Neural network modelling of coal pyrolysis

Yields of coal pyrolysis products obtained by four different techniques of pyrolysis were modelled using neural network analysis. Coals from five different sources were considered in this study. The modelling focused on effects as diverse as holding time of particles in bed, residence time of volatiles in freeboard, peak reaction zone temperature, pressure, pre-heating time of coal particles, rate of heating, secondary reactions in freeboard and bed, type of reactor, particle size, concentration of coal particles, bed depth, sample size, coal type, type of inert and type of carrier and its flow rate. A neural network model was trained and tested using the published experimental data. Even for the limited number of data points the network model was capable of approximating the experimental data precisely.