Remote anomaly detection for underwater gliders based on multi-feature fusion

As an autonomous platform for long-term observation in the complex ocean environment beyond visual line of sight, the underwater glider (UG) may suffer from some anomalies during the mission. Thus, the anomaly detection of UGs appears to be particularly important for finding malfunction in time and avoiding loss of prototypes. This study considers the limitation of big data transmission in the ocean environment and proposes a remote detection method for Ug anomalies based on multi-feature fusion, which only requires a small amount of data. First, the failure analysis of UGs is performed, and eight characteristic parameters are selected to reflect the anomalies of UGs. Then, the principal component analysis is adopted to fuse the characteristic parameters, which can reduce the dimensions of variables and retain the characteristic information of the original variables to the greatest extent. Finally, the Density-Based Spatial Clustering of Applications with Noise Algorithm is adopted to reduce the false alarm rate of the method. The feasibility of the proposed remote anomaly detection method is verified by the massive historical data of multiple acoustic UGs. It relys on a small amount of data to detect all the abnormal profiles and achieve a decrease in false alarm rate after optimization, which reduce the operational overhead and misson cost.

Automated summary

In this study, a remote anomaly detection method for underwater gliders (UGs) is proposed based on multi-feature fusion. By performing failure analysis and fusing characteristic parameters using principal component analysis, along with the adoption of the Density-Based Spatial Clustering of Applications with Noise Algorithm, the method achieves remote monitoring of UG anomalies while reducing false alarms. The feasibility of the proposed method is verified using historical data from multiple acoustic UGs, demonstrating its ability to detect all abnormal profiles with a decrease in false alarm rate after optimization, thereby reducing operational overhead and mission costs.