摘要
为解决复杂施工环境下基坑周边近接建(构)筑物监测数据难以及时获取的问题,建立了基坑施工诱发周边既有建(构)筑物沉降SVM-BP预测模型。以某基坑工程近接建筑物的沉降为例,描述了该预测方法在实施过程中关键难点,并以实测数据进行验证。构建基坑开挖诱发周边既有建(构)筑物沉降预测体系,体系影响因子均来源于《建筑基坑工程监测技术规范》规定的应测项目,预测体系的构建关键在于影响因子与周边既有建(构)筑物沉降关系的建立。为此,根据实际工程案例特点和力学机理,采用将定性问题转换为定量表达的方法。而后,选择机器学习中的SVM、BP神经网络算法分别建立建(构)筑物响应预测模型。在此基础上,建立基于最小二乘准则的SVM-BP组合算法的预测模型,该模型可以充分利用不同机器学习算法的优势,一定程度上减少信息丢失,降低不确定性,使得预测精度进一步提高。
In order to solve the problem that it is difficult to obtain the monitoring data of the adjacent buildings( structures) around the foundation pit under the complex construction environment in time,we established the SVMBP prediction model of settlement of the existing buildings( structures) around the foundation pit induced by the construction of the foundation pit. Taking the settlement of a building close to a foundation pit as an example,we describe the key difficulties in the implementation of the prediction method,and verify the method with the measured data. Firstly,we established the prediction system of settlement of existing buildings( structures) around the foundation pit induced by excavation. The influence factors of the system are all from the measured items specified in the Technical code for monitoring of building excavation engineering.,and the successful construction of it lies in the establishment of the relationship between the influence factors and the settlement of surrounding existing buildings( structures). Secondly,according to the characteristics and mechanical mechanism of this project,we proposed that a method of transforming qualitative problems into quantitative expressions. So we selected the Support vector machine( SVM) and the Back propagation( BP) neural network algorithm in machine learning to build the response prediction of buildings. Finally,on this basis,we proposed a prediction model of SVM-BP combined algorithm based on the least square criterion. This model can make full use of the advantages of different machine learning algorithms,reduce information loss,reduce uncertainty,and further improve the prediction accuracy to some extent.
作者
李立云
孙庆玺
LI Liyun;SUN Qingxi(Key Laboratory of Urban Security and Disaster Engineering of Ministry of Education,Beijing University of Technology,Beijing 100124,China)
出处
《防灾科技学院学报》
2020年第2期1-9,共9页
Journal of Institute of Disaster Prevention
基金
国家自然科学基金(51278017,51421005)
北京市科技计划资助项目(Z181100009018001)
关键词
近接施工
结构响应预测
机器学习
最小二乘法
close-space construction
response prediction of structure
machine learning
least-square method
