Condition Identification of Buried Drainage Pipeline Based on CEEMDAN-DE and Time-Frequency Images Recognition

Abstract

During the fault diagnosis of buried drainage pipeline based on active acoustic detection, regarding the difficulties in distinguishing the pipe components and different degrees of blocking. This paper proposes an intelligent pipeline fault diagnosis method based on smooth pseudo Wigner-Ville distribution (SPWVD) time frequency image features. Firstly, a noise reduction method based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) combined with Dispersion Entropy (DE) was applied to acoustical signals which were collected from the pipe. Then the SPWVD time-frequency analysis method was adopted to proceed further signal transformation and the time-frequency distribution images of pipeline under 6 working conditions were obtained. The segments of blockages and pipe components were derived from the time-frequency images through Otsu threshold segmentation method according to the time frequency energy arrogation points. Finally, the histogram and the gray-level co-occurrence matrix of the segmented grayscale image were calculated to form a 20-dimensional feature vector, A parameter-optimized support vector machine classifier was trained and tested by the feature sets. The experimental results have shown that the multi-dimensional features extracted from the time-frequency images can fully interpret the characteristics of blockages and pipe components, therefore are capable of providing sufficient information for effective identification of buried pipe conditions.