Damage Detection for Pressure Vessels Using a New Graph Neural Network Framework

Pressure vessels, under harsh operating conditions, are prone to damage from external forces, media interactions, and thermal effects, potentially leading to dangerous incidents like leaks, corrosion, and cracks. To detect damage without compromising equipment performance, the study first converts complex ultrasonic guided wave data into graph data, employing a sparse sampling method that utilizes data from only one excitation sensor to construct the graph. Subsequently, feature extraction module based on graph neural network combined with graph pooling techniques is designed to extract features and achieve precise damage localization and detection. Experimental results indicate that this method demonstrates high accuracy and robustness in the task of pressure vessel damage localization, offering a promising technique for intelligent monitoring of industrial equipment.