A study of froth flotation of coal using artificial neural networks

Abstract

Froth flotation has been widely used in concentration of few ores and Coals. Artificial Neural Networks (ANNs) based on models inspired by our understanding of the structure and function of the biological neural networks hold the key to the success of solving intelligent tasks by machines. A comparative study of various artificial neural networks for the efficient concentration of coal in froth floatation method is discussed. The main difficulty in each of the pattern recognition techniques is that of choosing an appropriate model and selecting appropriate parameters for the concentration process. The parameters used in computing chosen from a pool of available parameters so that the stability and convergence of the network is achieved with minimum effort. It must also be possible to determine the parameters with minimum cost at a speedy rate. This paper touches on models for activation and synaptic dynamics in each of the feed forward artificial neural networks. The feedforward network structure is found to be suitable for this pattern classification task as this structure supports mapping input parameter set to an output pattern. Samples of training and test data are taken and the network is trained on the training data. By computing the output for input test data in various networks, the effectiveness of various networks is analyzed.