This study investigates the efficacy of multiple machine learning (ML) strategies for optimizing the design of freeform Piezoelectric Micromachined Ultrasonic Transducers (PMUTs) by leveraging a data-centric methodology. We devise a comprehensive four-stage optimization framework comprising a freeform PMUT shape generator, a feature extractor, a finite element analyzer, and ML estimators. The ML evaluation compared to the finite element analysis reveals that the leading ML estimator accomplished over 95% prediction accuracy with notably low error rates. With this framework, a dataset comprising 30,000 samples was processed within 6 seconds, facilitating the rapid selection of optimal PMUT configurations. Our findings highlight the potential of ML methods to significantly accelerate and optimize PMUT design, resulting in improved sensitivity and precise operational frequency control.
«
This study investigates the efficacy of multiple machine learning (ML) strategies for optimizing the design of freeform Piezoelectric Micromachined Ultrasonic Transducers (PMUTs) by leveraging a data-centric methodology. We devise a comprehensive four-stage optimization framework comprising a freeform PMUT shape generator, a feature extractor, a finite element analyzer, and ML estimators. The ML evaluation compared to the finite element analysis reveals that the leading ML estimator accomplished...
»