Title: Visual Analytics in Process Engineering
Name: Jan-Tobias Sohns
Machine learning (ML) models in process engineering are attracting increasing interest due to their inherently data-driven approach and astonishing effectiveness compared to traditional simulations. While previous work mostly focused on improving the statistical performance of such models, the imparting of the thereby created knowledge has been taken for granted. However, beyond the direct application of ML models, the learned structures in model outputs are oftentimes equally informative, yet non-trivial to assess as they are high-dimensional. As such, the interpretable communication of this data is vital for its evaluation by domain experts.
Conveyance of complex domain data can be achieved through visualization. Specifically, visual analytics enables the interactive exploration of data sets and can thus reveal sophisticated structures in otherwise too large-scale data. The focus of this project is the development of application-fitting visualization techniques to provide detailed structural analysis of high-dimensional domain data.
The results of this project will enable the scientists to evaluate the structural findings of their current data models enabling new ways of further optimizing their process-level modeling.