Center-Frequency-Based Leak Localization in Long-Distance Pipelines via CPO-Optimized VMD and Automatic Frequency-Scanning Filtering

Accurate leak localization in long-distance pipelines remains a significant challenge due to multi-pulse interference and strong background noise in industrial environments. Addressing the requirements of non-destructive testing, this paper proposes a practical center-frequency-based leak localization strategy that integrates automatic frequency-scanning filtering with an adaptive Variational Mode Decomposition (VMD) framework, optimized via the Centroidal Porcupine Optimization (CPO) algorithm. The frequency-scanning filter automatically sweeps key frequency bands to suppress noise without manual intervention. Subsequently, the CPO-optimized VMD adaptively determines decomposition parameters, effectively avoiding mode mixing and over-decomposition. Combined with feature-driven signal reconstruction, the method significantly enhances the signal-to-noise ratio (SNR) and enables robust extraction of leak-induced waveforms under low-SNR conditions. Validation on both simulated and measured datasets demonstrates that the proposed CPO-VMD approach reduced the average relative localization error from 9.2% to 2.4% on a 209 m pipeline. Furthermore, the proposed method consistently outperforms benchmark methods, such as CEEMD and VMD-SVD, across different pipeline lengths, demonstrating superior accuracy, noise immunity, and engineering applicability for leak detection in real-world pipeline systems.