reactors by continuous ones. The reactor system which is investigated here comprises several inputs and due to the near plug fl ow characteristics between the inputs and the outputs, reacts to the changes of the infl ows with large time delays. Model-based optimizing control is the obvious candidate to control such a multi-input plants with significant delays. While the standard implementations of model predictive control penalize the deviations of the states or outputs from reference values and the control moves in the cost function, in this work the controller is formulated to account directly for the economics of the reactor system while the product quality parameters are imposed as constraints. [ 1 ] Some applications of optimizing control can be found in refs. [ 27 -30 ] . Limited computation power and lack of effi cient optimization algorithms have caused thatIn this contribution, a nonlinear model-based optimizing control for continuous polymerization of acrylic acid in tubular reactors with multiple side injections of monomer is developed. The background of this work is to transfer the production of polymers from semibatch to continuous operations. The confi guration of the tubular reactor, which imposes long delays between the inputs and the measurements, and the lack of intermediate measurements as well as the nonlinear reaction kinetics and sharp moving fronts of concentrations when the infl ows are changed, make the application of the optimizing control very challenging. In order to simulate the sharp fronts of the reactor system faithfully and fast, the spatial domain of the partial differential equations model of the tubular reactor is discretized by applying the weighted essentially non-oscillatory scheme. We formulate the controller such that it optimizes the productivity of the reactor directly, while the product quality parameters are imposed as constraints. A particle fi lter is implemented to estimate the states of the reactor system and to initialize the process model. The simulation results show that the controller can increase the product throughput considerably compared to an initial operating point and has a robust performance against the measurement noise. Furthermore, the effect of formulating the quality constraints as hard or soft constraints as well as a changeover scenario between the different grades of a polymer are studied.