Finite-time sliding mode fault-tolerant neural network control for nonstrict-feedback nonlinear systems

The article concentrates on the finite-time fault-tolerant control for a class of nonstrict-feedback nonlinear systems subject to uncertain control gains and multiple actuator faults. First, neural networks are employed to approximate system uncertainties. By means of a vital structural attribute of neural networks, algebraic loop problem in standard backstepping control design is excluded. Then, a sliding manifold with exponential monotonic attenuation is introduced to ensure chattering-free response and robust performance. Besides, the lumped uncertainty of multiple faulty actuators is handled via applying Nussbaum gain technique and a modified Nussbaum boundedness criterion. To circumvent the issue of “complexity explosion”, a second-order command filter is introduced in every step of recursive control design. Through the proposed adaptive finite-time fault-tolerant control scheme, the influence of actuator faults can be compensated effectively, and the tracking error converges into an arbitrarily small residuum in finite time. Meanwhile, the disc domain of convergent error as well as the upper bound of settling time can be estimated. Finally, two simulations concerned with practical models are discussed. It is expounded that the proposed scheme has more efficiency and less conservatism via comparing it with other existing methods.