AI-Based Turbocharger Identification for Automated Model Recognition and Repair Support
Master's Thesis — Ivana Petreva (2026)
Student: Ivana Petreva
Type: Master’s Thesis in Information Systems
Submitted: May 5, 2026
Supervision / advising team: Panagiotis Petropoulakis, Yonghan Kim, André Borrmann, Alois C. Knoll
This thesis develops an end-to-end computer-vision pipeline for identifying Garrett turbocharger part numbers from degraded industrial nameplates. The workflow combines spatial preprocessing, perspective and orientation correction, multimodal large language models, and deterministic fuzzy database matching.
The work evaluates robustness under realistic degradation such as rust, oil, glare, dot-peen engraving, noise, and scratches, with the goal of supporting automated model recognition and repair workflows in Maintenance, Repair, and Overhaul environments.