A hybrid approach to model predictive control of a polymer grade transition problem is developed The controller is posed as a mixed integer programming problem in which the discrete elements of the algorithm include both linear switching models used to approximate the nonlinear process and operating heuristics that are incorporated as propositional logic inequalities. The algorithm is tested in simulation on a polypropylene loop reactor system. Feed rates of monomer, hydrogen, and catalyst are manipulated to regulate slurry density, melt flow rate, and production rate, for various grades. Robust performance is demonstrated for grade transitional control. (C) 2004 American Institute of Chemical Engineers.