【摘要】 地震定位过程中，由于地球介质的不均一性以及台站局部地质条件的复杂性，使震中距和地震波走时呈非线性关系。利用通常地震定位方法所确定的爆炸地震震中位置和实际震中存在２０～３０ｋｍ偏差。人工神经网络具有高度非线性映射功能，可应用于地震震中定位。应用ＢＰ（反向传播）神经网络确定远场爆破地震震中的实例表明，所确定的震中位置和实际震中位置偏差在８ｋｍ以内，外延预测确定的震中位置和实际震中位置偏差小于１８ｋｍ。更多还原

【Abstract】 The algorithm principle of BP neural networks is briefly introduced in the paper, at the same time, the formation of networks and the study skills are provided. Because of the earth medium non homogeneity and complicity of geological conditions of seismic stations, the relationship between epicenter distance and seismic wave propagation time is not linear.It has been found that epicenters located by usual seismic location method deviate from real epicenters. The deviation of distance is about 20km to 30km. The neural networks can be sucessfully applied to the epicneter locating as they have a high non linear reflective function. The errors between epicenters calculated by neural networks and actual epicenters are less than 8km and 18km respectively in and out of the seismic network control area.更多还原