Investigating Performance Enhancements of CMUT Based Flow Measurement Using Regression Techniques

This study investigates the use of machine learning regression algorithms to analyze signals from Capacitive Micromachined Ultrasound Transducers (CMUTs) for enhanced flow rate estimation. CMUTs fabricated at Fraunhofer ENAS using wafer bonding technology, operating at a resonant frequency of 2 MHz, were integrated into a custom flow measurement setup based on the transit-time principle, utilizing two CMUTs in transmit and receive modes. To address the challenges in transittime analysis such as, fluctuations due to turbulence, signal overlap and low signal-to-noise ratio at low flow rates, we propose a data-driven approach that uses regression to extract flowdependent features directly from CMUT signals. Through careful feature selection and robust validation techniques, the study demonstrates that even simple models can effectively capture and predict flow rate variations, under constraints of limited dataset. These results highlight the potential of regression-based approaches to enhance the sensitivity and adaptability of CMUTbased flow measurement systems, without the need for additional sensor modifications.