AI-Based Denoising of High-Speed TSP Temperature Images for Boiling Heat Transfer

Institut

Lehrstuhl für Thermodynamik (TUM-ED)

Typ

Masterarbeit /

Inhalt

 

Beschreibung

Background

Temperature-Sensitive Paint (TSP) enables high-speed surface temperature measurements in boiling experiments, but the resulting image sequences are strongly affected by shot noise, camera noise, illumination fluctuations, and low signal-to-noise ratio. Since clean reference temperature fields are usually unavailable, this project will develop denoising methods that exploit spatial and temporal coherence in the measured TSP data.

Objectives

1. Build a reproducible processing pipeline for high-speed TSP temperature image sequences from pool boiling experiments.

2. Implement and benchmark classical methods such as Gaussian/median filtering and POD/SVD-based denoising.

3. Implement advanced methods inspired by recent PSP studies, e.g. projected MSSA and blind denoising autoencoders.

4. Evaluate how denoising influences physical analysis, including bubble footprint detection, transient wall temperature evolution, and local heat-flux estimation.

Methods

1. Work with real TSP image sequences measured at approximately 4 kHz from boiling experiments.

2. Use Python for image preprocessing, normalization, denoising, and quantitative evaluation.

3. Compare denoising quality using SNR improvement, temporal smoothness, feature preservation, and robustness near bubble interfaces.

4. Optional extension: combine denoising with automated bubble-region segmentation or heat-flux partitioning analysis.

Expected Outcomes

1. A documented TSP image-denoising workflow with reusable scripts and example datasets.

2. A comparison of conventional, time-series, and machine-learning-based denoising strategies.

3. Recommendations for future AI-assisted analysis of boiling heat-transfer experiments.

Voraussetzungen

• MSc student in Physics, Mechanical Engineering, Electrical Engineering, or related field.
• Basic programming experience in Python, or similar; interest in image processing and machine learning.
• Knowledge of heat transfer or fluid mechanics is helpful.

Möglicher Beginn

immediately

Kontakt

M.Sc. Zhongqi Liu
Raum: 5507.EG.729
Tel.: +49 89 289 16193
zhongqi_.liutum.de