Abstract: According to the microseismic monitoring results, four main influencing factors, that is, mining depth, coal seam dip angle, mining thick and structure impact level, were regarded as judgment indexes and used to establish the dynamic forecasting model of mining-induced failure depth of floor by unascertained clustering method. The mean value of training samples, which come from 18 datasets measured by microseismic monitoring, were set as cluster center, and the weight indexes of judgment were determined by information entropy theory. Through calculating the product sum of multi-index comprehensive measurement of sample and the corresponding sample average, the forecasting value of the mining-induced failure depth of floor was obtained, and then the model was identified by the whole samples. In addition, to further test its reliability, the method was applied to forecast the other five samples to compare the forecasted values with the measured values. The results show that the average of relative error between forecasted values and measured values is less than 1%. The dynamic forecasting model of mining-induced failure depth of floor is reliable and practical, and it can be popularized and applied to the similar mines.

Key words: microseismic monitoring, unascertained clustering, mining-induced failure depth of floor, forecast