基于回归分析和小波变换的边坡变形组合预测研究

发布时间：2018-06-14 14:38

本文选题：边坡变形 + 回归分析　；参考：《长江科学院院报》2017年04期

【摘要】：为有效地判断边坡变形的发展趋势,基于边坡变形的现场数据,首先利用回归分析和小波变换分解边坡变形数据的趋势项和误差项,并选取若干最优的分解数据进行组合确定边坡变形数据的趋势项和误差项,再利用BP和RBF神经网络对趋势项和误差项序列进行预测,得到单项预测的结果,最后研究分析了定权组合预测和非定权组合预测的效果。结果表明:在趋势项和误差项的分离过程中,不同分离方法的分离结果具有一定的差异,以6次多项式回归、5次及7次傅里叶回归和sym2小波变换的结果较好;同时,在单项预测中,分项预测的效果要优于传统的单项预测,验证了分项预测的有效性,并由组合预测的结果可知,2种组合预测的效果均较好,均很大程度上提高了预测精度,且非定权组合的预测精度要优于定权组合预测的精度。上述研究为边坡的变形预测提供一种新的思路。
[Abstract]:In order to judge the development trend of slope deformation effectively, based on the field data of slope deformation, the trend term and error term of slope deformation data are decomposed by regression analysis and wavelet transform. Some optimal decomposition data are selected to determine the trend and error items of slope deformation data, and then BP and RBF neural networks are used to predict the trend and error items, and the results of single prediction are obtained. Finally, the effects of weighted combination prediction and undetermined weight combination prediction are analyzed. The results show that in the process of separating trend term and error term, the separation results of different separation methods are different, and the results of 6 degree polynomial regression and 7 times Fourier regression and sym2 wavelet transform are better, at the same time, In the single item forecast, the effect of the item forecast is better than that of the traditional single item forecast, which verifies the validity of the item forecast. According to the result of the combination forecast, the effect of the two kinds of combination forecast is better, and the precision of the forecast is improved to a great extent. The prediction accuracy of unweighted combination is better than that of constant weight combination. The above research provides a new idea for slope deformation prediction.
【作者单位】：青海省水文地质工程地质环境地质调查院青海省水文地质及地热地质重点实验室;
【分类号】：TV223

【相似文献】