Safety State Evaluation of Operational Shield Tunnel Structures by Integrating Computer Vision and Performance Analysis

In the operation and maintenance of subway shield tunnels,deformation of tunnel structures and apparent lining damage are common indicators for evaluating structural safety states.To balance the subjectivity of expert assessments and the objectivity of physical-mechanical models,a tunnel safety state evaluation method was proposed,integrated computer vision-based structural damage detection information with structural performance analysis.The expert assessment of the weight ratios for different categorized damages was considered,and the weighted influence of refined indicators such as location,area,and volume on various types of damage also were incorporated.3D point clouds obtained from mobile laser scanning is used as the data foundation.The convergence deformation value,ellipticity and dislocation value of the shield tunnel section are calculated through ellipse fitting,and a deep learning model is used to automatically identify and quantify obvious lining leakage and spalling damage.The finite element numerical model was used to quantitatively analyze the safety state weights of leakage and spalling damage at different locations.The information entropy method was used to determine the weights of the two types of damage,lateral convergence deformation and ellipticity,and the tunnel structure safety status evaluation formula was obtained.The unsupervised machine learning method Kmeans++ clustering algorithm is used to determine the threshold for safety status classification.Field example verification results show that the evaluation method proposed in this article has advantages in efficiency,objectivity and comprehensiveness,and can provide a reference for maintenance and operation decisions of the shield tunnel department.