Abstract:With the increasing safety requirements for industrial equipment, the monitoring and anomaly detection of pressure vessels have become increasingly important; to address this issue, a deep learning method based on convolutional neural network (CNN) and long short-term memory (LSTM) was used to monitor and analyze the acoustic emission (AE) signals of pressure vessels, overcoming the limitations of traditional methods in processing complex signals; CNN was employed to extract the local spatial features of the signals, while LSTM captured the short-term and long-term dependencies in the time series to reconstruct the original time-domain signals; the presence of anomalous signals was determined by calculating the error between the reconstructed signal and the input signal; experimental results demonstrated that this method significantly improved the accuracy and robustness of anomaly detection, achieving an accuracy of 93.58% in detecting pressure vessel leaks; compared with existing methods such as AE, VAE, and other deep learning-based autoencoder methods, the proposed method showed greater advantages in capturing the characteristics of complex time-series signals; in addition, this method significantly reduced false positives and false negatives, providing reliable security monitoring for pressure vessels; meeting the practical needs of industrial engineering applications; and having significant engineering application value.
