The classification of seismic events is crucial for monitoring underground nuclear explosions and regional unnatural seismic events. To classify tectonic earthquakes, explosions, and mining-induced earthquakes, we established 36- and 60-dimensional network-averaged datasets and single-station datasets through feature extraction and spectral amplitude analysis. Using different artificial intelligence (AI) methods, including the support vector machine (SVM), extreme gradient boosting (XGBoost), long short-term memory network (LSTM), residual neural network, and long short-term memory fully convolutional network (LSTM-FCN), we constructed two-class and three-class models, analyzed the change in the classification with epicentral distances, and evaluated the generalizability of different classifiers. The results showed that the accuracies of different AI models with the feature extraction dataset were higher than those achieved with the spectral amplitude dataset, indicating that the feature extraction method can more clearly highlight the differences between different types of seismic events. The accuracies with the network-averaged dataset were 5%–8% higher than that achieved using the single-station dataset. The earthquake and mining-induced earthquake classifiers constructed by different AI methods had the best performance, followed by the earthquake and explosion classifier, and the explosion and mining-induced earthquake classifier, with average accuracies of 97.4%–98.4%, 96.5%–97.6%, and 88.8%–90.6%, respectively. In the model generalization evaluation, the test accuracies and F1-Scores of the two-class models with the 36-dimensional network-averaged dataset exceeded 90%. Among the five AI methods, XGBoost and LSTM both performed well in classification of different datasets, indicating that these models have good application prospects for seismic event classifications.

In response to the gaps in understanding the causal relationship between seismic waveform features and the types of seismic events, this research is focused on seismic events of low magnitude (ML≤3.0) in the North China region. Using the Bayesian network theory, we conduct an analysis to infer event types for natural earthquakes, artificial explosions, and mining collapses, and the outcomes achieved notable efficacy for the discrimination of seismic events. Through the analysis of seismic waveforms from 1818 events, we systematically extracted and quantified 55 features in temporal, spectral, and energy domains, which were then recoded as node variables for subsequent analysis. The new data set was subject to select nodes with strong associations to the node type. Subsequently, Bayesian network topologies were constructed using three different algorithms to reconstruct the custom network, calculating posterior probabilities and marginal probabilities. Simultaneously, an extensive evaluation with precision–recall curves of the network structure was carried out, encompassing accuracy, precision, recall, and F1-score. Ultimately, sensitivity analysis was performed on each node to reveal the extent of the influence of node variations on the inference of the node type. The findings showed that the sensitivity of discrimination of seismic events was notably high for several features, including high-frequency P/S spectral ratio values (11 to ∼20 Hz), central frequency, dominant frequency, average frequency, rise and decay average frequency, the real part of the complex cepstral coefficients, peak ground acceleration, and zero crossing. In the classification of natural earthquakes, artificial explosions, and mining collapses, it was observed that the probability of mining collapses was maximized when peak ground acceleration was less than 1526.08, and concurrently, the P/S spectral ratio (11 to ∼20 Hz) fell within the range of −0.25 to −0.02.

To further understand the rock damage zone, an approach based on microcrack distribution was proposed to characterize the crack space of rock specimens in this research. Acoustic emission (AE) technology was utilized on sandstone to obtain the spatial distribution of microcracks in which uniaxial compression forms the single-cracked fracture. The proposed theoretical distribution pattern space (TDPS), 3D convex hull, and the minimum volume enclosing ellipsoid (MVEE) algorithms were adopted to analyze the geometric features of the crack space. It was found that the 3D convex hull method returned the smallest results in both area and volume of the crack space, and the largest results were provided by the proposed TDPS method. The difference between the results of the proposed TDPS method and the MVEE method became smaller after 85%. The deviation angle of the principal axis of the cracked space gradually decreased as the spatial scale decreased, while the other two major axes exhibited a tendency to increase at the 65% scale. The results indicate that a spatial scale from 65% to 85% is a reliable range for the characterization of crack space.

Following North Korea’s sixth nuclear test in 2017, a series of small seismic events occurred around the Punggye-ri nuclear test site. Using continuous waveform data from 2016 to 2024, we employed the match-and-locate method and detected 647 new seismic events around the test site. These detected events all exhibit similar waveform characteristics, which are significantly different from the waveforms of the nuclear test and the collapse event. Statistical results indicate that seismic activity increased significantly near the nuclear test site after the sixth nuclear test on 3 September 2017. The magnitudes of these seismic events primarily range from 1.5 to 2.5. In September 2021 and February 2022, the frequency of such events exceeded 30 occurrences per month. Using the double-difference earthquake relocation algorithm (hypoDD) and the bootstrap method, we obtained precise locations for 491 seismic events and analyzed their location uncertainty. The relocated seismic events exhibit a north-northwest–south-southeast distribution trend. Most events are concentrated in the north and northwest of the test site, 4–9 km away. From 2017 to 2024, the seismic events show a trend of migrating southeastward toward the test site.