The prediction of caving sequence in bord and pillar workings using Random Forest algorithm
Depillaring of coal seams is of prime importance for coal mining industry in view of depleting superior quality coal reserve and increasing import of foreign coal. Depillaring in conjunction with caving is the most hazardous operation due to sudden roof fall. Some researchers have focused their work on roof fall risk assessment using statistical methods with a view to safety of men and machinery and to minimize accidents, down time and loss of production. Extensive research has not been done to predict roof caving sequence which is the basic requirement for successful caving operation for achieving production with zero harm potential. Roof caving is the result of interactions of all geotechnical and mining parameters including extraction area which is its main cause and contributory parameter. In this research, Random Forest, a supervised ensemble machine learning algorithm along with grid search and cross-validation is used to process the interactions among various parameters and to predict the sequential occurrence of roof caving and characterize the same as local or main fall with considerable and reliable accuracy