PiezoNet and Data-Driven Models for Time-Domain Characterization of Piezoelectric Resonators

This paper presents a fully automated data acquisition platform and the resulting database – PiezoNet<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> – for data-driven time-domain characterization of piezoelectric resonators used in power electronics. The platform measures the voltage and the current over piezoelectric resonators across a wide range of excitation waveforms and ambient temperatures. The power stage dynamically adjusts to efficient operating points for best operation under different load conditions. This system is mechanically versatile, accommodating crystals of diverse materials and dimensions. The platform enables comprehensive and precise characterization by automating the data collection process, thereby providing extensive datasets essential for training data-driven models (e.g., neural networks) to predict operating points and nonlinear behaviors, and to quantify the sample-to-sample variation of piezoelectric resonators. A family of sequence-to-sequence neural network models were trained and tested to validate the feasibility of time-domain data-driven models for piezoelectric resonators.