EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI-Powered Fault Detection in Compound Parabolic Concentrators

Amine Moulay Taj, Mohamed Boujoudar, Zoubir Barraz, Ibtissam Bouarfa, Masaab El Ydrissi, El Ghali Bennouna, Hicham Ghennioui

SolarPACES Conference Proceedings · 16.11.2026 · 4

DOI: 10.52825/solarpaces.v4i.3037 ↗

Kurz gefasst

Fachartikel in der Zeitschrift SolarPACES Conference Proceedings, online erschienen am 16.11.2026. Verfasst von Amine Moulay Taj und 6 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Compound parabolic concentrators (CPCs) represent a cost-effective solution for medium-temperature solar thermal applications; however, their efficiency is highly sensitive to optical and thermal faults, such as dust deposition, vacuum losses, and broken tubes. Conventional inspection approaches, based on manual surveys and handheld instruments, are time-consuming, subjective, and impractical for large-scale deployments. This study presents an AI-driven, UAV-based inspection framework for the automated detection of faults in CPC systems, implemented and experimentally validated at the Green Energy Park research facility in Morocco. The methodology integrates drone-based thermal and RGB imaging with a two-stage deep learning pipeline. In the first stage, YOLOv8 was used to detect CPC tube receivers and generate bounding boxes for object detection, achieving a validation precision of 95.1% and recall of 99.7%. In the second stage, a ResNet classifier was applied to localize and categorize faults into four classes (normal, dust, vacuum loss, and broken tube), reaching an overall classification accuracy of 92.1% and an F1-score of 0.92. Qualitative analysis demonstrated that thermal imagery is particularly effective for detecting vacuum anomalies, while RGB imagery enhances visual validation of dust and cracks, confirming the advantages of integrating thermal and RGB imagery. The results highlight the robustness of the YOLOv8–ResNet pipeline and its potential to support predictive maintenance strategies in CPC installations by enabling early and accurate fault detection.

Zitieren

APA

Taj, A. M., Boujoudar, M., Barraz, Z., Bouarfa, I., Ydrissi, M. E., Bennouna, E. G., & Ghennioui, H. (2026). AI-Powered Fault Detection in Compound Parabolic Concentrators. SolarPACES Conference Proceedings, 4. https://doi.org/10.52825/solarpaces.v4i.3037

Vancouver

Taj AM, Boujoudar M, Barraz Z, Bouarfa I, Ydrissi ME, Bennouna EG, et al. AI-Powered Fault Detection in Compound Parabolic Concentrators. SolarPACES Conference Proceedings. 2026;4. doi:10.52825/solarpaces.v4i.3037

BibTeX
@article{Taj2026,
  author = {Taj, Amine Moulay and Boujoudar, Mohamed and Barraz, Zoubir and Bouarfa, Ibtissam and Ydrissi, Masaab El and Bennouna, El Ghali and Ghennioui, Hicham},
  title = {AI-Powered Fault Detection in Compound Parabolic Concentrators},
  journal = {SolarPACES Conference Proceedings},
  year = {2026},
  volume = {4},
  doi = {10.52825/solarpaces.v4i.3037}
}
Kostenlos starten →