Methodology for quantitative hazard assessment for tunnel collapses based on case histories in Korea

• Published in: International journal of rock mechanics and mining sciences (Oxford, England : 1997) England : 1997), 2009-09, Vol.46 (6), p.1072-1087
• Main Authors: Shin, Hyu-Soung, Kwon, Young-Cheul, Jung, Yong-Soo, Bae, Gyu-Jin, Kim, Young-Geun
• Format: Article
• Language: English
• Subjects: Applied sciences; Artificial neural networks; Buildings. Public works; Exact sciences and technology; Interaction matrix; Sensitivity analysis; Tunnel collapse; Tunnels, galleries
• ISSN: 1365-1609; 1873-4545
• DOI: 10.1016/j.ijrmms.2009.02.009

In this paper we propose the KICT Tunnel Collapse Hazard Index (KTH-Index), an index system for assessing the hazard level of collapse at a tunnel face. To establish the index system, a logical process is proposed for “objectively” assessing the index based on a sensitivity analysis for a well-arranged database containing past collapse cases. For the sensitivity analysis, a neural network based technique is employed which has no restriction, in contrast to conventional sensitivity analysis based on mathematics or statistics. The proposed process is based on the concept of the interaction matrix, a tool for comprehensively expressing the influences among major factors of consideration. Each component of the interaction matrix represents the influential sensitivity between both factors horizontally and vertically linked on the matrix. Since the KTH-Index is assessed based on a database compiled and constructed from collapsed tunnel sites, reliability is compromised for collapse modes and characteristics of sites that are not included in the database. In addition, the characteristics of data available in the database in terms of each major factor must be understood prior to applying the proposed index. Accordingly, this paper includes a basic statistic of the data prepared for the study. The assessment system is applied to the section of the SYK tunnel where large-scale collapses have already occurred, and it is found that the predicted hazard levels are in good agreement with the field data already known.