Rock burst, which is exceptionally powerful and abrupt, is one of the most dangerous geological hazards in deep mines. Although it is difficult to prevent rock explosions, it can be predicted with continual monitoring. Traditional monitoring systems produce a large quantity of data and false signals. Most methods are affected by blasting and other mining activities. This study used modulated thermal wave imaging to locate the high damage zone in a rock sample containing artificially implanted subsurface microcracks prior to any rock bursting. Finite element (FE) simulations were used to examine the infrared thermal response of a rock sample to a fixed-frequency sinusoidal heat wave. To automate detection, a cutting-edge deep-learning technique was used to identify, localize, and segment cracks. A sizable dataset of images with a resolution of 640×480 was produced for the algorithm training and validation. The F1 score and precision of the applied method were considerably high.
Resource extraction depth is increasing exponentially as a result of technological advancements and increased demand for ores. When a mine is dug deeper, the most common problem is a rock burst (Budiansky et al. 1976), which occurs quickly and eventually results in the collapse of the entire structure. Sometimes it is so devastating that it won’t give the mine workers enough time to escape safely. As a result, research into early rock burst prediction has been ongoing for many years. It is exceedingly challenging to forecast because of the heterogeneity and high confining pressure in the underlying rock. Due to excavation, tensile stress increases in the principal plane of the tunnel, creating microcracks at weak spots (as shown in Figure 1) and softening the entire rock mass (Piane et al.,2015 and Griffiths, L et al., 2017). As these cracks spread, rock bursts occur. Therefore, rock bursts can be projected by continuous monitoring of precursors such as the formation of microcracks or weak spots. Researchers have used various predicting methods based on stresses present on rock (Miao et al., 2016), microcosmic activity (Lu et al., 2012), acoustic emission (He et al., 2010 and He et al., 2019), electromagnetic radiation (Li, Xuelong, et al., 2016) and optics (Luodes et al., 2008).
