Rock Precursor Signal Recognition Based on Machine Learning
Zongcheng Zhang, Jiaxu Jin · Advances in engineering research/Advances in Engineering Research · 2025
Due to the brittleness of hard rock, it is difficult to obtain the failure precursor signal, which endangers the safety of engineering.Therefore, this paper carried out uniaxial compression tests, monitored the compression process with the help of acoustic emission (AE) technology, and analyzed the characteristics of signal changes.In view of the fluctuation complexity of AE signals, it is difficult to effectively identify key information.Machine learning is introduced to identify AE precursor signals.The results show that the mutation point of AE is mainly concentrated in the yield failure stage, close to the peak point.Compared with ELM, RBF, and LSTM models, the Accuracy and AUC of CNN model are 0.90, which shows the excellent performance of the model.This method can provide insights into instability failure in rock engineering.