Friday, March 28, 2008

Role of SMB Chromatography Process Simulator

1. Chromatography as a manufacturing process

Traditionally a chromatographic separation refers a measurement technique rather than a manufacturing process. Concerning the driving force that enables a separation, it is achieved due to the different rates of migration for pure substance in a mixture – e.g. the smaller the affinity a molecule has for the stationary phase, the shorter the time spent in a column. Such different migration rates could be explained by various mechanisms in between molecule (mobile phase)–adsorbent (stationary phase) and molecule–molecule interactions. The followings are well-known interaction mechanisms driving a chromatographic separation:

- Elution chromatography
- Displacement chromatography
- Ligand-exchange chromatography
- Size-exclusion chromatography
- Ion-exclusion chromatography
- Reverse phase chromatography

It is not a surprise that we can see various chromatography applications from industry. High purity product is crucial for certain industry, for example, Life Sciences, and chromatographic separation guarantees very high product purity. However it should be noted that much of what is written about chromatographic separation as a process operation is somewhat misguided and written from an analytical measurement perspective.

Perhaps the most popular chromatographic separation practise in a research laboratory is analytical chromatography such as LC (Liquid Chromatography) or GC (Gas Chromatography). The analytical chromatography system consists of three major parts, injection, column, and detector. For LC system, depending on the injection method, the operation is classified as either isocratic elution (constant mobile phase concentration) or gradient elution (continuous change in mobile phase concentration). While analytical chromatography focuses on the measurement, the main goal of industrial chromatography is achievement of high purity and recovery of target product, and high throughput. Regardless of any difference in objectives, there is actually no difference in terms of separation principle for both applications. For instance, preparative elution chromatography is one of the most popular industrial applications (see Figure 1) because of its familiarity as an analytical approach.
In the realm of preparative and industrial chromatography, Simulated Moving Bed (SMB) chromatography has received continuously increasing attention since its introduction in the early 1960s. The SMB technology involves the simulated counter-current contact between the mobile phase and the stationary phase, which is the most efficient in terms of separation performance, solvent (eluent) consumption and productivity per unity mass of stationary phase. This can be accomplished in units constituted of a set of fixed bed chromatographic columns like that illustrated in Figure 2.



The SMB technology was originally developed for very large-scale applications in the petrochemical industry, such as the separation of para-xylene from alkyl aromatic C8 fraction. After its first commercialization, SMB application has been widely expanded to food and pharmaceutical, and nowadays to nearly all Life Sciences industries, because of its high profitability as a manufacturing process.

2. Computer Modelling of Chromatographic Separation Process

Perhaps the most basic requirements to model chromatography are the ability to characterize equilibrium and mass transfer. Also solving mathematical model for chromatographic separation requires good robustness and high numerical resolution due to the complex nature of underlying hyperbolic PDE system. It has been reported that simple finite differencing schemes (e.g. backward/forward differencing schemes) are not good enough to model a chromatography column because of the sharp front in concentration profile along the column. Therefore a simulation tool to model chromatography system should offer advanced PDE discretization methods, such as Orthogonal Collocation on Finite Elements.
Chromatography applications often involve sequences of operating schemes in boundary conditions to achieve good separation. This can be seen in Figure 3, which shows a gradient elution operation with a modulator, a common chromatography application in Life Science. Thus a good modelling tool should offers capabilities to handle dynamic events and quickly defined and edit operating scenarios.
On the other hand, the SMB technology is rather complex and deep understanding is needed to use it effectively. This suggests that a general model aimed at performing a parametric analysis of SMB behaviour would be very useful. However, several attempts in academia have shown that a more synthetic approach is required to guide the choices of parameter values, or such a parametric study becomes computationally intractable. This can be achieved by assuming the system is governed by equilibrium theory where mass transfer resistances and axial dispersion are neglected. In particular, assuming a Langmuirian type adsorption isotherm yields the so-called “Triangle Theory”, which helps determination of optimal and robust separation conditions. Figure 4 shows Aspen Chromatography’s user interface for applying the Triangle Theory to determine SMB operation conditions.
The Triangle Theory serves a reliable starting position for SMB process synthesis. Unfortunately, however, industrial SMB process often shows highly non-ideal behaviour which cannot be well explained by the ideal assumptions inherent in the Triangle Theory. These non-ideal effects in industrial scale SMB system are believed to be mainly due to (1) highly non-linear equilibrium (non-Langmuirian); (2) dead volumes which affect distribution in the system; (3) complex valve control methodology. In addition, recent progress in SMB process design, such as VariCol, PowerFeed, and ModiCon, revealed a gradient SMB operation is more profitable than the classical 4-zone SMB process under isocratic conditions. Without a doubt, more sophisticated operating regimes requires more solid mathematical framework in SMB modelling. And accordingly the Triangle Theory is being challenged to improve its capability for such complicated operations. Commercial SMB modelling tool, e.g. Aspen Chromatography®, is required to meet various demands from various industries. Key requirement items, as a reliable SMB process model, may be categorized as follows:

(1) Functional
- Rigorous chromatography column dynamics
- Ease of SMB zonal configuration
- Controllable valve stepping and port switching/shifting
- Rigorous internal submodels for delivery pipes and hold-ups
- Reactive SMB operation
- Modulation of feed rate
- Modulation of feed composition
- Asynchronous port switching
(2) Numerical
- High resolution PDE method and solid integrator
- Robustness in linear and non-linear solver
- Ease of dynamic event /scenario editing and tasking
In addition, a good modeling software should be able to be utilized for not only process operation but also design parameter estimation, and Table 1 shows the application of commercial chromatography model at different user level.
A screening process in order to determine either good solvent or good packing material is quite tedious job in R&D activity. Probably so called “Trial & Error” method may be a dominant rule of thumb for a screening process, however it requires many routine works to get result and even after result acquisition it is difficult to say one completed an optimal screening. Mathematical model based on the first principle guides logically and physically valid path for such instance and moreover one can be freed from a dilemma and remove a doubt about result. Very often, mathematical model helps one to have important design parameters by means of steady-state or dynamic estimation. For instance, one can estimate equilibrium isotherm parameters or kinetic information by using limited experimental data or real plant data. Since there are no universal model in adsorption equilibria and adsorption kinetics, these two design information (equilibrium and kinetics) are normally in veil at the beginning, and therefore in general they are supposed to be collected from direct measurement, and it is truly a rate limiting step in any design or revamp work.
Nowadays the enhancement of computation performance together with reliable data handling enables commercial software to equip very attractive features, especially for dynamic estimation. Using in-built user interface that designed for dynamic estimation, one can utilize real process model to extract design data from their experimental or plant data. Figure 5 is a good example for dynamic estimation by Aspen chromatography®. Against chromatogram data that measured from pulse injection operation on a pilot scale chromatography column, the dynamic estimation extracts the best matching equilibrium parameters with respect to the chromatogram data. Since this type of estimation work can be carried out against real plant data, the estimated parameters or information can provide more realistic insight on the process. In addition, such estimation feature can be applied to expose mass transfer and heat transfer mechanisms which may be affected by equipment geometry and operating conditions.
3. Industrial SMB Processes and Simulation
Industrial SMB processes can be classified with the following categories:

(1) By Industry
- Refining / Petrochemical
.UOP Sorbex™ processes (Parex™, Molex™, Ebex™, Olex™, etc)
.IFP p-Xylene SMB
.Cresols separation
- Food
.Sugars
.Proteins
.Etc
- Pharmaceutical / Life Science
.Antibiotics
.Insulin
.Chiral compounds .Pure enantiomers
(2) By Valve Stepping Method
- Rotary Valve: Simultaneous port switching only
.Single valve operation
.Less problematic in valve maintenance
.Poor or zero flexibility in operation
.For example, UOP Sorbex™ processes
- Sequential Valve: Flexible port switching available
.Multiple valves operation
.Periodic port shifting is possible (e.g. NOVASEP VariCol™ process)
.More problematic for maintenance than rotary valve
.For example, IFP p-Xylene SMB, Nearly all Life Science application

(3) By Equipment Style (see Figure 6)
- Multiple Columns in Series
.Mostly in pharmaceutical/life science application
- Carousel type
.Some application in food industry
- Single/Twin Stacks
.Refining/Petrochemical industry
Figure 6 Classification of Industrial SMB Process by Equipment Style
In general, industrial SMB separation units are operated with 8 – 36 ports, i.e. adsorption columns. Process performance is highly dependent on the physical property of packing adsorbent (or stationary phase) and desorbent (or so-called eluent or solvent), and typically process licensor keep details about the packing adsorbent a closely guarded secret from their licensee. In addition, industrial SMB process tends to use complex control method about valve stepping and throughput handling, in order to keep the process performance at optimal. Such a complex operation method often introduces a strong perturbation in SMB system therefore the actual operation is a little far away from the ideal SMB principle. By the reason, mathematical modelling and simulation for industrial SMB must require very solid features to meet the operational principle as well as the effects from equipment surroundings including process control. That is the reason why only a general model based on the first principle can properly model industrial SMB processes.

Figure 7 shows simulation interface for a pilot scale SMB process for amino acid separation. The process operates with 12 columns connected in series, each with an approximate volume of 500cm3 and packed with Chirosolve L-Proline with an overall voidage of 0.45. Figure 7 demonstrates the full flowsheeting capability of Aspen Chromatography® by explicitly placing 12 column models with mixers, splitters, pumps and feed and products diverters, all fully interconnected. A simulation scenario editor, so-called Cycle Organizer, is used to define dynamic event schedule and execute the switching cycle.
Figure 8 shows simulation result from a production scale 24 layer SMB unit for p-xylene separation, which separates a mixture of ethylbenzene (EB), m-xylene (MX), o-xylene (OX) and p-xylene (PX) using a para-di-ethylbenzene (PDEB) as desorbent. The SMB unit includes transfer line volumes that represent 1% of the total SMB unit volume. Due to a little extraordinary operation method such as line flushing with reflux-type recycling and additional desorbent feed, this system shows extra 3 zones hen making 7 zones SMB operation. Note that this example process and its process data are collected from open publications [1,2].
4. Remarks

Chromatographic separation is being received greater attentions from various industries due to the attractive characteristics. But at the same time, the related service company and process design company are being challenged by the complexities of the principle as well as the operation. Meanwhile operation companies are also being suffered by lack of resource to understand the process nature, to improve process performance and to maximize profit. In engineering point of view, good software tool acts like a good engineer. Utilizing good software we can minimize risks at every stage in process design and the optimisation.

Reference
[1] Minceva.M. and Rodrigues.A.R., “Modeling and Simulation of a Simulated Moving Bed for the Separation of p-Xylene”, , Ind Eng Chem Res, 41, 3453-3461 (2002)
[2] Minceva.M. and Rodrigues.A.R., “Influence of the Transfer Line Dead Volume on the Performance of an Industrial Scale Simulated Moving Bed for p-Xylene Separation”, Sep Sci & Tech, 38, 7, 1463-1497 (2003)

Process simulation in chemical engineering

The process industries are faced with an increasingly competitive environment, ever-changing market conditions, and government regulations, yet they still must increase productivity and profitability. These business objectives are achieved by reducing the time required to get new products to market, increasing the quality of product produced, operating the plant more safely and efficiently, and/or designing plants for optimum performance along their life cycle. In the process industries, operational efficiencies, production economics, product quality, and, ultimately, bottom line performance can be adversely affected by a multitude of factors. Many of these factors are extremely complex and are subject to varying degrees of unpredictability. To avoid production delays, downtime, or off-spec products, process manufacturers require cost-effective tools that help to identify and correct anticipated problems before they occur. Process engineers routinely address difficult manufacturing and production issues.

Unfortunately, experience alone is not always sufficient to answer the questions that continually arise, and trial and error efforts to provide meaningful insight are usually cost-prohibitive. Process simulation is recognized as a powerful tool that helps managers and engineers link business operations to process operations. Process simulation specifically enables resolution of common issues such as ensuring safety, ensuring environmental compliance, improving operability, improving start-ups and shut-downs, achieving consistent product quality, troubleshooting operational problems, and optimizing batch process operations.

We, chemical engineers, are often faced situations that requiring conceptual or detail modeling about chemical or bio-chemical processes. We want to suggest a best solution or at least a useful workaround to get closed to the best solution, thus a smart modeling method and also having right simulation tools are very important.

I wish I could make this blog as a valuable open resource for all of us, the Chemical Engineers.