神经网络在开采与矿山地震活动性关系研究中的应用

开采与矿山地震（矿震）活动性关系通常用理论公式来表达，在实际应用中需要靠经验或回归的办法来确定其参数，而用其预测的矿震活动性规律在准确性上不够理想。本文将神经网络引入开采与矿震活动性关系的研究中，以影响矿震活动性的开采强度、开采深度等因素为输入，分别以单位时间内矿震累积震级和矿震次数为输出，并将训练后的网络用于矿震活动性预测。通过将该法应用于老虎台矿矿震活动性预测实践，并与传统的回归方法作比较，结果表明神经网络能够充分表达开采与矿震活动性之间的高度非线性关系，用它预测的矿震活动性规律在准确性上优于回归方法所得结果。

NEURAL NETWORK''''S APPLICATION TO STUDY THE RELATION BETWEEN MINING ACTIVITIES AND SEISMIC VIBRATIONS

Normally,the relation between mining activities and seismic vibrations induced by them is expressed by theoretical formulas.In the practice,relevant parameters are predicted in the light of experiences or by means of regressive method that are not sufficiently exact.by using the neural network,the mining intensity and depth which influence mainly the occurence of seismic vibration are the input while the accumulative degrees and times of seismic vibrations induced by mining activities are the outputs.In this way,the highly non linear relation between mining activities and seismic vibrations induced by them is clearly expressed.The method has been successfully used in Laohutai Mine,and the results of prediction are more exact than those obtained from regressive method.