Industrial Internet of Things (IIOT) Physics-Based Dimensionality Reduction

N. Loychik
Los Alamos National Laboratory,
United States

Keywords: IIOT, dimensionality reduction, AI, ML


The industrial internet of things (IIOT) is fast becoming a reality with sensors increasingly deployed upon machinery to increase facility reliability, cybersecurity, and efficiency. Today, pumps, compressors, and fans are likely to have access to continuous data streams of raw high-dimensional vibration data to alert facility operators to equipment faults. Unfortunately, data collected is typically not fully utilized with sensor data mostly being used to triggering alerts for human inspection and diagnosis rather than providing real-time operational insight for facility efficiency, cybersecurity, and reliability. LANL is researching physics-based dimensionality reduction for edge-AI vibration sensors to provide a means to turn raw vibration data into process information (flow, pressure, run speed, etc.). This presentation shall introduce objectives of the sensor, the mathematical framework for dimensionality reduction, initial results, and plans for a prototype sensor.