Intelligence Unleashed: Elevating Navigation Structure Health Monitoring Through Advanced Anomaly Detection in Sensor Suites

This study involved an exploratory analysis of dam monitoring data collected from ten sites. Exploratory data analysis (EDA) was employed to gather an overarching understanding of the data and swiftly identify potential errors, preventing contamination of subsequent analyses. Three EDA tools, including histograms, heat plots, and box plots, were applied to generate charts illustrating the summary distribution of the dataset, visualizing outliers, and depicting correlations among various attributes collected for a lock/gate. The information is crucial for the United States Army Corps of Engineers (Corps) research agency, the Engineer Research and Development Center (ERDC), analysts to make decisions regarding additional instrumentation to enhance data coverage and interpret and understand anomalies in the data. Identifying performance anomalies was deemed ineffective, requiring hours of manual analysis, and risking the oversight of anomalies not adhering to conventional rules of thumb. Toward this end, a holistic (system-wide) anomaly detection methodology, addressing system-wide anomalies, is proposed. The methodology focuses on abnormalities within subsystems, potentially bypassing overarching system performance metrics like response time, end-to-end latency, and service level agreements. The methodology harness Shannon Entropy, providing a potent means to gauge the remaining uncertainty within the system after observation. A case study utilizing data supplied by ERDC demonstrated that the proposed approach achieved high precision in identifying anomalies in the monitoring data.