Nonlinear control of multi-actuator electrohydraulic systems based on feedback linearization with application to road simulators

by 1975- Ayalew, Beshahwired

iii Electrohydraulic actuators constitute important force generation and positioning elements in various industrial and testing applications. Their high power-to-weight ratio and high load stiffness make them better choices than their rival electromechanical actuators in multi-actuator service load simulation testing applications such as road simulators, flight simulators and shaker tables. However, electrohydraulic actuators exhibit significant nonlinearities in their dynamics. In order to obtain satisfactory performance in the presence of these nonlinearities, more elaborate control techniques than the ubiquitous PID loops may be necessary. In this thesis, nonlinear models of electrohydraulic systems are developed for a typical single actuator test system. This test system is such that detailed modeling of transmission line dynamics is found necessary. A useful result obtained from modal approximation of distributed transmission line dynamics is outlined for a specific causality case. Suitable system interconnection models are adopted and validated using experiments on the test system. The validated system model is then used to derive nonlinear pressure/force and position controllers based on feedback linearization and its robust enhancements. Feedback linearization can be applied to certain model structures which allow the cancellation, in real-time, of the measured and modeled nonlinearities of the system. It is shown in this thesis that a model of an electrohydraulic system can be configured as input-output (IO) linearizable (or partial feedback linearizable) under some basic assumptions. In fact, these assumptions are necessary, and yet not unduly restrictive, that the term Near IO linearization is used with the controllers so derived. A sliding mode controller is designed as a robust extension of the Near IO linearizing controller with pressure/force output. It is also shown that the Near IO linearizing controller with position output is equivalent to a cascade controller implementing the Near IO linearizing pressure/force controller as an inner-loop to a feedback plus feed forward outer-loop position controller. The cascade implementation has the convenient feature that the position control closed-loop error has a second-order iv linear dynamics driven by the pressure/force control closed-loop error, which in itself has a first-order linear dynamics. A consequence of the equivalence is that it gives insight into the choice of the linear gains for the Near IO linearizing position controller. Furthermore, the cascade form allows one to view the robustness issues for position control from a Lyapunov backstepping perspective. The performance of the nonlinear controller is compared against standard PID and linear state feedback with integral controllers using experiments and computer simulations of the nonlinear system model. It is shown that the nonlinear controllers have better tracking performance than the linear controllers, particularly in force control. It is demonstrated that there is more performance advantage for the nonlinear position controllers with suggested system layout changes and improved signal processing. The nonlinear position controllers are further considered for a multi-actuator application in road simulation. A nonlinear full-bus model of a transit bus is developed for computer simulations of a four-post road simulation system. Time domain interaction measures are derived to look at interactions between decentralized PID+?p and Near IO linearizing position control loops. It is shown that there is little interaction between either of the decentralized position control loops. However, a second cascaded decentralized controller considered for tracking a remote parameter like spindle vertical acceleration response faces significant and persistent interactions. Finally, the performance of the multi-actuator road simulation system under a decentralized Near IO linearizing controller and a decentralized PID+?p controller are compared for a typical rough road profile. The Near IO linearizing resulted in a more than 60% improvement in the tracking error metric across all four actuators and a more than 50% improvement in the response matching of the sprung mass acceleration power spectral density over that obtained with the PID+?p controller.