«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《建筑科学与工程学报》[ISSN:1673-2049/CN:61-1442/TU]

卷:

39卷

期数:

2022年04期

页码:

174-182

栏目:

出版日期:

2022-07-12

文章信息/Info

Title:

Research on Safety Assessment Model of Building Structure Based on Neural Network

文章编号:

1673-2049(2022)04-0174-09

作者:

王俊杰,焦 柯,彭子祥

(广东省建筑设计研究院有限公司,广东 广州 510010)

Author(s):

WANG Jun-jie, JIAO Ke, PENG Zi-xiang

(Guangdong Architectural Design & Research Institute Co., Ltd., Guangzhou 510010, Guangdong, China)

关键词:

结构安全评估;  神经网络;  小样本问题;  缺失数据插补;  查准率;  查全率

Keywords:

structural safety assessment;  neural network;  small sample problem;  missing value imputation;  precision rate;  recall rate

分类号:

TU312.3

DOI:

10.19815/j.jace.2021.09065

文献标志码:

A

摘要:

为实现建筑结构安全的快速评估,提出基于神经网络的建筑结构安全评估方法。基于《民用建筑可靠性鉴定标准》的调查与检测要求并考虑数据易获取性,选择45个涵盖承载力、耐久性、历史记录和环境情况等变量作为输入参数,以《民用建筑可靠性鉴定标准》中的安全等级作为输出参数,采用深度置信网络学习输入参数与输出参数间的非线性映射关系。对输入参数的选择、样本缺值问题、小样本问题和神经网络评估的可靠性进行探讨和验证。结果表明:在无法准确判断输入参数与输出参数相关性的前提下,采用全部输入参数的评估模型具有更高的鲁棒性; 迷失森林算法相较其他常用的缺值插补算法有更好的插补性能; 采用变分自编码器扩充训练样本集能有效提高神经网络的泛化能力和分类精度; 对深度置信网络引入加权交叉熵损失函数加以改进可增加训练时对不安全类别的敏感性,牺牲少量不安全类别的查准率可以大幅提高其查全率; 基于神经网络的结构安全评估模型能较好地预测结构的安全等级,具有快速且大批量运算的优势,是实现大范围建筑群结构安全监测的有效手段。

Abstract:

To achieve fast civil structure safety assessment, safety assessment model of building structure based on neural network was proposed. Based on the requirements of inspection and investigation of Standard for Appraisal of Reliability of Civil Buildings, 45 input parameters included bearing capacity, durability, historical condition, and environmental situation were selected as the input parameter. The safety level in the Standard for Appraisal of Reliability of Civil Buildings was selected as the output parameter. Deep belief network was adopted to study the nonlinear map relationship between the input and output parameters. The selection of input parameter, the problem of missing data, small sample problem and the reliability of neural network were discussed and validated. The results show that the evaluation model using all input parameters has higher robustness under the premise that the correlation between input parameters and output parameters cannot be accurately judged. Miss forest algorithm has better interpolation performance than other commonly used missing interpolation algorithms. Using variational autoencoder to expand training dataset can effectively improve the generalization ability and classification accuracy of neural network. Improved deep belief network by introducing weighted cross entropy loss function can increase the sensitivity of training to unsafe categories, and the recall rate can be greatly improved by sacrificing a small amount of precision rate of unsafe categories. The safety assessment model of building structure based on neural network is capable to predict the safety level of the structure. With the fast and mass operation ability, it is an effective mean to realize structural safety monitoring of large-scale buildings.

参考文献/References:

[1] 张协奎,成文山,李树丞.层次分析法在房屋完损等级评定中的应用[J].基建优化,1997,18(2):32-35.
ZHANG Xie-kui,CHENG Wen-shan,LI Shu-cheng.The Application of AHP in the Evaluation of Perfect Grade of Buildings[J].Optimization of Capital Construction,1997,18(2):32-35.
[2]李 静,陈龙珠,龙小梅.旧有建筑安全隐患及故障树分析方法[J].工业建筑,2005,35(增):46-49.
LI Jing,CHEN Long-zhu,LONG Xiao-mei.Hidden Troubles of Existing Buildings and Fault Tree Analysis Method[J].Industrial Construction,2005,35(S):46-49.
[3]袁春燕.城镇房屋安全管理与应急体系研究[D].西安:西安建筑科技大学,2008.
YUAN Chun-yan.Researches on the System of the Existing Building Safety and Crisis Management in City[D].Xi'an:Xi'an University of Architecture and Technology,2008.
[4]赵克俭.基于灰色理论的结构可靠性鉴定的研究[D].天津:天津大学,2005.
ZHAO Ke-jian.Study on Reliability Appraisal of the Structures Through Grey System Theory[D].Tianjin:Tianjin University,2005.
[5]邓聚龙.灰色系统理论教程[M].武汉:华中理工大学出版社,1990.
DENG Ju-long.Grey System Theory[M].Wuhan:Huazhong University of Science & Technology Press,1990.
[6]CAWLEY P,ADAMS R D.The Location of Defects in Structures from Measurements of Natural Frequencies[J].The Journal of Strain Analysis for Engineering Design,1979,14(2):49-57.
[7]YUEN M M F.A Numerical Study of the Eigenparameters of a Damaged Cantilever[J].Journal of Sound and Vibration,1985,103(3):301-310.
[8]PANDEY A K,BISWAS M,SAMMAN M M.Damage Detection from Changes in Curvature Mode Shapes[J].Journal of Sound and Vibration,1991,145(2):321-332.
[9]PANDEY A K,BISWAS M.Damage Detection in Structures Using Changes in Flexibility[J].Journal of Sound and Vibration,1994,169(1):3-17.
[10]PEREZ-RAMIREZ C A,AMEZQUITA-SANCHEZ J P,VALTIERRA-RODRIGUEZ M,et al.Recurrent Neural Network Model with Bayesian Training and Mutual Information for Response Prediction of Large Buildings[J].Engineering Structures,2019,178:603-615.
[11]XU Y J,LU X Z,CETINER B,et al.Real-time Regional Seismic Damage Assessment Framework Based on Long Short-term Memory Neural Network[J].Computer-Aided Civil and Infrastructure Engineering,2021,36(4):504-521.
[12]韩小雷,吴梓楠,杨明灿,等.基于深度学习的区域RC框架结构震损评估方法研究[J].建筑结构学报,2020,41(增2):27-35.
HAN Xiao-lei,WU Zi-nan,YANG Ming-can,et al.Research on Seismic Damage Assessment of Regional RC Frame Structures Based on Deep Learning[J].Journal of Building Structures,2020,41(S2):27-35.
[13]GOODFELLOW I,BENGIO Y,COURVILLE A.Deep Learning[M].Cambridge:MIT Press,2016.
[14]HORNIK K,STINCHCOMBE M,WHITE H.Multilayer Feedforward Networks Are Universal Approximators[J].Neural Networks,1989,2(5):359-366.
[15]BERGSTRA J,BENGIO Y.Random Search for Hyper-parameter Optimization[J].Journal of Machine Learning Research,2012,13(2):281-305.
[16]HINTON G E,OSINDERO S,TEH Y W.A Fast Learning Algorithm for Deep Belief Nets[J].Neural Computation,2006,18(7):1527-1554.
[17]SALAKHUTDINOV R,MNIH A,HINTON G.Restricted Boltzmann Machines for Collaborative Filtering[C]//GHAHRA MANI Z.ICML'07:Proceedings of the 24th International Conference on Machine Learning.New York:Association for Computing Machinery,2007:791-798.
[18]BARZI F,WOODWARD M.Imputations of Missing Values in Practice:Results from Imputations of Serum Cholesterol in 28 Cohort Studies[J].American Journal of Epidemiology,2004,160(1):34-45.
[19]TROYANSKAYA O,CANTOR M,SHERLOCK G,et al.Missing Value Estimation Methods for DNA Microarrays[J].Bioinformatics,2001,17(6):520-525.
[20]DEMPSTER A P,LAIRD N M,RUBIN D B.Maximum Likelihood from Incomplete Data via the EM Algorithm[J].Journal of the Royal Statistical Society:Series B(Methodological),1977,39(1):1-22.
[21]RUBIN D B.Multiple Imputation for Nonresponse in Surveys[M].Hoboken:John Wiley & Sons,Inc.,1987.
[22]BREIMAN L.Random Forests[J].Machine Learning,2001,45:5-32.
[23]STEKHOVEN D J,BUHLMANN P.MissForest-non-parametric Missing Value Imputation for Mixed-type Data[J].Bioinformatics,2011,28(1):112-118.
[24]STADLER N,BUHLMANN P.Pattern Alternating Maximization Algorithm for High-dimensional Missing Data[J].arXiv,2010:1005.0366.
[25]VAN BUUREN S,KARIN O.Flexible Multivariate Imputation by MICE[M].Leiden:TNO,1999.
[26]于 旭,杨 静,谢志强.虚拟样本生成技术研究[J].计算机科学,2011,38(3):16-19.
YU Xu,YANG Jing,XIE Zhi-qiang.Research on Virtual Sample Generation Technology[J].Computer Science,2011,38(3):16-19.
[27]KINGMA D P,WELLING M.Auto-encoding Variational Bayes[J].arXiv,2013:1312.6114.
[28]GOODFELLOW I J,POUGET-ABADIE J,MIRZA M,et al.Generative Adversarial Networks[J].Communications of the ACM,2020,63(11):139-144.

相似文献/References:

[1]杨佑发,熊丽,陈远.基于神经网络的框架结构损伤多重分步识别[J].建筑科学与工程学报,2011,28(01):106.
　YANG You-fa,XIONG Li,CHEN Yuan.Multi-stage Damage Identification for Frame Structures Based on Neural Network[J].Journal of Architecture and Civil Engineering,2011,28(04):106.
[2]李瞡,韩大建.索膜结构风振响应的神经网络辅助参数分析方法[J].建筑科学与工程学报,2008,25(04):98.
　LI Jing,HAN Da-jian.Neural Network Aided Parametric Analysis Method of Wind-induced Response for Cable-membrane Structures[J].Journal of Architecture and Civil Engineering,2008,25(04):98.
[3]叶爱文,谢慧才.混凝土中钢筋直径雷达检测的神经网络方法[J].建筑科学与工程学报,2008,25(04):105.
　YE Ai-wen,XIE Hui-cai.Neural Network Method of Diameter Detection of Rebar in Concrete by Using GPR[J].Journal of Architecture and Civil Engineering,2008,25(04):105.

备注/Memo

备注/Memo:

收稿日期:2021-09-03
基金项目:住房和城乡建设部科学技术计划项目(2019-K-157)
作者简介:王俊杰(1994-),男,广东广州人,助理工程师,工学硕士,E-mail:junjie.wang132@foxmail.com。

常用功能

更新日期/Last Update: 2022-07-10