High precision identification of dynamic model for accelerometer based on VMD

Abstract

The dynamic model of accelerometer reflects characteristics of dynamic output changes over time, directly affecting dynamic measurement performance. To solve the problem of low identification accuracy due to pure delay, nonlinearity, and noise, a high-precision closed-loop identification method for dynamic model based on variational mode decomposition (VMD) is proposed. In particular, emphasis is placed on highlighting the issue that equivalent stiffness parameter is more susceptible to noise due to its low parameter sensitivity. Additionally, to address the effects of pure delay and nonlinearity on model identification, a dynamic model of accelerometer containing both is proposed. In response to the impact of noise on model identification, it is proposed to use VMD for noise separation and data reconstruction. Finally, a test platform is established for validation. And experimental results indicate that the model identified using the proposed method shows an increase in model matching degree (MMD) from 90% to 98.8%.