Acoustic Detection Method for Outer Pipe Leakage in Subsea Pipelines Based on Active Excitation

Abstract Subsea pipelines are predominantly constructed with a double-layer insulated pipe structure. If the outer pipe of such a double-layer pipeline leaks, water ingress into the insulation layer can lead to pipeline condensation, blockage, and eventual failure, compromising flow assurance. Furthermore, direct exposure of the inner pipe wall to seawater can cause corrosion-induced perforation, potentially triggering oil and gas leaks that severely endanger offshore production. This paper proposes an internal inspection method for detecting water ingress into the insulation layer caused by outer pipe leakage, based on active excitation. The echo signals generated by an internal detector that actively knocks on the pipe wall to inject excitation signals are analyzed. The research includes: first, the design of a pipeline leak detection experiment; second, the development of a noise reduction method based on variational modal decomposition (VMD); and third, the combined use of time-frequency domain analysis and Mel spectrograms to accurately characterize the time-frequency and energy differences between acoustic signals. Finally, a random forest (RF) algorithm is used to extract and filter signal features, which are then input into a long short-term memory (LSTM) neural network. Compared with other machine learning models such as extreme gradient boosting (XGBoost), support vector machine (SVM), and convolutional neural network (CNN), the proposed method achieves the highest accuracy of 94.67%. Through theoretical analysis, model construction, and experimental verification, the feasibility and effectiveness of the proposed technical principle are preliminarily confirmed, providing a new research avenue for subsea pipeline outer pipe leakage detection.