Localization method for rock acoustic emission sources based on single sensor data and SVR-GBR machine learning models

•Developed a SVR-GBR machine learning localization model for rock fracture sources based on acoustic emission technology.•Proposed a method for rock fracture source localization using a single acoustic emission sensor.•Identified the optimal number and combination of acoustic emission parameters for rock fracture source localization using this method. In response to the challenges posed by traditional acoustic emission localization models, which typically rely on velocity and arrival time parameters and require signals from multiple sensors, this study introduces a novel approach. Leveraging machine learning techniques, a rock fracture source localization model termed SVR-GBR is proposed, utilizing basic parameters from a single acoustic emission sensor. The model’s efficacy is validated through both lead break experiments and real rock fracture localization experiments. Notably, this method circumvents the need to account for velocity, time synchronization, and arrival time differences, thereby simplifying signal processing and analysis. It relies solely on signals received by a single acoustic emission sensor to predict the location of rock fracture sources based on fundamental parameters. Experimental results from lead break and real fracture tests demonstrate the model’s high precision in localization, with stable positioning errors. In lead break experiments, the positioning errors for both the acoustic emission source plane and space were 1.28 mm and 6.9 mm, respectively, markedly lower than those of the employed acoustic emission equipment. Furthermore, in real granite fracture experiments, the predicted acoustic emission localization points closely matched the actual fracture areas of the rock. Additionally, it was found that for plane localization of acoustic emission sources, the optimal parameter combination comprised energy, rise time, and amplitude, while for spatial localization, the most effective combination included energy, absolute energy, count, rise time, and amplitude parameters.