Improved adaptive neural network motion control for an aero-engine hydraulic system

Generally, there are always many unknown disturbances in the aero-engine, such as the complex aerodynamic load force, the bandwidth and computation resource limitations. If these uncertainties are not effectively addressed, they will seriously affect the safety and performance of the hydraulic system. In this paper, the radial-basis-function (RBF) neural network (NN) combined with the event-triggered technique is used to address these unknowns, in which NN weights update only at necessary instants. This permits a greater number of NN nodes to be employed, augmenting the estimation accuracy of RBF-NN estimators without overconsuming limited computational resources. A new motion control scheme with an event-triggered communication protocol and the designed estimator is integrated to ensure the high performance and low occupation of bandwidth resources for hydraulic actuators. Using the proposed controller for hydraulic systems and analyzing from the viewpoint of impulsive dynamical systems, it is concluded that all closed-loop signals are ultimately bounded and the system output can accurately track the reference trajectory. Likewise, Zeno behavior no longer occurs. Finally, two comprehensive experiments, i.e., the unknown mass load and aerodynamic load force, demonstrate the remarkable validity and merits of the proposed control scheme in the aero-engine Hardware-in-the-Loop (HIL) networked control platform with hydraulic actuators.