Steady state simulations of hybrid reactive distillation provided new and valuable information for the design and optimisation of processes for ETBE and MTBE synthesis (Chapters 3-6). However, dynamic simulations are required to adequately investigate process dynamics and control issues. These areas are of interest due to the complexity and novelty of reactive distillation and the implications of poor operability and controllability: the steady state benefits will not be realised unless satisfactory and consistent operation near the design point is possible. Chapters 7-10 present a range of discussions and investigations that address this fundamental issue.
It was shown that the steady state reactive distillation model could be extended to the dynamic case if appropriate additions and modifications were made to the equation structure. The dynamic model must be fully closed and complete to prevent index problems arising in the numerical solution of the set of differential and algebraic equations. However, acceptable models were developed to incorporate reaction kinetics or by assuming chemical equilibrium. The dynamic model was initially used to predict the open-loop responses to a range of operating disturbances, and then to examine transient responses to parametric changes in the model (e.g. the catalyst activity and the heat transfer rate in the condenser).
A combination of steady state and dynamic simulations were used to investigate multiplicity in hybrid reactive distillation. It was concluded that input multiplicity would be present in all hybrid columns for ETBE or MTBE synthesis due to the VLE behaviour of these systems and the duality of effects that influence the reaction rate. Output multiplicity was also found in some columns and it was shown that physically realisable multiplicities are possible because of: (a) unit singularities; (b) the influence of the energy balance; (c) azeotropes; and (d) reaction hysteresis. Some output multiplicities (including several well known reports from the literature) were classified as pseudo-multiplicity since they were only possible if the system was constrained unrealisably (e.g. a constant molar flow). Where multiple steady states exist, it was found that all steady states are accessible since transitions between parallel steady states could be initiated by perturbations in either manipulated or disturbance variables. This behaviour has implications for equipment selection and start-up procedures, and also influences the control strategy.
Manual (open-loop) control of reactive distillation was found to be essentially infeasible due to the precision required and the capability of the process to shift between parallel steady states. One-point control can be effective and offers advantages where equipment constraints are important. However, care is required in selecting a controlled variable in order to avoid very non-linear and non-monotonic behaviour. Dynamic simulations were used to assess a range of control structures for their applicability for ETBE reactive distillation and the LV and the LB configurations were found to provide the best disturbance rejection and set-point sensitivity under a range of conditions.
The complexity of the reactive distillation process presents challenges for two-point control due to the range of interactions and non-linear responses that are present. However, several control strategies were developed and tested via dynamic simulations to determine a suitable approach. An inferential conversion controller was found to be effective in providing control of the reaction and could be implemented in conjunction with composition control using one of the schemes tested for one-point control. An integrated control scheme was also developed to provide for changing economic circumstances that affect the control objectives. Overall, the LV control structure was found to yield the best combination of control performance and stability for columns that can exhibit multiple steady states. The inventory control problem is simplified with this approach and appropriate control applications can be implemented to manage unwanted transitions between parallel steady states.
This page is intentionally blank
Was this article helpful?