【作者】 李伟；

【导师】 汪林兵；

【作者基本信息】 北京科技大学 ， 土木工程， 2025， 博士

【摘要】 地下排水管道基础设施是我国城市生命线工程的重要组成部分。其中,钢筋混凝土排水管道现有服役和未来需求的占比量相对于其他材质的排水管道尤为突出,在中国钢筋混凝土排水管道数据监测研究报告中提到,在2030年该类管道的需求量将预计达到2000万吨,约占据世界总额的41%。近年来随着我国地区传统自然灾害风险阈值的提升与新兴灾害风险的涌现,韧性城市建设也随之在维度和尺度上有了更高的需求。特别是暴雨洪涝灾害的频发,给地下排水管道基础设施的功能服役与结构服役方面带来了严峻的安全挑战。在当前“十四五”时期对城市地下排水管道基础设施的问题总结中发现,外部环境风险因素的耦合增加了地下管道服役安全的难度和运维作业的复杂性;地下排水管道的服役状态反馈与感知水平仍需进一步增强,特别是在新一代信息技术推动下,多方合作运维与智能协作平台的建设需进一步提升。因此,地下排水管道在服役期间的结构响应量化与辅助协作运维系统的研发具有重要的科研意义、工程意义和应用价值。本文针对基于结构响应代理模型的地下管道拓展现实辅助运维系统研究,在以下五个方面进行针对性的研究工作:(1)以实际工程中三种直径的地下足尺钢筋混凝土管道为研究对象,搭建基于物联网增强的结构加载试验系统。通过对管道在加载过程中结构响应特征参数的分析得到其在不同结构损伤阶段的特征数据,并形成结构化表格数据(试验数据集)。在人工测量和计算机视觉算法的基础上,得到管壁损伤裂缝的统计值;通过离散元-有限差分耦合数值模拟方法搭建整体管道的裂缝量化模型,以此得到地下管道在垂直荷载作用下的结构损伤量化模型与方法。试验与研究结果表明,本研究所提出的地下管道结构损伤量化模型可以为出厂服役管道的结构响应与损伤分析提供参考。(2)在足尺管道试验数据基础上,基于情景构建理论搭建地下管道的服役状态分析框架。对单一足尺管道和受到车荷载作用的服役管道进行数值模拟计算,得到地下管道常规服役状态下的结构响应分析模型。基于近年来暴雨灾害的案例情景分析,本研究将暴雨灾害特点分为两类典型的特征:“长时间-多峰值”与“短时间-单极大峰值”。为此,本研究结合北京地区的降雨特征,基于数值模拟方法构建暴雨灾害中多风险因素叠加下的地下管道承灾服役状态分析模型,通过控制变量法进行因素影响的对比分析,并形成结构化表格数据(数值模拟数据集)。研究结果表明,在“长时间-多峰值”特点的暴雨灾害中,考虑管道的直径、埋深、脱空信息和路基土渗透参数多风险因素叠加状态下,当管道承接口下方存在脱空时,相比于无脱空的竖向位移增幅约10%。并且随着路基土渗透性的改变,地下管道的位移和环向应变呈现复杂的服役状态;相关的数值模拟模型可以为地下管道承受暴雨灾害的服役状态与结构响应分析提供参数与数据集参考。(3)在Agent-Based Modeling理论的基础上,构建基于试验-数值数据驱动的地下管道服役状态代理模型,将其命名为“UP-ABM”。利用Transformer深度学习模型基于试验数据集搭建足尺管道结构响应的代理体模型,用以快速评估出厂后标准钢筋混凝土管道的结构状态。利用一维卷积神经网络模型基于数值模拟数据集搭建工程服役管线结构的承灾(暴雨)代理体模型,用以快速评估管线的变形响应。研究结果表明,试验数据集与数值模拟数据集中的样本特征之间具有较好的关联性;在试验数据集背景下,Transformer深度学习模型相比于传统的多层前馈神经网络模型在回归能力方面更加突出,在保证模型泛化能力和鲁棒性的前提下,抗过拟合与欠拟合的能力更优;在数值模拟数据集背景下,一维卷积核神经网络模型相比于多层感知机模型和KAN模型在模型泛化和鲁棒性方面更加突出。UP-ABM模型可以实现对地下管道在出厂后不同服役阶段中服役状态的快速感知,为地下管道的服役代理模型提供算法和模型参考。(4)为了解决地下管道在多方合作运维与协作平台的建设需求,本研究依托地下足尺管道试验系统,基于混合现实理论与技术搭建用于管道信息感知与人机交互的基础空间计算系统。通过开发MR-IoT技术实现传感器与混合现实头戴式计算机的数据通信,用户仅需通过混合现实头戴式计算机可以感知地下管道复杂的数字物理信息。研究结果表明,混合现实空间计算技术可以很好的提升地下管道的信息感知、空间状态理解和空间交互方面的能力,为地下管道的信息感知提供一种人-环境-计算机交互的技术参考。(5)在混合现实空间计算系统的基础上,针对当前地下管道运维作业中影响作业效率和信息共享的因素,本研究构建基于拓展现实技术的地下管道辅助运维系统。从现场运维端人员和后台管理端人员的作业特征出发,实现拓展现实辅助运维系统的分布式应用。研究结果表明,拓展现实作为一种集成性的空间计算技术框架,可以为未来地下管道的智能运维理论与方法提供新兴技术参考。更多还原

【Abstract】 The underground drainage pipeline infrastructure is an important component of urban lifeline engineering in China.Among them,the proportion of existing service and future demand for reinforced concrete drainage pipes is particularly prominent compared to drainage pipes made of other materials.According to the data monitoring research report on reinforced concrete drainage pipes in China,the demand is expected to reach 20 million tons by 2030,accounting for about 41%of the world’s total.In recent years,with the increase of traditional natural disaster risk thresholds and the emergence of emerging disaster risks in China,there is a higher demand for resilient city construction in terms of dimensions and scales.In particular,the frequent occurrence of rainstorm-flood disasters has brought severe safety challenges to the functional service and structural service of underground drainage pipeline infrastructure.In the summary of the problems in urban underground drainage pipeline infrastructure during the current "14th Five Year Plan" period,it is found that the coupling of external environmental risk factors has increased the difficulty of underground pipeline service safety and the complexity of operation and maintenance works;the feedback and perception level of the service status of underground drainage pipelines still need to be further enhanced,especially in the construction of multi-party cooperative operation and maintenance and intelligent collaboration platforms driven by the new generation of information technology.Therefore,the quantitative analysis of underground drainage pipeline service behavior and the development of collaborative operation and maintenance systems have important scientific research significance,engineering significance,and application value.This paper focuses on the research of the underground pipeline extended reality assisted operation and maintenance system based on the structural response agent model,and conducts targeted research in the following five aspects:(1)A structural loading experimental system based on Internet of Things enhancement is constructed for three types of underground full-scale reinforced concrete pipelines with different diameters in practical engineering.By analyzing the mechanical response characteristic parameters of the pipeline in the loading process,the characteristic data at different stages of structural damage are obtained,and structured tabular data(experimental dataset)is formed.The statistical values of pipe wall damage cracks are obtained based on manual measurement and computer image processing algorithm;the discrete element-finite difference coupled numerical simulation method is used to construct an internal crack damage model of the pipe.A quantitative model and method for the structural damage of underground pipeline under vertical loads are obtained.The experimental and research results indicate that the quantification model of underground pipeline structural damage proposed in this study can provide a reference for the structural damage analysis of newly manufactured pipelines.(2)Based on full-scale pipeline experimental data,a service behavior analysis framework for underground pipeline is constructed based on scenario construction theory.Numerical simulation calculations are conducted on a single full-scale pipe and a service pipeline subjected to vehicle loads to obtain a behavioral analysis model for underground pipeline under conventional service conditions.Based on the accident scenario analysis of rainstorm disasters in recent years,this study divides the characteristics of rainstorm disasters into two typical characteristics:"long time-multiple peak values" and "short time-single maximum peak";Therefore,based on the characteristics of rainfall in Beijing,this study constructs an analysis model for the service behavior of underground pipelines under the superposition of multiple risk factors in rainstorm disasters based on the numerical simulation method.Through the control variable method,the comparative analysis of the influence of factors is carried out,and a structured table data(numerical simulation dataset)is formed.The research results indicate that in the rainstorm disaster with the characteristics of "long time-multiple peak values",considering the superposition of multiple risk factors which include the diameter,buried depth,void information of pipelines,vertical vehicle load,and subgrade soil permeability parameters,when there is soil void below the pipes socket,the vertical displacement increases by about 10%compared with the vertical displacement without void.With the change of subgrade soil permeability,the displacement and circumferential strain of underground pipelines present a complex service state.The relevant numerical simulation models can provide a reference for the service state and behavior analysis of underground pipelines subjected to rainstorm disasters.(3)Based on Agent-Based Modeling theory,an experimental numerical datadriven underground pipeline agent model is constructed and named "UP-ABM".The Transformer deep learning model is used to build the agent model of a single pipeline structure based on the experimental dataset,to quickly evaluate the structural response of standard reinforced concrete pipelines after leaving the factory.Based on the numerical simulation data set,the one-dimensional convolutional neural network model is used to build a withstand disasters(rainstorm)agent model of the pipeline structure in service to quickly evaluate the deformation response of the pipeline.The research results indicate that there is a good correlation between the sample features in the experimental dataset and the numerical simulation dataset.In the context of the experimental dataset,the Transformer deep learning model is more prominent in terms of regression ability compared to traditional multi-layer feedforward neural network models.While ensuring the model’s generalization ability and robustness,it has better resistance to overfitting and underfitting.In the context of numerical simulation datasets,onedimensional convolutional kernel neural network model is more prominent in model generalization and robustness compared to multi-layer perceptron model and KAN model.The UP-ABM model can quickly obtain the service status of underground pipelines at different service stages,to provide the algorithms and model references for the service agent model of underground pipelines.(4)To solve the construction needs of multi-party cooperative operation and maintenance and intelligent platform for underground pipelines,this study relies on the underground full-scale pipeline experimental system and builds a basic spatial computing system for pipeline information perception and human-computer interaction based on mixed reality theory and technology.By developing MR-IoT technology to achieve data communication between sensors and mixed reality headmounted computer,users only need to use the mixed reality head-mounted computer to perceive the complex digital physical information of underground pipelines.The research results indicate that mixed reality spatial computing technology can effectively enhance the information perception,spatial state understanding,and spatial interaction capabilities of underground pipelines,providing a technical reference for human environment computer interaction in information perception of underground pipelines.(5)Based on the mixed reality spatial computing system,this study constructs an underground pipeline assisted operation and maintenance system based on extended reality technology to address the factors that affect operational efficiency and information sharing in current underground pipeline operation and maintenance works.From the job characteristics of on-site operation and maintenance personnel and backend management personnel,to realize the distributed application of extended reality assisted operation and maintenance system.The research results indicate that expanding reality as an integrated spatial computing technology can provide a emerging technological reference for the intelligent operation and maintenance theory and methods of underground pipelines in the future.更多还原