【摘要】 利用人工神经元网络方法,提出了一种从连续的地震数据中检测出地震事件的方法。该方法分两步,首先,低阈值的STA/LTA算法从连续的波形中检测出类似地震事件;其次利用神经元网络方法,区分事件是地震事件还是噪声事件。通过对数据检测结果比较,找出了适合地震检测的神经元网络训练方法和神经元传递函数。在对天山流动台阵其中两个台的检测结果表明,在连续约两个月数据中,39RLS台检测出地震75个,30RNA台检测出地震95个,证明该方法对地震事件检测来说是一种有效的方法。更多还原

【Abstract】 An automatic seismic event detection method based on artificial neural network(ANN)is presented in this paper.The algorithm is divided into two steps:firstly all possible events are extracted by a loosely confined STA/LTA algorithm from continuous waveforms,and then seismic events are distinguished from them by an ANN based algorithm.Optimal transfer function and number of nodes in hidden layers is found out by comparing the detecting results of several hundreds of sample events.Tested by the Tian Mountain portable seismic array records,75 and 95 earthquake events are detected for the station 39RLS and 30RNA respectively in a period of 2 months,which shows that the method presented in this paper is an effective method for seismic events detection.更多还原