Article Details

Abstract

To address the problem that blast yield estimation relies on a single magnitude parameter and suffers from limited accuracy, this study integrates multi-dimensional waveform features to improve estimation accuracy. Using 18 blasting events with known yields recorded during the “Deep Crustal Structure Experiment of Fujian and Taiwan Strait” from 2010 to 2012, we first employ orthogonal regression to establish an empirical relationship between the local magnitude ML and the blast yield Y in Fujian. Then, based on a random forest regression model, we introduce data augmentation techniques and fuse multi-dimensional waveform features—including amplitude, spectral, attenuation, and wavelet transform features—to construct a high-precision prediction model for blast yield. The results show that the derived magnitude–yield relationship is ML=0.348 lg Y+1.385. The constructed multi-feature fusion random forest model exhibits higher prediction accuracy on the test set (RMSE = 0.143 in logarithmic yield), and feature importance analysis indicates that duration is the key feature. The proposed model integrates the advantages of empirical relationships and machine learning, providing a more scientific basis for the rapid identification, yield estimation, and vibration effect evaluation of blasting events in the Fujian area. It has important application value for mine safety production and the intelligent monitoring of non-natural earthquakes.

Keywords

ML magnitude; blast yield; orthogonal regression; random forest; multi-feature fusion

References

1. [1] Kong Handong, Wang Tingting A Review of Research on Blasting Equivalent [J]. International Seismological Dynamics, 2018 (2): 9-25
2. [2] Zhang Ke, Ma Liang, Wang Yue, etc Analysis of Estimation Methods for Mine Explosion Equivalent: A Case Study of Heidaigou Open pit Coal Mine in Inner Mongolia [J]. China Earthquake, 2024, 40 (4): 819-830
3. [3] Guo Wei, Liu Fang, Fan Lingling Research on the correspondence between throwing blasting equivalent and earthquake magnitude, as well as waveform characteristics: taking the Heidaigou open-pit coal mine as an example [J]. Shanxi Earthquake, 2024 (4): 11-15, 20
4. [4] Jiang Wenbin, Chen Yong, Peng Fei. Estimation of explosion equivalent at Jiangsu Xiangshui Chemical Plant in March 2019 [J]. Journal of Geophysics, 2020, 63 (2): 541-550
5. [5] Lay T. Estimating Explosion Yield by Analytical Waveform Comparison [J].Geophysical Journal International，1985，82（1）：1-30.
6. [6] Li Yue, Dong Miao, Li Min Research Progress on the Application of Machine Learning in Geothermal Flow [J]. Advances in Geophysics, 2025, 40 (6): 2460-2475
7. [7] Liang Xiaonan, Wang Jianjun, Yin Xinxin Classification of mining seismic and blasting signals based on machine learning [J]. Geodesy and Geodynamics, 2025, 45 (10): 1093-1100
8. [8] Ji Donglin, Sun Suyi, Guo Yubo, etc Research on Optimization of Single Frequency Observation Pseudo Range Accuracy Based on Multi Layer Perceptron Neural Network [J]. Modern Information Technology, 2025, 9 (17): 38-44
9. [9] Liu Ruifeng, Li Zan, Zhang Ling, etc Preliminary Study on Estimation of Explosive Equivalent [J]. Earthquake and Geomagnetic Observation and Research, 2019, 40 (4): 1-7
10. [10] Carroll R J，Ruppert D. The Use and Misuse of Orthogonal Regression in Linear Errors-in-Variables Models [J].The American Statistician，1996，50（1）：1-6.

Funding

China Earthquake Administration Earthquake Monitoring and Early Warning Work Task "Development and Business Application of Non Natural Earthquake Intelligent Recognition System Based on Ensemble Learning Algorithm" (CEA-JCYJ-202602003)

About the authors

Cai Xinghui (1976-), male, Han ethnicity, from Yong'an, Fujian, is a senior engineer with a bachelor's degree. His research interests include earthquake monitoring and intelligent recognition of non natural earthquakes.