Mechanical parameter identification for rotor systems with application to active vibration control

Abstract

This Ph.D. thesis proposes and validates novel frameworks for real-time mechanical parameter identification in rotor systems, with active vibration suppression serving as the primary application and validation domain. The research is motivated by a fundamental limitation in high-performance rotating machinery: the performance of model-based controllers is critically dependent on the accuracy of the underlying dynamic model, yet key parameters such as inertia, damping, stiffness, and load torque are inherently time-varying and uncertain during operation due to factors like thermal effects, wear, and changing loads. This parametric uncertainty severely limits the robustness and adaptability of conventional vibration control strategies.

To address this core challenge, this thesis develops Lyapunov-based adaptive observers and integrated estimation architectures that enable accurate, online identification of these critical parameters. The core methodological contribution lies in transitioning the control paradigm from reliance on a fixed, nominal model to one driven by a continuously updated and identified model of the plant. The practical significance of these identification frameworks is rigorously demonstrated through their application to two prominent vibration problems: torsional vibration suppression via motor control and lateral vibration mitigation via active piezoelectric bearings.

The core contributions, which bridge the gap between parameter identification and high-performance control, are summarized as follows:
(1) A Lyapunov-based real-time parameter identification framework for adaptive torsional vibration suppression.
An observer is developed to continuously estimate key mechanical parameters. By integrating these estimates into the motor speed control loop, the framework enables parameter-adaptive torsional vibration suppression under varying loads and conditions. Experimental validation shows that this parameter-aware strategy reduces steady-state error by over 20% compared to conventional non-adaptive designs, demonstrating the direct performance benefit of accurate real-time identification.

(2) Parameter-aware active control for lateral vibration suppression using identified rotor dynamics.
The real-time identification framework is extended to lateral vibration control via active bearings. The controller dynamically adjusts piezoelectric actuator forces based on identified rotor parameters (e.g., imbalance, eccentricity), effectively compensating for geometric variations. Experiments confirm that this identification-informed actuation reduces lateral vibration amplitudes by more than 80% under eccentric operation, highlighting its robustness against real-time parameter changes.

(3) Integration of predefined-time disturbance observation with parameter identification to enhance convergence and robustness.
A Predefined-Time Extended State Observer (PTESO) is integrated with the parameter identification module. This co-design ensures rapid, predictable estimation of disturbances while accelerating parameter adaptation, addressing the trade-off between speed and stability. Compared to an ESO-based baseline, the integrated framework reduces residual vibrations by 15–16% and shortens convergence time, validating the synergy between fast disturbance observation and adaptive parameter identification.

In summary, this thesis primarily contributes a systematic parameter identification methodology for rotor systems. Its efficacy and practical impact are demonstrated through significant performance improvements in active vibration suppression, confirming that real-time parameter identification is a critical enabler for next-generation adaptive controllers in high-performance rotating machinery.

Date of Award	1 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Dunant Halim (Supervisor), John Xu (Supervisor) & Chung Ket Thein (Supervisor)