Original scientific paper

https://doi.org/10.17559/TV-20240831001958

Innovative Vibration Signal Recognition for Enhanced Pipeline Safety Using Optimized Wavelet Denoising and Particle Swarm-Tuned Neural Networks

Fang Wang ; Fujian Boiler and Pressure Vessel Inspection and Research Institute NO. 370 Lubin Road, Cangshan District, Fuzhou 350008, Fujian Province, China *
Caijun Xu ; Fujian Boiler and Pressure Vessel Inspection and Research Institute NO. 370 Lubin Road, Cangshan District, Fuzhou 350008, Fujian Province, China
Bin Liu ; Shijiazhuang Tiedao University, Shijiazhuang, Hebei 050043, China No.17, North Second Ring East Road Shijiazhuang 050043, Hebei Province, China
Wenjun Chen ; Shijiazhuang Tiedao University, Shijiazhuang, Hebei 050043, China No.17, North Second Ring East Road Shijiazhuang 050043, Hebei Province, China
Shaodong Yan ; Shijiazhuang Tiedao University, Shijiazhuang, Hebei 050043, China No.17, North Second Ring East Road Shijiazhuang 050043, Hebei Province, China

* Corresponding author.

Abstract

The precision of vibration signal recognition is critical in the safety monitoring of long-distance pipelines. This paper introduces an advanced methodology for signal classification that integrates wavelet denoising with particle swarm optimization (PSO), to enhance the accuracy of signal recognition and restoration. A novel wavelet parameter selection technique is proposed, which significantly refines the denoising process. Additionally, PSO is employed to fine-tune the initial weights and thresholds of a BP neural network, serving as the classifier in this study. The input data, collected from a long-distance pipeline simulation system under various vibration scenarios, resulted in a dataset of xx samples. Each signal is processed using wavelet denoising and MATLAB computation, resulting in 22 feature values used as inputs for the classifier. The findings demonstrate that the optimized BP neural network achieves a recognition accuracy of 97.5%, with a performance improvement of 13.53%. This methodology aligns well with the future direction of intelligent pipeline systems, providing substantial support for enhancing pipeline safety monitoring. The practical industrial significance of this research is its potential to significantly enhance the accuracy of vibration signal processing in long-distance pipeline transportation systems. With the implementation of optimized BP neural networks, pipeline operators can more accurately detect and identify potential safety hazards, such as cracks, enabling earlier warning and intervention. This improvement in detection accuracy is crucial for maintaining the structural integrity of pipelines, minimizing downtime, and preventing environmental contamination, thus offering a powerful solution for enhancing operational safety and reliability in pipeline systems.

Keywords

BP neural network; feature extraction; Particle swarm optimization algorithm; Wavelet denoising

Hrčak ID:

335062

URI

https://hrcak.srce.hr/335062

Publication date:

30.8.2025.

Visits: 265 *