An expert technique for optimization of underground mine support system

Exploitation of coal with safe roof control in India has been a challenging problem for many years. In terms of the method of coal excavation, the share of opencast mining, which was as low as 14% in 1951, increased to current high level of above 80%, whereas the share of underground mining declined from 77% in 1971, to its current 20%. Even though, we can't ignore the underground mining due to its superior quality of coal as well as for societal reasons i.e. rehabilitation of people in mining areas. According to statistics of accident data fall of roof/sides is one of the major causes of underground mine accidents. In many cases, our experiences and knowledge about soil and rock behavior still falls short of being able to predict how the ground will behave. Presently, empirical relations to design are widely used in estimating mine support parameters. Under these circumstances, expert technique plays an important role, and such accidents can be minimized with collection and optimization of parametric data and analysis using the fuzzy-neuro technique of artificial intelligence. The availability of data and knowledge are two important considerations in implementing such techniques i.e expert technique. In this paper we have discussed the different techniques of Artificial Intelligence (AI) that are adopted here include Artificial Neural Network (ANN), Fuzzy Logic Technique, and its hybridization, Fuzzy-Neuro Technique. Fuzzy-Neuro technique has come out with comparatively better solution to approximate the amount of pre-load causing roof fall which is applied to the standing props to support the roof during excavation. We have taken 12 input parameters of mining operation. Output obtained with fuzzy technique with 5 triangular membership function they were again trained with neural network using sigmoidal function. After having trained the network actual underground excavation parameters data were simulated through MATLAB to find the final output i. e pre-load to be applied on props so as not to dislodge during blasting. It was found satisfactory. Fuzzy-Neuro technique has better performance over individual fuzzy logic or neural networks technique.