Zakaria Omarar completes his Master's thesis on multimodal fault localization in hydropower

Zakaria Omarar completed his Master’s thesis, Multimodal Sensing for Acoustic–Vibration Based Fault Localization in Hydropower (PDF). The thesis combines microphone arrays and vibration sensors to detect and localize faults, and asks an adversarial question: when does jointly trained multimodal fusion actually earn its complexity? The answer is an honest map of operating points, with negative results treated as first-class findings. The work is part of our hydropower condition-monitoring line. Congratulations, Zakaria!

The problem is real and confirmed on data from the industrial Rodundwerk II machine: when the operating regime shifts, the acoustic baseline moves with it, and unimodal detectors raise synchronized false alarms. The thesis evaluates jointly trained fusion against simpler unimodal, late-fusion, and classical alternatives at realistic data scale.