Modeling, Control and Optimization of a Plate Reactor

Abstract: A new chemical reactor, the Alfa Laval Plate Reactor, is being developed by Alfa Laval, a Swedish world-leading heat exchanger company. The plate reactor combines the high-heat-transfer capabilities of plate heat exchangers with the efficient mixing and reaction control typical of microreactors. With this new concept, highly exothermic reactions can be produced using more concentrated reactants and more accurate temperature control. This will reduce the reaction time and the need for downstream separation, thus saving energy and reducing the impact on the environment. The focus of this thesis is to develop and apply control methods to take advantage of the full potential of the novel plate reactor concept. A nonlinear model of the reactor is derived based on first principles to conduct a system analysis and enable model-based control. The physical model allows a detailed investigation of the potential control inputs and how the process design and choice of inputs may affect the control design. Two control concepts are examined, decentralized control using multi-loop PID controllers and centralized control using Model Predictive Control. The concepts are evaluated and compared in terms of design methods, performance and practical aspects. A cooling system is designed and experimentally verified, where a mid-ranging control structure is implemented to increase the operating range of the hydraulic equipment. The start-up control problem is challenging due to process uncertainty, highly nonlinear dynamics and input and temperature constraints. The dynamics and the constraints are easily captured by the process model in the optimization problem. The open question is how to address the process uncertainty. Here, robustness to uncertainty is achieved by introducing state-space constraints in the optimization formulation, which decrease the sensitivity of the optimal solution. The start-up control problem has been approached from two sides, a time-driven continuous approach and an event-driven hybrid approach. Some of the results are specific for the plate reactor, but many parts may be generalized to other applications, for example the decentralized and centralized control design, the start-up/transition control design and use of mid-ranging control to handle two-input one-output systems.