The simulation of indoor air flows using computational fluid dynamics (CFD) is a key tool in HVAC engineering. However, high-resolution simulations are very computationally intensive. Super-resolution methods offer the possibility of reconstructing finer flow structures from low-resolution simulations using machine learning techniques (e.g., autoencoders, convolutional neural networks (CNN)). This could significantly improve computational efficiency and open up new fields of application.

The aim of this thesis is to identify freely available Python code for flow field super-resolution through a literature review. The most promising code will be implemented and applied to CFD data generated for this purpose for indoor air flows, and systematically tested. In doing so, various influencing factors (e.g., magnification factor, Reynolds number of the flow) can be varied, and their effect on the quality of the results can be investigated.

This project requires programming skills as well as a basic understanding of computational fluid dynamics (CFD) and fluid mechanics. The scope and specific focus of the project can be adjusted.

Since our internships and the supervision of student projects are primarily intended to provide practical experience and career guidance, we ask for your understanding that no compensation is provided. However, we would be happy to issue you a formal internship certificate and nominate your student project for the annual academic award presented by the Association for the Promotion of Air Conditioning and Refrigeration Technology (Verein zur Förderung der Luft- und Kältetechnik e.V.).