A neural control strategy for nonlinear processes with time-variant time-delay is proposed in this paper. In this strategy, a dynamic neural network based nonlinear Smith predictor is constructed to compensate for the effect of time-delay of a class of nonlinear processes. An on-line optimizing controller is illustrated based on the neural Smith predictor. It is known that the performance of the Smith predictor may be deteriorated if the time-delay of the process changes with time. In order to improve the performance of the Smith predictor, a time-delay adaptation mechanism is introduced into the control structure to track the variation of the time-delay. The simulation, comparing with the classical Smith predictive control, on a continuous-stirred-tank-reactor (CSTR), where the time-delay of the manipulating flow changes with time, is used for the test.