【摘要】 地震滑坡是一种有着严重危害的次生地震灾害形式,形成机制复杂,涉及因素较多。地震滑坡在空间上不是完全随机分布的,换言之,地震滑坡的影响因素和它的分布规律之间存在着相关性。利用径向基概率神经网络自学习的特性,通过对样本训练、检测,得到一个稳定可靠的模式识别网络,并用其对工作区进行潜在地震滑坡危险性区划,通过结果对比,在本例中识别精度达到89.9%以上,显示是一次有效的尝试。更多还原

【Abstract】 As we know, the landslide caused by earthquake does not distribute at random but has its destinations. This means that there are some inner relationships between the distribution and the factors which affect the happening of the landslide. But due to the complex mechanism of the earthquake-induced landslide, it’s difficult to describe the relationship clearly. For expressing the non-linear relationship, we create the artificial network with the Radial Basis Probabilistic Neural Network Algorithm by using the toolbox of MATLAB. At first, we select fault, river, rock, slope angle, earthquake intensity as the landslide affecting factors after studying the research results of other scholars, then use GIS to model the research area and grid it. From the cells, we select neural network training samples and testing samples. After repeated training, the neural network reaches its steadiness. Then we use this network to simulate the whole research area. In this paper, the studying area is between 98.5°～99.0°E, 24.2°～24.9°N, where 2 strong earthquakes took place in 1976 and caused many landslides. We selected two different sets of training samples from the research area, and by contrasting the result of RBPNN with the facts, we find that the predicted hazard area is generally in accordance with the fact in the distribution of earthquake-induced landslide. In conclusion, we think RBPNN is a useful method for the research of earthquake-induced landslide, especially for the regional earthquake-induced landslide zoning.更多还原