• 中国科学引文数据库(CSCD)来源期刊
  • 中文核心期刊中文科技核心期刊
  • Scopus RCCSE中国核心学术期刊
  • 美国EBSCO数据库 俄罗斯《文摘杂志》
  • 《日本科学技术振兴机构数据库(中国)》
二维码

隧道建设(中英文) ›› 2020, Vol. 40 ›› Issue (1): 1-8.DOI: 10.3973/j.issn.2096-4498.2020.01.001

• 专家论坛 • 上一篇    下一篇

人工智能科学在软土地下工程施工变形预测与控制中的应用实践——理论基础、方法实施、精细化智能管理(示例)

孙钧1, 2, 温海洋1   

  1. (1. 同济大学隧道及地下工程研究所, 上海 200092; 2. 上海市隧道(股份)集团公司院士工作室, 上海 200032)
  • 收稿日期:2019-10-23 出版日期:2020-01-20 发布日期:2020-04-04
  • 作者简介:孙钧(1926—),男,江苏苏州人,1949 年毕业于上海国立交通大学土木工程系结构工程专业,国内外隧道与地下工程知名学者、专家,同济大学荣誉、终身、一级教授,中国科学院(技术科学部)资深院士,前国际岩石力学学会副主席暨中国国家小组主席,中国岩石力学与工程学会名誉理事长(前理事长),中国土木工程学会顾问、名誉理事(前副理事长),中国土木工程学会、中国公路学会、上海市土木工程学会等学会隧道与地下工程分会前副理事长、理事长。 长江三峡工程、南水北调和港珠澳大桥等技术专家组专家。 E-mail: junsunk@163.com。

Application of Artificial Intelligence Science to Construction Deformation Prediction and Control of Underground Engineering in Soft Soil: Cases Study on Theoretical Foundation, Method Application and Fine Intelligent Technical Management

SUN Jun1, 2, WEN Haiyang1   

  1. (1. Institute of Tunnel and Underground Engineering, Tongji University, Shanghai 200092, China; 2. Academician Working Station of Shanghai Tunnel Engineering Co., Ltd., Shanghai 200032, China)
  • Received:2019-10-23 Online:2020-01-20 Published:2020-04-04

摘要:

: 首先,介绍了基于人工神经网络的智能预测方法(多步滚动预测)和基于智能模糊逻辑法则的施工变形控制方法对策;其次,介绍了基坑施工和盾构掘进施工变形智能预测与控制案例。经过应用实践,认为智能方法的优点是: 对于结构变形位移和周边地表沉降/隆起,智能方法所得的预测值(3~5 d)与其相应实测值的精度偏差一般为5%~10%;不只是可以了解到当天已发生的信息,还可预见3~5 d将要发生的变形位移和沉降/隆起等的预测定量值;在施工变形达到超限阈值前,采用智能模糊逻辑控制法则作处理,通过调整相应的施工技术参数,即可使后续变形始终处于允许的限值之内,而无需附加额外的巨大花费,节约造价,节省工期,还可实现远程、无线、视频监控。在探讨地铁施工变形智能预测与控制的基础上,开发了盾构掘进施工中工程周边地表沉降/隆起变形的多媒体三维动态可视化仿真程序软件,研制了盾构掘进施工计算机智能管理系统。目前,上海隧道工程有限公司已在上海市沿江通道盾构施工中进行试验性应用,取得了良好的技术效益。最后,对人工智能科学发展的前景及存在的一些问题进行了探讨。

关键词: 人工智能, 神经网络, 机器学习, 轨道交通/地铁, 地下车站深大基坑, 盾构法区间隧道, 施工技术参数, 施工变形智能预测, 智能模糊逻辑控制, 精细化智能技术管理, 5G网络系统

Abstract:

In this paper, the construction deformation control method based on artificial neural network intelligent prediction method(multistep fluctuant prediction) and intelligent fuzzy logic is put forward firstly. And then the intelligent deformation prediction and control of foundation pit and shield tunneling are presented by cases study. The applicable results show that the advantages of the intelligent method are as follows: (1) the precision error between 3 to 5 day predicted values of structural deformation and surface settlement and monitored values is within 5%~10%; (2) the current data can be seen and the deformation/displacement and settlement in 3 to 5 days can be predicted; (3) the subsequent construction deformation can be effectively controlled within limit by adopting intelligent logic control and adjusting construction technical parameters; (4) and a great amount of money can be saved, the construction schedule can be shorted, and the remote, wireless and video monitoring can be realized. Meanwhile, the multimedia 3D dynamic visualized simulation software and the computer intelligent management system are developed based on the discussion of metro construction deformation prediction and control. The software and the system have been successfully applied to the Yanjiang Tunnel shield construction in Shanghai by Shanghai Tunnel Engineering Co.〖KG-*3〗, Ltd. which resulting in good technical benefits. Finally, the prospect and issues of artificial intelligence science are discussed.

Key words: artificial intelligence, neural network, machine learning, rail transit/metro, deep and large foundation pit of underground station, shield tunnel, construction technical parameters, construction deformation intelligent prediction, intelligent fuzzy logic control, fine intelligent technical management, 5G network system

中图分类号: