University of Thessaly (UTH)
Ανάλυση τυρβώδους ροής με χρήση δεδομένων άμεσης αριθμητικής προσομοίωσης και μεθόδων μηχανικής μάθησης
Abstract
dc:descriptionThis doctoral dissertation studies turbulent flows near solid walls, with emphasis on the zero-pressure-gradient turbulent boundary layer and the fully developed turbulent channel flow. Its purpose is to connect classical turbulence modelling with modern machine learning methods. Direct numerical simulation data are used as the main reference, allowing the evaluation of classical and new models for the description of the flow. First, semi-empirical composite models of the mean velocity profile are examined, and the validity of the logarithmic law and the power law is analysed for different Reynolds numbers. At the same time, second-order statistical quantities, as well as viscous stresses and Reynolds turbulent stresses, are evaluated. Then, a hybrid approach is developed, combining parameter optimization with physics-informed neural networks, in order to improve the prediction of the time-averaged velocity profile and of key turbulent quantities. Finally, symbolic regression is employed to derive interpretable mathematical expressions from high-fidelity data. The results show that the combination of the governing physics, DNS data and machine learning methods can lead to more accurate, computationally efficient and interpretable models for the description and prediction of turbulent flows near solid walls.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Palasis, Apostolos
- Παλάσης, Απόστολος
Subjects
dc:subject × 25- Τύρβη κοντά σε στερεά τοιχώματα
- Άμεση αριθμητική προσομοίωση
- Τυρβώδης ροή σε κανάλι
- Τυρβώδες οριακό στρώμα
- Λογαριθμικός νόμος ταχύτητας
- Νευρωνικά δίκτυα με ενσωματωμένη φυσική πληροφορία
- Συμβολική παλινδρόμηση
- Υπολογιστική ρευστομηχανική
- Wall-bounded turbulence
- Direct numerical simulation
- Turbulent channel flow
- Turbulent boundary layer
- Logarithmic law
- Physics-informed neural networks
- Symbolic regression
- Computational fluid dynamics
- Άλλες Επιστήμες Μηχανικού και Τεχνολογίες
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανική, διεπιστημονική προσέγγιση
- Επιστήμη Πολιτικού Μηχανικού
- Πολιτική μηχανική
- Other Engineering and Technologies
- Engineering and Technology
- Engineering, interdisciplinary
- Civil Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/62187