Pilot-Scale Distillation Column

A pilot-scale distillation column located at the University of Sydney, Australia is used as the case study [60]. The 12-tray distillation column separates a 36% mixture of ethanol and water. The following process variables are monitored temperatures at trays 12, 10, 8, 6, 4, and the reflux stream, bottom and top levels (condenser), and the flow rates of bottoms, feed, steam, distillate and reflux streams. The column is operated at atmospheric pressure using feedback control. Three variables are controlled during the operation top product temperature, condenser level, and bottom level. Temperature at tray 8 is considered as the inferential variable for top product composition. To maintain a desired product composition, PI controllers cascaded on flow were used to manipulate the reflux, top product and bottom product streams. 

The column was operated four times at various operating conditions. The first three data sets corresponding to a total of 12.8 hr of operation were used to train the 0-NLPCA network, and the fourth one was used for model validation. However, prior to building a calibration model, both the training and the testing data were processed through the robust tandem filter to remove noise and suppress possible outliers. 

The 0-NLPCA network has 8-6-10-12 neurons in each layer, yielding a prototype model with 6 principal components (PCs). For comparison, the linear PCA was also applied to the same data. As a performance criterion, the root mean square of error (RMSE) was evaluated to compare the prediction ability of the developed PCA and O-NLPCA models on the training and validation data. While the linear PCA gave 0.3021 and 0.3227 RMSE on training and validation data sets, respectively, the O-NLPCA provided 0.2526 and 0.2244 RMSE. This suggests that to capture the same amount of information, the linear PCA entails utilization of more principal components than its nonlinear counterpart. As a result, the information embedded in the nonlinear principal components addresses the underlying events more efficiently than the linear ones. 

To define the NO region (NOR) of the piant, kernei density estimation (KDE) is used. The joint probabiiity density of the first and second, and 

To filter incoming process data, a window length of 500 samples was used. The window size is maintained constant by forgetting the first entry of the data vector and appending the new measurement vector to the end. To test the strategy for tracking sensor failures, a complete failure case 

---

The experimental facility is a pilot-scale distillation column connected to an industrial ABB MOD 300 distributed control system, which in turn is connected to a VAX cluster. The control system consists of a turbo node (configuration, history, console) remote I/O, and an Ethernet gateway, which allows communication with the VAX-station cluster through the network. This connection allows time-consuming and complex calculations to be performed in the VAX environment. Figure 10 shows the complete setup. 

Nooraii, A., Barton, G., and Romagnoli, J. A. (1994). On-line data reconciliation, simulation and optimising control of a pilot-scale distillation column. Process Syst. Eng. PSE 94 2, 1265-1268. 

To address this concern, one needs to have some sort of experimental procedure that allows one to measure column profiles at finite reflux. One could decide to erect a pilot scale distillation column to do this, but this has the drawback that it may be difficult to start-up and control, potentially wasting valuable engineering time without getting the needed insight, not to mention the huge costs involved. In... 

Consider the following empirical model of a pilot-scale distillation column (Wood and Berry, 1973)... 

Supercritical fractionation of a liquid lipid feed material is usually carried out in a packed column. Standard columns are not available commercially and have to be custom built either in-house or by manufacturers of extraction units. Lab-scale and pilot-scale supercritical columns, 0.6-13.6 m high with internal diameters of 14.3-68 mm are available in research labs around the world and have been used for the processing of deodorizer distillates (56, 57, 86-90), vegetable and fish oils (91-105), cocoa butter, and milkfat (106-109). A schematic diagram of a typical SCCO2 fractionation column (2.8 m, 2.54 cm o.d.), which was designed and built... 

RGA Example In order to illustrate use of the RGA method, consider the following steady-state version of a transfer function model for a pilot-scale, methanol-water distillation column (Wood and Berry, Terminal Composition Control of a Binaiy Distillation Column, Chem. Eng. Sci, 28, 1707, 1973) Ku = 12.8, K = -18.9, K. i = 6.6, and Koo = —19.4. It follows that A = 2 and... 

Bench-scale tests were conducted to determine design parameters and operability of Che distillation process for the pilot plant. The column, reboller, and condenser used in the bench-scale system were constructed of fiberglass-reinforced ethylene-chlorotrifluoroethylene. The heat exchangers were spaghetti tube bundles constructed of fluorinated ethylene-propylene copolymer. 

In the mid Eighties, 99 batch processes within 74 UK companies were identified (Parakrama, 1985). In the last decade or more a continuous shift towards batch processes has been noticed and many small-scale companies are using their existing continuous columns for batch distillation (Willet, 1995) without much realising the implications of such practice. Also R D sections of many multi-national chemical companies do the pilot plant study in CBD columns for their continuous distillation columns which are in operation in the plant (Chen, 1995 Jenkins, 2000 Greaves, 2003). Because of confidentiality the results of such studies are not available in public. 

Aittamaa (1981) simulated a number of experiments with the systems ethanol-benzene-n-heptane, and chloroform-benzene-n-heptane (data were obtained at Hoffmann-La Roche in a fair size pilot scale column) and 1-butanol-ethanol-water in a 12 sieve-tray column. The Hoffman-La Roche data were taken in a column having 24 sieve trays and 30-cm inside diameter. Unlike many studies of distillation efficiency, these experiments were not carried out at total reflux. The measured flow rates and compositions of the feed, distillate, and... 

The best known and most used empirical method is that of O Connell (Figure 12.60), for distillation columns and absorbers. The curves are based on plant data for several bubble-cap columns plus a few pilot-scale units. Efficiency is related to two properties of the feed mixture liquid viscosity and relative volatility a. Higher values of the p a product indicate larger liquid-side mass transfer resistance and hence a lower efficiency. For a vapor feed or a mixed vapor-liquid feed, the correlating viscosity should be that of the feed tray liquid. 

Analyse the feasibility of different control structures by a controllability analysis for a distillation column methanol-water. Consider the data of the EboAkademi pilot-scale column (Haggblom Waller, 1992). 

When chromatographic separations (7) are operated in a batch mode, a portion of the mixture to be separated is introduced at the column inlet. A solute-free carrier fluid is then fed continually through the column, the solutes separating into bands or zones. Some industrial operations such as mixed-vapor solvent recovery and sorption of the less volatile hydrocarbons in natural gas or natural gasoline plants are being carried out on pilot plant and semiworks scales. Continuous countercurrent systems designed along the basic principles of distillation columns have been constructed. 

Figure 10.17 represents a comparison of the concentration profiles in two different ethyl acetate synthesis modes, with and without a decanter. The investigation is performed for the pilot-scale column, with a molar feed ratio acetic acid/ethanol equal to 1.2, reflux ratio equal to 3, and a total feed rate equal to 30kg/h. The distillate-to-feed ratio is set to 0.9. The simulations reveal that the conversion with the liquid-liquid separator is about 5% higher that without a decanter, since there is less water and more acetic acid in the catalytic section. Improved conversion and product enrichment due to liquid-liquid separation result in a significant (29%) improvement of the product purity. Finally, because there is less condensed water in the reflux to be evaporated, the heat duty is reduced by up to 26%. 

My own technical experience has pretty much followed this history of distillation simulation. My practical experience started back in a high-school chemistry class in which we performed batch distillations. Next came an exposure to some distillation theory and running a pilot-scale batch distillation column as an undergraduate at Penn State, learning from Arthur Rose and Black Mike Cannon. Then, there were five years of industrial experience in Exxon refineries as a technical service engineer on pipestills, vacuum columns, light-ends units, and alkylation units, all of which used distillation extensively. 

Niesbach, A., Fuhrmeister, R., Keller, T., Lutze, P., Gorak, A. (2012). Esterification of acrylic acid and n-butanol in a pilot-scale reactive distillation column-experimental investigation, model validation, and process analysis. Industrial Engineering Chemistry Research, 51, 16444-16456. 

In this paper, we focused on the relevance of the commercial software HYSYS for the simulation of catalytic distillation columns. As in the current version of HYSYS the built-in RD module is not directly suitable for the simulation of the heterogeneous catalytic distillation process, this study is concentrated to develop a model for heterogeneous RD and to implement it in the HYSYS simulation environment. The objectives of this work are to develop a suitable simulation module for heterogeneous reactive distillation compatible with HYSYS and to apply it to an intermediary scale pilot plant unit. 

Shah verified the effectiveness of the nonlinear estimator by both simulation studies and experimental tests on a pilot-scale column. The work of Weber and Mosler should also be noted. They proposed several control schemes using the sum of two or more temperatures, the difference between two temperatures, and the distillate-to-feed ratio to adjust manipulated variables. 

The CWS obtained the clue which led them to vacuum distillation in November 1943 when Capt. J. W. Eastes visited the University of Illinois to confer with NDRC chemists. He learned that they had distilled at low pressure mustard which had been washed with water, and that the temperatures in the distillation column indicated that fairly pure jS,/ -dichloro-ethyl sulfide could be prepared in this way. In other words, water removed certain impurities, and distillation removed the remainder. The CWS had investigated vacuum distillation earlier, but had never washed the crude mustard before distilling. The Technical Division investigated the process and found that it produced a purer and more stable j8,j8 -di-chloroethyl sulfide than the other methods and that it was quite practical so far as apparatus was concerned. A pilot plant was first set up and then a full-scale plant. In 1945 the service switched to the new process at Edgewood and at Rocky Mountain Arsenal. By the end of the year 9,218,-357 pounds of distilled mustard (symbol, HD) had been produced. With the successful production of HD, production of the old Levinstein mustard was halted. 

---