Recognition of process conditions of a coal column flotation circuit using computer vision and machine learning

Abstract

It is a well-established fact that the flotation performance is reflected in the structure of the froth surface. A machine vision is able to extract the froth features and present them to the plant operators or the process control system. In this communication, computer vision and machine learning techniques are integrated for recognition of process conditions of a coal column flotation circuit. An industrial flotation column is operated under various process conditions. The metallurgical parameters (combustible recovery and concentrate ash content) are measured and the froth visual (bubble size, froth velocity and color) and textural (energy, entropy, contrast, homogeneity and correlation) features are extracted by a machine vision system. The principle component analysis (PCA) is applied to reduce the input space. The relationship between the froth characteristics and the metallurgical parameters is modeled using different intelligent algorithms and a predictive model is built. The froth images are classified based on the froth features using the K-means data-clustering algorithm. The predictive and classification models are eventually integrated to diagnose the process conditions of the flotation column.