论文部分内容阅读
由于降雨入渗过程及其对滑坡的作用机理非常复杂,获得完整的小时间尺度的孔隙水压力时间序列对于研究降雨滑坡机理意义重大。但在对滑坡孔隙水压力进行监测时,由于观测方法、传感器质量、环境干扰等原因会造成部分传感器暂时或永久失效,获得的时间序列在时间尺度上差异很大,给实际应用带来了困难。本文将应用基于遗传算法的BP神经网络对孔隙水压力时间序列进行内插和外延预测,不仅预测效果良好,而且能较好地反映降雨过程与孔隙水压力变化之间的内在物理关系。
Because of the complicated infiltration process and its mechanism of action on landslides, it is of great significance to study the mechanism of rainfall-landslide when obtaining the complete time series of pore-water pressure on a small time scale. However, when monitoring the pore-water pressure of landslide, some sensors will temporarily or permanently fail due to the observation methods, sensor quality and environmental interference, and the obtained time series vary greatly in time scale, which brings difficulties to the practical application . In this paper, BP neural network based on genetic algorithm is used to interpolate and predict the pore water pressure time series. It not only has a good prediction effect, but also can well reflect the intrinsic physical relationship between rainfall process and pore water pressure change.