Signal-driven System Identification and Model Predictive Control for Multi-DOF Micro-vibration Isolation Platforms

With the escalating demand for environmental stability in high-end manufacturing and precision metrology equipment, micro-vibration isolation systems face two persistent challenges. Insufficient model identification accuracy and limited real-time control performance. To address these issues, this paper introduces a signal-driven system identification approach. Furthermore, a model-predictive control (MPC) strategy is devised. The proposed method is based on a discrete extended Kalman filter (DEKF), which iteratively suppresses measurement noise and modeling errors to achieve efficient estimation of key parameters. A bespoke MPC scheme is subsequently formulated, incorporating an exponentially weighted cost function to avoid matrix ill-conditioning during optimization. The applicable frequency band of the proposed method is 0.5-100 Hz. Experimental results show that the average parameter-identification error is restricted to 1.05%. With active control engaged, the system achieves vibration attenuation levels of -20 dB at 2 Hz and -40 dB at 10 Hz.