In order to catch more process details in chemical processes, a dynamic model for prediction of process trends is proposed by modifying traditional time-series ANN (artificial neural networks) model with impulse response identification means. The application result of the model is briefly discussed.