One of the main concerns of strips producers is to measure strip thickness accurately as it is produced. Correct modelling of the sensitivity of output variables to input variables in a rolling mill model is one of the keys to obtaining more accurate data. An adaptive control system that uses an artificial neural network (ANN) creates a model of the process directly from measurement data. Using the model, the control system can predict how the process will react to control actions. The creation of the model and the computation of the control strategy are carried out automatically by the control system. The proportional-integral-derivative controller is used in this method to increase accuracy of final estimated variables and to increase accuracy of control of the system. To determine the correct tuning for thickness control, three control parameters are considered: the roll gap, and front and back tensions. A predictive model is used, based on the sensitivity equations of the process, where the sensitivity factors are computed by differentiating a previously trained neural network. Results of a case study in a real plant show that this online-offline model is effective in reducing thickness variations in produced strips.