A novel fault tolerant control strategy based on radial basis function (RBF) neural network adaptive compensation is presented for a class of networked control system with random time delay. The networked control system is modeled as stochastic switched system by estimating time delay online and an RBF neural network is designed to compensate effect of the system faults and disturbances dynamically by using model reference adaptive control technology. Furthermore the weights tuning law of neural network is given through Lyapunov stability theory to ensure the tracking performance and system state uniformly ultimately bounded when failures occur. Finally, the simulation results are given to illustrate the feasibility of the proposed method.