Predictive models for pot-hole depth in underground coal mining—some Indian experiences

Abstract

Ground subsidence induced by extraction of coal seam belowground brings about changes in territorial environment. This occurs in two forms, namely, trough and pot-hole subsidence. Pot-hole subsidence is extremely hazardous as it does not give any prior indication before its occurrence. Several pot-holes have occurred in the recent past in the coal mines of South Eastern Coalfields Limited and called for a specific study to develop an in-depth understanding of various parameters influencing the pot-hole occurrence for formulating the basis of different predictive models. These critical parameters have been compiled and analysed for seven mines located in different areas of SECL, a subsidiary of Coal India Limited. Multiple regression and artificial neural network (ANN) techniques were used for the preparation of the predictive models to calculate pot-hole depth under different conditions. Different conditions considered in the study are development and depillaring, presence and absence of faults and water bodies. This paper presents the results of the studies carried out in Indian mines representing different geo-mining conditions along with the pot-hole depth prediction models developed. The developed models were validated for a few new cases with the model results matching (within 10 % error in the case of ANN model) with the actual pot-hole depth measured. More varied data sets can fine tune the developed models further.