Control of an experimental in-line pH process exhibiting varying nonlinearity and deadtime is described. A radial basis function (RBF) artificial neural network is used to model the nonlinear dynamics of the process. Accommodation of the varying process deadtime in the neural model is achieved by the generation of a feed-forward signal, for input to the neural network, from a downstream pH measurement. The feedforward signal is derived from a variable delay model based on process knowledge and a flow measurement. The neural model is then used to realise a predictive control scheme for the process. Development of the neural process model is described and results are presented to illustrate the performance of the neural predictive control scheme which is tested as a regulator at different setpoints.