基于灰色系统及BP神经网络算法的边坡变形预测精确度影响分析

Influence Analysis of Slope Deformation Prediction Accuracy Based on Grey System and BP Neural Network Algorithm

在边坡变形数字化分析中,数据算法的选取已成为影响预测结果的重要因素之一。本文旨在排除工程地质条件等其他现场因素的影响,研究分析工具中常见的窗口平滑多项式、小波分析,对灰色系统GM(1,1)及BP神经网络在预测边坡变形时的精确度影响。结果表明:(1)当边坡变形稳定,边坡位移时序图呈直线抬升趋势时,此时基于窗口平移多项式的BP神经网络能够获得较为精确的预测值;(2)当边坡变化异常,位移时序图呈阶跃式曲线时,此时基于小波分析的BP神经网络能够获得较为精确的预测值。

In the digital analysis of slope deformation,the selection of data algorithm has become one of the important factors affecting prediction results.The purpose of the paper is to research the influence of window smoothing polynomials and wavelet analysis,which are common in analysis tools,on the accuracy of grey system GM(1,1)and BP neural network in prediction of slope deformation by eliminating the influence of engineering geological conditions and other field factors.The results show that:(1)When the slope deformation is stable and the slope displacement sequence chart shows a straight upward trend,the BP neural network based on window smoothing polynomials can obtain more accurate predicted values;(2)When the slope changes abnormally and the displacement sequence chart shows a step curve,the BP neural network based on wavelet analysis can obtain more accurate predicted values.