Time Delay Control (TDC) that uses a multilayer neural network as a nonlinear plant modeler is proposed in this paper. In the proposed controller structure, TDC is used to compensate for the changes of the plant and/or uncertainties including neural network modeling errors. The neural network adapts to learn the uncertainties and the changes in the system, which are eventually embedded into the neural network. In this way, the proposed method exhibits short-term adaptability through TDC and long-term adaptability through neural network adaptation. Because the method uses a neural network as a modeler, it can be effective for the control of nonlinear systems which are hard to model in an analytic way; it also has the ability to cancel out unmodeled dynamics. It is proved that neural network learning error and control error is uniformly bounded. Computer experiments reveal that the proposed algorithm is effective in controlling nonlinear systems. (C) 1997 Elsevier Science Ltd.