Radar-Based Condition Monitoring for Enhanced Efficiency and Safety in Industrial Processes

• Published in: 2024 International Radar Symposium (IRS) pp. 303 - 307
• Main Authors: Schmitz, Robin, Vogt, Michael, Roitzheim, Maximilian, Hammes, Markus, Schulz, Christian, Barowski, Jan, Rolfes, Ilona
• Format: Conference Proceeding
• Language: English
• Published: Warsaw University of Technology 02-07-2024
• Subjects: Accuracy; Condition monitoring; Feature extraction; Filling; Heating systems; Liquids; machine learning; millimeter-wave radar; Radar measurements; radar signal processing; range-doppler measurements
• ISSN: 2155-5753

In the diverse field of process industry, a plethora of technical processes are leveraged. Among these, liquids are often involved, being stored, transported and processed in tanks and reactors. Tailored to such applications, this contribution seeks to introduce and evaluate a novel technique for condition monitoring (CM). Within this technique, radar systems are employed and their functionality is augmented, enabling not only precise and accurate level measurements of liquids, but also distinction between different process conditions (PCs). For this purpose, distinguishing quantitative features are extracted from range-Doppler radar measurements. Thereafter, a classification is performed applying methods from machine learning. The potential of our proposed technique has been experimentally evaluated with respect to commonly encountered PCs. A frequency-modulated continuous wave (FMCW) Doppler radar system, operating in the V-/W-band, has been utilized for the measurements. It is demonstrated that the selected PCs can be distinguished with almost 98% accuracy, by extreme gradient boosting (XGB) along with quantitative features such as the peak-to-averaae power ratio.