, production)
  • Implementation of manufacturing expertise in the selection of training data and model development
Kontakt
contact Ingolf Lepenies
  • FEM methods