A Spatiotemporal Correlation-Based Low-Power Task Scheduling and Anomaly Detection for HM-FW Sensing Systems

Health monitoring and fault warning (HM-FW) sensing systems are vital for ensuring the operational safety of industrial equipment and infrastructure. However, their sensing efficiency and data security are limited in low-power scenarios. In this article, we propose a novel computational model named smart energy and guard assistant (SEGA). It aims to optimize low-power task efficiency and data security in HM-FW sensing systems by integrating spatiotemporal neighborhood information with adaptive sensor-task scheduling. To achieve this, we first design a novel neighborhood-aware model to generate reference secure regions for anomaly detection and correction of the sensed data. In addition, to optimize low-power sensing efficiency, we design multiorder derivative analysis for dynamic sensor-task scheduling, which efficiently captures the trends of the sensing information with minimized energy consumption. We also build an adaptive window to enhance anomaly detection performance and develop a dynamic missing data imputation model for asynchronous sensor data fusion that ensures consistent and accurate data processing. To evaluate our design, we use a self-collected industrial temperature and humidity dataset with 4700 samples, collected every second under varied industrial conditions. The results indicate that SEGA can improve the power efficiency of the sensor task by 70% and reduce data redundancy by 79%. In addition, it achieves a precision and recall of 0.98, along with an ${F}1$ score of 0.97 in anomaly detection to ensure data security. Benefiting from low power consumption and high-security features, we believe that SEGA can help HM-FW sensing systems operate more efficiently and provide more reliable monitoring for industrial equipment.