Research On Rock Fracture Monitoring Instruments Based On Deep Learning

In rock fracture monitoring,tracking and predicting the development direction of rock surface fractures is a way to avoid some geotechnical disasters.A method for tracking and predicting the development direction of cracks on the rock surface is proposed for the rock cracks produced in the uniaxial compression experiment of rock.The details are as follows:1)A method of applying deep learning to the prediction of the development direction of rock fractures in rock fracture monitoring is proposed.In order to verify the effectiveness of the method,we firstly collect the infrared thermal image of rock fractures through uniaxial compression experiments of granite that meet the requirements of rock mechanics experiments.And establish a data set.2)Propose a rock fissure target detection algorithm based on Faster R-CNN(Faster Regions with Convolutional Neural Networks Features),and change the original feature extraction network to Res Net-50(Residual Neural Network),so that it can have a stronger fitting ability.Then,aiming at the problem of a certain deviation in the location of relatively small rock fractures,a structure using ROIAlign(Region of Interest Align)is proposed,so that the fracture characteristics and the space have a more accurate correspondence.In consideration of the large difference in the size of rock fissures,the FPN(Feature Pyramid Networks)is added to optimize the features of the extracted rock fissures,thereby improving the detection effect of different sizes of fissures.Considering that the accuracy of the monitoring area has a certain impact on the subsequent analysis results,so as to improve the positioning accuracy of the detection frame,a Cascade structure is added to improve the accuracy of the detection.The final accuracy rate reached 93.8% and 84.12% respectively.3)Using the proposed rock fracture target detection model combined with the average infrared radiation temperature-time curve,a rock fracture monitoring system for tracking and predicting the development of rock fractures is designed and simulated.The two problems encountering in the system design are analyzed and solutions are proposed.4)Analyze the graph drawn in the simulation to verify the effectiveness of the method.Figure 45;Table 6;Reference 51...