Intelligent Optimization Design of a Phononic Crystal Air-Coupled Ultrasound Transducer

Abstract

To further improve the operational performance of a phononic crystal air-coupled ultrasonic transducer while reducing the number of simulations, an intelligent optimization design strategy is proposed by combining finite element simulation analysis and artificial intelligence (AI) methods. In the proposed strategy, the structural design parameters of 1–3 piezoelectric composites and acoustic impedance gradient matching layer are sampled using the optimal Latin hypercube sampling (OLHS) method. Moreover, the COMSOL software is utilized to calculate the performance parameters of the transducer. Based on the simulation data, a radial basis function neural network (RBFNN) model is trained to establish the relationship between the design parameters and the performance parameters. The accuracy of the approximation model is verified through linear regression plots and statistical methods. Finally, the NSGA-II algorithm is used to determine the design parameters of the transducer. After optimization, the band gap widths of the piezoelectric composites and acoustic impedance gradient matching layer are increased by 16 kHz and 13.5 kHz, respectively. Additionally, the −6 dB bandwidth of the transducer is expanded by 11.5%. The simulation results and experimental results are consistent with the design objectives, which confirms the effectiveness of the design strategy. This work provides a feasible strategy for the design of high-performance air-coupled ultrasonic transducers, which is of great significance for the development of non-destructive testing technology.

Integrative BNN-LHS Surrogate Modeling and Thermo-Mechanical-EM Analysis for Enhanced Characterization of High-Frequency Low-Pass Filters in COMSOL

“…At the heart of our analysis is the S21 parameter, a crucial metric of power transmission in filters, whose variability under diverse geometric alterations is meticulously evaluated to underscore our model's sophisticated understanding [10,11]. Through an exhaustive parameter sweep, our research unveils the subtle impacts of minute dimensional adjustments, offering a rich, comprehensive analysis that is both thorough and computationally savvy.…”