Soft Computing Analysis Approach on Spontaneous Heating of Indian Coals

Abstract

Mine fire is one of the important aspect for the loss of production in coal mines. Once the coal catches fire, then it is challenging to defeat it. Spontaneous heating of coal causes due to auto-oxidation, when coal interacts with oxygen existing in the environment, then auto-oxidation process takes place. The interaction of coal and oxygen liberates heat, and the continuous accumulation of heat accelerate the rate of oxidation. If heat releases throughout the oxidation process are more than the heat dissipated, it causes to increase in the temperature. When the temperature of coal reaches up to the ignition point, then it catches fire it known as spontaneous heating. Spontaneous heating depends on the quality of coal. It is crucial to investigate their degree of proneness to the spontaneous heating; if we are able to find out, the nit helps the mine management to take action towards the mine fire to avoid loss of lives and property, coal reserves, environmental pollution and financial elements of mine.This M.Tech.work deals with designing of the prediction models using the various soft computing techniques. In my project, data collected from the previous researcher's work, it includes the data of South Eastern Coalfields Limited (SECL), Singareni Collieries Company Limited (SCCL), Mahanadi Coalfields Limited (MCL),Western Coalfields Limited (WCL),Eastern coalfield limited(ECL), North Eastern Coalfields (NEC), Northern Coalfields Limited (NCL), Indian Iron and Steel Company (IISCO), Bharat Coking Coal Limited (BCCL) and Tata Iron and Steel Company Limited (TISCO).Fifty-eight coal sample data have been used, which includes proximate analysis, ultimate analysis and susceptibility indices, i.e., crossing point temperature and wet oxidation potential difference of samples. This work deals with designing various soft computing models for prediction of spontaneous heating susceptibility indices of coal samples. To decide the input parameter for the various model statistical regression analysis carried out between the intrinsic parameters and the susceptibility indices and the parameters which shows a significant statistical result was taken as attributes for the models Further different soft computing techniques such as ANN Multilayer Perceptron Network (MLP),Fuzzy logic, Multivariate Adaptive Regression Splines (MARS),support vector regression(SVR)and hybrid soft computing technique, i.e., an adaptive neuro-fuzzy interface system (ANFIS) applied for the prediction of susceptibility indices to assess the spontaneous heating of coal.Input parameters for the proposed models selected by regression analysis. After the successful completion of all the proposed models, comparative studies were made between the models iv based on the coefficient of regression and root mean square error. From the various designed models, it was observed that wet oxidation provided a better result with ANFIS model than MLP, SVR, MARS and Fuzzy logic