Smart Sensor Mode Switching for Real-Time Anomaly Detection

The patent segments the sensor data processing into two distinct parts: (1) smart sensor devices perform real-time preprocessing, anomaly detection, and validation locally; (2) the control system performs batch processing only on validated data subsets. This segmentation allows both real-time responsiveness and accurate modeling to coexist, resolving the contradiction between processing speed and model accuracy.

The patent implements dynamic operational modes for smart sensor devices that can switch between low-power monitoring and high-processing anomaly detection based on operational conditions. The processor adjusts its processing intensity dynamically, performing comprehensive analysis only when anomalies are detected or during critical operational phases, thereby maintaining reliability while reducing overall energy consumption.

The patent extracts and removes invalid or low-quality sensor data at the source through validation algorithms implemented in smart sensor devices. By filtering out erroneous data before transmission, the system reduces the burden on external processing systems and communication networks, thereby reducing overall system complexity while maintaining measurement precision for valid data.

The patent applies partial processing in real-time at the smart sensor level, performing only essential functions such as data validation, basic anomaly detection, and filtering. More comprehensive analysis is performed selectively on subsets of validated data by the control system. This partial real-time processing maintains productivity while avoiding the excessive energy consumption that would result from full real-time processing at all levels.