Research on mid-to-high frequency vibration control strategies for structures considering parameter errors in active control systems

To address the performance degradation caused by electromagnetic actuator nonlinearity and secondary-path model mismatch in 400–800 Hz mid-to-high frequency active vibration control of electric drive assembly reducers, an adaptive weight adjustment robust compensation-feedback FxNLMS algorithm (AWR-FxNLMS) is proposed. An actuator surrogate model with Bouc-Wen hysteresis and an electromechanical coupling model are established, and the stochastic characteristics of multiplicative amplitude and frequency uncertainties in the control system are analyzed. The error-to-disturbance energy ratio is used to adjust online the weights of the FxNLMS adaptive feedforward branch and the H ∞ robust feedback branch. It enables online switching/fusion between the adaptive feedforward and H ∞ robust feedback. The controller does not rely on an accurate secondary path; it maintains the optimal convergence of FxNLMS under low mismatch and automatically increases the weight of the robust branch under high mismatch. 50 independent repeated experiments are conducted on a cRIO experimental platform. The results show that the two algorithms are comparable at 100–300 Hz; at 700 Hz, AWR-FxNLMS improves the attenuation rate by 41.53% relative to F-FxNLMS and shortens the convergence time by 2.503 s; and at 800 Hz under severe model mismatch, the startup overshoot is suppressed.