Gas pipeline defect identification using improved singular spectrum decomposition and resonance energy concentration

Non-destructive testing of pipeline structures is crucial for ensuring industrial safety. To address the challenges of insufficient separation of multi-modal frequency signals and difficulty in feature extraction under noise interference in acoustic resonance technology, this study establishes a mathematical model of resonance frequencies for pipeline defects based on acoustic resonance mechanisms. Additionally, an improved singular spectrum decomposition (ISSD) method with a frequency-domain concentration characteristic (FDCC) indicator. ISSD adaptively optimises the fixed embedding dimension using a spectral kurtosis-to-spectral entropy ratio factor, effectively suppressing mode mixing. The FDCC indicator quantitatively evaluates resonant energy concentration, enabling precise component selection and accurate defect characterisation. Simulation and experimental results demonstrate that, compared to other methods, ISSD exhibits superior purity and noise robustness, clearly distinguishing between pipe wall and defect resonance frequencies. Using FDCC, two-dimensional defect localisation and contour quantification are achieved, while depth is determined via the resonance-thickness relationship. Tests on four defects show depth errors under 0.2 mm and in-plane errors within 4 mm.This study provides an effective means for the identification and quantitative assessment of pipeline defects.