PMUT-Based Devices and Machine Learning-Assisted Noninvasive-Bubble Recognition for Small Pipes

The detection of bubbles in small pipes is of significant importance for the stable operation of pipeline systems and large equipment. Existing detection methods are invasive and struggle to effectively identify bubbles in small pipes ( ≤ 5 mm). This paper proposes a non-invasive bubble detection method based on piezoelectric micromachined ultrasonic transducers (PMUTs) and machine learning assistance. The study utilizes the principle of ultrasonic wave transmission combined with a support vector machine classification model to process the transmitted ultrasonic signals. The PMUTs frequency is optimized through a finite element method to ensure the stability of bubble detection. The fabricated PMUTs successfully enable the determination of bubbles in pipes with an inner diameter of ≤ 3 mm and a minimum detectable bubble diameter of 1 mm. Further research demonstrates that the method achieves recognition accuracy above 92% across various temperatures, pipe inner diameters, and flow rates, showing great potential for engineering applications.