To investigate the role of gritstone after hydro-thermal treatment on the mechanical properties and deformation failure behavior, the Brazilian test are carried out on gritstone specimens. Using load-displacement curves, the peak load and peak displacement of the gritstone specimens are analyzed in detail. The mechanical parameters are closely related to the high temperature, high pressure of water and the cooling down methods. The static splitting tensile strength (STS S ) of specimens are decreasing with the increase of water pressure. In the cooling down group, the STS S and peak displacement of specimen WP1 are the lowest one. While, the STS S of specimen W1 is the biggest one. The differences in mechanical properties of the gritstone specimen are mainly caused by the ablation effect in the microscale under different HTHP and cooling down conditions. The surface deformation characteristics of the tested gritstone specimens are investigated by analyzing the full strain field and the local strain concentration. In the gritstone specimen, the major principal strain is first concentrated in the bottom or the top of gritstone specimens where the crack is initiated. The small jump of local strain means crack initiation and propagation, while the fracture of strain gauge leads to strain mutation. The SEM results are discussed to describe the fracture mechanism of brittle gritstone after HTHP treatment.

The acoustic emission (AE) characteristics of rock during loading can reflect the law of crack propagation and evolution in the rock. In order to study the fracture mode in the process of rock fracture, the AE characteristics and crack types of red sandstone during fracture were investigated by conducting Brazilian indirect tensile tests (BITT), direct shear tests (DST), and uniaxial compression tests (UCT). The evolution law of AE event rate, RA and AF values, and the distribution law of RA–AF data of red sandstone samples in three test types were analyzed. Based on the kernel density estimation (KDE) function and the coupling AE parameters (RA–AF values) in DST and BITT, the relatively objective dividing line for classifying tensile and shear cracks was discussed, and the dividing line was applied to the analysis of fracture source evolution and the failure precursor of red sandstone. The results show that the dividing line for classifying tensile and shear cracks of red sandstone is AF = 93RA + 75. Under uniaxial compression loading, the fracture source of red sandstone is primarily shear source in the initial phase of loading and tensile source in the critical failure phase, and the number is far greater than shear source. K = AF/(93RA + 75) can be defined as the AE parameter index, and its coefficient of variation CV (k) can be used as the failure judgment index of red sandstone. When CV (k) < 1, it can be considered that red sandstone enters the instability failure phase.

In order to solve the problem that the slope surface diseases cannot be accurately identified, which cannot be repaired in time and cause serious slope disasters, a slope intelligent recognition technology based on deep neural network is proposed. Based on convolutional neural network (CNN) theory, the technology adopts the transfer learning method to solve the overfitting problem of slope surface samples, which is difficult to obtain a large number of marked samples, and verifies the proposed model by experiment. The results are as follows: the recognition results of various slope surface diseases by ResNet-18 network are higher than AlexNet and VGG-16, with an average accuracy of 84.1%, and the recognition effect of cracks is the best. Under the same migration strategy, the detection accuracy of ResNet-18 is 96.3%, which is much higher than the other two, and the detection time is reduced by 15% on average. It is proved that the ResNet-18 model proposed can identify slope changes very effectively, so that workers can be timely dispatched for maintenance, reducing the possibility of disaster, which has great significance.

No abstract

No abstract

The high redundancy among keyframes is a critical issue for the existing summarizing methods in dealing with user-created videos. To address the critical issue, we present an unsupervised learning method, Spatial Attention Model guided Bi-directional Long Short-term Memory network (Bi-LSTM), on the combination of visual and semantic features. As for the visual feature, we design a Salient-Area-Size-based spatial attention model on the observation that humans tend to focus on sizable and moving objects in videos. Moreover, the Bi-LSTM network is leveraged to exploit the semantic feature. Afterward, the Soft Selected Probability generated from the spatial attention and semantic feature is fused to obtain the final probability for keyframe selection. The reinforcement learning framework, trained by the Deep Deterministic Policy Gradient algorithm, is adopted to do unsupervised training. Extensive experiments on the SumMe and TVSum datasets demonstrate that our method outperforms the state-of-the-art methods in terms of F-score.

No abstract