Analysis of Identification of Seismic Event Properties Based On the permutation entropy of the Variational mode decomposition and support vector machine optimized by pelican optimization algorithm

Discriminating natural earthquakes from artificial explosions is an important aspect of seismic monitoring. In order to effectively extract waveform features and improve discrimination accuracy, a novel method for seismic type identification is proposed, combining Variational Mode Decomposition (VMD), Permutation Entropy (PE), and Pelican Optimization Algorithm (POA) optimized Support Vector Machine (SVM). The event samples are analyzed using VMD, and the optimal number of decomposition layers is determined using the center frequency method combined with Pearson product-moment correlation coefficient. After applying steps such as Wiener filtering, Hilbert transformation, and frequency mixing, the event waveforms are decomposed into 6 Intrinsic Mode Functions (IMFs). IMF components with low Variance Contribution Rate (VCR) are removed, and the remaining IMF components are used to calculate the permutation entropy as the feature set. The POA is used to optimize the penalty coefficient in SVM and the kernel radius in the Gaussian kernel function to construct the new POA-SVM classifier for event discrimination. Experimental results demonstrate that the POA-SVM classifier outperforms SVM, Random Forest, Decision Tree, Naive Bayes, and other classifiers, with an average accuracy of 99.2% in 100 rounds of discrimination experiments, indicating excellent discrimination performance.