基于机器学习技术的隧道掘进机性状的预测模型研究

发布时间：2021-07-11 16:14

　　盾构隧道掘进机器性能和刀具磨损的预测是一个非线性和多变量的复杂问题。为解决这个问题,本研究旨在:i)建立确定隧道掘进机性能的智能分析框架,ii)预测隧道掘进过程中的机器性能（即盾构掘进效率和盾构切入速率）,iii)建立预测盾构刀盘寿命的智能化统计模型,iv)分析隧道施工过程中每个参数的作用效应,特征和影响因素。研究过程中,应用统计分析,机器学习技术,智能分析和现场实测数据验证等手段研究这一系列问题。首先,通过确定隧道掘进过程性能预测中最有效的参数,提出盾构掘进效率和盾构切入速率的新预测模型;然后,建立盾构刀盘寿命的智能化新模型来预测刀盘寿命;为了获得更为可靠的施工操作,基于地层力学参数与盾构施工参数两个方面,提出一种智能分析方法;最后,讨论分析盾构刀盘寿命预测中最重要影响参数的作用,确定预测模型。研究的创新成果总结如下:（1）提出了新的机器学习模型预测盾构机的掘进效率提出的机器学习模型集成了改进的粒子群优化（PSO）合法自适应神经模糊推理系统（ANFIS）在一起。提出的改进模型组合了基于模糊规则的系统和PSO算法,可以同时调整先行变量和后续变量。提出的模型与当前广泛使用模型,如神经网络... 

【文章来源】：上海交通大学上海市 211工程院校 985工程院校 教育部直属院校

【文章页数】：212 页

【学位级别】：博士

【文章目录】：
Abstract
摘要
List of acronyms and abbreviations
Chapter1 Introduction
    1.1 Background and motivation
    1.2 Role of models in tunneling
    1.3 Definition of the problem
    1.4 Objectives of this study
    1.5 Research strategy/design
    1.6 Structure of this dissertation
Chapter2 Literature review
    2.1 Introduction
    2.2 TBM Tunneling
        2.2.1 Working principle
        2.2.2 Earth pressure balance(EPB)tunnel boring machine
    2.3 Parameters influencing excavation performance and tool wear
        2.3.1 Face pressure
        2.3.2 Screw conveyor
        2.3.3 Thrust and torque of cutter wheel
        2.3.4 Soil conditioning agent
        2.3.5 Cutter wheel rotation speed
        2.3.6 Penetration rate,utilization factor,and advance rate
    2.4 Current state of disc cutter design and development direction
        2.4.1 Disc cutter life(Hf)prediction model
    2.5 TBM prediction models
        2.5.1 Artificial intelligent techniques
        2.5.2 Optimization techniques
        2.5.3 Evaluation TBM performance through AI techniques
    2.6 Summary
Chapter3 Data-driven framework for improving shield performance
    3.1 Introduction
    3.2 Visual analysis of the data
        3.2.1 Statistical modeling
        3.2.2 Principal component analysis
        3.2.3 Simple regression analysis
        3.2.4 Non-linear multiple regression analysis
    3.3 Advance rate prediction through neural network model
        3.3.1 Neural network architecture selection
        3.3.2 Analysis of neural network
    3.4 Advance rate prediction through fuzzy logic model
        3.4.1 Fuzzification part
        3.4.2 Knowledge base
        3.4.3 Fuzzy inference system(FIS)
        3.4.4 Defuzzification process
        3.4.5 Analysis of fuzzy logic
    3.5 Advance rate prediction through ANFIS techniques
        3.5.1 ANFIS Architecture
        3.5.2 Hybrid learning algorithm
        3.5.3 Structure identification methods
    3.6 Summary
Chapter4 Machine performance using optimization models
    4.1 Introduction
    4.2 Estimating TBM performance
    4.3 Proposed technique for advance rate prediction
        4.3.1 Original PSO algorithm
        4.3.2 Improvement of inertia weight
        4.3.3 Improvement of constriction factor
        4.3.4 Synchronously inertia weight and constriction factor
        4.3.5 Hybrid improved IPSO-ANFIS model
        4.3.6 Model evaluations
        4.3.7 Comparison of IPSO-ANFIS model with other techniques
    4.4 Proposed technique for penetration rate prediction
        4.4.1 Genetic algorithm
        4.4.2 Improving ANFIS using GA model
        4.4.3 Multi-objective fitness function
        4.4.4 Model evaluation
        4.4.5 Comparison of multi-objective optimization model with other technique
    4.5 Discussion
    4.6 Summary
Chapter5 Intelligent approach to estimate disc cutter life
    5.1 Introduction
    5.2 Visualization to estimate the consumption of disc cutter
    5.3 Developing model for estimating disc cutter life
        5.3.1 Statistical analysis
        5.3.2 Simple regression analysis
        5.3.3 Non-linear multiple regression analysis
    5.4 An intelligence technique
        5.4.1 Group method of data handling polynomial neural network
        5.4.2 Hybrid GMDH-GA technique
        5.4.3 Evaluation methodology of cutter life using optimized GMDH-GA
        5.4.4 Model validation
    5.5 Analyze the efficiency of parameters to predict cutter life
    5.6 Summary
Chapter6 Case studies:prediction of tunnel performance
    6.1 Introduction
    6.2 Guangzhou Metro Line no.9(Case study)
        6.2.1 Project description
        6.2.2 Geological conditions
        6.2.3 Rock percentage encountered the tunnel face
        6.2.4 Cutter wear and its effect on shield advancement rate
        6.2.5 Effect of TBM field database on advance rate
    6.3 Guangzhou-Shenzhen intercity railway project
        6.3.1 Project description
        6.3.2 Geological conditions
        6.3.3 Disc cutter consumption
        6.3.4 Analysis of shield parameters
    6.4 Discussion
        6.4.1 Visualization of the evolving models for TBM performance
        6.4.2 Visualization of the evolving models for disc cutter life
    6.5 Summary
Chapter7 Concluding remarks
    7.1 A brief summary
    7.2 Limitations
    7.3 Perspective
Appendix A
Appendix B
References
Acknowledgements
Curriculum vitae
Publications during my Ph D study


【参考文献】：
期刊论文
[1]The comparative analysis of rocks’ resistance to forward-slanting disc cutters and traditionally installed disc cutters[J]. Zhao-Huang Zhang,Sun Fei,Meng Liang.  Acta Mechanica Sinica. 2016(04)
[2]巨斑状花岗岩条件下TBM大直径盘形滚刀磨耗规律[J]. 杜立杰,纪珊珊,左立富,孔海峡,许金林,杜彦良.  煤炭学报. 2015(12)
[3]上软下硬地层碴土改良试验及应用研究[J]. 叶新宇,王树英,肖超,阳军生,周纯择.  现代隧道技术. 2015(06)
[4]TBM盘形滚刀在山岭隧道掘进过程中的磨损研究[J]. 赵战欣.  地下空间与工程学报. 2015(S1)
[5]惯性权值对粒子群算法收敛性的影响及改进[J]. 黄翀鹏,熊伟丽,徐保国.  计算机工程. 2008(12)
[6]粒子群优化算法的收敛性分析及其混沌改进算法[J]. 刘洪波,王秀坤,谭国真.  控制与决策. 2006(06)
[7]盘形滚刀的使用与研究（1）——TB880E型掘进机在秦岭隧道施工中的应用[J]. 万治昌,沙明元,周雁领.  现代隧道技术. 2002(05)
[8]隧道掘进机在中国地下工程中应用现状及前景展望[J]. 钱七虎,李朝甫,傅德明.  地下空间. 2002(01)
[9]基于粒子群优化的文档聚类算法[J]. 魏建香,孙越泓,苏新宁.  情报学报. 2010 (03)



本文编号：3278422