Roof fall threat analysis using fractal pattern recognition and neural network over mine microseismicity in a Central Indian longwall panel overlain by massive sandstone roof

Abstract

• Roof fall threat analysis using fractal pattern recognition and neural network. • Identification of stressed zones through spatial distribution of microseismic events. • Fast and automated identification of roof stress level through neural network. • Spatio-temporal forecasting of roof fall for enhancing safety in underground mines. Due to its high degree of mechanization, caved longwall mining is the most sophisticated and advanced mining method, and it is widely accepted across world's major coal producing countries. The overlying strata dynamics above goaf (void spaces behind powered supports) acts as the deciding parameter towards smooth and safe longwall operation. Usually, this overhanging load above goaf undergoes regular caving for releasing the excess accumulated stress, but sometimes the presence of massive hard strata (mostly sandstone beds) reduces the cavablity of immediate roof. The delayed caving tends to increase the hanging time and load weight of strata. The excess accumulated stress is released in form of sudden roof falls, which are massive by nature and also considered to be the most serious threat to the underground mines. The aftereffects of these untimely roof falls are often associated with catastrophic consequences in form of fatalities, non-fatal injuries, instrument loss and reduction in productivity. Therefore, fast-automated monitoring of roof stress level with higher degree of precision in real time is highly desired for safe operation of mines, which will have the ability to detect the location (in 3D) and threat level of stressed zones in the overlying strata for setting a precursory alarm towards roof fall forecasting. Similar work has been carried out in the present study, which tends to use the passive microseismicity associated with coal mine for monitoring the stress level in the overlying strata and roof fall forecasting through fractal pattern recognition and neural network over microseismic event distribution. The prime objective of the study was to enhance the safety in the working area of the underground mines through: (a) fast and automated identification of roof fall threat level in real time and to categorized it under five threat classes using fractal pattern recognition of microseismic event distribution, (b) developing a suitable predictive model in order to identify high hazard event windows, which acted as precursory signatures towards roof fall forecasting, and (c) timely identification of stressed areas in the overlying strata for providing prior information towards spatio-temporal forecasting of roof fall.