Graduate School

Material and simulation models for the optimization of manufacturing processes

Person in Charge:M.Sc. Lukas Morand

Motivation and aims:

Optimization is crucial for the design of new manufacturing processes. Optimization tasks are for instance to determine optimal process parameters and to determine the base material to obtain certain component properties in an efficient way. Such kind of problems are of inverse nature. In contrast to the direct problem, inverse problems do not hold a unique relation between cause and effect, which pose a major challenge for a solution approach. The task of A16 is to develop simple but efficient process models in order to investigate and to solve optimization problems focusing the process steps rolling and deep drawing.
Untersuchungen Ergebnisse 
Solution approach for optimization problems:
  • Further development and implementation of methods from literature, that allow fast process simulations
  • Definition of optimization problems to study and, based on that, the generation of adequate data
  • Statistical analysis of the data and identification of significant parameters
  • Mapping of correlations in the data using methods of machine learning (e.g. neural networks)
Simple models to solve the direct problem:
  • Implementation of Taylor model for crystal plasticity for fast data generation
  • Direct mapping of correlated parameters and material properties of a rolling simulation via neural networks (surrogate model)
First results for optimization problems:
  • Model to infer skin pass level in rolling for given Rp₀₂
  • Determination of simple single crystal orientation for given material properties (e.g. elasticity) using ensemble techniques
Scheme of a simple ensemble technique (mixture model)
Determining ϕ for given E
Skinn pass level for given Rp₀₂