Face stability analysis of mechanized shield tunneling: An objective systems approach to the problem

• Published in: Engineering geology 2019-11, Vol.262, p.105307, Article 105307
• Main Authors: Zare Naghadehi, Masoud, Thewes, Markus, Alimardani Lavasan, Arash
• Format: Article
• Language: English
• Subjects: Artificial neural networks (ANN); Deep learning; Machine learning; Mechanized tunneling; Objective systems approach; Tunnel face stability
• ISSN: 0013-7952; 1872-6917
• DOI: 10.1016/j.enggeo.2019.105307

•An objective system of TBM face stability is developed and quantified.•A new face vulnerability index is introduced for tunnel face stability analysis.•ANN with a mirror configuration is utilized for interactions coding.•Field applicability of the new index is approved by actual measured values.•Design charts are presented to estimate required face-support pressure in the future. The stability of the tunnel face is one of the most critical issues having to be secured for a successful tunneling practice. It becomes more crucial for the tunneling in an urban environment and even more when large diameters are contemplated, where catastrophic and costly consequences can happen due to the excessive settlements and ground deformations. In this research, an objective systems methodology is incorporated into this problem for the first time, and through the application of machine learning, a Face Vulnerability Index (FVI) is presented to assess the stability conditions of tunnels. To this end, seven parameters that are important for tunnel face stability in the subsoil – including engineering geological, geotechnical and environmental factors – are employed for the FVI definition, and a comprehensive worldwide database of mechanized tunneling case histories is developed. The interaction matrix in the framework of the systems approach is then objectively coded by using the database and a deep learning technique (deep Artificial Neural Networks- ANN) capabilities. The results (FVI predictions) are compared with a number of well-known analytical methods and the actual applied face-support pressures. A good agreement between predictions and observations has been found that proves the field applicability of the new index to a great extent, which has led to the suggestion of design graphs for future applications.