Observer Based Adaptive Neural Networks Fault-Tolerant Control for Pneumatic Active Suspension With Vertical Constraint and Sensor Fault

This paper treats the problem of a pneumatic active suspension considering uncertain parameters and displacement constraints under the presence of sensor failure and unmeasured states. A quarter car model is established using an air spring to provide flexible stiffness and an active force to suppress the chassis vibrations. To approximate unknown nonlinear functions of pneumatic actuator dynamic, neural networks are employed as a function approximator. The sensor fault is investigated while all system states of the controlled suspension are assumed to be unmeasurable. Then, a neural state observer is constructed to estimate unmeasured states and compensate for the partial loss of effectiveness of sensor failure simultaneously. An adaptive fault tolerant control is proposed via the command filter backstepping technique to alleviate the effects of sensor failure and solve the explosion of complexity problem. In addition, to improve the tracking accuracy and get the ride comfort, this study is concerned with a prescribed performance function so that the vertical displacement of sprung mass does not violate the predefined boundary. The Lyapunov theorem is then applied to demonstrate that all system signals are semi-globally uniformly ultimately bounded while error variables can converge to the small neighborhood of zero. Finally, comparative simulation examples are analyzed to verify the effectiveness of the designed control with sensor noise and unmeasurable variables.