An image processing system for char combustion reactivity characterisation

Coal is the most used fuel source to generate electricity by pulverised coal combustion. During this process, volatile compounds are liberated giving rise to the formation of a variety of char particles. Char particles morphology can be classified into groups reflecting different coal reactivity levels which may be used to evaluate the effect of coal on the performance of burner. Char particles morphological classification may be automatically done with benefits in terms of speed, consistency and accuracy. However, the classification performance relies on correct identification of char particles. Moreover, broken walls, created during char generation process, blurriness and low contrast are factors that make the classification task a challenging problem. In this paper, we propose a system for particle detection and particle classification into two reactive groups. Initially, a set of candidate regions, that may contain particles, is selected by combining regions and edges. Then, regions containing particles are detected using texture features and a Support Vector Machine classifier. The particle classification is done based on the International Commission for Coal Petrology criteria. Experiments using coals from two Colombian regions-Valle and Antioquia-showed that the proposed system, in most cases, correctly detect char particles. Regarding the classification of detected particles, analysed char samples were automatically classified similarly as manual classification did. Consequently, the system is found to be a successful first approach for char combustion reactivity characterisation. (C) 2018 Elsevier B.V. All rights reserved.