Chrissi Zachariades
Research Mentor(s): Kate Kusiak Galvin
Mentor Department: CEE
Authors: Meagan Tobias, Chrissi Zachariades
Session: Session 5 (2:00pm – 2:50pm)
Presentation Type: Poster 5
Abstract
With increasingly variable weather patterns and rising flood risks, knowing environmental conditions in real-time is crucial for emergency response and resource management. Sensing networks are promising to deliver vast amounts of data for environmental monitoring, but scaling these networks presents significant challenges. As networks grow, maintaining dependable data streams becomes resource-intensive, with traditional manual quality assurance and quality control (QAQC) approaches becoming unsustainable. The key to successful expansion lies in automating portions of the QAQC workflow while maintaining data integrity. This research addresses QAQC challenges in water depth sensor systems deployed throughout the Huron River watershed. These sensors operate in unstable environmental conditions where physical obstructions, failing batteries, and other factors frequently compromise data integrity. The current QAQC procedures have significant limitations: they rely on manual processes, lack standardization, depend on individual consistency, and offer no systematic approach to ticket prioritization or resolution tracking. Stakeholder interviews identified key workflow pain points, guiding the development of priority-based alerting features. A new solution implements an automated QAQC procedure that analyzes incoming sensor data to identify anomalies indicative of system failures. The algorithm applies classification techniques to detect specific problems in the data, including physical obstructions and battery charging issues. Upon detection of compromised data quality, it generates immediate notifications to Slack (a channel-based messaging platform for work), enabling rapid response to critical issues. By eliminating the dependence on manual inspection and individual consistency, the system standardizes the QAQC process and reduces response times. Preliminary implementation shows workflow efficiency improvements and enhanced data integrity, ultimately supporting more reliable sensor management decisions.
