Advanced process control of distillation
The introduction of model predictive control enabled a bio-products plant to raise throughput and automate product variation.
HANS AALTO and STEFAN TÖTTERMAN
Viewed : 3124
Crude tall oil (CTO) is a by-product of pulp mills that can be further processed in tall oil distillation plants to provide valuable bio based chemical products such as rosin and fatty acids. CTO is normally distilled in vacuum columns to avoid high temperatures whereby undesirable side reactions will occur.
Forchem Oyj has recently invested in an advanced process control (APC) package with the aim of increased production of its plant in Rauma, Finland. The APC package chosen is a multivariable model predictive control (MPC) solution with optimising features such as maximisation of CTO feed while respecting multiple dynamically varying process constraints. The MPC application is equipped with product grade change logic because three types of rosin are produced, and because an automatic calibration application for product quality variables measured on-line is known to show some differences with respect to laboratory measured values.
The MPC application package was in guarantee runs demonstrated to be capable of keeping production at a level 8% higher than is reasonably achievable by manual controls. Product quality is more uniform and operation of the plant is more stable than before. Furthermore, the work of the control room operators has become much easier than before the installation of MPC.
One of the processing options for CTO is to separate tall oil rosins (TOR) and tall oil fatty acids (TOFA) in a distillation process. Tall oil products are produced all over the world and the world market in year 2014 was estimated to be 400000 t/y TOR and 380000 t/y TOFA.1 One of the biggest and also most modern tall oil distillation plants is owned and operated by Forchem Oyj. The plant has a maximum capacity of 200000 t/y CTO feed. Forchem recently finished an investment programme that targeted elevating the plant’s capacity to the number mentioned above with low capital expenditure investments. As one of those investments, Forchem wanted to consider whether on-line real-time optimisation (RTO) of the process could give some benefits. The operators were skilled but operating the process included many manual control actions, although most control loops were permanently on automatic mode. The grade changes, CTO total feed rate changes, overall material balance management, and finally product quality control were performed by manual adjustments of setpoints of flow and level controllers.
This article provides a short process description, followed by an outline of the control and optimisation strategy as well as some implementation highlights, and finally the results achieved.
Water is removed from CTO feed pumped from feed storage tanks by heating, followed by a pitch removal stage which is based on vaporising the CTO. Pitch is composed of heavy compounds that degrade the quality of TOR and TOFA and must be removed, but pitch has a (low) market value because it can be burned to produce energy. After depitching, the feed enters a distillation section, which at Forchem consists of four vacuum distillation columns. Vacuum distillation is used to keep the boiling points in the columns at lower values, because at high temperatures CTO, TOR and TOFA start to participate in harmful chemical side reactions such as degradation (decarboxylation) of TOR into lighter rosins which are not desirable in the TOR product and, on the other hand, formation of bigger rosin molecules (dimerisation) which are heavier and end up in the bottoms streams of the distillation columns instead of in the product streams. In the columns where TOFA concentration is higher, there is a risk of polymerisation at higher temperatures.
The first distillation column (see Figure 1) separates ‘crude TOR’ and ‘crude TOFA’, the latter being the lighter product and thus is removed near the top of the column. Crude TOFA is sent to the heads column which removes the heads oil that is partly used to dilute the pitch and partly burned in the hot oil furnace. The bottoms from the heads column is fed to the fatty acids column from which TOFA is drawn.
The bottoms of the first column is sent to the rosin redistillation column from which TOR is drawn close to the bottoms and distillate is drawn from a tray near the top. The bottoms of the redistillation column is pitch.
All columns are equipped with reboilers using hot oil as the heating medium. The circulating hot oil is heated up in a furnace that burns heads oil and ejector oil (oil from the vacuum system), so in terms of thermal energy the plant is self-sufficient.
Three types of TOR with different product characteristics (softening point) are produced in sequence, which means that product grade changes need to be performed. The grade change involves adjustment of the TOR softening point target and multiple adjustments of the plant’s overall material balance.
The qualities of crude TOFA rosin content – which is an impurity – and TOFA product rosin content are measured on-line by a near infrared (NIR) analyser. The softening point of the TOR product is measured on-line by density correlation. Density correlations are also available for other streams in the plant, including the CTO feed. The plant has a high degree of automation and is operated using an Emerson Delta V distributed control system (DCS) which uses fieldbus instrumentation.
Real-time optimisation of distillation
Real-time optimisation and model predictive control
RTO is traditionally placed as a separate application above MPC and/or DCS in the automation hierarchy. This is also true in cases where dynamic real-time optimisation (DRTO) is being used instead of the steady-state based RTO.2 MPC has always contained optimising features like pushing process variables towards constraints from which more optimal operation follows, and more can be done in order to achieve full-blown economic optimisation by MPC.3 MPC has nowadays become a flexible tool with a selectable degree of real-time optimisation of process economics. The application at Forchem is close to a standard MPC with field-proven constraint handling capability and flexibility to handle changing process conditions
It is highly desirable to study a plant by performance analysis prior to implementing MPC. The most important results of a performance analysis are estimates of the benefits achievable (in terms of profit increase, increased efficiency, decreased energy consumption, and so on) and a cost estimate of implementation. This should identify not only the costs directly related to MPC but also costs related to renewal of instrumentation, on-line analysers, DCS, and even process equipment. This valuable information reduces the risks of costly surprises during implementation. The performance analysis also typically reveals attractive small improvements, for instance in DCS control loops, which may provide a surprising benefit. Finally, the performance analysis should define the preliminary control strategy because this is also the basis of the MPC package’s cost estimate.
The performance analysis carried out for Forchem concluded that the process and instrumentation was in good shape and no other costs would be incurred other than the MPC package’s implementation cost. A process test performed during the performance analysis revealed that careful manual operation, where process constraints are logically relaxed, could give some 8% increased production. Another 8% production increase was estimated to be achievable by MPC.
Add your rating:
Current Rating: 2