典型盾构刀具离散元模拟与磨损预测研究

易早月

摘要

随着近年我国城市化进程的逐步加快，城市地上交通拥挤成为常态，开发地下空间是解决问题的有效手段之一。盾构法由于其自身安全、高效以及施工时不影响地面活动的优点，被广泛应用于各大城市地下空间的开发中，成为地下空间开发主要的工法之一。然而，盾构掘进时，刀盘刀具会出现磨损。特别在复合地层中，磨损现象更为显著。在面对建设水下或复合地层隧道等更复杂工况时，每次磨损换刀都承担极高风险。因此，对盾构刀具的磨损情况进行准确的预测，以确定合适的换刀时机十分必要。

本文依据不同城市两个盾构隧道工程案例的地质条件与盾构换刀数据，对典型盾构刀具的磨损情况进行分析。采用离散元建模方法，研究不同刀具类型在不同地层中掘进时的磨损变化规律，而后基于磨损机理和人工智能算法对刀具的磨损情况进行预测。完成的主要研究工作如下：

（1）在某市某盾构隧道工程背景下，运用SolidWorks和EDEM建模与分析软件，构建了盾构单刃/双刃滚刀和刮刀在软、硬、上软下硬以及上硬下软地层中服役的离散元模型，研究了同一刀具在不同地层中与三种刀具在同一地层中切削岩土体的磨损规律，为盾构刀具选型提供可供参考的量化依据。

（2）基于地质条件、刀具材质以及施工参数，结合刀具的破岩力学模型与破岩轨迹公式，推导了基于磨粒磨损的复合地层大直径盾构刀具的磨损解析计算公式，对刀具在单一岩层和复合地层中的磨损量与速率进行了预测，并与工程中的实际磨损量进行对比，验证了该公式用于刀具磨损预测的可行性与准确性。

（3）由于前文刀具磨损解析计算公式主要用来计算刀具破岩时的磨损情况，不适用于复合砂卵石地层，故本文在另一城市某盾构工程案例的基础上，探究了复合地层中磨损系数的计算方法，分别采用人工智能AdaBoost增强学习算法和BP神经网络算法对砂卵石地层大直径盾构刀具的磨损情况进行预测，同样也与工程中的实际磨损量进行了对比，验证了AdaBoost增强学习算法应用于刀具磨损预测中的可行性，发现其预测精度高于BP神经网络算法，而后对比算法预测中数据集不同划分比例对预测结果的影响，得出样本训练-预测验证的划分比例为7:3时，模型预测精度最高。考虑AdaBoost增强学习算法本身的不确定性，对其进行分析得知该算法较为稳定。

本文对典型盾构刀具切削岩土体的服役工况进行了离散元建模，量化分析了多型刀具在不同地层条件下的磨损规律，从磨损机理的角度推导了刀具的磨损计算公式；利用AdaBoost增强学习算法和BP神经网络算法建立刀具磨损预测模型，在案例工程中结合实际的刀具磨损检测与换刀记录，进行了模型有效性的应用验证，为在不同地层中盾构刀具选型以及换刀决策研究，提供了量化依据。

关键词：盾构刀具；离散元；磨损规律；人工智能；磨损预测

Abstract

With the acceleration of urbanization in China in recent years, overground traffic congestion has become a normal situation. Developing underground space is one of the effective means to solve the problem. Shield tunneling is widely used in underground space development in major cities because of its advantages of safety, high efficiency and no influence on ground activities during construction. It has become one of the main construction methods for underground space development. However, when shield tunneling, the cutter head will wear. Especially in the composite formation, the wear phenomenon is more significant. In the face of more complex working conditions such as underwater or composite stratum tunnel construction, every wear and tool change will bear a high risk. Therefore, it is necessary to accurately predict the wear condition of shield cutting tool to determine the appropriate tool changing time.

According to the geological conditions and shield cutter changing data of two shield tunnel engineering cases in different cities, the wear condition of typical shield tool was analyzed in this paper. The discrete element modeling method was adopted to study the wear change rule of different tool types in different strata, then, the wear condition of the tool has been predicted based on the wear mechanism and artificial intelligence algorithm. The main research work completed is as follows:

(1) In the context of a shield tunnel project, using SolidWorks and EDEM modeling and analysis software, the discrete element model of shield single-edge/double-edge hob and scraper in soft, hard, upper soft and lower hard and upper hard and lower soft strata was constructed, and the wear law of the cutter in different strata and three kinds of cutters in the same stratum has been studied, which provides quantitative basis for the selection of shield cutter.

(2) Based on the geological conditions, the cutter material and construction parameters, combined with the rock breaking mechanical model and rock breaking trajectory formula of the tool, the analytic calculation formula of large diameter shield tool wear in composite stratum based on abrasive wear has been derived. The prediction of tool wear quantity and rate in single rock and composite stratum has been carried out, and the comparison with the actual wear quantity in engineering verified the feasibility and accuracy of the formula used in tool wear prediction.

(3) Since the analytical calculation formula of cutter wear mentioned above is mainly used to calculate the wear condition of cutter when breaking rock, and is not applicable to composite sand and pebble stratum. Therefore, based on a shield construction case in another city, the calculation method of wear coefficient in composite stratum was explored, Artificial intelligence AdaBoost enhanced learning and BP neural network algorithm were respectively used to predict the wear condition of large diameter shield cutter in sand and pebble stratum, and the predicted results were also compared with the actual wear quantity in engineering, which verifies the feasibility of AdaBoost enhanced learning algorithm in tool wear prediction, found that the prediction accuracy was higher than the BP neural network algorithm, then the influence of different partition ratio of data set on the prediction result was compared, and it was concluded that when the partition ratio of sample training-prediction verification was 7:3, the prediction accuracy of the model was the highest, considering the uncertainty of AdaBoost enhanced learning algorithm, the analysis showed that the algorithm was relatively stable.

In this paper, discrete element modeling was carried out for the service condition of typical shield cutters cutting rock and soil mass, and the wear law of multi-type cutters under different stratum conditions was quantitatively analyzed, the calculation formula of cutter wear was derived from the perspective of wear mechanism, AdaBoost enhanced learning algorithm and BP neural network algorithm were used to establish a cutter wear prediction model, and the validity of the model was verified by combining the actual cutter wear detection and tool change records in a case project, which provides a quantitative basis for the selection of shield cutting tools and the decision to change tools in different formations.

Key Words：Shield cutter; Discrete element; Wear rules; Artificial intelligence; Wear prediction