LUBE OIL DISTILLATION REVAMPING AND HETP EVALUATION

LUBE OIL DISTILLATION REVAMPING AND HETP EVALUATION a Machado, R. S. 1 ; b Orlando Jr., A. E.; c Medina L. C.; d Mendes M. F.; e Nicolaiewsky, E. M. A. a UFRRJ PPGEQ b PETROBRAS TRANSPORTE S.A. c CENPES/PETROBRAS d DEQ/IT/UFRRJ e Centro de Tecnologia Bloco E Sala 211 Cidade Universitária Ilha do Fundão Rio de Janeiro/RJ ABSTRACT This paper deals with experimental tests in a lab-scale distillation column and the Height Equivalent to the Theoretical Plate (HETP) evaluation using two distinct base lube oil mixtures, a heavier one composed of neutral medium and bright stock (NM-BS), and another composed of spindle and neutral light (S-NL). Simulation studies using the PRO II software have been performed in order to establish the best operating conditions in the distillation column. The feed and products analysis have been performed by CENPES/PETROBRAS and successfully compared with the simulation results. Some models for HETP evaluation, both empirical and theoretical, were chosen for comparison. The results have shown that empirical models give unrealistic HETP values when pressure is lower than 30 mbar. The theoretical models show two situations: lower values for effective interfacial area of packed columns with higher deviations for HETP or correct effective area, but are not fit for lube oil column performance analysis. KEYWORDS base lube oils; HETP; laboratory columns; revamp 1 To whom all correspondence should be addressed. Address: M.Sc. student UFRRJ PPGEQ BR 465 km 7 E-mail: rdmach@gmail.com 35

1. INTRODUCTION Column performance evaluation is a matter of great importance for the design of structured packed columns, either new or revamped, especially when operating with heavy feedstocks. As lighter crude oils have become scarcer worldwide, research for new technologies should be encouraged in oil refining. The aim of this work is to shed some light on base lube oil distillation in a laboratory column in order to establish the best conditions to process heavier feedstock (Brazilian crude oil), concerning Group II lube oil hydrocracking route. In Brazil, lube oil is produced in two refineries, by liquid-liquid extraction, a technology which is dependent either on light paraffinic crude oil (from Bahia, whose reserves are very low) or on imported crude oil. If Brazil wants to praise its independency from crude oil price fluctuations, lube oil processing should be conducted via the hydrocracking route, which can use very heavy feedstock, mostly encountered in deep waters, as in Brazil. With the objective of studying Group II base lube oil separation, concerning the hydrocracking route for Brazilian heavy crude oil processing, a laboratory scale distillation column was designed and constructed (Nicolaiewsky et al., 2002). The column is presently located at the Laboratory of the Department of Chemical Engineering, at Escola de Química of Universidade Federal do Rio de Janeiro (LADEQ EQ/UFRJ). This project was sponsored by PETROBRAS and FINEP. In its first version, the column (40 mm nominal diameter and 2.2 m high), filled with Sulzer DX gauze metallic packing, operated continuously, with great flexibility in terms of feed position and pressure, ranging from atmospheric to 10 mbar. During the first assays, the heavier mixture (Neutral Medium and Bright Stock NM-BS) was processed. However, as the column had been designed to process a hydrocracked feedstock, which was not available at the moment, the column was not able to satisfactorily process NM- BS oil mixture. Therefore, in order to fulfill the objectives of the project, the distillation unit had to undergo several modifications (revamp), which will be described later. After revamping, the lighter mixture (Spindle and Neutral Light S-NL) was processed. Prior to the experimental tests, simulation studies have been performed using the PRO II software, enabling to establish the best operating conditions in the distillation column. After the experimental tests, all the products were analyzed by simulated distillation (chromatographic analysis HT-750), also performed in the Petroleum Evaluation Laboratory (PEL) of CENPES/PETROBRAS. The agreement between the experimental and the simulation curves indicates that the operating conditions established by the simulation studies are correct and that the output data from the software reports could be used on further calculations, as HETP (Height Equivalent to theoretical Plate) evaluation. Concerning column performance evaluation, two approaches were chosen: a more practical one, dealing with empirical models and suppliers curves, and a more theoretical one, regarding mass transfer models. In the first category, models proposed by Lockett (1998) and Carrillo et al. (2000) were chosen. In order to represent the mass transfer models, the model proposed by Olujić et al. (2004) and its modified version were investigated. 2. PREVIOUS WORK Lockett (1998) has proposed a simplified relation for HETP evaluation, based on the rigorous correlation reported by Rocha et al. (1996), although disregarding diffusivities and packing geometrical characteristics. Carrillo et al. (2000) have modified Lockett s equation (Lockett, 1998), introducing the effect of pressure and the F-factor, testing other types of structured packing besides Flexipak, which was previously tested by Lockett (1998), and also using Sulzer BX similar to the gauze packing of a QVF column (Sulzer DX). 36

Among the rigorous models, the one proposed by Olujić et al. (2004) and an adaptation of that model proposed by Orlando Jr. (2007) were used, in which the effective area was calculated by using Rocha, Bravo and Fair s correlation (Rocha et al., 1993; Rocha et al., 1996). That modification had been tested by Orlando Jr. (2007) and Orlando Jr. and coworkers (2009), when operating the same distillation unit, although employing a different and lighter mixture, from a kerosene blend. Additionally, the former model by Bravo et al. (1985), with the modifications proposed in their work, in terms of effective area calculation, has achieved the best performance (8% average deviation), still remaining a good choice for gauze packing HETP evaluation. Orlando Jr. et al. (2009) have tested several theoretical models and Carrillo et al. (2000) empirical correlation for HETP evaluation. The authors values obtained with the empirical correlation were much better than the ones using the theoretical models (only 12% average deviation). That empirical correlation has the advantage of not requiring any calculations of mass transfer coefficients and effective area. The effective interfacial area is a parameter of great influence in HETP evaluation. In their work, by comparing models from University of Delft (Netherlands) and University of Texas at Austin (model developed by Fair s group, which will be referred to as the RBF Model), Fair and coworkers (2000) concluded that the Olujic et al. (2004) model overestimated the effective areas. Therefore, aiming to correct the effective superficial area, Olujic et al. (2004) have adapted their model using Onda s correlation (Onda et al., 1968). In the present work, a comparison has been conducted between Olujić and coworkers original correlation (Olujic et al., 2004), and their wetted area evaluation based on Onda s correlation (Onda et al., 1968) and the effective area calculated by the RBF Model (Rocha et al., 1993; Rocha et al., 1996). Moreover, the results from the empirical models will be also compared with the experimental results and with the values claimed by the packing supplier. 3. SIMULATION STUDIES As mentioned before, simulation studies have been performed using the PRO II software, in order to establish the best operating conditions in the distillation column. Those studies were conducted for both lube mixtures: NM-BS, the heavier mixture, tested before the column revamping and S-NL, the lighter one, after the column modifications. One of the input variables in PRO II software was the feed characterization: for the heavier mixture (HT-750 curve, %weight analysis) and for the lighter one (ASTM D1160, % volume), both performed at PEL (see Appendix, Table A1). PRO II output reports show results in several forms, which can be compared with the original curves given by CENPES/PETROBRAS. Again, the comparison between the simulated and the analytical curves will tell whether the simulations were correct or not. 3.1 Before Revamping NM-BS mixture The simulation studies with the NM-BS mixture before revamping were performed by Orlando Jr. (2007), who used parameters like reflux ratio, operational pressure, feed location, temperature and flow rate of the feed as input variables, using PRO II software. The Soave- Redlich-Kwong (SRK) equation of state was used, as it describes satisfactorily the thermodynamic behavior of hydrocarbon mixtures. The simulation plan for the NM-BS mixture can be observed in Table A2, in Appendix A. In total, Orlando Jr. (2007) performed 24 simulations by varying the parameters mentioned above, and the maximum deviation between the simulation results and the original products curves (HT-750) was 10%. In Figure 1, results are shown for two simulations, cases 4 and 5, in which there was good agreement (only 5% deviation). 3.2 After Revamping S-NL Mixture Whilst the unit was being revamped, simulation studies with a lighter S-NL mixture 37

Figure 1. Comparison between NM-BS HT-750 curve and simulation results using PRO II. were performed, also using the PRO II TM software, already taking into account the modifications which are described in next section (Experimental Tests). As in NM-BS simulations, characterization curves were needed as input variables. However, when testing S-NL mixtures, the ASTM D1160 curve, also called TBP curve, was used instead of the HT-750 curve. The simulation plan for the lighter mixture (S- NL) can be found in Table B1, in Appendix B. The most influential variable for the S-NL mixture separation was the reflux ratio, as can be observed in Figure 2. From the simulation results, it was shown that a reflux ratio of 0.5 would be Figure 2. Influence of reflux ratio (RR) on S-NL separation, according to simulation results. 38

the best condition. However, that condition was tried experimentally, when the QVF column had been dried off. 4. EXPERIMENTAL TESTS This section is divided in two parts. In the first one, the tests performed in the QVF distillation unit before revamping, using the NM-BS mixture, will be described. The second part deals with the tests after the revamping unit, with the S-NL mixture. A) Before revamping A.1) Distillation column description The distillation unit has been designed and constructed by QVF, a German company, subsidiary of De Dietrich. The design and optimization of the distillation unit were part of a large project with PETROBRAS, concerning highquality lube oil production. The column and the reboiler were built in stainless steel, but the containers for feed and products and the condenser were made of borosilicate glass. The QVF distillation unit before revamping is shown in Figure 4. The unit is controlled by the WinErs software, which enables start up conditions (total reflux) and continuous operation. Moreover, the WinErs software was able to control the operating pressure and the temperature of the top product, as well as the reflux ratio. More technical The experimental tests have been performed in a laboratory distillation column, with a 40 mm nominal diameter, having 4 sections of 550 mm each, containing Sulzer DX (gauze) structured packings, as the contacting device. Figure 3 shows the Sulzer DX packing and its geometric characteristics are presented in Table 1 (Orlando Jr., 2007). Figure 4. Picture of the Sulzer DX packing. Figure 3. Picture of the distillation unit before revamping from QVF Engineering GmbH. Table 1. Geometric characteristics of the Sulzer DX structured packing. Corrugation height (h) 2.9 mm Corrugation base (B) 6.4 mm Corrugation angle 60 Specific surface area (a p ) 900 m 2 /m 3 Porosity 93.7% Height of packing element 0.055 m Recommended diameter range 30-125 mm Reference: Orlando Jr. (2007) 39

specifications of the distillation column and its controlling system can be found elsewhere (Orlando Jr., 2007; Orlando Jr. et al., 2009). A.2) Experimental tests with the NM-BS mixture Experimental tests were performed by Orlando Jr. (2007) with the simulation conditions established previously. From the 24 simulations, the author picked 16 different operational conditions to test the QVF column. In order to be used for performance evaluation, some of those data were chosen to be analyzed in the present work, especially the ones which had shown good agreement between simulation and experimental results (less then 16% average deviation for the bottom product, for example). However, when comparing the experimental bottom product results with the original BS HT- 750 curve (%w), one can observe that the former is lighter in all points, showing that there had been contamination of the bottom product with the feed stream, on the experiments performed by Orlando Jr. (2007), before the QVF distillation unit revamping. It was then decided that the unit needed revamping, in order to continue with the whole project. Therefore, the German company s office in Brazil (De Dietrich of Brazil) was in charge of the project this time. The top product (Neutral Medium) has not shown good agreement with the simulation results, mostly because the column never reached equilibrium. Consequently, the top product collected is much lighter than the expected in all assays, as can be observed in Figures 5 and 6. The bottom product, as already mentioned, has undergone contamination with the feed, also producing a lighter bottom product than the original BS. The main objective of the lube oil laboratory unit revamping, besides solving the experimental limitations of the QVF column, was to add a side stream withdrawal in the distillation column enriching section. The aim of such modification was to assure flexibility, not only in terms of base lube oil separations but also concerning heavy feedstock vacuum distillations. B) After revamping B.1) Distillation Column Description In order to operate with heavier feedstock, some modifications were necessary, such as: a) Substitution of the solenoid valve (stainless steel): the original valve was not made of borosilicate glass, therefore it could not endure Figure 5. Comparison between NM-BS HT-750 curve, simulation and experimental results (case 4). 40

Figure 6. Comparison between NM-BS HT-750 curve, simulation and experimental results (case 5). higher temperatures; b) New bottom product withdrawal: originally, it was collected from the reboiler (partial collection), mixing up with the feed. Then, with the new collector, the reboiler turned into a total device; c) New side stream in the middle of the enriching section: when collecting a side stream product, that feed entrance cannot be used. In Figure 7, one may observe the differences between the QVF unit situations, before and after revamping. The column before revamping (2.2 m high) was simulated by Orlando Jr. (2007) as having 16 theoretical stages, in order to process the NM-BS mixture, resulting in an experimental HETP of 0.138 m. For the lighter mixture (S-NL), the best number of equilibrium stages for the separation was 18, for the same column (before revamping). (a) Figure 7. Comparison between the QFV distillation unit: (a) before and (b) after revamping for simulation studies using PRO II. (b) 41

After revamping, due to the new bottom product withdrawal, there was a loss of 0.2 m packing height, which reduced the number of stages to 17. Thus, the experimental HETP, after revamping, turned to 0.138 m for the top section and 0.129 m for the bottom section. B.2) Experimental tests with the S-NL mixture After the modifications (revamp) described above, tests with the lighter S-NL mixture have been conducted mostly to evaluate how the column was going to perform after revamping. Prior to the tests, some time had to be spent in order to get the unit in shape in terms of leakage problems, in order to reach the desired vacuum. After the introduction of two side withdrawals, the column head loss had increased so much that the vacuum pump was not able to reduce the pressure to the desired levels. During the tests, the column differential pressure raised to 88 mbar. Prior to revamping, that was around 8 mbar. Another more robust vacuum pump was acquired for the project. With both pumps working in series, the pressure inside the column could reach 1 mbar, but the second pump could not be controlled by the WinErs software. Therefore, it was not useful for research purposes. Working with only one pump, the best vacuum achieved was 50 mbar, due to the higher head loss, even when all the leaking problems were solved. As a consequence, only lighter lube oil mixtures could be tested. The choice was then made for the S-NL mixture. After revamping, the QVF column did not perform well during the experimental tests. One of the hugest problems was the solenoid valve for the reflux divider. The stainless steel device was much too heavy to be worked on the original automation system. There were many trial-anderror attempts to fit the reflux divider into place, but it did not work properly. Therefore, the total reflux condition was never properly observed, after the unit had been revamped. Only three assays have been conducted with the S-NL mixture, with reflux ratios of 0.5, 2 and 3, and operational pressure of 50 mbar. Those conditions were chosen because they had provided the best simulation results. Unfortunately, since total reflux conditions were never properly attained in the distillation column, there is no certainty about the results. As expected on observing Figure 8, the experimental Figure 8. Comparison between experimental curves and simulations results by PRO IITM. 42

results turned out very different (much lighter) from the simulated curves, due to the operational problems mentioned before. This is an example of an unsuccessful revamp study case, which is very rarely described in the literature. 5. RESULTS AND DISCUSSIONS As mentioned before, concerning column performance evaluation, two approaches were chosen: a more practical one, dealing with empirical models and suppliers curves, and a more rigorous one. However, before choosing models to work with, it is necessary to establish agreement between the simulation results and the original characterization curves given by CENPES/ PETROBRAS. Only after the points have matched, one should proceed with the experimental steps and then with the modeling stage. The experimental part does not give all data and parameters necessary for HETP calculations, at least not for the theoretical models. Therefore, one has to use some of the data available from the simulation output reports. This is why the simulation results have to match not only the original characterization curves but also the experimental curves, as well. In Tables C1 and C2, examples are shown for the simulation deviations for both mixtures (case 4, with NM-BS, and case 8, with S-NL). It is not easy to model and describe complex mixtures like base lube oil products; however, this was achieved by the pseudo-components properties, generated by PRO II library. Concerning the empirical models, a comparison between those proposed by Lockett (1998) and Carrillo et al. (2000) was conducted for the NM-BS mixture and the results can be observed in Tables 2 and 3. The same calculations were carried out for the S-NL mixture and the results are shown in Tables 4 and 5. When analyzing those results, it is evident that Lockett s correlation (Lockett, 1998) remains a good choice among the empirical models to Assay Table 2. Deviation between experimental (NM-BS) and Lockett (1998) correlation results. EXP HETP = 0.138 m EXP HETP=0.129 m HETP (m) LOCKETT HETP AVG AVG % 1 0.129 0.094 0.110 6.52 31.88 20.00 2 0.132 0.097 0.113 4.35 29.71 17.82 3 0.141 0.092 0.114 2.17 33.33 17.09 4 0.141 0.088 0.112 2.17 36.23 18.55 5 0.141 0.093 0.115 2.17 32.61 16.36 6 0.100 0.070 0.083 27.54 49.28 39.64 7 0.102 0.070 0.085 26.09 49.28 38.18 Assay Table 3. Deviation between experimental (NM-BS) and Carrillo et al. (2000) correlation results. EXP HETP=0.138 m EXP HETP=0.129 m HETP (m) CARRILLO et al. HETP AVG AVG % 1 0.044 0.044 0.042 68.12 70.29 69.46 2 0.050 0.050 0.047 63.77 67.39 65.82 3 0.037 0.037 0.038 73.19 71.01 72.36 4 0.037 0.037 0.035 73.19 75.36 74.54 5 0.044 0.044 0.042 68.12 71.74 69.46 6 0.008 0.008 0.007 94.20 99.28 94.91 7 0.073 0.073 0.062 47.10 62.32 54.91 43

Assay Table 4. Deviation between experimental (S-NL) and Lockett (1998) correlation results. EXP HETP=0.138 m EXP HETP=0.129 m HETP (m) LOCKETT HETP AVG AVG % 8 0.213 0.149 0.178 54.35 15.50 33.61 9 0.221 0.153 0.184 60.14 18.60 37.86 10 0.227 0.154 0.187 64.49 19.38 40.33 11 0.213 0.149 0.178 54.35 15.50 33.63 12 0.220 0.152 0.183 59.42 17.83 36.97 Assay Table 5. Deviation between experimental (S-NL) and Carrillo et al. (2000) correlation results. EXP HETP=0.138 m EXP HETP=0.129 m HETP (m) CARRILLO et al. HETP AVG AVG % 8 0.075 0.032 0.051 45.50 73.08 61.70 9 0.115 0.087 0.100 16.47 32.47 24.89 10 0.078 0.059 0.068 43.16 53.98 48.86 11 0.101 0.047 0.069 27.12 63.68 48.55 12 0.102 0.063 0.080 26.31 50.80 39.79 evaluate HETP of heavy hydrocarbon mixtures, since it does not require calculations of diffusivity and mass transfer coefficients. The correlation proposed by Carrillo et al. (2000) works better at higher pressures; therefore it has not shown good agreement with the experimental results, which were obtained at 10 mbar (Table 3). Since Lockett s correlation (Lockett, 1998) claims that HETP is inversely dependent on liquid viscosity, a heavier and more viscous mixture (NM-BS) produces lower HETP values than a lighter one, like the S-NL mixture, as demonstrated in Table 4. pointing out that the low vapor flowrate inside the column was the most influential variable. For both mixtures, large deviations have been obtained, although some points could not be extrapolated for S-NL mixtures (bottom product). Among the theoretical models, the one proposed by Olujić et al. (2004) was chosen for being one of the most recent and robust. Concerning Carrillo and coworkers correlation (Carrillo et al., 2000), which is strongly dependent on pressure, the deviations between the experimental and the calculated HETP were smaller for the lighter mixture, especially for the top product, which operated at higher pressure conditions (50-70 mbar), as shown in Table 5. Figure 9 presents the relation between the F- factor with the Sulzer DX HETP supplier. The results from Tables 6 and 7 have shown that the nature of the mixtures had no influence on HETP deviations from Sulzer DX values, Figure 9. Sulzer DX HETP versus F. 44

Table 6. Comparison between experimental and Sulzer DX HETP for NM-BS mixtures. Assay Fv Fv MIN. MAX. MIN. MAX. 1 0.937 0.806 63.77 71.01 65.22 72.46 2 1.307 1.024 60.14 69.57 63.77 71.01 3 0.624 0.773 68.84 74.64 68.12 73.19 4 0.620 0.512 68.84 74.64 72.46 78.26 5 0.976 0.746 63.77 71.01 68.12 73.19 6 2.084 0.725 49.28 56.52 68.12 73.19 7 2.474 1.153 27.54 49.28 63.77 71.01 Table 7. Comparison between experimental and Sulzer DX HETP for S-NL mixtures. Assay Fv Fv MIN. MAX. MIN. MAX. 8 0.330 0.053 48.00 53.33 - - 9 0.923 0.480 52.17 64.35 52.87 58.62 10 0.282 0.148 51.28 55.13 - - 11 0.659 0.108 52.48 60.40 - - 12 0.683 0.225 52.94 60.78 41.27 46.03 Unfortunately, neither mass transfer model was able to properly describe the base lube oil distillation. For instance, Olujić and coworkers original model (Olujić et al., 2004) yielded underestimated area values, by using Onda s correlation (Onda et al., 1968), thus producing increased HETP values and being conservative for distillation column design. On the other hand, its modified version, proposed by Orlando Jr and coworkers (2009), although providing more realistic values for the effective areas (Tables 8 and 9), have produced large deviations for HETP. As shown in Tables 10 and 11, large deviations are observed between experimental and calculated HETP values, especially with the modified version of model by Olujic et al. (2004). Table 8. Theoretical models parameters NM-BS mixtures. Case 1 ugs uls kl kg ae model top bottom top bottom top bottom top bottom top bottom Olujić (Onda) 120.59 141.43 0.548 0.123 8.5 10-5 1.6 10-4 8.9 10-5 1.2 10-4 0.018 0.010 Olujić* 695.17 663.63 Case 2 model top bottom top bottom top bottom top bottom top bottom Olujić (Onda) 177.33 184.99 Olujić* 1.128 0.704 3.7 10-4 4.7 10-4 1.5 10-4 1.7 10-4 0.027 0.028 615.78 601.18 model top bottom top bottom top bottom top bottom top bottom Olujić (Onda) Case 3 1.519 1.093 5.7 10-4 6.7 10-4 1.7 10-4 2.3 10-4 182.64 190.54 0.034 0.048 Olujić* 587.65 576.17 * Olujić et al. (2004) with the effective areas calculated by the Rocha et al. (1993) and Rocha et al. (1996) models 45

Table 9. Theoretical models parameters for S-NL mixtures. Case 1 ugs uls kl kg ae model top bottom top bottom top bottom top bottom top bottom Olujić (Onda) 100.46 115.22 3.015 3.685 3.7 10-4 7.5 10-4 8.5 10-5 1.1 10-4 0.086 0.145 Olujić* 615.40 567.30 model top bottom top bottom top bottom top bottom top bottom Olujić Case 2 (Onda) 1.932 2.570 2.0 10-4 5.7 10-4 2.5 10-5 3.9 10-5 0.119 0.462 58.26 75.17 Olujić* 645.08 573.69 * Olujić et al. (2004) with the effective areas calculated by the Rocha et al. (1993) and Rocha et al. (1996) models Table 10. Deviations between experimental HETP and Olujic et al. (2004) model for NM-BS mixtures. NM-BS mixture EXP HETP = 0.138m Olujic et al.(2004) Olujic et al. (2004)* Deviation (%) HETP (m) HETP (m) Olujic Olujic* Top 0.176 0.031 27.54 77.54 Case 4 Bottom 0.102 0.021 26.09 84.78 Average 0.134 0.025 2.90 81.88 Top 0.295 0.052 113.77 62.32 Case 5 Bottom 0.102 0.021 26.09 84.78 Average 0.173 0.033 25.36 76.09 * Effective area calculated by the Rocha et al. (1993) and Rocha et al. (1996) models Case 8 Case 9 Case 12 Table 11. Deviations between experimental HETP and Olujić et al. (2004) model for S-NL mixtures. S-NL mixture Olujić et al.(2004) Olujić et al. (2004)* Deviation (%) HETP (m) HETP (m) Olujic Olujic* Top 0.109 0.019 20.85 86.27 Bottom 0.045 0.010 65.39 92.62 Average 0.070 0.013 47.66 89.94 Top 0.196 0.061 41.71 56.03 Bottom 0.117 0.039 9.65 70.11 Average 0.151 0.048 13.15 63.74 Top 0.174 0.048 26.16 65.59 Bottom 0.097 0.029 24.83 77.70 Average 0.130 0.037 2.62 72.30 * Effective area calculated by the Rocha et al. (1993) and Rocha et al. (1996) models Those large deviations were similar for both mixtures, regardless of the operational conditions applied. This behavior could be explained by the effect of superficial area on HETP evaluation, as previously mentioned. 46

6. CONCLUSIONS The present work aimed to present a revamp study case and a HETP evaluation of a laboratory distillation column, when operating with complex lube oil mixtures. Concerning the simulation results, very good agreement between the simulation and original curves has been achieved (approximately 5%), allowing that fluid properties generated by PRO II reports be used on HETP calculations. Due to that excellent agreement, the experimental HETP was considered the same as the simulated one. Among the models tested, Lockett s correlation (Lockett, 1998) has provided a fair estimate for column performance evaluation, having the advantage of not requiring calculations of diffusivity and mass transfer coefficients. Moreover, among the theoretical models investigated, the modified version by Olujić and coworkers (2004) was not able to properly evaluate the performance of base lube oil in the laboratory distillation column tested, although yielding corrected values for effective interfacial area. It could be observed from the results of this work that the deviations obtained between the experimental and evaluated HETP models were high. Therefore, it can be concluded that more studies should be encouraged on distillation of complex mixtures in order to enhance packed column design and HETP evaluation. ACKNOWLEDGMENTS The authors would like to acknowledge financial support from CAPES and CENPES/PETROBRAS in the development of this work. 7. REFERENCES Bravo, J. L.; Rocha, J.A.; Fair, J. R. Mass-transfer in gauze packings. Hydrocarbon Processing, v. 64(1), p. 91-95, 1985. Carrillo, F.; Martin, A.; Rosello, A. A shortcut method for the estimation of structured packings HEPT in distillation. Chemical Engineering & Technology, v. 23(5), p. 425-428, 2000. Fair, J. R.; Seibert, A. F.; Behrens, M.; Saraber, P. P.; Olujic, Z. Structured packing performance experimental evaluation of two predictive models. Industrial & Engineering Chemistry Research, v. 39(6), p. 1788-1796, 2000. Lockett, M. J. Easily predict structured-packing HETP. Chemical Engineering Progress, v. 94(1), p. 60-66, 1998. Nicolaiewsky, E. M. A.; Ahón, V. R.; Mendes, M. F. Projeto CTPETRO/OTIMDEST, Otimização de destilação piloto/laboratorial para obtenção de óleos lubrificantes pela rota de hidrorrefino. 2002. (in Portuguese) Olujić, Z.; Behrens, M.; Collo, L.; Paglianti, A. Predicting the efficiency of corrugated sheet structured packings with large specific surface area. Chemical and Biochemical Engineering Quarterly, v. 18(89-96), 2004. Onda, K., Sada, E., Takeuchi, H. Gas absorption with chemical reaction in packed columns. Journal of Chemical Engineering of Japan, v. 1(1), p. 62-66, 1968. Orlando Jr., A. E. Análise de Desempenho de Coluna de Destilação Contendo Recheio Estruturado. 203p. Dissertação de Mestrado. Escola de Química, Universidade Federal do Rio de Janeiro, 2007. (in Portuguese) Orlando Jr., A.E.; Medina, L.C.; Mendes, M.F.; Nicolaiewsky, E.M.A. HETP evaluation of structured packing distillation column. Brazilian Journal of Chemical Engineering, accepted for publication, in press, 2009. Rocha, J. A.; Bravo, J. L.; Fair, J. R. Distillation columns containing structured packings: a comprehensive model for their performance. 1. Hydraulic models. Industrial & Engineering Chemistry Research, v. 32(4), p. 641-651, 1993. Rocha, J. A.; Bravo, J. L.; Fair, J. R. Distillation columns containing structured packings: a comprehensive model for their performance. 2. Mass-transfer model. Industrial & Engineering Chemistry Research, v. 35(5), p. 1660-1667, 1996. 47

Appendix Table A1. Feed characterization for NM-BS and S-NL and individual lube oil characterization. HT-750 (%w) ASTM D1160 (%v) NM-BS mixture ORIGINAL LUBE OIL ORIGINAL LUBE OIL S-NL mixture Neutral Medium Bright Stock Spindle Light Neutral IBP* ( C) 333.5 332.0 437 333.0 340.1 378.9 5 403 388.5 482.5 355.6 364.9 417.9 10 420.5 405.0 499.5 362.7 372.0 428.5 20 440.5 422.5 521 371.9 384.6 441.4 30 456 433.5 538.5 381.1 394.3 446.8 40 471 443.0 554.5 390.4 404.2 453.8 50 497.5 451.0 570.5 399.8 413.2 459.0 60 531 458.0 587.5 408.8 423.7 466.6 70 560.5 465.0 606 418.9 432.9 476.1 80 591.5 472.5 629.5 429.7 441.8 497.8 90 630.5 484.5 660.5 441.8 451.5 562.5 94.1 - - - - - 598.3 95 657.5 498.5 686 451.1 458.1 - FBP* 99% 699 577.5 723 464.9 464.6 - * IBP initial boiling point, FBP final boiling point Table A2. PRO II planning for NM-BS simulation. NM-BS Case P / mbar T / C F / (kg/h) RR 1* 10 230 1 1 2* 10 230 1 2 3 10 230 1 1 4 10 280 1 1 5 10 280 1 2 6 1 280 1 2 7 10 280 2 1 * Feed in top section 48

Table B1. PRO II planning for S-NL simulation. S-NL Case P (mbar) T ( C) F (kg/h) RR 1 50 280 1 0.5 2 50 280 1 3 3 70 280 1 0.5 4 50 280 2 0.5 5 50 300 1 2 Table C1. Deviation between original, experimental and simulation curves (case 4, NM BS). ORIGINAL CURVES CENPES / PETROBRAS EXPERIMENTAL SIMULATION SIM CENPES (%) SIM EXP (%) % w NM BS top bottom top bottom top bottom top bottom 1.0 349.0 437.0 184.8 407.0 324.5 483.0 7.55 9.52 43.05 15.73 5.0 388.5 482.5 253.4 423.4 342.4 494.6 13.46 2.45 25.99 14.40 10.0 405.5 499.5 291.4 435.1 374.5 512.5 8.28 2.54 22.19 15.10 30.0 433.5 538.5 357.2 466.1 412.1 559.0 5.19 3.67 13.32 16.62 50.0 451.0 570.5 380.6 511.9 436.2 604.4 3.39 5.61 12.75 15.30 70.0 465.0 606.0 394.6 571.3 456.0 666.0 1.97 9.01 13.46 14.22 90.0 484.5 660.5 404.8 640.6 470.7 707.0 2.93 6.58 14.00 9.39 95.0 498.5 686.0 407.4 670.7 478.5 721.9 4.18 4.97 14.86 7.09 Average 5.87 5.54 19.95 13.48 Table C2. Deviation between original, experimental and simulation curves (case 8, S-NL). ORIGINAL CURVES CENPES / PETROBRAS EXPERIMENTAL SIMULATION SIM CENPES (%) SIM EXP (%) % w Spindle NL top bottom top bottom top bottom top bottom 1 324.1 359.9 253.6 324.6 334.46 389.51 3.1 7.6 27.80 10.87 5 346.3 391.3 305.7 349.4 343.67 398.02 0.77 1.69 13.28 11.99 10 346.5 401.3 315.2 358.1 350.47 402.51 1.13 0.3 9.93 12.06 30 358.5 417.4 333.4 375.8 367.01 417.93 2.32 0.13 7.53 11.07 50 375 428.2 346.1 388.4 378.23 427.86 0.85 0.08 8.35 10.25 70 386.4 438.4 356.5 402.6 394.81 439.78 2.13 0.31 8.39 8.89 90 413.8 452.2 369.2 425.9 415.29 456.88 0.36 1.02 12.08 6.18 95 425.3 458.8 374.2 437.7 424.47 461.16 0.19 0.51 13.66 4.82 Average 1.49 1.44 12.63 9.52 49