Process design and advanced control strategies for model based reactived distillation columns

Abstract

Reactive distillation, which is a novel alternative to the sequential operation in reactor and distillation for certain reactive systems, is gradually becoming an important unit operation in chemical process industries. It offers advantages in chemical reaction through lifting the chemical equilibrium limitations and separation by overcoming the distillation boundaries. Economic advantages can also result from the direct energy integration and the reduction of the number equipments due to simpler production flow sheet. Therefore, it can reduce in investment and operational costs Reactive distillation is a highly non-linear system due to the interactions between vapour-liquid phase equilibrium, chemical kinetics, vapour-liquid mass transfer and diffusion inside the particle catalysts. It causes the existence of multiplicity phenomena and complex dynamics, which have been verified using experimental laboratory and pilot plant unit. A high quality-rigorous mathematical process model is required to investigate and obtain the optimal design and control of the reactive distillation. For modelling the reactive distillation, a rate-based model can represent physically closer to the real system. However, it requires estimation of more empirical and semi-empirical parameters and offers no improvement in accuracy for most reactive systems. An equilibrium-stage model with proper efficiency (for tray column) or height equivalent to a theoretical stage, HETP (for packed column) is satisfactory and still a preferable alternative model. This work focuses on the application of reactive distillation for ethyl tert-butyl ether (ETBE) production. Compared to methyl tert-butyl ether (MTBE) reactive distillation, the ETBE reactive distillation is still limited and yet to be commercialised. Current research has focused on MTBE, while recent studies reveal that MTBE :]as severe ingress problem, which pollutes underground water. The pilot scale of reactive distillation column for ETBE production, which is available in the Department of Chemical Engineering, provides experimental data to verify the mathematical model and serve as the case study for control design and optimisation. The conceptual design of reactive distillation estimates the number of separation and reactive stages, feed flows and locations, energy loads, catalyst requirements, etc. It has been investigated extensively and several methods are proposed. However, reactive distillation design is still an open research area due to the complex interactions between the vapour-liquid transfers and reaction rates. Some unusual responses have been reported and should be fully understood to avoid poor design and sub-optimal performances. Besides, the market price of the products, the costs of the reactants, and the utility costs determine the optimisation process. A rapid method, which can optimise a reactive distillation column, is challenging to develop. Despite the presence of multiplicity phenomena and complex dynamics, conservative approach of adding a few extra stages to the calculated theoretical stages can be applied in designing reactive distillation columns. The additional stages do not degrade the column performance if the operating conditions are chosen appropriately. However, shorter columns are recommended for both single-feed and double-feed because they can produce comparable performance with less energy consumptions. If related problems of the internal catalyst such as flexibility for loading and unloading and supporting techniques for adequate liquid-vapour contacts can be overcome, a pre-reactor can be eliminated from the ETBE production flowsheet employing a reactive distillation column. The concept of side reactors is introduced to overcome hardware design limitations especially those in the reactive section. The side reactor can reduce the requirement of catalyst loading in the reactive section. The reactive section can be shortened, leading to potential reduction of the costs of the column if the amount of catalyst forces a decrease in the diameter or height of the column. The side reactor also offers convenient procedure for shut down operation and catalyst replacement.Control system design basically include· the selection of control algorithms and control schemes. The concepts of established linear control were initially applied for reactive distillation column studied in this thesis. The linear controllers with proper standard control schemes can be successfully implemented to avoid control problems associated with the presence of multiplicity and complex dynamics while achieving the control objectives. However, the nonlinear-multi variable nature of the reactive distillation combined with higher product competitiveness, tighter safety and environmental regulations would certainly increase the growing nonlinear control applications on reactive distillation. Therefore, advanced control algorithms such as adaptive controllers, which are based on a linear controller integrated with a tuning method, were considered. Firstly, nonlinear PI controller was designed allowing its controller gain to vary to accommodate the nonlinearity of the process gain. Secondly, model gain-scheduling controller was designed using multi-simplified models, which cover relevant operating conditions and cope with nonlinear characteristics, and a switching scheme to integrate the models. The results clearly show that the proposed adaptive controllers can improve the control performance for both set-point tracking and disturbance rejection. However, the model gain-scheduling controller may destabilise the reactive distillation due to fast switching resulting from large values of noise. An estimator is also recommended for controlling the primary controlled variable (product purity) especially due to changes in the feed composition. Multivariable inferential controller was then designed to overcome the limitations. The inferential models were constructed using steady state least square regression method for the estimation of the product purity and reactant conversion from multisecondary variables. The models were then implemented on reactive distillation using several control schemes employing standard linear controllers. The results indicate that a proper control scheme (cascade two-point control scheme) with an appropriate controller tuning can offer an alternative of model-based nonlinear controller for reactive distillation. Operability and optimisation of the existing reactive distillation column were investigated using set-point optimisation method. Despite a fixed column configuration (number of stages and feed point) and a fixed control configuration (pairing of controlled and manipulated variables), the operating conditions were optimised to maximise profitability while a set of safety, operating and product quality constraints and specifications are satisfied. The set-poiE~ optimisation framework was then used to provide a basis for a supervisory control system.