Advancing Particle Size Detection in Mineral Processing: Exploring Edge AI Solutions

Effective monitoring and control of particle sizes play a pivotal role in minimizing variability and enhancing energy efficiency within mineral processing plants. The conventional industry approach relies on laboratory procedures for particle size analysis, yet challenges arise in obtaining representative samples from bulk materials and rapidly assessing particle size. To address this, we introduce a novel machine vision framework based on Edge AI architecture and deep convolutional neural algorithms. This framework enables real-time particle size analysis, offering an alternative to traditional offline laboratory methods. This paper is a crucial component of our proposed concept and focuses solely on validating a deep convolutional neural network algorithm trained using synthetic datasets. The introduced model achieved an impressive mean Average Precision (mAP) score of 0.96, with processing times of under 1 s. These outcomes underscore the potential of deep convolutional neural networks for real-time particle size segmentation, representing a significant step toward pioneering an innovative Edge AI system for particle size assessment within mineral processing plants.