【摘要】 根据诱发地质灾害的内外因素建立地质灾害预警模型,评价河南省地质灾害的危险性。选取地形地貌、地质构造、岩土体类型、植被分布、水土流失、降雨和人类生产活动7个影响地质灾害的因素,建立了地质灾害预警的BP神经网络模型。根据监测数据进行训练和检验后,采用该模型对河南省汛期地质灾害进行预测,发现预测结果与实际情况基本一致。研究表明,建立的BP神经网络模型作为一种灾害预警的探索和尝试,具有一定的适用性和推广价值,可以作为地质灾害危险性评价预测方法的补充。更多还原

【Abstract】 The given paper takes as its aim to introduce a discriminative analysis and forecast model intended to give warnings and predictions of geological hazards known as BP neural network model.The model developed by us comes from the need that the hazards of geological nature have to be evaluated and predicted through the internal and external influential factors on the hazards.And, theoretically speaking, seven main factors can be thought to account for such hazards, as the topographical and geomorphological factors, geological structure, types of rock and soil quality, vegetations, soil erosion, the local rainfall as well as the interference of the human before and during the geological events.Based on the analysis of such key factors, we have established an discriminative analysis model for prediction and evaluation of such hazards mainly caused by all the said seven or just some of these.Furthermore, BP neural network method has been adopted for the actual prediction of such hazards in flood seasons in Henan, a province, which is often hit by such disasters due to its complicated and intrigue landform types and climatic features.Large amounts of the monitoring data on such hazards collected and examined prove that the forecast and evaluation results are in close conformity with the actual situations.Thus, the results encourage us to believe that BP neural network model can be taken to serve as a powerful tool for predicting and warning such geographical disasters.In addition, the geological hazards, though their influencing factors are full of complication and uncertainty, can be well disclosed by using the non-linear information of BP neural network.And, therefore, the proposed method proves to be a reliable practical weapon in predicting and evaluating the geological hazards.更多还原