Abstract
We present a novel method for characterizing the average size and mass concentration of coal particles within droplets in a coal monoethylene glycol slurry spray. The method is based on the analysis of continuous light scattering signals from individual droplets interacting with a shaped light beam. The experimental setup employs four detectors that collect scattered light at different angles, generating four distinct data streams corresponding to the droplet flow. These signals are transformed into Particle-Tensor-AI (PTA) representations, which can be constructed in various formats. In this work, we implement a format in which detector signals are mapped into RGB images, with each color channel representing a specific detector. The PTAs, labeled with the average coal particle size and mass concentration, are used to train convolutional neural networks (CNNs) for classification tasks. Various detector combinations, downsampling strategies, and material conditions were investigated. Validation experiments demonstrated proof-of-principle feasibility, achieving prediction accuracies up to 95%. This approach enables trigger-free, information-preserving analysis of light scattering data, offering strong potential for advanced diagnostics not only in coal monoethylene glycol sprays but also in other droplet and particle flows.