Markus Duss, Sulzer Chemtech , firstname.lastname@example.org
Concerning the accuracy of simulating efficiencies, historically, when Sulzer developed structured packing, it had to guarantee not only the hydraulic behaviour but also the purity. Therefore, it built a pilot plant to measure not just test mixtures but also to carry out tests on real feedstock at design conditions. With these test results, a lot of information concerning HETP was accumulated. This holds true not only for distillation columns but also for many absorption and stripping applications.
On the other hand, structured packing allows for the description of the flow pattern for the liquid and vapour phase quite accurately. This does not hold for random packing, where the flow (vapour and liquid) is substantially more complex. A good example of the difference between structured and random packing is as follows. We can easily distinguish between X and Y types of Mellapak, where the angle of inclination is different. We can even pretty accurately predict the efficiencies any other angles. However, for random packing, all published correlations do not care about the actual geometry of the ring – all rings are treated more or less the same – which cannot be an accurate approach.
For trays, the correct description of vapour and liquid behaviour is very difficult, so reliable models to calculate HETP are not yet available.
Published correlations for kG, kL and the interfacial area are available for packings and trays. However, the systems measured to correlate these properties are predominantly organic test mixtures in distillation applications. Sulzer has built its own correlation for structured packings (Mellapak) and we strongly believe the accuracy is better than that found in public literature. We think the correlations are reliable for all our Mellapak types, not only for M250.Y, which was the basis for most correlations published in the literature. The basis for our proprietary correlation was the huge database available from pilot tests. Our model uses the penetration model for the liquid-side mass transfer and a Sherwood correlation for the vapour side. To model the interfacial area, we use an empirical correlation.
Since many physical properties are required to calculate the efficiency, Sulzer has developed its own program to calculate HETP, which is linked with a flowsheet simulator. We see that for most applications we can determine the mass transfer within 10% accuracy for the vapour side and 20% for the liquid side. Of course, there might be other issues that can influence the measured HETP, such as maldistribution. Maldistribution does not harm the mass-transfer coefficient or the interfacial area, but reduces "only" the driving force and the local L/V-ratio. However, the overall outcome is that the column might not perform as expected and the overall measured HETP is lower than calculated. Sulzer tries to take this into account by identifying the real root cause for this, which might be Reynolds numbers and physical properties such as surface tension, viscosity and density ratios of vapour and liquid. However, this approach is heuristic and we still try to develop a more rigorous way to account for maldistribution. A first step to judge the hydraulic behaviour will be published soon. We have not included any correlations for random packing yet, because all these correlations do not really consider the individual geometry, and the database to make one’s own correlations is not enough. For trays the situation is similar: correlations are available but the insecurity remains very high.
In summary, based on simulations carried out by Sulzer using SimSci’s PROII, we are in the position to make a “statement” about the HETP for structured packing (Mellapak). Furthermore, based on a hydraulic analysis, we can apply a heuristic method to judge, whether other phenomena (like maldistribution) might be an important consideration.
We do not give a statement about HETP for rings and trays, when no industrial experience is available.
We have built many columns based on the calculated HETP, but only for Mellapak. This includes many absorber and strippers, where mass transfer might be quite different from the typical HETP values measured with organic distillation systems. A good example is a MDI plant, where we have designed two phosgene absorbers, a phosgene stripper with Hcl, as well as an Aniline and water stripper. Some of the systems are unusual due to the high liquid viscosity (over 20cP). The operator built the columns without any pilot testing (and without prior experience) and the columns are fulfilling the task as required.
In general, all absorption systems are designed using the proprietary correlations, since the HETP depends on the sensitivity of the chosen operating conditions (stripping factor, pressure, temperature). Furthermore, we design all columns with high relative volatility using our proprietary correlation. Please note that for systems with a chemical reaction involved, a rigorous rate-based simulation approach is required. Therefore, we cannot calculate such systems, unless they can be clearly categorised as a vapour-side controlled system (with chemical reaction in the liquid phase). This holds for most systems with an instantaneous reaction.
It is important to note that the approach chosen to calculate HETP is based on the Chilton and Colburn (NTU-HTU-method). A so-called rigorous rate-based simulation is not carried out; instead, an equilibrium simulation is made and we then apply a rate-base calculation for the design components. This works well for almost all physical systems; however, you must be able to assign these design components. This is normally not a problem in absorption and stripping applications. In distillation applications, the design components are the light and the heavy key respectively. In refinery applications, with assay data, this might cause some problems.