Development of Mathematical Models for the Assessment of Fire Risk of Some Indian Coals using Soft Computing Techniques

Coal is the dominant energy source in India and meets 56% of the country’s primary commercial energy supply. In the light of the realization of the supremacy of coal to meet the future energy demands, rapid mechanization of mines is taking place to augment the Indian coal production from 643.75 million tons (MT) per annum in 2014-15 to an expected level of 1086 MT per annum by 2024-25. Most of the coals in India are obtained from low-rank coal seams. Fires have been raging in several coal mines in Indian coalfields. Spontaneous heating of coal is a major problem in the global mining industry. Different researchers have reported that a majority (75%) of these fires owe their origin to spontaneous combustion of coal. Fires, whether surface or underground, pose serious and environmental problems are causing huge loss of coal due to burning and loss of lives, sterilization of coal reserves and environmental pollution on a massive scale. Over the years, the number of active mine fires in India has increased to an alarming 70 locations covering a cumulative area of 17 km2. In Indian coalfield, the fire has engulfed more than 50 million tons of prime coking coal, and about 200 million tons of coals are locked up due to fires. The seriousness of the problem has been realized by the Ministry of Coal, the Ministry of Labour, various statutory agencies and mining companies. The recommendations made in the 10th Conference on Safety in Mine held at New Delhi in 2007 as well as in the Indian Chamber of Commerce (ICC)-2006, New Delhi, it was stated that all the coal mining companies should rank their coal mines on a uniform scale according to their fire risk on scientific basis. This will help the mine planners/engineers to adopt precautionary measures/steps in advance against the occurrence and spread of coal mine fire. Most of the research work carried out in India focused on the assessment of spontaneous combustion liabilities of coals based on limited conventional experimental techniques. The investigators have proposed/established statistical models to establish correlation between various coal parameters, but limited work was done on the development of soft computing techniques to predict the propensity of coal to self-heating that is yet to get due attention. Also, the classifications that have been made earlier are based on limited works which were empirical in nature, without adequate and sound mathematical base. Keeping this in view, an attempt was made in this research work to study forty-nine coal samples of various ranks covering the majority of the Indian coalfields. The experimental/analytical methods that were used to assess the tendencies of coals to spontaneous heating were: proximate analysis, ultimate analysis, petrographic analysis, crossing point temperature, Olpinski index, flammability temperature, wet oxidation potential analysis and differential thermal analysis (DTA). The statistical regression analysis was carried out between the parameters of intrinsic properties and the susceptibility indices and the best-correlated parameters were used as inputs to the soft computing models. Further different ANN models such as Multilayer Perceptron Network (MLP), Functional Link Artificial Neural Network (FLANN) and Radial Basis Function (RBF) were applied for the assessment of fire risk potential of Indian coals. The proposed appropriate ANN fire risk prediction models were designed based on the best-correlated parameters (ultimate analysis) selected as inputs after rigorous statistical analysis. After the successful application of all the proposed ANN models, comparative studies were made based on Mean Magnitude of Relative Error (MMRE) as the performance parameter, model performance curves and Pearson residual boxplots. From the proposed ANN techniques, it was observed that Szb provided better fire risk prediction with RBF model vis-à-vis MLP and FLANN. The results of the proposed RBF network model was closely matching with the field records of the investigated Indian coals and can help the mine management to adopt appropriate strategies and effective action plans in advance to prevent occurrence and spread of fire