Abstract:Highlights • We proposed unsupervised graph anomaly detection with discriminative embedding similarity for viscoelastic sandwich cylindrical structures. • The graph topological structures are constructed to comprehensively capture the attribute characteristics and similarity correlations within normal samples. • The graph representation learning process facilitates grasping the information across diverse nodes and enhances the expression of graph data. • We integrate reconstruction errors with embedding similarity errors for the final detection indicators to enhance the anomaly detection ability. In the detection of slipping anomalies in viscoelastic sandwich cylindrical structures (VSCS), conventional methods may encounter challenges due to the extremely rare and weak nature of slipping signals. This study focuses on normal signals and introduces an unsupervised graph representation learning (UGRL) with discriminative embedding similarity for VSCS's detection. UGRL involves data preprocessing, model embedding, and matrix reconstructing. Association graphs are constructed based on sample similarities for yielding adjacency and attribute matrices. Subsequently, the matrices undergo embedding and reconstruction via various network modules to enhance graph data characterization. Detection indicators are derived by calculating embedding similarities and reconstruction errors, and thresholds are constructed using these indicators to enable efficient anomaly detection. The experiments in VSCS slipping dataset effectively indicate the superiority of the proposed method.

Unsupervised Graph Anomaly Detection with Discriminative Embedding Similarity for Viscoelastic Sandwich Cylindrical Structures