Neural network-based inverse design for multi-objective performance optimization of 70 kHz ultrasonic transducers
Measurement Science and Technology · 08.05.2026 · 37 (19) · 196002
DOI: 10.1088/1361-6501/ae5dee ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 12.05.2026. Verfasst von Yali Yi und 5 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Gefördert von: State Key Laboratory of New Ceramic Materials Tsinghua University, National Natural Science Foundation of China.
Abstract (Originalsprache)
Seventy-kilohertz ultrasonic transducers are essential in advanced manufacturing, especially in precision welding and micromachining. Key performance metrics, including resonant frequency, nodal deviation, and amplitude ratio, are strongly coupled through the structural parameters, which makes coordinated design difficult using conventional approaches. To address this issue, this paper presents a neural-network-based inverse design method. First, a theoretical performance-structure relationship is established using the Mason equivalent model. Subsequently, high-fidelity finite element analysis (FEA) is used to generate the training dataset, with resonant frequency further checked by comparison with three sets of impedance-based experiments. A seven-factor orthogonal design is then used to identify the structural parameters that most strongly influence the target performance. Finally, a particle swarm optimization–backpropagation inverse model is developed to predict key structural parameters directly from target performance inputs. FEA verification shows that the predicted design satisfies the preset specifications for resonant frequency, nodal deviation, and amplitude ratio, while direct experimental validation in the present study is provided only for resonant frequency. This study provides an efficient inverse-design framework for coordinated performance-oriented design of 70 kHz ultrasonic transducers.
Förderung
- State Key Laboratory of New Ceramic Materials Tsinghua University
- National Natural Science Foundation of China
Zitieren
APA
Yi, Y., Yang, Q., Zhou, X., Han, X., Chu, X., & Jin, H. (2026). Neural network-based inverse design for multi-objective performance optimization of 70 kHz ultrasonic transducers. Measurement Science and Technology, 37(19), 196002. https://doi.org/10.1088/1361-6501/ae5dee
Vancouver
Yi Y, Yang Q, Zhou X, Han X, Chu X, Jin H. Neural network-based inverse design for multi-objective performance optimization of 70 kHz ultrasonic transducers. Measurement Science and Technology. 2026;37(19):196002. doi:10.1088/1361-6501/ae5dee
BibTeX
@article{Yi2026,
author = {Yi, Yali and Yang, Qingxin and Zhou, Xukang and Han, Xueyan and Chu, Xiangcheng and Jin, Herong},
title = {Neural network-based inverse design for multi-objective performance optimization of 70 kHz ultrasonic transducers},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {19},
pages = {196002},
doi = {10.1088/1361-6501/ae5dee}
}
EUVIMED