Prediction model of natural gas pipeline crack evolution based on optimized DCNN-LSTM. (1st December 2022)

• Record Type: Journal Article
• Authors: Wang, Bin; Guo, Yanbao; Wang, Deguo; Zhang, Yuansheng; He, Renyang; Chen, Jinzhong
• Is Part Of: Mechanical systems and signal processing. Volume 181(2022)
• Publication Date: 2022-12-01
• Subjects: Acoustic emission -- Time series prediction -- CNN-LSTM -- Attention mechanism -- Original CEEMD algorithm; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2022.109557
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: This paper presents an intelligent crack evolution prediction modal by the analysis of the time series. Aiming at static equipment of natural gas pipeline, an optimized DCNN-LSTM prediction model based on deep learning was proposed. By combining multiple algorithm models, this model makes up for the deficiency of single model and improves the accuracy and reliability of the algorithm. Experimental results show that the proposed method has high prediction accuracy in predicting the crack evolution of natural gas pipeline. Abstract: Pipelines are one of the most important tools for natural gas transportation. To avoid accidents caused by local cracks in the pipeline, it is necessary to develop a model that can predict crack evolution. A time series prediction method for the evolution of natural gas pipeline crack based on acoustic emission signals is proposed in this paper based on the convolutional neural network (CNN), and long-short-term memory (LSTM) models (CNN-LSTM), on attention mechanism (AM), and optimized noise reduction. The model includes three structures. First, based on the original CEEMD algorithm combined with wavelet threshold denoising, the crack evolution signal features are effectively enhanced by improving the denoising threshold function. Then, the crack evolution features are extracted and predicted for the noise-reduced acoustic emission signal through an AM-based bidirectional CNN-LSTM network. Lastly, the final prediction result of the model is output by the fully connected layer. The experimental results show that the method effectively improves the prediction accuracy of crack evolution and achieves end-to-end prediction. Compared with other prediction models, the results obtained in this paper demonstrate that the proposed model is characterized by higher effectiveness and superiority in predicting time series failures. The aforementioned is significant for improving the long-term safe operation of natural gas projects.