基于改进的支持向量机隧道变形预测模型  

Tunnel Deformation Prediction Based on Modified Support Vector Machine

作　　者：周奇才[1] 范思遐[1] 赵炯[1] 熊肖磊[1] 

机构地区：[1]同济大学,上海201804

出　　处：《铁道工程学报》2015年第3期67-72,共6页Journal of Railway Engineering Society

基　　金：国家自然科学基金项目(51375345);上海科学技术委员会资助项目(08201202103)

摘　　要：研究目的:用于消除地铁隧道变形监测数据受环境等因素引起的噪声影响,为提高变形数据的外推预测能力,提出基于离散小波消噪和动态在线滑移窗的支持向量机预测方法。研究结论:(1)利用离散小波变换对地铁隧道变形数据进行消噪处理,提取低频有效工程信息;(2)引入动态滑移窗技术,通过二阶双重滑移窗口的动态调整、更新,确定隧道变形预测模型的最终参训样本和实训样本数据,提高变形数据的有效利用率;(3)利用上海地铁隧道变形数据,从稳态与非稳态两个工况下分别对预测模型进行验证,检测模型的预测效果,结果表明该方法具有消噪效果好、预测精度高等特点,同时该预测模型可转为工程化应用,对隧道预测模型的搭建具有指导意义。Research purposes： In order to remove the noise of metro tunnel deformation monitoring data which affected by environmental factors and increase the extrapolated predictive ability of deformation data, a new prediction method based on discrete wavelets and a modified support vector machine （SVM） is proposed. Research conclusions： （ 1 ） The noise of tunnel deformation data can be removed by discrete wavelet analysis and the low frequency effective information can be acquired. （2） A dynamic on -line sliding window technique is introdueed. Additionally, the final training samples and the training samples of the tunnel deformation prediction model could be determined by utilizing the two orders of sliding windows which aiming at dynamically adjusting and updating the number of samples. This could improve data utilization rate. （3） The performance of prediction model has been validated with Shanghai metro deformation data under steady state and unsteady state. The results show that this method has good de -noising effect and high precision. Also, the prediction model could be transferred to engineering application and may give references to the construction of tunnel prediction model.

关 键 词：隧道变形 支持向量机 小波分析 滑移窗口 预测 

分 类 号：TU452[建筑科学—岩土工程]