The accurate detection of tunnel lining cracks and prompt identification of their primary causes are critical for maintaining tunnel availability. The advancement of deep learning, particularly in the domain of convolutional neural network (CNN) for image segmentation, has made tunnel lining crack detection more feasible. However, the CNN-based technique for tunnel lining crack detection commonly prioritizes increasing algorithmic complexity to enhance detection accuracy, posing a challenge in balancing the accuracy of detection and the efficiency of the algorithm. Motivated by the superior performance of Unet in image segmentation, this paper proposes a lightweight tunnel lining crack detection model named Mini-Unet, which refined the Unet architecture and utilized depthwise separable convolutions (DSConv) to replace some standard convolution layers. In the optimization of the proposed model parameters, applying a hybrid loss function that integrated dice loss and cross-entropy loss effectively tackled the imbalance between crack and background categories. Several models were set up to contrast with Mini-Unet and the experimental results were analyzed. Mini-Unet achieves a mean intersection over union (MIoU) of 60.76%, a mean precision of 84.18%, and a frame per second (FPS) of 5.635, respectively. Mini-Unet outperforms several mainstream models, enabling rapid detection while maintaining identified accuracy and facilitating the practical application of AI power for real-time tunnel lining crack detection.

Natural aeolian sand has the characteristics of low cohesion and poor water stability. In order to improve its crack resistance properties in the process of freeze-thaw cycles, P.O 42.5 ordinary Portland cement was added to form a mixture called cement improved aeolian sand (CIAS). SEM was used to analyze the microscopic micro-structure of CIAS at different times (7 days and 28 days). The mechanical properties of CIAS samples affected by freeze-thaw cycles were tested in a triaxial instrument, and gray-scale images of the three-phase distribution in the CIAS after freeze-thaw cycling were obtained by computed tomography (CT) scanning technology. The pore characteristic parameters (pore area, fractal dimension, and crack length) were studied by digital image process technique. Based on classical Griffith fracture theory, the development of the crack length and crack width with increasing freeze-thaw cycles is determined. Assuming that the pore area subordinates to the Weibull distribution, the parameters of the Weibull distribution, the damage evolution defined by the elastic modulus attenuation, and the pore area development of CIAS were determined. Research shows the cohesion decreases and internal friction angle increases with increasing cycle numbers. Three development patterns are observed: crack growth, crack closure, and crack merging, and the three patterns interact during freeze-thaw cycling. Furthermore, the fractal dimension of the pore edge fluctuates with the increasing number of freeze-thaw cycles. This work provides a theoretical basis for the application of aeolian sand and develops a method for disaster prevention in applications of freeze-thaw cycling.

The experiment and numerical simulation were used to study the entrainment of path material by the granular flow. The effects of several factors on the entrainment were observed through the experiment. Research results show that when the granular flow passes on the loose material, the entrainment and deposition occur simultaneously. The final mass reduction in the entrainment area is a comprehensive performance of the entrainment and deposition. In general, the mass reduction in the entrainment area is directly proportional to the relative position of the source area and inversely proportional to the length of the entrainment area. The longer the entrainment area, the stronger the retarding effect on the granular flow Within the ability of the entrainment area to support grains, the deposition mass increases as the mass of the material from the source area increases. The numerical results show that the entrainment mainly passes through three main stages. In the first stage, the shear friction and collision between grains form a limited entrainment. In the second stage, the granular flow mainly scales up the entrainment by scraping. In the last stage, the rear part of the granular flow is deposited in the formed pit due to resistance.