Automated crack classification for underground tunnel infrastructure using deep learning
University College Cork
One early sign of tunnel structure deterioration originates in the form of cracking, and therefore crack detection and resultant classification is integral for tunnel structural inspection and maintenance. Conventionally tunnel cracks are manually recorded and classified by trained professionals, which is costly, time-consuming and inevitably subjective. Recent advances in the deep learning space have allowed for automatic cracks detection algorithms to be developed and subsequently utilized in surface structural health assessment of surface buildings, bridges, roads and other civil infrastructure. Nevertheless, these methods of development underperform when implemented for a tunnel structure in an underground environment due to the disparity of illumination combined with the congested image data caused by pipes, steel mesh, wires, and other tunnel amenities. This thesis develops an intuitive crack directional classification approach that increases the accuracy, reduces time and subjectivity in comparison to traditional inspection methods. The detection of cracks by utilising CNN’s is antiquity investigated by in literature however little of the writings develop the algorithm further for classification purposes. The novel of this research is centred on the development of a crack classification algorithm that adheres to the directional classification rationale. The output information of the crack classification is correlated to the structural movement of the lining providing a deeper understanding of the tunnel behaviors. To surmount these challenges, this thesis constructs a Convolution Neural Network (CNN) image-based crack detection method accompanied by an innovative crack classification for underground infrastructure environment. Conventional CNN’s are developed from scratch, the proposed CNN incorporates transfer learning in the form of the VGG16 model with weights transferred from ImageNet. The transfer model was trained under various scenarios to determine the optimal model for the operational task in the tunnel environment. The various models are trained using over 10’000 images validated on 2’500 images all of which are 256 x 256 pixels in size, these models are all subsequently tested using 30 images 3072 x 4096 pixels in size. The transfer learning model used outperforms that of the traditional CNN training method of training from scratch. The optimum transfer model accomplished testing metrics of 96.6%,87.3%,92.4%,89.3% for Accuracy, Precision, Recall and F1 score respectively. The proposed CNN appraises images regarding the existence and subsequent location of cracks. Detected cracks are subjected to the secondary classification CNN where the crack is categorized into one of the four crack classes which include the three directional classes of Horizontal, vertical and diagonal with the last crack classes incorporated to represent complex crack regions. The secondary classification CNN attains an Accuracy of 92.3% a Precision of 83.9% a Recall value of 82.3 % and an F1 score of 81.5%. The performance of the manufactured integrated detection and classification method is analysed by performing a field test to evolve the research from a controlled theoretical setting into a realistic tunnel environment. The field test is performed on three separate tunnel sections with an amassed distance of 150 meters with the section testing the robustness, speed and ultimately prospect of application in the CERN inspection scenario. The outcome from this testing demonstrates that the established CNN crack detector/classifier can effectively overwhelm the unfavourable tunnel environment and accomplish results to a high standard.
European Organization for Nuclear Research , Convolutional neural networks , Transfer learning , Crack detection , Crack classification
O'Brien, D. 2021. Automated crack classification for underground tunnel infrastructure using deep learning. MSc Thesis, University College Cork.