Vision-based air-flow monitoring in an industrial flare system design using deep convolutional neural networks

Said Boumaraf, Muaz Al Radi, Fares Oussama Abdelhafez, Pengfei Li, Khalid Yousef Al Awadhi, Hamad Karki, Sahraoui Dhelim, Naoufel Werghi

Abstract: Highlights•We design a novel lab-scale flare stack for controlled industrial conditions simulations.•We propose a vision-based system for air-flow monitoring in flare stacks using deep learning.•We achieve 99.04% accuracy and 98.85% F1-score using EfficientNetB7 with adaptive histogram equalization preprocessing to recognise three air-flow levels.•The proposed approach provides a real-time solution to optimize flares and minimize emissions.

External IDs:dblp:journals/eswa/BoumarafRALAKDW25