Model-free adaptive fault-tolerant control for subway trains with speed and traction/braking force constraints

This study investigates the subway train fault-tolerant control problem for the actuator fault with constraints of speed and traction/braking force. The complex subway train dynamics is first transformed into a compact form dynamic linearization data model with the help of the concept pseudo-partial derivative (PPD) proposed under the framework of model-free adaptive control. By using the approximation of the uncertainty fault with radial basis function neural network (RBFNN), then a model-free adaptive fault-tolerant control (MFAFTC) scheme is designed by only using saturated input/output data of a subway train. The proposed MFAFTC scheme consists of two data driven on-line learning updating algorithms for RBFNN-weights and PPD, and its convergence is strictly proved by rigorous theoretical analysis, and whose correctness and effectiveness are further verified by a numerical simulation.