[Bug]: 实验性功能精准翻译pdf文档（nougat）不可用

1 Introduction

At the present, TBM has become a preferred tool in the construction of long and large tunnels due to its high excavation rate and safety, low labor intensity as well as environmentally-friendly and cost-effective nature (Qian et al., 2002; Wang et al., 2004; Zhang and Fu, 2007). However, TBM is extremely sensitive to geological changes and excessively dependent on the operator's own experience. In case of geology condition changes, it is very difficult for TBM to timely monitor rock mass and equipment status and then make scientific decisions, resulting in frequent occurrence of geological disasters or construction accidents during excavation, including water and mud inrush, collapse and machine blockage, rock burst, squeezing and deformation of surrounding rock, damage of cutterhead or main bearing, and bringing substantial delay in construction and huge economic loss to the project and even causing equipment damage and personal death (Liu et al., 2016; Yin et al., 2005).

Many TBM projects have demonstrated that if safe and efficient construction is to be ensured, TMB shall be able to (i) monitor multi-source information in a real-time manner, including hydrology, geology, rock mass mechanics and machine operation, (ii) promptly and comprehensively integrate such information and make accurate judgement on the interaction between rock and TBM, (iii) make scientific decision and immediate regulation (Qian, 2017). With the help of information and intelligence technology such as Internet of things, big data and artificial intelligence, TBM can, during construction, solve problems such as delayed collection and transmission of information, incomplete analysis and processing of information and insufficient sharing and usage of information, to achieve safe and efficient TBM construction.

Through field research and literature review, this paper summarizes application conditions of existing information and intelligence technology in TBM projects. Additionally, combined with progress of the program called Basic Research on Whole-process Intelligent Control and Support Software for Safe and Efficient TBM Tumelling which is listed in National Key Basic Research Program of China, proposes an information and intelligence technology application system for TBM and its construction, and introduces key technologies in this system and their prospect so as to provide valuable references for TBM's future technical development and its safe and efficient construction.

现如今，隧道掘进机（TBM）由于其高速挖掘、安全性高、劳动强度低以及环保和经济效益的特点，已成为长距离、大型隧道施工中的首选工具（Qian等，2002;王等，2004;张和付，2007）。然而，TBM对地质变化极为敏感，并过度依赖操作者的经验。一旦地质条件发生变化，TBM很难及时监测岩体和设备状态，并做出科学决策，结果导致频繁发生地质灾害或施工事故，包括涌水和涌泥、塌陷和机械堵塞、岩爆、围岩挤压和变形、刀盘或主轴承损坏，并给项目带来严重的施工延误和巨大的经济损失，甚至导致设备损坏和人员死亡（刘等，2016;尹等，2005）。

许多TBM项目已经证明，如果要确保安全高效施工，TBM应能够（i）实时监测多源信息，包括水文、地质、岩体力学和机械运行等；（ii）及时全面地整合这些信息，并对岩石与TBM之间的相互作用进行准确判断；（iii）做出科学决策和立即调整（钱，2017）。借助物联网、大数据和人工智能等信息和智能技术，TBM能够在施工过程中解决信息采集和传输延迟、信息分析和处理不完整以及信息共享和使用不足等问题，以实现安全高效的TBM施工。

通过现场研究和文献回顾，本文总结了现有信息和智能技术在TBM项目中的应用条件。此外，结合列入中国国家重点基础研究计划的“全过程智能控制与支撑软件的安全高效TBM掘进基础研究”项目的进展，提出了一种TBM信息和智能技术应用系统及其施工方案，并介绍了该系统中的关键技术及其前景，以为TBM未来的技术发展和安全高效施工提供有价值的参考。

2 Information and intelligence technology application system for TBM construction

TBM is a very complicated and integrated equipment comprising various systems. Although the master computer in the main control room can provide more than one hundred kinds of TBM parameters, a TBM operator still has an insufficient knowledge of current geologicalconditions and TBM operating status and whether such geological conditions match such operating status, so the operator may, based on his personal experience, only perform manual control without scientific basis for decision-making when facing geological conditions with a high abrasion, strength or in-situ stress and complicated geological conditions such as water-bearing zone, fault fractured zone and leakage of toxic and hazardous gas. Meanwhile, that the inability to record and mine massive information and accident lessons generated from previous construction due to lack of proper information management platform leads to prevalence of unsafe and uneconomical TBM construction and frequent and repeated occurrence of various geological disasters and project accidents.

Aiming at the above problem which refers to lack of perception, decision-making and platform, this paper introduces an integrated technology application system to enable TBM informatized and intelligentized, which takes geology and TMB information perception as the basis, multi-source massive information fusion platform construction as the center and intelligent TBM excavation control as the application target, as shown in Fig. 1, and proposes an ecological technology framework for future TBM, which consists of information perception, information transmission, information fusion and intelligent decision, as shown in Fig. 2.

information perception: It aims to (a) promptly and accurately acquire forward geological information and mechanics information of rock mass being excavated, (b) collect real-time information related to TBM operating and main control parameters and (c) monitor health status of key TBM components;
information transmission: It intends to build an information transmission system from underground TBM construction site to the data center, so as to meet speedy and reliable transmission of multi-source heterogeneous and massive information and achieve real-time interaction between information from tunnel interior and exterior as well as between information from construction site and the data center;
information fusion: By using cloud computing technology, it aims to establish a standardized and formatted data storage warehouse and comprehensive information management platform, which not only provides the participants involved in the project with basic information service such as remote monitoring, construction management and data sharing, but also provides scientific decision basis for TBM health management and intelligent excavation by performing further analysis, calculation and mining of multi-source massive TBM information;
intelligent decision: By mining these rules about the interaction between rock mechanics properties and machine characteristics as well as equipment maintenance and fault diagnosis on the basis of information platform, it aims to study judgement and recognition algorithm or control model for different purposes to accomplish such functions as data driven based real-time safety warning, intelligent deviation correction and intelligent excavation.
TBM施工的信息和智能技术应用系统

TBM是一个非常复杂和综合的设备，包括多个系统。尽管主控室中的主计算机可以提供100多种TBM参数，但TBM操作员仍然缺乏对当前地质条件和TBM操作状态的了解，以及这些地质条件是否与操作状态匹配。因此，当面对高磨损、强度或原位应力以及包括含水带、断层裂缝带和有毒有害气体泄漏等复杂地质条件时，操作员可能只依靠个人经验进行手动控制，缺乏科学决策的依据。同时，由于缺乏适当的信息管理平台，无法记录和挖掘由于先前施工而产生的大量信息和事故教训，使得不安全、经济效益低的TBM施工普遍存在，并频繁发生各种地质灾害和工程事故。

针对上述缺乏感知、决策和平台的问题，本文介绍了一种集成的技术应用系统，使得TBM实现信息化和智能化，以地质和TBM信息感知为基础，多源大规模信息融合平台构建为中心，智能TBM挖掘控制为应用目标，如图1所示，并提出了未来TBM的生态技术框架，包括信息感知、信息传输、信息融合和智能决策，如图2所示。

信息感知：旨在(a)及时准确地获取正在开挖的岩体前向地质信息和力学信息，(b)收集与TBM操作和主控参数相关的实时信息，和(c)监测关键TBM部件的健康状态；
信息传输：旨在建立一个从地下TBM施工现场到数据中心的信息传输系统，以满足多源异构和大规模信息的快速可靠传输，实现隧道内外信息以及施工现场和数据中心之间的实时交互；
信息融合：通过使用云计算技术，旨在建立一个标准化和格式化的数据存储仓库和综合信息管理平台，不仅为项目参与者提供远程监控、施工管理和数据共享等基本信息服务，还通过对多源大规模TBM信息进行进一步分析、计算和挖掘，为TBM健康管理和智能挖掘提供科学决策依据；
智能决策：通过在信息平台基础上挖掘有关岩石力学特性和机器特性以及设备维护和故障诊断之间相互作用的规律，旨在研究不同目的的判断和识别算法或控制模型，实现基于数据驱动的实时安全预警、智能偏差修正和智能挖掘等功能。

3 Adverse geology and rock mass mechanics information perception

The beginning of 3 Adverse geology and rock mass mechanics information perception section.

第3章 "不良地质和岩体力学信息感知" 的开始部分如下所示：

Forward geological prospecting in TBM tunnel

Vereina railway tunnel in Switzerland is the first project to conduct forward geological prospecting by using TSP (Tunnel Seismic Prediction) system during TMB excavation (Xiao and Wu, 2004). Thereafter, many TBM tunnels all over the word, e.g. West Qinling tunnel, Zhongtinashunnel and a water diversion tunnel in China, headache tunnel of Neelum-Jhelum hydroeletic project in Pakistan, Tsukui tunnel in Japan, adopted different methods such as HSP (Horizontal Sound Probing), TRT (Tunnel Reflection Tomography), TSP (Tunnel Seismic Prediction) and TST (Tunnel Seismic Tomography) to perform forward geological prospecting, which had a positive effect on safe TBM construction(Ye, 2011; Gong et al., 2017; Xiong and Zhu, 2017; Yokota et al., 2016).

Compared with conventional tunnel construction methods, tunnel face is fully covered by TBM cutterhead and main TBM structures occupy most space and TBM's own metal structure and electrical system

Figure 1: Information and intelligence technology application system for safe and efficient TBM construction.

result in complicated electromagnetic environment in the tunnel and strong interference, so forward geological prospecting applicable to TBM is more difficult. In addition, TBM tunnelling speed is fast and under an average daily advance of 20-60 m, the demand for rapid and continuous detection of adverse geological information ahead of TBM is stronger.

In the early stages, forward geological detection technology was not effectively carried with TBM and such disadvantages as numerous arrangement workload of signal source and receiver points, time-consuming detection and interrupted detection exist. The BEAM system developed by GET (Geo Exploration Technologies) company in Germany utilizes focused electricity-induced polarization to achieve detection of water-bearing body ahead of TBM, the ISIS system developed by Geo Forschungs Zentrum uses seismic reflection imaging method to detect adverse geological structure and a team from Shandong University in China has also independently developed induced polarization system and 3D seismic wave geological prospecting system as shown in Fig. 3. All the above methods has developed into mature technique which can be integrated into TBM and has been applied in many tunnels with good results (Kaus and Boening, 2008; Lutth et al., 2008; Li et al., 2017, 2018).

In summary, forward geological detection technique mainly includes seismic wave method and electrical resistivity method. The seismic wave method has a better detection result for adverse geological structure of surrounding rock, e.g. fault and fractured zone, but is not sensitive to water. The electrical resistivity method has a good detection result for water, but cannot achieve three-dimension positioning and characterize water-bearing structures. Therefore, forward geological detection in the future needs to combine advantages of the above two kinds of detection methods to predictively locate and quantify adverse geology ahead of TBM by joint detection and information fusion. Moreover, rapid interpretation of detection results by means of machine learning or expert system, 3D imaging of prediction results as well as real-time and continuous detection by using vibration caused by disc cutter breaking rock as the signal source at the time of excavation will be an important development trend for forward geological detection in TBM tunnel in the future.

TBM隧道中的前向地质勘探

瑞士的Vereina铁路隧道是首个在TBM开挖过程中使用TSP（隧道地震预测）系统进行前向地质勘探的项目（Xiao and Wu, 2004）。此后，世界各地的许多TBM隧道，如中国的西秦岭隧道、中铁隧道，以及一条引水隧道，巴基斯坦的尼勒姆-哲鲁姆水电项目的头痛隧道，日本的津久井隧道，采用了不同的方法，如HSP（水平声波探测）、TRT（隧道反射层析成像）、TSP（隧道地震预测）和TST（隧道地震层析成像），进行前向地质勘探，对TBM安全施工产生了积极影响（Ye, 2011; Gong et al., 2017; Xiong and Zhu, 2017; Yokota et al., 2016）。

与传统的隧道施工方法相比，TBM结构占据了大部分空间，其刀盘完全覆盖了隧道的顶部，而TBM自身的金属结构和电气系统导致隧道内电磁环境复杂且干扰强烈，因此，适用于TBM的前向地质勘探更加困难。此外，TBM开挖速度快，平均日进尺达20-60米，对TBM前方不良地质信息的快速连续检测需求更加强烈。

在早期阶段，前向地质勘探技术未能有效地与TBM配合，存在信号源和接收点的大量设置工作量、耗时的检测和中断检测等缺点。德国GET（地质勘探技术）公司开发的BEAM系统利用聚焦电感极化检测TBM前方的含水层，Geo Forschungs Zentrum开发的ISIS系统采用地震反射成像方法探测不良地质结构，中国山东大学的一个团队也独立开发了感应极化系统和3D地震波地质勘探系统，如图3所示。所有上述方法已经发展成为成熟的技术，可以集成到TBM中，并在许多隧道中应用并取得了良好的效果（Kaus and Boening, 2008; Lutth et al., 2008; Li et al., 2017, 2018）。

总之，前向地质勘探技术主要包括地震波方法和电阻率方法。地震波方法对周围岩体的不良地质结构（如断层和裂隙带）有较好的检测结果，但对水不敏感。电阻率方法对水具有较好的检测结果，但无法实现三维定位和表征含水结构。因此，未来的前向地质勘探需要将上述两种检测方法的优势结合起来，通过联合检测和信息融合，对TBM前方的不良地质进行预测定位和量化。此外，利用机器学习或专家系统进行快速解读检测结果、预测结果的三维成像，以及使用盘刀破岩所引起的振动作为开挖时的信号源来进行实时连续检测，将是未来TBM隧道前向地质勘探的重要发展趋势。

Evaluation of mechanical parameters of rock mass being excavated by TBM

Internationally-accepted rock mass quality classification methods include such systems as Q, RMR, RQD and GIS, etc. On the other hand, domestically-recognized methods in China include national standard BQ and water resource industry standard HC. These methods are all for the surrounding rock stability analysis under conventional construction

Figure 3: Schematic diagram of induced polarization and 3D seismic wave detection system developed by Shandong University.

Figure 2: Ecological framework integrating information and intelligence technology for TBM construction.

method, but cannot provide a better evaluation on the rock mass boreability (Liu et al., 2016). RMC method proposed by Laughton (1998), QTBM method derived from improvement of Q system by Barton (2000) and the new RME rating system proposed by Bieniawski et al. (2007) can be used to evaluate rock mass excavability, predict average advance rate and complete TBM model selection. On the basis of RQ method, He et al. (2002) and Li and Peng (2006) respectively revised different rock mass mechanical parameters to put forward surrounding rock quality classification methods based on TBM construction. By using fuzzy mathematics method, Qi and Wu (2011) adopted uniaxial compressive strength and rock mass integrity index to classify rock mass quality for TBM construction. All the above methods are, based on case study, used to evaluate rock mass boreability and classify the rock mass quality by summarizing effect of geology and rock mass characteristics on actual TBM penetration rate. In the future, by accumulating and mining massive data of TBM project, we can obtain more accurate TBM performance and utilization prediction models, which will significantly improve scientificity of rock mass quality classification.

As TBM tunnelling is distinct from conventional construction method, the site personnel cannot directly observe rock mass conditions of tunnel face and shield TBM advances in the dark, leading to frequent occurrence of project accidents and inefficient excavation. Specific to difficulty in real-time evaluating parameters of rock mass being excavated, a new idea is to compare TBM excavation to a large-scale torsional shear test, which means that massive mechanical, electrical and hydraulic data generated during TBM excavation will be recorded by the monitoring system and the interaction rules between rock mechanics properties and machine characteristics will be mined by using the big data method to establish a new model for evaluating rock mass parameter. Fukui and Okubo (2006) suggested a method for evaluating the strength of rock mass being excavated, based on thrust force, torque, number of disc cutter and cutterhead diameter as well as penetration depth. Sun et al. (2008) suggested using statistical regression and artificial neural network to determine rock mass quality. Jing et al. (2019) suggested using TBM performance prediction model obtained from stepwise regression to reversely predict rock mass parameters based on TBM tunneling parameters.

Furthermore, in the previous engineering practices, the geological engineers generally determine rock mass conditions at the tunnel face by observing such features as geometrical morphology of muck excavated by TBM and particle size distribution, which is more practical in shield TBM tunnelling projects (Jin et al., 2001; Zhang, 2008). However, by depending on subjective experience, muck information cannot be expressed quantitatively and its timeliness and sustainability is very poor. Based on machine vision and by means of standardized image collection, intelligent detection and segmentation of muck edge and statistics of particle size distribution, a database in which muck characteristics correspond to rock mass parameters will be established to conduct research on intelligent algorithm for quality evaluation and classification of rock mass being excavated. Similar method has been applied in mine blasting result evaluation, but its application in TBM still needs to break through intelligent image segmentation technology, as shown in Fig. 4.

In the future, relying on the integrated information management platform or intelligent terminal for TBM and based on rock mass condition evaluation model and intelligent algorithm for muck-rock mass quality classification, a software applicable to real-time evaluation of mechanics information of rock mass being excavated by TBM will be developed to provide an important basis for safe and intelligent TBM excavation, as shown in Fig. 5.

TBM开挖的岩体力学参数评价

国际上公认的岩体质量分类方法包括Q、RMR、RQD和GIS等系统。另一方面，中国国内承认的方法包括国家标准BQ和水资源工业标准HC。这些方法都用于常规建设下的围岩稳定性分析，但无法更好地评估岩体的钻掘性能（Liu et al.，2016）。

图3：山东大学开发的感应极化和3D地震波检测系统示意图。

图2：信息和智能技术与TBM施工的生态框架集成。

Laughton（1998）提出的RMC方法，Barton（2000）改进Q系统导出的QTBM方法和Bieniawski等人（2007）提出的新的RME评级系统可以用于评估岩体的可钻性，预测平均进度速度和完成TBM模型选择。在RQ方法的基础上，He et al.（2002）和Li and Peng（2006）分别修正了不同的岩体力学参数，提出了基于TBM施工的围岩质量分类方法。Qi和Wu（2011）采用模糊数学方法，采用抗压强度和岩体完整性指数来对TBM施工的岩体质量进行分类。上述所有方法都是基于案例研究，通过总结地质和岩体特征对实际TBM穿透速度的影响，用于评估岩体的可钻性和分类岩体质量。将来，通过积累和挖掘TBM项目的大量数据，我们可以得到更准确的TBM性能和利用预测模型，从而显著提高岩体质量分类的科学性。

由于TBM隧道工程与常规建设方法有所不同，现场人员无法直接观察隧道掌子面的岩体条件和盾构机在黑暗中的推进情况，导致项目事故频发和开挖效率低下。针对实时评估正在开挖的岩体参数的困难，新的想法是将TBM开挖与大规模扭剪试验进行比较，这意味着TBM开挖过程中生成的大量机械、电气和液压数据将由监测系统记录，利用大数据方法挖掘岩石力学特性与机械特性之间的相互作用规律，建立一个评估岩体参数的新模型。福井和大久保（2006）建议采用推力、扭矩、刀盘切削器数量和刀盘直径以及穿透深度来评估正在开挖的岩体强度。孙等人（2008）建议使用统计回归和人工神经网络来确定岩体质量。荆等人（2019）建议使用由逐步回归获得的TBM性能预测模型，根据TBM掘进参数反向预测岩体参数。

此外，在以往的工程实践中，地质工程师通常通过观察TBM开采挖出的废矿的几何形态和颗粒粒度分布等特征来确定隧道掌子面的岩体条件，这在盾构TBM隧道工程中更为实用（Jin et al.，2001；Zhang，2008）。然而，凭借主观经验，废矿信息无法以定量方式表达，其即时性和可持续性非常差。基于机器视觉，并借助标准化图像采集，废矿边缘的智能检测和分割以及颗粒粒度分布的统计，将建立一个废矿特征对应岩体参数的数据库，进行正在开挖的岩体质量评价和分类的智能算法研究。类似的方法已经在矿山爆破的结果评价中应用，但在TBM中的应用还需要突破智能图像分割技术，如图4所示。

将来，依靠集成的信息管理平台或TBM的智能终端，并基于岩体条件评估模型和废矿-岩体质量分类的智能算法，将开发一个适用于实时评估TBM开挖的岩体力学信息的软件，为安全和智能的TBM开挖提供重要依据，如图5所示。

4 TBM equipment condition monitoring and evaluation

The beginning of 4 TBM equipment condition monitoring and evaluation section.

第四章 TBM设备条件监测和评估的开始部分。

TBM disc cutter condition monitoring system

By means of monitoring TBM condition, the TBM operator can effectively avoid frequent failure or further deterioration of TBM equipment to enhance maintenance timeliness and increase equipment availability, meeting requirements for long distance TBM excavation under harsh geological conditions.

Disc cutter directly contact and cut rock mass and its operating status is one of the key indicators to describe current TBM conditions. At present, disc cutter wear or damage conditions can be observed only during TBM downtime. Due to lack of real-time monitoring on disc cutter condition, it is very difficult for the operator to timely adjust tunneling parameters for improving service life of disc cutter and damaged disc cutter cannot be replaced in time, causing more serious damage. In some projects, the cost of disc cutter consumed accounts for around 30% of total TBM construction cost and the time consumed for replacement of disc cutter accounts for 20-40% of total construction period(Wan et al., 2002; Zhang, 2007). The disc cutter condition monitoring system can achieve the following functions as shown in Table 1.

Samuel and Seow (1984) monitored forces acting on disc cutter by measuring cutter shaft strain at the construction site. Zhang et al. (2003) developed a disc cutter force and temperature monitoring system to obtain parameters for disc cutter working status. Beer (2009) developed the Mbydic system to monitor such parameters as disc cutter force, temperature and rotational speed, which was applied in slurry shield machine. Moulin and Vallon (2010) made predictive description for geological condition ahead of TBM by means of monitored data such cutter force and rotational speed. Shanahan and Box (2011) developed a disc cutter monitoring system with additional monitoring of cutter vibration parameters. Entacher et al. (2013) developed a cutter force monitoring system, achieving visual perception of geological formation at TBM tunnel face. Robbins company developed the Samtructurer system which was applied in Rossidga project in Norway and upgraded in AMR project in India (Crawford, 2017).

Working conditions for disc cutter is extremely harsh under the effect of such factors as cutterhead vibration, ground water and dust. Additionally, the disc cutter condition monitoring system is not applied maturely due to closed structure of mell cutterhead, small installation space and extreme difficulty in supplying power and transmitting signal. At the moment, only DCRM (2018), Smartcutter and Mbydic systems are employed in actual projects.

In the future, efforts shall be made continuously to conduct in-depth research on microsensor and communication techniques, power batteries with a high capacity so as to achieve long-term real-time monitoring on information such as disc cutter force, temperature and wear, which is important for improving TBM intelligence. Moreover, it is also very necessary to add cutterhead vibration sensor for prevention of its crack and to perform strain monitoring in shield area for early warning of TBM blockage (Huo et al., 2017; Huang et al., 2018).

通过监测TBM（隧道掘进机）的状态，TBM操作员可以有效地避免TBM设备的频繁故障或进一步恶化，提高维护的及时性并增加设备的可用性，满足在恶劣地质条件下进行长距离TBM隧道掘进的要求。

刀盘刀具直接接触和切割岩石，并且其工作状态是描述当前TBM状态的关键指标之一。目前，刀盘刀具的磨损或损坏状态只能在TBM停机期间观察到。由于缺乏对刀盘刀具状态的实时监测，操作员很难及时调整掘进参数以延长刀盘刀具的使用寿命，而损坏的刀盘刀具无法及时更换，导致进一步的严重损坏。在某些项目中，刀盘刀具消耗的成本约占总TBM施工成本的30%，而更换刀盘的时间占总施工周期的20-40%（Wan et al.，2002; Zhang, 2007）。刀盘刀具状态监测系统可以实现如下表格1所示的功能。

Samuel和Seow（1984）通过测量施工现场刀盘轴应变来监测刀盘刀具受力情况。Zhang等人（2003）开发了一个刀盘刀具力和温度监测系统，以获取刀盘刀具工作状态的参数。Beer（2009）开发了Mbydic系统，用于监测刀盘刀具力、温度和旋转速度等参数，该系统应用于泥浆盾构机。Moulin和Vallon（2010）通过监测刀盘刀具力和旋转速度等数据，对TBM前方的地质条件进行预测描述。Shanahan和Box（2011）开发了一个刀盘刀具监测系统，同时监测刀盘振动参数。Entacher等人（2013）开发了一个刀盘力监测系统，实现了TBM隧道掘进面地质形成的可视感知。Robbins公司开发了Samtructurer系统，该系统应用于挪威的Rossidga项目，并在印度的AMR项目中升级（Crawford，2017）。

刀盘刀具在刀盘振动、地下水和灰尘等因素的影响下，工作条件极其恶劣。此外，由于刀盘结构封闭、安装空间有限以及供电和信号传输的极大困难，刀盘刀具状态监测系统的应用仍不成熟。目前，实际项目中只有DCRM（2018）、Smartcutter和Mbydic系统得到了应用。

未来，我们应不断深入研究微传感器和通信技术、高容量电池等方面，实现对刀盘刀具力、温度和磨损等信息的长期实时监测，这对于提高TBM的智能化至关重要。此外，为防止刀盘开裂，添加刀盘振动传感器，以及在盾构区域进行应变监测以提前预警TBM堵塞同样非常必要（Huo et al.，2017; Huang et al.，2018）。

TBM equipment health management

As TBM faults are invisible and caused by other faults, fault diagnosis and maintenance are not so easy. A previous research showed that when TBM fails to work correctly, 70-90% out of TBM downtime is spent on diagnosing and locating faults (Huang et al., 2012). Zhao et al. (2003) and Han (2003) suggested performing condition monitoring and fault diagnosis on key components such as main TBM bearing and gearbox by means of vibration monitoring and oil analysis technologies. Based on importance of each system of TBM, Liu (2007) classified monitored components into three levels, namely key, significant and general, and suggested implementing different monitoring strategies. Wang and Xu (2003) established a TBM fault database to develop an expert system for fault diagnosis. All the above researches have a positive effect on evaluating TBM health status and shortening the time of fault diagnosis.

With the continuous development of maintenance concept, complicated equipment health management has gone far beyond the scope of fault diagnosis and maintenance. In particular, the implementation of predictive maintenance technology has a significant effect on rationally arranging maintenance cycles, reducing downtime and maintaining high production efficiency, as shown in Fig. 6.

In the future, TBM equipment health management needs to be strengthened in the following aspects (i) vibration, sound wave, temperature and other non-destructive monitoring sensors shall be provided additionally to cover key parameters of key components, (ii) a TBM health data collection, storage and analysis system shall be built, including hand-held point and tour inspection, online monitoring of multi-source data, big database and information platform and (iii) an equipment health condition evaluation system and intelligent fault diagnosis model shall, based on multi-source information fusion, be established to provide scientific decision basis for equipment condition evaluation, predictive maintenance and control parameters optimization, and its technical framework is shown in Fig. 7.

隧道掘进机设备健康管理

由于隧道掘进机(TBM)的故障是隐形的且由其他故障引起的，因此故障诊断和维护并不容易。一项先前的研究显示，当TBM无法正常工作时，70-90%的停机时间用于故障诊断和定位(Huang et al., 2012)。Zhao et al. (2003)和Han(2003)建议通过振动监测和油分析技术对主要TBM轴承和齿轮箱等关键部件进行状态监测和故障诊断。根据TBM每个系统的重要性，Liu (2007)将被监测的组件分为关键、重要和一般三个级别，并建议实施不同的监测策略。Wang和Xu (2003)建立了一个TBM故障数据库，开发了一个专家系统用于故障诊断。上述研究都对评估TBM健康状况和缩短故障诊断时间起到了积极的作用。

随着维护理念的不断发展，复杂的设备健康管理已超出了故障诊断和维护的范畴。特别是预测性维护技术的实施，对合理安排维护周期、减少停机时间和保持高生产效率具有显著的影响，如图6所示。

未来，TBM设备健康管理需要加强以下方面的工作：(i)额外提供振动、声波、温度等无损监测传感器，以覆盖关键部件的重要参数；(ii)建立TBM健康数据的采集、存储和分析系统，包括手持点检和巡视、多源数据的在线监测、大型数据库和信息平台；(iii)基于多源信息融合，建立设备健康状况评估系统和智能故障诊断模型，为设备状况评估、预测性维护和控制参数优化提供科学决策依据，其技术框架如图7所示。

5 Integrated TBM construction information management platform

The beginning of 5 Integrated TBM construction information management platform section.

"5 Integrated TBM construction information management platform" 章节的开篇部分如下:

该章节介绍了5个集成型TBM（隧道掘进机）施工信息管理平台。这些平台旨在通过集成不同的信息系统和软件工具，实现对TBM施工过程中的关键信息进行管理和协调。这些平台具有集中管理施工数据、提高施工效率和质量控制的功能。此外，还介绍了这些平台的架构、功能以及它们在现实中的应用案例。

Application status

Large-scale shield manufacturers and construction contractors develop their own shield construction information management platforms, including IRIS (2018) by ITC Engineering, TPC (2018) by Babenderrede Engineer, KRONOS by GeoDATA (Chmelina et al., 2013), TIM (2018) by London Bridge Associates Ltd, TRTMS by CUMT (Jiang et al., 2007), to provide basic information services such as online monitoring, construction management and data sharing. Nevertheless, relevant literature relating to research, development and application of integrated information management platform for hard rock TBM is rare. Qian et al. (2004) and Ren (2008) proposed functional requirements and framework design for hard TBM information management system, but its practical application has not been retrieved. With the wide-spread application of shield and TBM, each party involved in a project has an increasing demand for information management service platform and such platform has an obvious trend of popularity remarkably.

The value contained in multi-source massive information of TBM is far from being limited to meeting onsite construction management and rock-machine interaction law obtained through in-depth information mining can be widely applied, for instance, to provide basis for design of new TBM and guidance for safe and efficient construction, precisely estimate construction duration and cost and evaluate life of key components for maintenance or remanufacturing. For a long time, TBM industry has attached less importance to each type of data during construction and has not established an agreed standard and method to store such data. Data relating to completed projects is distributed among different participants such as contractor, manufacturer and employer, causing barriers to data sharing (Cho et al., 2013; Shang et al., 2007). As a result, an integrated information management platform for TBM shall not only provide each party involved in a project with basic information service, but also provide an effective means for storing, analyzing and mining multi-source massive information of

Figure 4: Image collection, segmentation and feature extraction of muck excavated by TBM.

Figure 5: Technical roadmap for real-time evaluation of mechanics parameters of rock mass being excavated.

TBM, achieving maximization of data value.

大型盾构制造商和建筑承包商开发了自己的盾构施工信息管理平台，包括ITC Engineering的IRIS（2018年）、Babenderrede Engineer的TPC（2018年）、GeoDATA的KRONOS（Chmelina et al.，2013年）、London Bridge Associates Ltd的TIM（2018年）以及中国矿业大学的TRTMS（Jiang et al.，2007年），以提供在线监测、施工管理和数据共享等基本信息服务。然而，关于硬岩隧道掘进机（TBM）集成信息管理平台的研究、开发和应用的相关文献很少。钱等人（2004年）和任（2008年）提出了硬岩TBM信息管理系统的功能要求和框架设计，但其实际应用尚未找到。随着盾构和TBM的广泛应用，项目中各方对信息管理服务平台的需求越来越大，这种平台有明显的普及趋势。

TBM的多源大量信息所包含的价值远不仅限于满足现场施工管理，通过深度信息挖掘获得的岩石机械相互作用规律可以广泛应用，例如为新TBM的设计提供基础和对安全高效施工提供指导，精确评估施工时间和成本，并评估维护或再制造的关键部件的使用寿命。长期以来，TBM行业对施工过程中的每一类数据重视不足，并未建立统一的标准和存储这些数据的方法。与已完成项目相关的数据分布在承包商、制造商和雇主等不同参与者之间，导致数据共享存在障碍（Cho et al., 2013; Shang et al., 2007）。因此，TBM的集成信息管理平台不仅应为项目中的每一方提供基本的信息服务，还应为存储、分析和挖掘TBM的多源大量信息提供有效手段。

图4：TBM挖掘出的泥土图像采集、分割和特征提取。

图5：TBM挖掘中岩体力学参数实时评估的技术路线图。

TBM的集成信息管理平台实现了数据价值最大化。

Comparison of features between big data and massive TBM information

Big data is not solely dependent on the size of data set and is mainly featured by large volume, various types, rapid generation, high value and low density (He and He, 2014; Li, 2012). In addition, the correlation degree among data and difficulty in data mining are also important factors to judge big data.

TBM construction information types are diverse with complicated relations and their specific features include the following: (i) spatiality, which means that TBM parameters vary with geological changes along the tunnel axis, (ii) real time, which means that TBM parameters are real time feedback of geological information changes and adjustment of control parameters may also lead to real time response of the equipment, (iii) multi-source heterogeneous, which means that structural data such as geological information and equipment parameters differs much from unstructured data including construction video and pictures and (iv) massive, which means that TBM construction information will be more and more with the number increase of TBM tunnel projects.

By comparing Fig. 8, it is found that massive TBM information is characterized by multi-source and multi-dimension and is naturally compatible with big data (Qian, 2017). Therefore, utilization of big data provides an effective means for exploring rock-machine interaction relations.

\begin{table} \begin{tabular}{l l} \hline \hline \multicolumn{1}{l}{Monitored Parameters} & \multicolumn{1}{l}{Force, temperature, rotation speed, vibration, wear} \ \hline Function description & (1) Evaluate cutter working status and equipment operating status; \ & (2) Enhance maintenance timeliness to extend the service life of disc cutter; \ & (3) Set early warning threshold to ensure safe excavation; \ & (4) Describe geological condition ahead of TBM based on monitored data to optimize tunneling parameters. \ \hline \hline \end{tabular} \end{table} Table 1: Main parameters and function description for real-time monitoring of disc cutter status.

Figure 8: Comparison of features between big data and massive TBM information.

Figure 6: Downtime comparison between predictive and corrective maintenance.

Figure 7: TBM equipment health management information system.

大数据不仅仅取决于数据集的大小，主要特点是数据量大、类型多样、生成快速、价值高和密度低(He和He, 2014; Li, 2012)。此外，数据之间的相关程度和数据挖掘的困难程度也是判断大数据的重要因素。

TBM施工信息的类型多样，关系复杂，具体特点包括以下几点：(i) 空间性，即TBM参数随着隧道轴线沿线地质变化而变化，(ii) 实时性，即TBM参数是地质信息变化的实时反馈，调整控制参数也可能导致设备的实时响应，(iii) 多源异构性，即结构化数据如地质信息和设备参数与施工视频和图片等非结构化数据差异较大，以及(iv) 大规模性，即随着TBM隧道项目数量的增加，TBM施工信息将越来越多。

通过比较图8，可以发现大量的TBM信息具有多源和多维特点，与大数据自然兼容(Qian, 2017)。因此，利用大数据提供了一种探索岩石-机器相互作用关系的有效手段。

表1: 实时监测刀盘状态的主要参数和功能描述。

图8: 大数据与大量的TBM信息的特征比较。

图6: 预测性维护和纠正性维护的停机时间比较。

图7: TBM设备健康管理信息系统。

Development of integrated TBM construction information management platform

A single TBM can yearly generate a large quantity of data and conventional data platforms and databases is incapable of storing and calculating such large-scale data. As a result, it is a must to rely on a new working model to deal with massive and diverse information with a high growth rate.

Cloud computing is an emerging information technology based on distributed computation and can allocate virtualized resource in reliance on calculated load resilience to meet user's demand for storage and real-time computation of massive data (Liu, 2015). Cloud computing can be classified into infrastructure service, platform service and software service by service level and its service model is shown in Fig. 9.

This research builds a private cloud computing platform based on OpenStack technology, and realizes big data storage, management, analysis and visual display of TBM construction by means of Hadoop cluster and its subsystems and overall structure of this platform is shown in Fig. 10. Main functions of this platform are to (i) realize standardized collection and management of massive TBM data, (ii) provide each party involved in a project with basic information service such as remote monitoring, data sharing and construction management for TBM and (iii) provide encapsulated algorithm components to conduct data cleaning, processing and mining of massive TBM construction information.

Multi-source data collected and stored by the Integrated TBM Construction Information Management Platform includes project profile, geological and rock mass parameters, TBM operating parameters, construction technology, maintenance record, cutter change record, video, picture and test reports. These multi-source data is stored in MongoDB database after data standardization and correlation, as shown in Fig. 11.

综合TBM施工信息管理平台的开发一台TBM每年能够产生大量数据，传统的数据平台和数据库无法存储和计算如此大规模的数据。因此，必须依赖一种新的工作模式来处理高速增长的海量和多样化的信息。

云计算是一种基于分布式计算的新兴信息技术，它可以根据计算负载的弹性分配虚拟化资源，以满足用户对海量数据存储和实时计算的需求（Liu，2015）。云计算根据服务级别可以分为基础设施服务、平台服务和软件服务，其服务模型如图9所示。

本研究基于OpenStack技术构建了一个私有云计算平台，并通过Hadoop集群及其子系统实现了TBM施工的大数据存储、管理、分析和可视化展示，该平台的总体结构如图10所示。该平台的主要功能包括：（i）实现对海量TBM数据的规范化收集和管理；（ii）为项目相关方提供远程监控、数据共享和施工管理等基础信息服务；（iii）提供封装的算法组件，用于对海量TBM施工信息进行数据清洗、处理和挖掘。

综合TBM施工信息管理平台收集和存储的多源数据包括项目概况、地质和岩石参数、TBM操作参数、施工技术、维护记录、刀具更换记录、视频、图片和测试报告。这些多源数据经过数据标准化和相关性处理后，存储在MongoDB数据库中，如图11所示。

6 Intelligent control of TBM tunnelling process

The beginning of 6 Intelligent control of TBM tunnelling process section.

智能控制技术在TBM隧道掘进过程中的应用日益广泛。这一节将介绍智能控制技术在TBM掘进过程中的关键方面。

Current development of intelligent control of TBM

Although informatization and intelligence level in TBM tunnelling process has been greatly improved compared with that in the early days, it is still limited to automatic coordination of logical relations among various systems of TBM, as shown in Fig. 12. Nakayama et al. (1996) and Okubo and Fukui (2003) put forward conceptual design for automatic TBM tunnelling. Hereknecht AG developed a TBM remote assisting and control system for Mecca TBM project (2015). With the advent of artificial intelligence, research and application of intelligent TBM construction control for different functions is developing at a growing rate. For instance, key parameters are selected by combining model and historical data to set TBM control threshold to trigger safety alarm or emergency stop, and the real-time evaluation of rock mass parameters based on equipment parameters as described in Section 3.2 is used as the basis for decision making and optimization of TBM tunnelling control parameters, and the European Underground Engineering Construction Strategy Research Institution has set 'unmanned TBM tunnelling' as a long-goal for 2030.

尽管与早期相比，TBM隧道掘进过程中的信息化和智能水平已经得到了极大的提高，但仍然局限于TBM各个系统之间的自动协调关系，如图12所示。Nakayama等人（1996年）和Okubo和Fukui（2003年）提出了自动TBM隧道掘进的概念设计。Hereknecht AG为麦加TBM项目开发了一套TBM远程辅助和控制系统（2015年）。随着人工智能的出现，针对不同功能的智能TBM施工控制的研究与应用正在以不断增长的速度发展。例如，通过模型和历史数据的结合选择关键参数，设置TBM控制阈值以触发安全警报或紧急停机，并基于设备参数对岩体参数进行实时评估，作为决策和优化TBM隧道掘进控制参数的依据，欧洲地下工程建设战略研究机构已将“无人TBM隧道掘进”设定为2030年的长期目标。

Intelligent guidance of TBM

Nowadays, a guidance system of TBM has some functions such as tunnel design axis fitting, TBM position detection and corrective trajectory planning, visually present TBM tunnelling trends and deviation. In view of current problems existing in the application of TBM guidance, development of intelligence guidance in the future will focus on two aspects, namely, intelligent TBM posture control and multi-system coordination.

At the present time, TBM position control is adjusted by the operator based on his own experience and operation specification and such semi-automatic adjustment method is largely affected by personal factors. In order to achieve self-adaptive control of TBM posture, Peng et al. (2016) and Liu et al. (2014) established kinematics and dynamics models for the orientation adjustment mechanism to map and correlate TBM posture and displacement of orientation adjustment mechanism. Nakayama et al. (1996) introduced the fuzzy control theory into TBM posture control and established control rules by learning operator inference and decision process and proved system reliability in excavation. Due to lack by fuzzy control of self-learning and self-adaptation ability (Sun et al., 2011), an important aspect of intelligent guidance system in the future is to establish, by combining respective advantages of machine learning and fuzzy control, a posture control model with excellent drivers' experience as learning object to automatically track TBM tunnelling trajectory based on correction rules, as shown in Fig. 13.

TBM posture and tunnelling trajectory is not only affected by conditions of surrounding rock, but also closely related to many factors such as side cutter wear, segment spacing and thrust force (Guo et al., 2011). Manufacturers of guidance system has started to study joint effect of multiple factors on TBM tunnel axis and developed control modules accordingly. In the future, multi-system coordination and control through sensor monitoring and information fusion is the inevitable development trend of intelligent guidance system.

Figure 10: Schematic structure of integrated TBM construction management platform based on cloud computing.

Figure 9: Schematic diagram of cloud computing service level.

如今，TBM的导向系统具有一些功能，如隧道设计轴线拟合、TBM位置检测和纠偏轨迹规划，可直观呈现TBM的掘进趋势和偏差。针对TBM导向应用中存在的问题，未来智能导向的发展将侧重于两个方面，即智能TBM姿态控制和多系统协调。

目前，TBM的位置控制是根据操作员的经验和操作规范进行调整的，这种半自动调整方法在很大程度上受个人因素的影响。为实现TBM姿态的自适应控制，彭等人(2016)和刘等人(2014)建立了动力学和力学模型，将姿态调整机构的位移与TBM姿态进行映射和相关。中山等人(1996)将模糊控制理论引入TBM姿态控制，并通过学习操作员推理和决策过程建立了控制规则，并证明了挖掘的系统可靠性。由于缺乏模糊控制的自学习和自适应能力(孙等人，2011)，未来智能导向系统的一个重要方面是通过结合机器学习和模糊控制的各自优势，建立一个以优秀司机经验为学习对象的姿态控制模型，以根据修正规则自动跟踪TBM掘进轨迹，如图13所示。

TBM的姿态和掘进轨迹不仅受到围岩条件的影响，还与侧刀磨损、分段间距和推力等许多因素密切相关(郭等人，2011)。导向系统制造商已开始研究多个因素对TBM隧道轴线的联合影响，并相应地开发控制模块。未来，通过传感器监控和信息融合进行多系统协调和控制，是智能导向系统的必然发展趋势。

图10：基于云计算的集成TBM施工管理平台的示意结构。

图9：云计算服务层的示意图。

Intelligent control in TBM tunnelling

Intelligent control in TBM tunnelling shall pursue comprehensive optimization of multiple objectives such as safety, high efficiency and cost effectiveness, and the indicators such as rock breakage specific energy, utilization rate, cutterhead wear and truck characteristics have been employed to evaluate performance of TBM tunnelling (Rostami, 1997; Gong, 2005). The precondition to achieve intelligent control is to establish a TBM control parameter decision method that includes multiple evaluation indicators by fully considering incommensurability among various evaluation indicators.

Basic principles of intelligent control in TBM tunnelling are: (i) in adverse formations, disaster warning and risk evaluation is achieved by combining intelligent control system and forward geological prospecting and safe passage of TBM through such formations is ensured under the guidance of disaster prevention and relief experience of expert system and (ii) in normal formations, the intelligent control system adjusts, controls and optimize tunnelling parameter based on rock mass mechanics information and TBM excavation status monitored by it in real time and by incorporating operator's driving behavior model. The foregoing is shown in Fig. 14.

The terminal software for Intelligent TBM tunnelling not only include the above-mentioned intelligent control modules, but also integrate other function modules, such as rock mass perception, early fault and risk warning, intelligent support and guidance. The intelligent terminal software will be loaded into the main control room of TBM to work together with the master computer system and finally achieve unmanned TBM tunnelling with the improvement of automation level of construction procedures such as support and transportation.

TBM隧道掘进智能控制应追求安全、高效和成本效益的综合优化。为评价TBM隧道掘进的性能，常采用岩石破碎比能、利用率、铣头磨损和卡车特性等指标（Rostami, 1997; Gong, 2005）。实现智能控制的前提是建立一种包括多个评价指标的TBM控制参数决策方法，充分考虑各种评价指标之间的不可比性。

TBM隧道掘进智能控制的基本原理之一是，在恶劣地质条件下，通过智能控制系统与前向地质勘探相结合，实现灾害预警和风险评估，并在专家系统的灾害预防和救援经验指导下确保TBM安全通过这些地质条件。另一方面，在正常地质条件下，智能控制系统根据实时监测的岩体力学信息和TBM挖掘状态，并结合操作员的驾驶行为模型，对掘进参数进行调整、控制和优化。上述内容如图14所示。

智能TBM隧道掘进的终端软件不仅包括上述智能控制模块，还整合了其他功能模块，如岩体感知、早期故障和风险预警、智能支护和导向。智能终端软件将加载到TBM的主控室中，与主机计算机系统一起工作，最终实现无人化TBM隧道掘进，并提高支护和运输等施工程序的自动化水平。

7 Conclusions

Information and intelligence technology employed in TBM tunneling can effectively avoid or reduce TBM construction disasters, shorten downtime and improve tunnelling efficiency. The main contribution of

Figure 11: Storage structure of TBM construction information in MongoDB database.

Figure 12: Comparison of intelligence and informatization level of TBM tunnelling control (Babendererde, 2017).

Figure 13: Schematic diagram of intelligent guidance control of TBM.

this paper is to put forward an informatization and intelligence technology application system for TBM and its excavation process and introduces and predicts research and application of key techniques in this system, providing a reference for development of this filed. We firmly believe that informatization and intelligence technology remains one of main aspects of TBM development in the future before the advent of a new generation of rock breaking technology.

Construction of large-scale infrastructures around the world especially in China provide a sustainable and huge market for TBM and Internet of things, artificial intelligence and big data provides a strong support to enable TBM informatized and intelligent. Any person engaged in research and development of TBM shall seize this historic opportunity and continuously enhance scientific innovation to bring a completely intelligent and unmanned TBM.

信息和智能技术在TBM隧道施工中的应用可以有效避免或减少TBM施工事故，缩短停工时间，提高隧道掘进效率。本文的主要贡献是提出了一种TBM及其掘进过程的信息化和智能化技术应用系统，并介绍和预测了该系统中的关键技术的研究和应用，为该领域的发展提供了参考。我们坚信，在新一代破岩技术出现之前，信息化和智能技术仍将是TBM未来发展的主要方面。

全球特别是中国大规模基础设施建设为TBM和物联网、人工智能和大数据提供了一个持续的巨大市场，为实现TBM信息化和智能化提供了有力支持。从事TBM研究开发的任何人都应抓住这一历史机遇，不断增强科学创新，实现完全智能化和无人化的TBM。

Acknowledgements

This research was supported by the National Basic Research Program of China (973 Program, Grant No. 2015CB058103). The author is very grateful for the huge support of all members of this research project.

本研究得到了中国国家重点基础研究发展计划（973计划，批准号：2015CB058103）的支持。作者对本研究项目所有成员的大力支持表示非常感激。

binary-husky / gpt_academic