Intelligent identification of acoustic emission Kaiser effect points and its application in efficiently acquiring in-situ stress

Large-scale underground projects require precise in-situ stress information, and the acoustic emission (AE) Kaiser effect method currently provide lower costs and streamlined procedures. In this method, the accuracy and speed of Kaiser point identification are crucial. Thus, the integration of chaos theory and machine learning for the precise and rapid identification of Kaiser points constitutes the objective of the study. The intelligent model of the AE partitioned areas identification was established by phase space reconstruction (PSR), genetic algorithm (GA), and support vector machine (SVM). Then, the plots of model classification results were made to identify Kaiser points. We refer to this method of identifying Kaiser points as “The Partitioning Plot Method based on PSR-GA-SVM” (PPPGS). The PSR-GA-SVM model demonstrated outstanding performance, achieving a 94.37% accuracy rate on the test set, with other evaluation metrics also indicating exceptional performance. The PPPGS identified Kaiser points similar to the tangent-intersection method, with greater accuracy. Furthermore, in the classification model's feature importance score, the fractal dimension extracted by PSR ranked second after accumulated AE counts, confirming its importance and reliability as a classification feature. To validate practicability, the PPPGS were applied to in-situ stress measurement at a phosphate mine in Guizhou Weng'an, China, demonstrating good performance.