Inverse design of digital nonlinear piezoelectric metamaterial using Trans-conv model

Abstract In applications of precision machining and measurement, it is critical to suspend the ambient vibrations. Due to the extraordinary vibration isolation capability of metamaterials, the exploration of vibration suspension based on piezoelectric metamaterials becomes highly significant. This paper proposes an inverse design method for digital piezoelectric nonlinear metamaterials based on the Trans-conv model and provides case studies demonstrating its application in structural vibration suppression. We first designed a programmable digital nonlinear piezoelectric metamaterial with adaptiveness. An automated inverse design method using the Trans-conv model is then developed to design metamaterials with specific user-defined properties. The proposed model combines the advantages of transformer networks and convolutional neural networks. It captures both local and global data features effectively and demonstrates strong inverse design capabilities. Moreover, a pipeline training method is used to solve the common issues of network instability caused by data inconsistency and non-unique mappings. The network was tested on both single-frequency and mixed-frequency datasets. The proposed method enhances the design efficiency of metamaterials, improves the robustness and stability of the design network, and provides an effective solution for vibration suppression in high-precision applications. It holds significant potential for applications in demanding environments such as aerospace, semiconductor manufacturing, and medical devices.