Rock lithology recognition plays a fundamental role in geological survey research, mineral resource exploration, mining engineering, etc. However, the objectivity of researchers, rock variable natures, and tedious experimental processes make it difficult to ensure the accurate and effective identification of rock lithology. Additionally, multitype hybrid rock lithology identification is challenging, and few studies on this issue are available. In this paper, a novel multitype hybrid rock lithology detection method was proposed based on convolutional neural network (CNN), and neural network model compression technology was adopted to guarantee the model inference efficiency. Four fundamental single class rock datasets: sandstone, shale, monzogranite, and tuff were collected. At the same time, multitype hybrid rock lithologies datasets were obtained based on data augmentation method. The proposed model was then trained on multitype hybrid rock lithologies datasets. Besides, for comparison purposes, the other three algorithms, were trained and evaluated. Experimental results revealed that our method exhibited the best performance in terms of precision, recall, and efficiency compared with the other three algorithms. Furthermore, the inference time of the proposed model is twice as fast as the other three methods. It only needs 11 milliseconds for single image detection, making it possible to be applied to the industry by transforming the algorithm to an embedded hardware device or Android platform.
As metal mines advance into deep mining, the increase in tectonic stress and horizontal stress leads to a higher degree of joint and fissure development in roadway surrounding rocks, along with a significant rise in both the fragmentation degree of the rock mass and the support cost. This paper adopts field monitoring and numerical simulation methods to analyze the characteristics of mining-induced stress rotation after unloading due to deep roadway excavation in the Jinchuan mining area, and proposes corresponding surrounding rock control countermeasures and optimized schemes for the original support. The research results show that after the unloading caused by the excavation of deep roadway surrounding rock, the magnitudes and directions of the maximum, intermediate, and minimum principal stresses all exhibit a trend of slow change, followed by drastic change, and finally gradual stabilization. When the roadway advances to 4 m in front of the monitor section, the adjustment of the magnitude of principal stress of the surrounding rock is the most drastic. Moreover, as the working face moves away from the monitor section, the principal stress gradually stabilizes and becomes lower than the initial stress value. When the roadway advances to 6 m in front of the monitor section, the adjustment of the direction of the principal stress of the surrounding rock is the most drastic. The rotation angle of the maximum principal stress shows a trend of first increasing and then decreasing with the increase in the excavation step, while the rotation angles of the intermediate and minimum principal stresses show a trend of first decreasing and then increasing as the excavation step increases. Based on the spatial distribution characteristics of joints and fissures in the roadway surrounding rock, the sensitive area for the rotation of mining-induced stress direction is defined. By changing the advancing direction of the roadway, the rotation trajectory of the principal stress can be deviated from the sensitive area, thereby improving the self-stabilization ability of the roadway surrounding rock. It is proposed that asymmetric coupling support be adopted to reinforce the positions where the principal stress rotation of the rock mass around the anchorage is severe, which can effectively reduce the range of the plastic zone in the roadway surrounding rock. The research results provide new ideas for the surrounding rock control of deep roadways, as well as a theoretical basis for the design and optimization of roadway support parameters in similar mines.
Experimental studies were carried out to analyze the impact of optimizers and learning rate on the performance of deep learning-based algorithms for rock thin-section image classification. A total of 2634 rock thin-section images including three rock types—metamorphic, sedimentary, and volcanic rocks—were acquired from an online open-source science data bank. Four CNNs using three different optimizer algorithms (Adam, SGD, RMSprop) under two learning-rate decay schedules (lambda and cosine decay modes) were trained and validated. Then, a systematic comparison was conducted based on the performance of the trained model. Precision, f1-scores, and confusion matrix were adopted as the evaluation indicators. Trials revealed that deep learning-based approaches for rock thin-section image classification were highly effective and stable. Meanwhile, the experimental results showed that the cosine learning-rate decay mode was the better option for learning-rate adjustment during the training process. In addition, the performance of the four neural networks was confirmed and ranked as VGG16, GoogLeNet, MobileNetV2, and ShuffleNetV2. In the last step, the influence of optimization algorithms was evaluated based on VGG16 and GoogLeNet, and the results demonstrated that the capabilities of the model using Adam and RMSprop optimizers were more robust than that of SGD. The experimental study in this paper provides important practical value for training a high-precision rock thin-section image classification model, which can also be transferred to other similar image classification tasks.
The deformation and failure of a rock is closely related to the strain energy consumption during the load process of rock. To investigate the effect of water on energy evolution and damage characteristics of dolomite samples from a deep mine, the uniaxial compression tests were carried out on dry and water-saturated dolomite samples at different burial depths (900 m–1200 m). The effects of water on the evolution characteristics of elastic and dissipative energy ratios ( U e / U and U d / U) during rock deformation and failure was analyzed. Based on the variation rate of damage factor ( D f ), a new brittleness index is proposed, which can effectively characterize the brittleness characteristics of water-bearing dolomite. The results show that the uniaxial compressive strength and elastic modulus of the water-saturated dolomite are significantly reduced compared to dry sample. The energy and damage evolution process of dolomite can be divided into four stages: initial damage stage, stable damage stage, pre-peak accelerated damage stage and post-damage stage. The variation rate of damage factor of the rock samples in the stable damage stage and the pre-peak accelerated damage stage appeared to increase significantly after water saturation treatment. Compared with water-saturated samples, more pronounced energy hardening characteristics and brittleness characteristics were observed in dry samples. In addition, the possible impact on the stability of deep rock engineering after the deterioration of rock mechanical properties and energy storage properties caused by water was analyzed. Groundwater can somewhat reduce rock burst proneness. However, it also has the potential to lead to greater rock engineering destabilization and failure hazards.
Abstract Hard rock failure and rockburst hazards under high in situ stresses have been the subject of deep rock mechanics and engineering. Previous investigations employed cubic rock specimens with a central hole for simulation of rock fracturing around deep tunnels at a laboratory scale, while the failure characteristics and crack evolution behavior around different shapes of holes induced by excavation unloading response have been barely considered. A commercially combined finite‐discrete element method (combined FEM/DEM) was used to investigate the failure characteristics and crack propagation process of typical hard rock specimens (marble) via the unloading of central hole with different shapes. Rock heterogeneity was also considered in the model in combination with the engineering reality. The combined FEM/DEM approach was first validated by simulating uniaxial compression and Brazilian tests. Then, the parametrical analysis was conducted in detail on the basis of five different sectional shapes of central holes, including a circle, ellipse, U‐shape, trapezoid, and cube, and two lateral pressure coefficients. Analysis of crack propagation paths, released strain energy, displacement, and average velocity distribution of the monitoring points around the central hole suggests that the failure degree and destruction intensity are strongly related to the sectional shape and lateral pressure coefficients. Hard and brittle rock failure induced by the excavation unloading effect can be classified as stable failure (slabbing failure) and unstable failure (strain rockburst). The cubic, trapezoidal, and U‐shaped holes within the specimen are the most likely to induce unstable failure, while stable failure is the dominant failure pattern around circular and elliptical holes. The lateral pressure coefficient λ was also found to affect failure position and intensity (only for the axisymmetric section) around the central hole. The influence of rock heterogeneity on failure intensity and range around the central hole within the specimen was also discussed. This study emphasizes the importance and necessity of the excavation unloading effect when evaluating surrounding rock failure around deep tunnels.
