一种改进SVM优化模型的沉降预测方法  

A settlement prediction method based on improved support vector machine model

作　　者：容静 王凯 王文贯 范亚军 刘立龙 谢劭峰 Rong Jing;Wang Kai;Wang Wenguan;Fan Yajun;Liu Lilong;Xie Shaofeng(Guangxi Polytechnic of Construction,Nanning 530007,China;Guangxi Key Laboratory of Spatial Information and Geomatics,Guilin 541004,China;College of Geomatics and Geoinformation,Guilin University of Technology,Guilin,541004,China)

机构地区：[1]广西建设职业技术学院土木工程系,南宁530007 [2]广西空间信息与测绘重点实验室,广西桂林541004 [3]桂林理工大学测绘地理信息学院,广西桂林541004

出　　处：《工程勘察》2021年第9期46-49,59,共5页Geotechnical Investigation & Surveying

基　　金：广西高校中青年教师科研基础能力提升项目(2019KY1376,2020KY35022);广西自然科学基金项目(编号:2018GXNSFAA294045,2015GXNSFAA139230);广西空间信息与测绘重点实验室课题(16-380-25-11)。

摘　　要：筑路建设里一项至关重要的技术就是控制路基沉降,而动态、复杂的路基沉降预测系统导致其沉降数据呈现出非线性特征和"小量级、大波动"的敏感性,采用单一、传统的预测手段难以实现对路基沉降预测精确度的提升。针对这一问题,本文从常用的、针对路基沉降序列时序性和偶然性的角度出发,比较多种方法的属性原理、优缺点和普适性后,引入数据预处理来提取固有潜在信息与残差修正最优化的概念,结合相空间重构改变时序数据的关系结构来构造嵌入维数以及改进的灰狼算法优化支持向量机强大的跟踪、追捕功能,建立了小波与时间序列分析基础上的PSR-IGWO-SVM变形预测模型。工程实例分析表明,该模型使单项模型原本暴露出的形态稳固、指向性强、适应性弱等普遍预测缺陷得到了一定程度的完善,预测精度处于较高水平,优化模型符合建模工况原型,可以作为路基中长期沉降预测的有效辅助手段。Controlling subgrade settlement is a crucial technology in railway engineering.The dynamic and complex subgrade settlement prediction system results innonlinear characteristics and to be sensitivity as“small magnitude and large fluctuations”.Therefore,it is difficult to improve the accuracy of subgrade settlement prediction with a single,traditional prediction method.For solving this problem,this paper starts from the perspective of the timing and contingency of the subgrade settlement sequence.After comparing the advantages and disadvantages of various methods,the scope of application and related principles,introducing data preprocessing to extract the concepts of inherent potential information and residual correction with optimization models,combining with the phase space reconstruction to change the structure of time series data to construct the embedded dimension and the improved gray wolf algorithm is used to optimize the support vector machine for powerful tracking and pursuit,a phase space reconstruction IGWO-SVM with deformation prediction model based on wavelet analysis and time series analysis is established.Analysis of engineering examples shows that,the model makes up for the traditional single-item model whish shows the sensitivity prediction defects such as fixed form and strong pertinence,and obtains high prediction accuracy when the optimized model is adapted to the modeling conditions.It can be used as an effective auxiliary means for predicting long-term settlement of roadbed.

关 键 词：铁路路基 沉降预测方法 模型试验 沉降数据 改进的灰狼算法 

分 类 号：O141.4[理学—数学]