Accelerated inverse design of vibration isolators with customizable low dynamic stiffness characteristics via deep neural network

Abstract

The quasi-zero stiffness (QZS) isolator composed of curved beams is considered to be an effective way to address the contradiction between high load-bearing capacity and low-frequency vibration isolation. However, finding geometries with target QZS characteristics is not simple. In this study, we present a framework for designing customizable QZS isolators. We employ a deep neural network to accurately learn the relationship between the geometry of the curved beam and its nonlinear mechanical response. Furthermore, we combine the network with genetic algorithm to inverse-design isolators that exhibit the targeted QZS characteristics, thereby achieving a two-order-of-magnitude improvement in speed compared to traditional method. Static experiments demonstrate the reliability and customizability of the proposed design strategy for QZS isolators. Dynamic analysis shows that the isolator has a low resonant frequency, enabling ultra-low-frequency vibration isolation. Notably, series-parallel arrangements can significantly improve the load-bearing capacity or vibration isolation performance of the isolator. Our design framework addresses efficiency issues in traditional QZS designs, enabling faster iterations and calculations. It has broad applicability and potential in systems requiring customized nonlinear mechanical responses.