A hybrid model of a polyethylene production process is developed. The mechanistic model utilizes fundamental material and energy balances to predict important process conditions, such as the reactor temperatures, conversions, and the molecular-weight distribution (MWD) of the polymer. Using plant data, it is shown that accurate MWD predictions are not obtained from the mechanistic model alone, despite efforts to accurately model the system and improve the accuracy of the input data. Because an accurate prediction of the MWD is required to predict end-use properties, a hybrid model was developed by adding an empirical layer to the mechanistic model. The empirical layer was developed by using an optimization algorithm to adjust the predicted MWD by manipulating multipliers of the key descriptors (states or functions of states) of the distribution. These multipliers were then predicted from plant data using feedforward artificial neural networks (FANNs). They are then combined with the mechanistic model to allow accurate MWD prediction.