I n d . Eng. Chem. Res. 1987,26, 1003-1010 Herbst, J. A.; Rajamani, K. In Design and Installation of Comminution Circuits; Mular, A. L., Jergensen, G. V., Eds.; SME-AIME; New York, 1982; p 325. Herbst, J. A.; Grandy, G. A.; Fuerstenau, D. W. Proceedings of the Tenth International Mineral Processing Congress; Institute of Mining and Metallurgy; London, 1974; p 23. Herbst, J. A.; Rajamani, K.; Lo, Y. C. "Ball Mill Scale-up Testing at Carol Lake Pellet Plant", report of University of Utah Grinding Project, Jan 1982; University of Utah, Salt Lake City. Klimpel, R. R.; Austin, L. G. Powder Technol. 1984, 38, 77. Lynch, A. J. Mineral Crushing and Grinding Circuits; Elsevier: New York, 1977; pp 45-85. Rogers, R. S. C.; Austin, L. G. Particul. Sci. Technol. 1984, 2, 193.
1003
Rogers, R. S. C.; Austin, L. G.; Brame, K. A.; Tangsathitkulchai, C. Proceeding of the 114th Annual Meeting; SME-AIME: New York, 1985; Preprint 85-141. Rogers, R. S. C.; Gardner, R. P. J. Am. Inst. Chem. Eng. 1979,24, 229. Rowland, C. A., Jr.; Kjos, D. J. In Mineral Processing Plant Design, 2nd ed.; Mular, A. L., Bhappu, R. B., Eds.; SME-AIME: New York, 1980; pp 239-278. Shoji, K.; Austin, L. G.; Smaila, F.; Brame, K. A.; Luckie, P. T. Powder Technol. 1982, 31, 121.
Received for review December 9, 1985 Accepted November 18, 1986
Coking Rates in a Laboratory Pyrolysis Furnace: Liquid Petroleum Feedstocks Harry P. Leftin* and David S.Newsome T h e M . W. Kellogg Company, Technology Development Center, Houston, Texas 77084
Integral rates of coking for 14 feedstocks (light naphtha t o vacuum gas oil) and mixtures of these were determined in a laboratory pyrolysis furnace between 815 and 943 "C and between 70 and 340 ms. These can be ranked as severe coking (S.C.) and low coking (L.C.) feedstocks and are characterized by production of filamentous and amorphous (encapsulating) coke, respectively. Admixture of a L.C. feedstock in greater than a critical minimum amount of a S.C. feedstock imparts a natural inhibition on the coking rate of the S.C. feedstock, so t h a t the coking rate of the mixture mimics that of the L.C. component. The inhibition potential of the L.C. feedstock is attributed to its ability t o form a n amorphous coke deposit which deactivates the metal surface by encapsulation of catalytically active metal sites. An empirical correlation has been derived which relates the coking propensity with physical properties of the liquid feedstocks. Although formation of carbonaceous deposits is a minor reaction in the steam pyrolysis of hydrocarbons, it is of major importance in that it leads to significant losses in process efficiency. Coke formation on the inner walls of a thermal cracking reactor retards heat flow through the tube wall and also increases the pressure drop across the reactor. Periodically, it becomes necessary to decoke the furnace tubes which requires feed interruption to the cracking furnace with a concomitant loss in production. Both the frequency of decoking and the associated penalties are greater when cracking normally liquid feedstocks, as opposed to gaseous feedstocks, aiid increase with increasing cracking severity. Coke formation in the steam pyrolysis of hydrocarbons is a complex process which can lead to several types of carbonaceous deposits ranging from amorphous to highly crystalline (Lehaye et al., 1977). The morphology of the coke as well as the kinetics of its formation is known to depend upon the operating variables used in the pyrolysis process and the type of feedstock (Leftin and Newsome, 1979), furnace configuration (Chen and Maddock, 1973), materials of construction (Dunkleman and Albright, 1976), and pretreatment of the inner walls of the pyrolysis tubes (Crynes and Albright, 1969). For any specific feedstock, the rate of coke deposition is strongly influenced by residence time, temperature, degree of conversion, hydrocarbon partial pressure, steam partial pressure, and run duration (Newsome and Leftin, 1980). Response to these variables differs with feedstock properties. Temperature affects not only the rate but also the mechanism of coke deposition. Trimm (1977) and Lobo et al. (1972) found that low- and high-temperature mechanisms were char-
* Present address:
2314 Lexford Lane, Houston, TX 77080.
acteristic of coking for several pure hydrocarbons. In the intermediate temperature region, very small and even negative activation energies were observed. Similar results were observed by Martens et al. (1979) with several petroleum fractions. Coking rates also vary with residence time and exhibit maxima which occur at shorter residence times as the reactor surface/volume increases (Shah et al., 1976). While the literature on pyrolysis coking is extensive (e.g.: Palmer and Cullis, 1965) and has shed much fundamental insight into this complex phenomenon, most experimental studies have been at conditions that are related to, but not directly comparable with, those of commercial practice. These studies usually are restricted to pure hydrocarbons and relatively low temperatures. No direct measurements of coking in commercial furnaces are available in the literature; such studies as exist are based on macroscopic effects such as pressure losses and tube wall temperature. The purpose of this study was to evaluate, experimentally, the coking characteristics during the steam pyrolysis of commercially used liquid petroleum feedstocks in a laboratory pyrolysis furnace under a wide range of conditions including the high temperatures and short residence times actually employed in the modern industrial steam pyrolysis furnaces. Experimental Section Apparatus. With the exception of the carbon burnout and collection system, which is described below, the bench-scale pyrolysis apparatus used in this study has been described previously (Leftin and Cortes, 1972) and has been shown (Leftin et al., 1976) to provide product yields and conversions which are in good agreement with industrial furnace data.
0888-5885/87/2626-lOO3$01.50/0 0 1987 American Chemical Society
1004 Ind. Eng. Chem. Res., Vol. 26, No. 5 , 1987 PREHEATER
TRAVELING T.C.
OMETER WINDOW
BOILER
0 HEATER
FEED
FEED
COPPER OXIDE
G C
ASCARITE
Figure 1. Bench-scale pyrolysis coking apparatus.
Figure 1 shows the bench-scale appartus in the configuration used for measurement of the amount of coke deposited during a pyrolysis run. The pyrolysis furnace is electrically heated by passing low-voltage,high-amperage power through a split cylindrical graphite heating element, which is coaxial with the reactor tube over its entire length. The reactor is an annulus between a reactor tube and coaxial thermowell, both of 310 stainless steel and having a surface-to-volume ratio (S/V) of 19.36 cm-'. These are contained in a water-cooled cylindrical shell which is nitrogen filled (to protect the heating element) and is provided with glass windows so that the outer wall of the reactor can be sighted with an optical pyrometer during a run. Thermal expansion of the reactor is accommodated by free movement within a packing gland where the reactor outlet enters the quench system. This packing gland serves as a gas seal between the nitrogen-filled furnace shell and the reactor outlet. A positive nitrogen pressure differential is maintained between the shell and the outlet so that a small quantity of nitrogen leaks continuously into the product stream where it is separately determined and subtracted from the total gas yield in calculating the overall material balance for a run. With the exception of those runs where excessive reactor coking rates forced early run termination, material balances normally fall between 95% and 100%. Fluid temperature profiles are measured with a Calibrated chromel-alumel thermocouple manually driven over the entire length of the reactor. Wall temperatures are measured with an optical pyrometer whose movement is synchronized with the traveling thermocouple. Both inlet and outlet pressures are measured with calibrated Bourdon gauges sensitive to h 3 torr. After a pyrolysis is completed, the carbon in the reactor is burned out by a steam-air mixture fed into the reactor through the preheater section. Only air free of carbon dioxide and degassed water are used in the burnout procedure. Unreacted water is removed from the combustion products with a room-temperature trap followed by a drying agent. The dried gases are then passed over a heated copper oxide bed to completely convert carbon monoxide to carbon dioxide and finally into preweighed ascarite/magnesium perchlorate tubes. Total carbon deposited in the reactor is calculated from the weight of C02 produced during the reactor burnout. A gas chromatograph in the system is used for determination when a burnout is completed. On the completion of a pyrolysis run/burnout cycle, the reactor tube is conditioned for the next run by reducing and sulfiding the reactor walls with a flow of hydrogen containing 2500 ppm hydrogen sulfide. To maintain sufficient sulfur to deactivate the reactor walls for steam reforming, an undesirable side reaction during steam pyrolysis, ethyl mercaptan is added to the boiler feed
water to provide 50 ppm sulfur during each pyrolysis run. By use of the bench-scale apparatus described above, coking tests can be made rapidly and reproducibly. Product Analysis. For purposes of determining material balances for the runs, two gas samples were taken for duplicate analysis by mass spectrometry during the first third and final third of the run period. Results of duplicate analysis fell within the established limits for this analytical method, and the averaged values were then used. Liquid products, after separation from process steam condensate, were assayed into gasoline (36-218 "C), light fuel oil (218-343 "C), and heavy fuel oil (343+ "C) by gas chromatographic simulated distillation (G.C.S.D.) as defined in ASTM method D-2887. Although yields and conversion data, which were obtained for all runs where run lengths permitted good material balances, are not reported in this paper, they were in close agreement with pilot plant data for all cases where direct comparison was possible. Feedstocks. With the exception of H-GO and H-VGO (which were partially desulfurized and severely hydrotreated gas oils), all of the feedstocks used in this study were virgin petroleum fractions which previously had been used in client feedstock evaluation studies in either the pilot plant or bench pyrolysis units. Inspection data for these feedstocks are summarized in Table I along with their US Bureau of Mines Correlation Index (BMCI) (Nelson, 1953), Coke Inhibition Index (CII) (Leftin and Newsome, 1979),and coking propensity as determined in this study. Data and Discussion The temperature profile provided by the laboratory furnace was parabolic in shape as described by Leftin and Cortes (1972). Pyrolysis run temperatures reported here are the maximum temperatures (T,) of the parabolic temperature profile and ranged between 815 and 949 "C. All runs were isobaric a t a total pressure of 15 psig (202 kPa) and were made with steam dilution corresponding to a steam/hydrocarbon weight ratio of 0.51 f 0.05. Contact times between 0.07 and 0.34 s were examined, covering the range of both the millisecond and conventional pyrolysis furnaces. A 310 stainless steel (SS) (25Cr, 20Ni) reactor was used in these measurements, as its chemical composition is closely similar to that for the HK alloys used in industrial pyrolysis furnace tubes. The effective reactor wall area was 32.8 cmz, including the thermowell. Pyrolysis runs were terminated voluntarily after preselected run lengths, or involuntarily when excessive reactor coking, evidenced by reactor pressure buildup, forced early termination. At the completion of each run, the total carbon deposited in the reactor during the run was determined by the burnout procedure. Integral coking rates were then calculated as milligrams of coke/minute using the total carbon deposited and the total run time. Coking Propensity of Neat Feedstocks. Coking data for neat feedstocks are summarized in Table 11. With each feedstock, runs were made a t various combinations of contact time and temperature to define the relative coking propensity for that feedstock. For many feedstocks, coking rates are known to change with run length (Shah et al., 1976; Kopinke et al., 1981), such that a rapid initial coking rate is followed by a slower rate which either decreases with run length or remains more or less constant (Trimm, 1977). Similar behavior has been noted in industrial furnaces from changes in tubewall temperatures and pressure drop measurements (Ranzi et al., 1985).
Ind. Eng. Chem. Res., Vol. 26, No. 5, 1987 1005 Table I. Feedstock Inspections H-GO! feed
LN-1"
R-lb VGO'
LN-2d WRN-1"
GOr
WRN-2e
H-VGO,'
R-2b WRN-38 K-lh HN' LGOi EX.GOk EX.VG0
ASTM Dist'n, OC
IBP 10% 30 % 50% 70% 90 70 EP OAPI
M"
45.6 51.1 52.8 55.6 59.4 65.0 69.4 80.9 80.7
elemental anal. wt % C 84.1 H. 15.7 S 0.014 type anal. vol % paraffins 84.8 8.8 cycloparaffins aromatics 6.4 BMCI 6.1 CII -21.1 coking severe classification
60.6 70.0 75.0 80.0 86.1 96.1 137.2 74.9 92
366.1
497.8 23.3 366
71.1 76.7 79.4 82.8 88.9 100.0 108.9 69.9 92
37.2 55.0 67.8 82.8 101.7 138.3 191.1 71.0 93
233.9 301.7 352.8 386.1 416.1 442.8 458.9 26.3 322
60.0 81.7 95.0 114.4 136.7 166.1 193.3 58.7 106
68.3 83.9 95.0 111.7 130.6 151.7 228.9 57.4 103
84.7 15.3 0.010
84.6 11.95 3.01
84.5 15.5
