AbstractReal-time hybrid testing allows engineering systems to be tested using a combination of modelling and experimental methods. By separating the system into numerical and physical substructures the scheme enables the benefits of both schemes to be attained such as low cost due to minimal component requirements and the ability to test complex systems that cannot be simulated. The numerical and physical substructures of the hybrid test are coupled and run together using actuators and force sensors to transfer data at the interface in real-time. However, actuator dynamics lead to poor stability and tracking errors which undermine the reliability of hybrid tests. Existing techniques to mitigate actuator dynamics are largely based on linear models of actuation hardware and are thus less effective in the face of nonlinearity. Moreover, such schemes require information of actuator dynamics to be known prior to testing.
This thesis explores the application of several passivity controllers in real-time hybrid testing to improve stability and performance. The work here covers the first applications of passivity control in hybrid tests and uses a combination of modelling and experimental techniques to validate the effectiveness of the schemes proposed. Performance of the compensated hybrid tests are assessed in comparison to a simulation of the true system to be emulated. The preliminary passivity controller designed is based on damping the numerical substructure using measurements of energy flowing in the system and was found to restore unstable systems with phase margins of up to -210. Performance of the scheme with state-of-the-art model-based lag compensators were found to further improve performance allowing targeted improvements in tracking to be achieved with improved stability. Experimental results show the effectiveness of passivity control and 2nd order transfer function based lag compensation in mitigating phase lags of up to 220 from a closed loop hybrid test.
However, high dependency on the magnitude of energy flow required considerable tuning of the passivity controller when operating conditions shifted. Thus, a modified passivity control scheme acting on a normalised power flow measurement was designed which alleviates this limitation whilst still allowing stability gains to be achieved. Unlike its predecessor, the modified scheme was seen to enable identical performance for a range of excitation amplitudes resulting in the same system natural frequency, damping ratio and response distortion for step excitations of magnitudes 0.5mm up to 500mm.
Besides linear hybrid tests, this thesis also focusses strongly on nonlinear systems and all schemes presented require no information of actuator dynamics to function. Nonlinearities tested in the numerical and physical substructures include stiffening behaviour and discontinuity whilst nonlinearity in the actuator in the form of nonlinear friction has also been tested. In all cases, the passivity controller was seen to improve the response by stabilizing diverging systems and eliminating oscillation caused by periodic instability in the actuator due to friction.
A limitation of passivity control used by itself however is its inability to eliminate the phase lag in the actuator. Thus, the use of passivity control together with a novel adaptive feedforward filter to mitigate actuator phase lag was analysed. The two schemes were found to complement each other with passivity control allowing stability to be maintained in otherwise unstable tests while adaptive filtering converged the substructure position error towards zero overtime. Finally, to end with, a passivity based adaptive delay compensation scheme was designed which measures the energy flow in the hybrid test to quantify the actuator delay. This scheme was seen to enable phase lags of up to 360 in the actuator to be cancelled.
This thesis is written in the alterative format, with each passivity controller presented in a research paper. Each paper precedes a context section which outlines the motivation and purpose and the chapter ends with a summary linking the main findings to the overall research question. The work in this thesis comprises of a conference paper, four journal papers and a technical article.
|Date of Award||13 May 2020|
|Supervisor||Andrew Plummer (Supervisor) & Jonathan Du Bois (Supervisor)|