Prediction of spontaneous heating susceptibility of Indian coals using fuzzy logic and artificial neural network models

Coal mine fires due to spontaneous heating are a major concern worldwide. Most of these fires could be averted if suitable preventive measures are taken. Since the spontaneous heating potential of all types of coals are not the same, its accurate prediction is essential to plan efficient preventive measures and improve production and storage capabilities of a mine. The current paper presents a comparison of two approaches viz. fuzzy expert system and the commonly used artificial neural networks (ANN) for forecasting the self heating of coals. To apply these techniques, 30 coal samples of varying ranks were collected from different coalfields of the country. The intrinsic properties of the coal seams were determined by proximate, ultimate and petrographic analyses. The spontaneous heating proneness of the samples was studied using crossing point temperature (CPT), which is treated as an important measure for fire susceptibility of coal seams in Indian mines. Correlation studies between the intrinsic properties and CPT was carried out to identify the parameters for prediction purpose. Using moisture, volatile matter, and ash content as input parameters, CPT is predicted using fuzzy logic based on Takagi-Sugeno-Kang (TSK) model and ANN based on back propagation algorithm. Triangular fuzzy membership function is adopted for describing input variables. The results show both the models predict CPT with reasonable accuracy. Fuzzy modelling being a simpler approach, it can be utilized in the field where experimental data on coal properties are not precisely available but human judgement and intuition can be adopted for prediction purpose.