A control system using a neural net is applied for product composition control of a distillation plant. The neural network controller design is based on the process inverse dynamic modeling. Once the inverse dynamic model is available then it can be used for control. The backpropagation algorithm of the Generalized Delta Rule is used to train the network minimizing the sum of squares of the residual. The algorithm is applied to dynamic nonlinear relationship between product composition and reflux flow rate. The obtained results ilustrate the feasibility of using neural net for learning nonlinear dynamic model distillation column from plant input-output data and control. These results demonstrate the importance to take the time-delay of the plant into account.