The resin anchored rock bolt is used extensively in the mining industry to stabilize the roof and prevent it from collapsing. However, there are different defects associated with a resin anchored rock bolt. Examples are partially encapsulated bolts, over-spinned bolts and corroded bolts. These defects reduce the integrity of the roof, and thereby have an effect on the safety and productivity of the mines. The integrity of the rock bolts is a critical issue for the mining industry because of its influence on the safety of mining operations. Different research groups around the world have addressed the problem of determining rock bolt integrity. The most promising technique found in the literature study was based on guided ultrasonic waves (Beard and Lowe, 2003). This study extended the previous work by Beard and Lowe (2003) using guided ultrasonic waves, to investigate damage in more realistic embedded bolts which deviate from pure cylinders. The fundamental L(0,1) mode in its lower frequency range, as suggested by Beard and Lowe was utilized. This was done through the use of finite element model simulations of various defect scenarios, which were compared to experimental measurements on bolts. Defects like loss of resin encapsulation, voids and local corrosion cracks were addressed. The time traces of the different finite element defect scenarios could be directly compared to experimental time traces which distinguish this study from the analytical approach. Some finite element modelling issues were investigated and it was found that the time step is critical if an implicit solver is used, whereas for an explicit solver the element size is critical if accurate answers are needed. Furthermore it was also apparent that the boundary of the mortar has an influence on the results. The method used in the study was to move the boundaries far enough to prevent interference. This however increases the model size and thereby the computer resources required. Axisymmetric defects were modelled using axisymmetric finite elements to reduce the problem size. These models gave results comparable to the measured bolts. Three-dimensional finite element models seemed to be promising for simulating non-axisymmetric defects. It was found that it is not possible to solve large three-dimensional models without energy absorbing boundaries. Axisymmetrical and three dimensional finite element models of a partially encapsulated bolt and a bolt with a local corrosion crack were built. It was possible to detect simulated local corrosion cracks with the finite element models. Clear reflections for the crack in the bolt could be seen. If the bolt, resin and rock are cracked, different reflections will be detected. These different reflections complicated the interpretation of the results. Once the integrity of models such as these has been established, the models could in principle be used to train neural networks for use in commercial equipment. The present study was limited to lower frequencies because of computer resource limitations. Basic principles and modelling issues could however be addressed and it may be expected that these principles could soon be extended to higher frequencies with a new generation of computers.
Dissertation (MEng)--University of Pretoria, 2009.