Residual Swin transformer-based weld crack leakage monitoring of pressure pipeline

In recent years, acoustic emission (AE) has been widely used in pipeline operation safety monitoring and pipeline integrity maintenance. However, the diversity of cracks in the weld joints of pipelines leads to the complexity of time-varying acoustic emission signals, which limits the monitoring accuracy in practical applications. Therefore, a pipeline weld crack leakage monitoring system based on acoustic emission data image coding and deep learning model is proposed in this paper. Specifically, firstly, based on Markov transition field, the leakage signal collected by the AE monitoring system is encoded into two-dimensional image data, and the multi-dimensional phase space trajectory of the signal is revealed while strengthening the correlation and time dependence between the time series sampling points. Then, a residual Swin transformer network model is constructed to obtain useful information from AE coding images and identify different leakage conditions. Finally, experiments with different leakage states are designed to verify the superiority of the proposed method in multiple evaluation indexes, and the recognition accuracy rate reaches 97.86%. The comparison experiment with other methods further proves that the proposed monitoring strategy can be deployed online to maintain the safety of weld pipeline operation.