Abstract

Shield construction is widely used in the construction of urban subways. The shield condition is a reflection of its construction status, and it has an important impact on construction efficiency and safety risks. However, due to the deep underground construction of shield construction, it is difficult to directly observe its working conditions. Therefore, this study uses the shield parameters as the characterization of the shield condition, and uses visual and data clustering methods to classify it, and studies the influencing factors of the shield condition and the correlation between them. At the same time to provide reference for the actual project.

In this study, with the ring as the research object, the thrust and other 12 shield parameters were selected as the characterization of the working conditions. In order to achieve visual processing, the high-dimensional parameter set is transformed into a complex network by using the visibility method and the Euclidean distance-similarity matrix method. Then, a spectral clustering algorithm based on complex networks is constructed as the main clustering method. And introduced multiple data clustering and complex network community mining algorithms as a control method, and composed of four indicators such as modularity to form a clustering effect evaluation system. Finally, taking the shield construction process of a certain river crossing tunnel as an engineering case, the 751 rings were selected. The clustering algorithm based on complex networks was found to be significantly better than the other four by cluster analysis. The analysis found that the factors that caused the different working conditions were mainly the geological adaptability of the formation and the working status of the machinery, and the risk degree of different working conditions was different.

Through research, we have found that the shield condition is a complex system that is affected by a variety of factors including the formation of the cavern body, mechanical wear, etc., mainly from two aspects of the geological adaptability of the formation and the working state of the machine. Spectral clustering based on complex networks can be well applied to shield condition clustering. At the same time, the shield condition is also a comprehensive reflection of the construction risk. It can infer the current construction risk by observing the shield condition. The relevant results of this research also provide a theoretical basis for the control of working conditions in the actual shield. In addition, this study also lays the foundation for the future research of the shield condition classifier.

Key words: Shield condition； Cluster analysis； Complex network；Spectral clustering；Geological adaptability