Research and Advances in the Characteristics of Blast-Induced Vibration Frequencies

Engineering blasting technology is widely applied in mining operations and the construction of buildings and structures, but ground vibrations caused by explosions are a major environmental concern. Peak particle velocity (PPV), frequency, and duration are the primary parameters for blast-induced vibrations, with PPV often used to assess the safety of vibrations. However, frequency attenuation or resonance can lead to building collapse and equipment damage, and relying solely on PPV-based safety standards is insufficient to fully evaluate the safety of blast-induced vibrations. Studying dominant frequency characteristics is crucial for protecting residents, buildings, and equipment. This paper systematically discusses the safety criteria for blast-induced vibrations in various countries, summarizes the classification of dominant frequencies, and points out that there is no clear scope of application for different dominant frequencies, with varying criteria across countries and the absence of a unified standard. Furthermore, the paper analyzes the influence of blasting parameters, explosive types, and geological conditions on dominant frequency, emphasizing the lack of research on other factors such as borehole parameters, free faces, burden, charge structure, and delay time. It also proposes the need for further exploration of factors such as charge coefficients, decoupling coefficients, borehole density coefficients, and specific explosive consumption. For dominant frequency prediction, the machine learning (ML) models proposed in this study have performed excellently in multiple experiments, especially on large-scale datasets. The experimental results show that the correlation coefficients between the predicted values of the ANN and ANFIS models and the measured data are 0.95 and 0.9988, respectively, indicating high prediction accuracy. In addition, the SVM model, when predicting the dominant frequency, generally keeps the relative error within 10%, demonstrating its efficiency and accuracy in predictions. These methods fully validate the prediction capability of the proposed models, highlighting the significant advantages of ML methods in this study and providing strong support for applications in related fields. Although ML methods can significantly improve prediction accuracy, issues such as insufficient sample size and poor generalization ability may lead to reduced prediction accuracy.