Search

To investigate the mechanical properties and fracture mechanism of rocks containing multiple flaws after freeze–thaw (F-T) treatment, sandstone specimens with multiple flaws were prepared and subjected to F-T treatment and uniaxial compression tests. The global strain field on the specimen surface during the loading process was calculated using the digital image correlation (DIC) method, and four grayscale parameters were introduced to quantitatively characterize the evolution of the strain field. Then, the differentiation rates (DRs) of these parameters were defined, and the correlation between these parameters and the rock fracture state was analyzed. The F-T cycling deteriorated the mechanical properties and increased the strain concentration of the specimens. The calculation results of the horizontal strains at the measurement points along the crack propagation path showed that the crack initiation of the specimens after F-T treatment occurred earlier, and their brittle failure characteristics were less pronounced. The specimen fracture process was successfully described by combining DIC and statistical grayscale parameters. The evolution pattern of each grayscale parameter of the specimens was generally similar before and after F-T treatment in that it exhibited disorderly fluctuation first, then steady development, and finally accelerated change. The grayscale parameters changed abruptly at the initiation of wing cracks and antiwing cracks, resulting in sharp points in the DR vs. strain curves. These points can be defined as precursor points P1 and P2 for identifying the deformation and fracture state of the rock. The F-T cycling led to the delayed occurrence of P1 and the advanced occurrence of P2. [ABSTRACT FROM AUTHOR]

Displacement prediction is critical for the early detection of landslides, and the empirical, statistical, and machine learning models have been commonly used. In the Three Gorges reservoir area (TGRA), many landslides experience step-like deformations due to the periodic change of influencing factors. In this study, a novel and dynamic model is proposed to predict the displacements of step-like landslides. Two typical landslides in the TGRA are taken as case studies. Variational mode decomposition (VMD) is used to decompose the cumulative displacements into stochastic, periodic, and trend components. The influencing factors are decomposed into low-frequency and high-frequency components. Two principles, including the physical connotation and minimum sample entropy, are employed to optimize the VMD parameters. The trend displacement is fitted and predicted by a polynomial expression with an optimized order, and the periodic and stochastic displacements are dynamically modeled by the bidirectional long short-term memory (LSTM) model. The cumulative displacement prediction is the addition of the three displacement components. The proposed model has been shown to exhibit superior performance in the displacement prediction of step-like landslides. To achieve acceptable prediction, a size ratio between the training and testing datasets greater than or equal to five is recommended. The min–max and zero-mean normalizations are applicable to the data preprocessing of this work. [ABSTRACT FROM AUTHOR]

Physical modeling of rock masses containing complex fracture geometries is an important but very challenging task in rock mechanics and rock engineering. In this work, a novel integration of 3D printing and water-soluble casting is presented for the preparation of rock-model specimens with fracture networks, which is similar to the lost-wax casting process. The fracture network is created with water-soluble polyvinyl alcohol (PVA) material by 3D printing. The intact rock matrix is prepared with cement paste. Uniaxial compressive tests combined with the digital image correlation (DIC) technique are conducted to investigate the mechanical properties and failure process of the prepared rock-model specimens. The differentiation rate of the effective variance is defined to recognize the precursory anomalies during the loading process. The mechanical anisotropy is further investigated by a series of rotated models. The results show that the proposed methodology has distinct advantages for replicating rocks with complex fracture geometries in terms of high accuracy for fabricating fracture networks, high repeatability of mechanical properties, and high strength and brittleness of rock matrices. Due to the presence of complex fracture geometries, the nonlinear mechanical responses of the fractured rock-model specimens, including strain-softening and anisotropy, are observed in our experiments. [ABSTRACT FROM AUTHOR]