Double-Feedback Design Using an Estimation Network for Servo Resonance Suppression

Mechanical resonance remains a critical bottleneck in high-performance dual-inertia servo systems, where its destabilizing effects are amplified by external torque disturbances, ultimately compromising control precision and dynamic response. To achieve more competitive resonance suppression, this paper proposes a novel mechanical resonance suppression scheme based on an estimation network employing five extended sliding-mode observers (ESMOs). The proposed scheme constructs a dual-feedback structure, where the differential speed feedback and the shaft torque feedback can effectively suppress mechanical resonance through zero-pole placement, while the additional load torque compensation term mitigates the impact of load torque variations on system resonance. Since the double-feedback structure relies on accurate mechanical parameter, the estimation network is designed to provide the mechanical parameters, and consists of five ESMOs to estimate the load speed, the motor inertia, the load inertia, the stiffness coefficient, and the load torque in parallel. Interestingly, the equivalent-changing rate and the inline gain of the estimated value designed in each ESMO provide the possibility for the convergence of the estimation network. Finally, compared with existing methods, simulation and experimental results verify the effectiveness.