Answers are grounded in this paper and cite their sources.
Detection of gas pipeline leaks is essential for energy security and environmental preservation. Non-contact detection can be accomplished by analyzing acoustic signals produced by gas leakage. Nonetheless, accurate leakage identification continues to pose a considerable difficulty owing to external noise interference and inadequate signal features. An intelligent leakage detection method is proposed by utilizing adaptive Variational Mode Decomposition (VMD) and multi-feature extraction (MFE) to overcome these challenges. Firstly, the Crested Porcupine Optimizer (CPO) is introduced to effectively optimize two critical parameters of Variational Mode Decomposition (VMD): the penalty factor and the number of decomposition modes. A fitness function is formulated using the ratio of mean to variance of permutation entropy, facilitating the decomposition of the original signal into several narrow-band Intrinsic Mode Functions (IMFs). Secondly, effective IMFs are chosen based on correlation coefficients and reconstructed to mitigate noise. Thirdly, features from the frequency domain, time domain, and waveforms of the reconstructed signals are extracted to create a feature matrix. Finally, the feature matrix is input into a Support Vector Machine (SVM) for the detection and classification of gas leakage. Experimental results indicate that the CPO-VMD method surpasses the PSO-VMD and SABO-VMD methods for robustness and convergence. The proposed method achieves a classification accuracy of 98.19% across 12 categories in the GPLA-12 public dataset, and a classification accuracy of 94.20% for 3 categories in the experimental dataset with differing leakage hole diameters. In comparison to other methods, the suggested approach exhibits enhanced performance in signal decomposition and classification accuracy.
No takes yet. Share an insight, caveat, or question.
Lai et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: