| The natural rock mass itself contains many primary micro-fractures,which interact with each other to form uneven and irregular fracture network groups,and the fracture network groups within the rock mass will interact with each other to cause instability damage and rock collapse under the external action of rock mass engineering such as tunnel boring and foundation excavation.Therefore,it is important to monitor and identify the damage process of fractured rock masses to effectively avoid engineering disasters.This paper carries out the preparation of rock-like specimens containing 20 random prefabricated fissures,and analyses the crack expansion pattern during the damage process of complex fissured rock-like specimens through uniaxial compression tests and digital image correlation techniques,and finally proposes a new method for the rapid and accurate identification of cracks in fissured rock masses,using YOLOv5 target detection The algorithm uses the YOLOv5 target detection algorithm to obtain the strain evolution cloud map during the fractured rock damage process for dynamic fracture detection and identification,and the results of its dynamic fracture target detection verify the effectiveness of this method.The main research results are as follows.(1)Uniaxial compression tests were carried out on complex fracture-like rock specimens with fracture network models prepared by 3D printing technology.The analysis and research revealed that the damage occurring on the specimen surface was accompanied by crack sprouting and expansion at the tip of the prefabricated primary fracture,causing the fractures within the specimen to cross-lap each other to form a larger fracture group,which in turn affected the overall strength and mechanical properties of the complex fracture-like rock specimens.The expansion of each fracture on the surface of the specimen can be assessed semi-quantitatively to determine the damage process of the specimen.(2)The DIC technique is used to process the deformation characteristics of the specimen surface during the loading damage process and obtain a strain evolution cloud.The strain cloud can obviously characterise the damage state of the specimen by combining the macroscopic damage characteristics of the specimen surface,and a deep red area of strain concentration will appear around each primary fracture when crack initiation occurs,so the area of strain concentration can be used as an early warning of crack initiation and expansion of the specimen,and an early indication of the overall damage of complex fractured rock specimens.(3)ABAQUS finite element software was used to numerically simulate the uniaxial compression test of the fractured rock mass,and the global insertion of cohesive units was used to simulate the crack initiation,expansion and penetration of the complex fractured rock mass specimen.The numerical model simulation results show that the fracture tip and local crack emergence and extension during the model loading damage are similar to the macroscopic test crack damage,further verifying the correctness and effectiveness of the test and numerical correlation techniques.(4)The strain evolution cloud map of the specimen surface during the damage process processed by the DCI technique was used as a data set and input to the established YOLOv5 algorithm target detection model for dynamic fracture detection and identification of complex fractured rock specimens.The testing results show that the established dynamic fracture intelligent detection model is fast and accurate.The model can intelligently and accurately identify dynamic fractures during the damage process,and can intelligently determine the status of each fracture based on the fracture type,and can also intelligently count the number of dynamic fractures.This method can provide new ideas and new methods for intelligent monitoring and damage warning of fractured rock damage in actual engineering.Figure 44 Table 5 Reference 82... |
PDF Full Text Request