EARTHQUAKE EARLY DETECTING SYSTEM HAVING SELF-LEARNING FUNCTION BY NEURAL NETWORK

摘要

PROBLEM TO BE SOLVED: To improve evaluation precision by applying a neural network to the evaluation of hypocentral parameter (magnitude, hypocentral distance and depth) performed within an observation point earthquake detecting device. SOLUTION: The evaluation of hypocentral parameter of an earthquake early detecting system having a self-learning function by a neural network is shown by flowcharts, wherein (a) is performed when an earthquake is present, and (b) is performed in learning which is performed sometimes when no earthquake is present. This system is basically the same as a conventional. system, and all evaluations are instantaneously ended after detection of an S-wave only in one observation point. In this system, the evaluation is performed by use of a neural network having all conceivably influential parameters as inputs. The neural network is an analyzing tool modeled after human neutron which has two functions of a learning function and an evaluating function using the network obtained therefrom. The learning is ordinarily performed, the hypocentral information of the Japan Meteorological Agency extending from the initial information of earthquake wave to the detected point to derive a network for evaluating the optimum value of the hypocentral parameter.