Role of pressure in coking of thin films of bitumen.

Gray, Murray R. ; Le, Tuyet ; Wu, Xin A. 等

INTRODUCTION

Cracking processes are important for conversion of the vacuum residue fractions of oil sands bitumen and petroleum. The most commonly used process is delayed coking, but several processes based on moving beds and fluidized beds are in use or are under development in order to increase liquid yields. The fluid coking process is used at a very large scale by Syncrude Canada to convert up to 5.6 x [10.sup.4] [m.sup.3]/d. Liquid feed is cracked by spraying it into a fluidized bed of fine coke particles. Other processes under development propose the use of fluidized sand or other reactor geometries (Gray, 2002). All of these hot-particle processes operate at temperatures in excess of 500[degrees]C, which gives some vaporization of the vacuum residue components (524[degrees]C+), in addition to the gas oil and lighter fractions (< 524[degrees]C). For example, Olmstead and Freund (1998) found that components with boiling points up to approximately 650[degrees]C were subject to a combination of volatilization and cracking. The reactors operating in this high-temperature regime usually involve thin liquid films reacting on the surfaces of solid particles that serve as a heat medium (Gray, 2002), therefore, the geometry of the reacting liquid phase promotes the transport of high-boiling components to the liquid phase. During the cracking reactions, light components and gases form in the liquid phase. Depending on the rate of reaction and the diffusion transport of volatile out of the liquid phase, bubbles may form in the liquid (Gray et al., 2001). This combination of transport and reaction is fundamentally different from processes such as evaporation or boiling in thin films. The formation of bubbles by reaction changes the transport processes in the liquid phase, and may have mechanical implications when the feed liquid is aggregated with fine solid particles, as would normally occur when spraying liquid in to a fluidized bed (Gray, 2002). Some prior reports considered reaction and product evolution from softening coal (Attar, 1978; Oh et al., 1989), a system that offers some analogies, but the physical properties were dramatically different from bitumen fractions and the temperature arrange of interest was much higher, up to 1000[degrees]C.

Although coking processes are considered low-pressure operations, they can operate at a range of pressures above ambient, up to circa 700 kPa. Higher operating pressure will increase the residence time of the vapours inside the reactor, but the impact on the reacting liquid phase is more subtle. The objective of this research was to measure the kinetics of coking over a range of pressure, using carefully controlled conditions of temperature and film thickness. This approach ensured that all of the relevant transport processes (heat transfer and product evolution from the liquid phase by diffusion and bubble formation) are properly controlled so that their role in the observed kinetics is fully defined. Our hypotheses were:

a) An increase in reactor pressure will reduce the evolution of bubbles, when liquid films are over 20-30 [micro]m in thickness, and thereby increase the yield of coke;

b) The rate and extent of vaporization of liquid components into the vapour phase will depend on temperature, liquid composition, and the partial pressure of the components in the vapour phase, rather than total pressure.

[FIGURE 1 OMITTED]

MATERIALS AND METHODS

Materials

The experiments used Athabasca vacuum residue, supplied by Syncrude Canada Ltd. The properties of this material are given in Table 1. Solvents (methylene chloride and toluene) were obtained from Fisher Scientific (Mississauga, ON) and used as received. Compressed gases (nitrogen and carbon dioxide) were supplied by Praxair.

Equipment

The experimental apparatus is illustrated in Figure 1. For the kinetics experiments, measuring yields as a function of time and pressure at 503 and 530[degrees]C, the reactor consisted of inductively heated strips of Curie-point alloy (25 cm x 2 cm x 0.04 cm), each spray-coated with residue to give a film thickness of circa 20 [micro]m and an area of 46 [cm.sup.2]. Up to six strips were held in ceramic holders to form an annulus within a glass reactor tube. In order to study the role of film thickness, thin films of bitumen were deposited in shallow pans of Curie-point alloy then reacted in a controlled environment using rapid induction heating. The basic design of the reactor followed Gray et al. (2004); except that the vacuum residue material was coated onto pans of Ni/Fe alloy (Ametek Special Metals, Wallingford, CT) with Curie-point temperatures in the range of 503 and 530[degrees]C. For reactions of thicker films, strips of Curie-point alloy (25 cm x 2 cm x 0.04 cm) were formed into pans by bending the edges to a depth of approximately 2 mm. This design ensured that the liquid phase did not drip off of the alloy surface during reaction of thick films. The coated area of residue on each pan was approximately 46 [cm.sup.2]. The Curie-point temperatures of the alloys were verified using fine-gauge thermocouples that were spot-welded to the metal surface. Thin films of vacuum residue material were created by dissolving the feed in methylene chloride, then spraying the solution onto the strips or pans. The strips were dried overnight and reweighed to determine the amount of vacuum residue. The minimum mean film thickness was 20.0 [+ or -] 1.5 [micro]m, calculated as volume of residue divided by the area coated. Thicker films of up to circa 140 [micro]m were prepared by repeated applications of the methylene chloride-vacuum residue solution.

The pans of vacuum residue were maintained in a controlled gas environment during reaction. Solenoid valves controlled the gas flow through the apparatus. Pressure relief valves were installed before and after the reactor for safety reason, and a pressure control (Tescom, Elk River, MN) was installed downstream to provide backpressure control to the reactor system. The reactors were 5 cm and 2.5 cm diameter beaded glass (Fred S. Hickey Corp., Schiller Park, IL). The 5 cm tube had a safe operating pressure limit of 446 kPa, and was operated at a maximum pressure of 377 kPa. The 2.5 cm tube had a maximum pressure limit of 790 kPa, and was operated at a maximum pressure of 652 kPa. This high-pressure tube was coated in resin, except for the central portion, as a safety measure in case of breakage. The pans holding the vacuum residue were supported inside the glass tube by ceramic holders.

Both the alloy strips and the glass tubes were held within an induction coil to allow rapid heating (Ameritherm Inc., Scottsville, NY, Model XP-30). The induction coil was in turn surrounded by a Faraday cage of perforated metal sheet to control radio-frequency noise from the apparatus, and to provide an additional safety barrier in case of breakage of the glass tubes. A fibre-optic viewer (Titan Tool Supply Inc., Buffalo, NY) was threaded through the cage and the induction coil and placed on top of the glass reactor to allow observation of the formation of bubbles during reaction.

A dry test meter was located down stream of the reactor tube, after the pressure control unit, to determine the gas flow rate for the experiment. The condenser was immersed in liquid nitrogen to condense the gas and cracked liquids produced from the reactor.

Experimental Procedure

Nitrogen sweep gas was introduced at one end of the glass tube through a sintered metal diffuser and exited the opposite end through a conical section. The gas flow rate was set at 10 sL/min at ambient pressure. The pressure was slowly increased to 377 or 470 kPa by the pressure controller. A leak test was then performed. Once the system was leak free, the pressure was relieved through the vent, and the reactor was purged at 10 sL/ min for 15 min for the 5 cm diameter reactor and 10 min for the 2.5 cm diameter reactor.

When the purging step was completed, a computer program would run the experiment with a standard 5 min purge, followed by a period of induction heating and an immediate cooling at the end of the reaction with a purge of liquid carbon dioxide. For the kinetic experiments with constant film thickness of 20 [micro]m, the reaction time varied from 5 s to 60 s. For experiments conducted with thicker films, higher than 20[micro]m, the reaction temperature was maintained for 240 s to ensure that the ultimate coke yield had been reached. For these experiments with complete conversion of the liquid film to coke and products, the coke yield was determined gravimetrically by weighing the pan before and after reaction.

For the experiments where the reaction time was less than 240 s, the products were recovered as coke, extract, and overhead product. The non-condensable gases were not measured in this study, due to the low yield and poor experimental accuracy of measuring these components in the diluted gas stream (Gray et al., 2001). The Curie-point alloy strips or pans were sonicated in toluene. The coke yield was the increase in the mass of the alloy pieces after cleaning plus the mass of insoluble solids removed from the toluene. These coke solids were recovered by filtering the wash solvent through a 0.2 [micro]m Millipore filter. The extract was the toluene-soluble portion of the material washed off of the alloy pieces, recovered by evaporating the toluene. The overhead product was the condensed liquids, which were recovered from the reactor components and downstream filter by washing with methylene chloride and then evaporating the solvent. Liquid products were sent to the Syncrude Research laboratory for simulated distillation (SDA) and microcarbon residue (MCR) analysis.

RESULTS

Kinetic Results from 20 [micro]m Thick Films

In order to verify that the reactor performance was unaffected by the pressure of the gas in contact with the Curie-point strips, the temperature was measured directly by attaching a thermocouple. At 101.3 kPa, the strip reached 95% of its final temperature in 3.96 s, while at 377 kPa the time to heat to 95% of the final temperature was 3.95 s. Consequently, the heating rate of the alloy strips was unaffected by the gas pressure. As a further check, the gas density was varied from 0.1-1.6 g/L by using helium and nitrogen at 101.3 and 377 kPa. The yield of extractable liquid, from strips coated with a 20 [micro]m film of vacuum residue, after 20 s of reaction at 503[degrees]C was 18.8 [+ or -] 3.5 wt.% ([+ or -] standard deviation), independent of the gas density. Similarly, the amount of overhead product recovered after 20 and 240 s of reaction at 503[degrees]C was unaffected by gas density. Consequently, the heating of the strips and the amount of liquid remaining on the strips were independent of the gas-phase environment.

At a constant film thickness of 20 [micro]m, the yields of coke and extractable liquids as a function of time at 503[degrees]C and 530[degrees]C were almost identical at 101.3 kPa and 377 kPa (Figures 2 and 3). The yields of extractables after 20 s of reaction were the most variable, likely due to the sensitivity of the results to small errors in time or temperature. The differences in extractable yield were no significant given a standard deviation of 3.5 wt.% on replicate experiments. The final coke yield was insensitive to pressure at 503[degrees]C, as illustrated in Figure 4. The coke yield increased significantly with temperature (95% confidence) at both 101.3 kPa and at 377 kPa. The yield of coke increased with pressure at 530[degrees]C, from 14.09 [+ or -] 0.18 wt.% at 101.3 kPa to 14.74 [+ or -] 0.14 wt.% at 377 kPa (95% confidence limits). Although these changes in the coke yield were statistically significant, they indicate a weak dependence on temperature and pressure. This result is consistent with the thermodynamics and mass transfer of product release. In the absence of bubble formation, the diffusion of products out of the liquid film will depend on the thickness and composition of the liquid film and the partial pressure of product in the gas phase adjacent to the film surface. The purge of nitrogen gas would keep the product concentration low, regardless of pressure, therefore the yield of coke should be insensitive to pressure. In contrast, an increase in reactor pressure should increase the coke yield in thick liquid films because the high pressure will suppress the evolution of bubbles. A longer residence time for cracked products in the liquid film will enhance recombination reactions that increase the yield of coke.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

The yields of overhead product were less repeatable than the yields of extractable liquid, due to the larger errors in recovering the material from the reactor components. The material balances (calculated as coke + extract + overhead product) were better than 95%, and the overhead product yields were not sensitive to pressure within experimental variation. The overhead products were collected and analyzed by simulated distillation and MCR content. Within the accuracy of the methods, the composition of the overhead was insensitive to pressure, for example, as illustrated in Figures 5 and 6. At least three replicate samples from independent reactor experiments were analyzed at 101 and 377 kPa, which gave yields of overhead products of 81.2 [+ or -] 1.8% and 82.5 [+ or -] 1.1%, respectively. The error bars show the standard deviation. The yields of overhead product were similar at 652 kPa, but the amounts of sample were much smaller because only one pan could fit in the reactor. The overhead product from several pans at identical reaction conditions were, therefore, pooled for analysis. The data of Figure 6 clearly show that the heavy material that contributes the MCR in the products increases through the first 30 s of reaction, reaching a constant level of 12-14% once the evolution of products from the liquid film is complete. This evolution pattern was insensitive to pressure.

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

Coke Yield as a Function of Temperature and Film Thickness

To investigate the effect of liquid film thickness on ultimate coke yield (coke yield after 240 s of reaction time) a series of experiments were conducted at 503[degrees]C and 530[degrees]C at pressures from 101.3 to 652 kPa and with film thicknesses from 22 [micro]m to 120 [micro]m. The data of Figure 7 show that at 503[degrees]C, the ultimate coke yield increased with increasing film thickness and pressure. The increase in coke yield with film thickness was consistent with previous observations by Gray et al. (2001). The yield was insensitive to pressure for any film thickness up to 377 kPa, then increased significantly at the higher pressure of 652 kPa as illustrated in Figure 7. In contrast, at 530[degrees]C, the coke yield increased with pressure for thin films in the range of 20-40 [micro]m, then the yield from films > 40 [micro]m was the same at 377 and 652 kPa. Experiments with films thicker than 40 [micro]m were not conducted due to significant bubble formation (discussed below). The results at 503 and 530[degrees]C suggest that coke yield may be significantly altered by pressure depending on the balance between product formation and evolution from the liquid film.

[FIGURE 8 OMITTED]

[FIGURE 9 OMITTED]

The MCR content of the overhead from the experiments as a function of film thickness at 503 and 530[degrees]C is shown in Figure 8. This reduction in the MCR yield in the overhead with increased film thickness is consistent with higher coke formation. Within the variation of the data, the MCR in the overhead was not sensitive to pressure. The data from simulated distillation showed a significant increase in the fraction of distillate (< 524[degrees]C) as a function of film thickness (Figure 9). This trend was in agreement with the trends for MCR and coke yield, indicating that less high-boiling material was released from the thicker films.

[FIGURE 10 OMITTED]

Observation of Bubble Formation

The formation of bubbles on the surface of the film of liquid in the reactor pan was observed with a fibre-optic viewer. This viewer allowed safe observation of bubble formation in situ, but did not provide enough brightness for recording images. Consequently, the results are reported from written notes for each observation. As soon as the induction heat was turned on, for films of bitumen thicker than 20 [micro]m, the layer of bitumen swelled due to bubble formation under the liquid surface. Each bubble then broke up to give zones of exposed metal surface as the reaction progressed. The duration of bubbling depended on the thickness of the bitumen film; the thicker the film, the longer the bubbling time. Whenever bubbles formed, they were apparent very early in the reaction before the fluid properties had time to change significantly. At the end of the reaction, the evidence for the bubbles could often be observed in the coke layer, as illustrated in Figure 10. The area of exposed metal surface provided a measure of the extent of bubble formation, as illustrated in Figure 11, but direct observation was more accurate. The evaporation of liquid film at 503[degrees]C was slower than that at 530[degrees]C, so the onset of bubbling was easier to observe at the lower temperature. Once the film thickness was greater than 30[micro]m, the bubbling was very obvious at both temperatures both during and after reaction (Figure 10 shows examples at 503[degrees]C). Visual observation was more difficult at higher pressures, since the efficiency of nitrogen carrier gas for sweeping the product aerosol out of the field of view was reduced.

As the pressure increased the formation of visible bubbles was suppressed in successively thicker and thicker liquid films. Direct observation showed that at 503[degrees]C, bubbles first formed at film thicknesses of 26 [micro]m at 101.3 kPa, increasing to 78 [micro]m at 652 kPa (Figure 12). At a reaction temperature of 530[degrees]C, the onset of visible bubbling started at much lower film thicknesses and was less sensitive to pressure, increasing from 22 [micro]m at 101.3 kPa to 43 [micro]m at 652 kPa.

DISCUSSION

The results of this study suggest that the yield of coke was at most weakly dependent on pressure and temperature, for the range 500-530[degrees]C and 101.3-652 kPa, when bitumen was reacted in thin films (20 [micro]m) to minimize mass transfer resistance in the liquid phase. The total pressure did not affect the vaporization from the thin film in the absence of bubble formation. The yield of cracked products and the MCR content of the products did not vary significantly for thin films over this range of conditions. Over the same range of operating conditions, the yield of coke increased from circa 12% to 17% as the film thickness increased from 20 [micro]m to circa 120 [micro]m. At a reactor temperature of 503[degrees]C, a further increase in the pressure to 652 kPa increased the yield of coke over a wide range of film thicknesses, with a maximum value of 19.7 wt.% from a 120 [micro]m-thick film. The content of distillate fractions in the overhead product was insensitive to film thickness within experimental error, but the total concentration of distillable material in the overhead products (i.e., the < 524[degrees]C + fraction) increased with film thickness. The trend of MCR content and coke yield was consistent with this observation; thin films gave more vaporization of heavy material, which reduced the coke yield.

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

The most obvious conclusion from this study is that the total operating pressure of fluidized-bed cokers will have little impact on the liquid phase reactions over a wide range of operating conditions. The total operating pressure is distinct from the partial pressure of hydrocarbons, which could have an impact through vapour-liquid equilibrium. The reactor in these experiments was swept with nitrogen to ensure a low partial pressure of hydrocarbon products at the liquid surface. This sweep gas also ensured that the vapour phase was much cooler than the reacting liquid. In an actual reactor, an increase in the pressure also increases the residence time of the cracked vapours in the reactor, thereby increasing the severity of the gas-phase reactions. This feature of commercial cokers was not included in this study, nor did we determine the coke formation from a recycled residue stream at various pressures.

The most important impact of total pressure observed in this study was the suppression of bubble evolution. As the pressure on a bubble of vapour increases, the size of the bubble decreases so it is less able to rise and break out of the liquid phase. In the context of coking reactions in a fluidized bed of particles, increased operating pressure would suppress bubble development in both thick layers of liquid feed and in agglomerates of coke and liquid feed. In both cases, the increase in operating pressure would give smaller bubbles at a given time, temperature and feed composition. The volume of a bubble of vapour product will diminish in proportion to the absolute pressure. The data of Figure 12 indicate that at 503[degrees]C, a decrease in operating pressure enhances bubble evolution significantly. Pressure was less significant in changing the threshold for bubble evolution at 530[degrees]C, likely due to the more vigorous reaction rate.

The threshold for bubble formation in Figure 12 is potentially important in the context of agglomerates of liquid and bed particles. If we assume spherical uniform bed particles, then the smallest agglomerate of interest contains 5 particles in close-packed configuration. For a mean particle diameter of 150 [micro]m, then the liquid at the centre of the agglomerate is over 40 [micro]m from the vapour phase, over the threshold for bubble evolution at 530[degrees]C and 377 kPa. Larger agglomerates, or agglomerates with thicker liquid films, would have much larger characteristic lengths, making bubble formation inevitable as the liquid reaches reactor temperature and products are evolved. For the liquid inside agglomerates, therefore, the initial effective film thickness can be quite large even if the ratio of liquid to coke in the agglomerate is low. If the agglomerate is not broken up in the reactor by fluid-dynamic forces, then the yield of coke from the liquid inside the agglomerate will tend to be at the top end of the data sets in Figure 7. Higher operating pressure, which would suppress bubble expansion inside coke-liquid agglomerates, will tend to make these structures more physically stable.

The data of Figure 4 are quite striking in showing the insensitivity of coke yield to operating conditions in the thin film limit. As the film thickness is increased, the yield of coke increases as illustrated in Figure 7, with an accompanying decrease in MCR content of the overhead product. The work of Aminu et al. (2004) showed that the viscosity of the liquid phase increases rapidly with time. This increase in viscosity would make the escape of products more difficult during the latter stages of reaction, due to lower diffusion rates and less favourable bubble evolution. In such a viscous mixture, the thickness of the initial liquid film would have a major impact on the release of material to the vapour phase.

The observations of bubble evolution as a function of pressure and film thickness, summarized in Figure 12, are not conclusive on the role of vapour bubbles in removing products from the liquid phase. Increasing pressure definitely suppressed bubble evolution, but the resulting impact on coke yield was inconsistent. The data for coke yield at 503[degrees]C (Figure 7) did not exhibit any clear change in slope due to the onset of bubble evolution at a film thickness of 60 [micro]m at 377 kPa or at 80 [micro]m at 652 kPa. If bubbles were an important path for removal of cracked products, we would expect the coke yield to remain relatively constant with increasing film thickness after the onset of bubble evolution a given temperature and pressure. The increase in coke yield with film thickness, regardless of the onset of bubbling, suggests that diffusion of products from the liquid film was the main transport mechanism over the entire range of film thickness. An important limitation to the apparatus does need to be considered very carefully. When the liquid film is uniform, the temperature will also be uniform and very close to the Curie-point alloy, based on heat transfer calculations for very thin films. As soon as bubbles become visible, the heat transfer conditions change. The liquid in the liquid film covering the bubble is no longer in contact with the metal substrate and, therefore, it will be cooler because the convective losses to the exterior gas will remain significant even as the vapour within the bubble insulates the liquid cap from the hot metal. The results for the films with bubbling, therefore, are for non-isothermal experiments where portions of the liquid were cooler than the rest of the film. Given the insensitivity of the coke yield to temperature in the thin-film limit, this effect may not be significant.

CONCLUSIONS

In the case of thin liquid films (ca. 20 [micro]m), the effect of pressure on reaction kinetics was insignificant and the coke yield depended only weakly on pressure at 530[degrees]C. Thicker liquid films gave an increase in final coke yield increased with increasing film thickness. At 503[degrees]C reaction temperature, the final coke yields at ambient and 377 kPa were almost the same. There was a significant increase in ultimate coke yield as the pressure was raised to 652 kPa, especially when film thickness was greater than 40 [micro]m. At 530[degrees]C reaction temperature, the effect of pressure on ultimate coke yield was minimal. The liquid film evaporated so rapidly that the effect of pressure on mass transfer was insignificant.

Visible bubbling of the liquid due to cracking reactions during coking was suppressed by increasing pressure. The transition from the quiescent film to bubbling increased from circa 26 [micro]m at 101.3 kPa to 78 [micro]m at 652 kPa when reaction was done at 503[degrees]C. Higher reaction temperature reduced the critical film thickness for bubbling and gave less sensitivity to pressure.

ACKNOWLEDGEMENT

The authors are grateful for financial support from Syncrude Canada Ltd.

Manuscript received August 23, 2006; revised manuscript received October 17, 2006; accepted for publication January 5, 2007.

REFERENCES

Aminu, M. O., J. A. W. Elliott, W. C. McCaffrey and M. R. Gray, "Fluid Properties at Coking Process Conditions," Ind. Eng. Chem. Res. 12, 2929-2935 (2004).

Attar, A., "Bubble Nucleation in Viscous Material Due to Gas Formation by a Chemical Reaction: Application to Coal Pyrolysis," AIChE J. 24, 106-115 (1978).

Gray, M. R., T. Le, W. C. McCaffrey, F. Berruti, S. Soundararajan, E. Chan and I. Huq, "Coupling of Mass Transfer and Reaction in Coking of Thin Films of Athabasca Vacuum Residue," Ind. Eng. Chem. Res. 40, 3317-3324 (2001).

Gray, M. R., "Fundamentals of Bitumen Coking Processes Analogous to Granulation: A Critical Review," Can. J. Chem. Eng. 80, 393-401 (2002).

Gray, M. R., W. C. McCaffrey, I. Huq and T. Le, "Kinetics of Cracking and Devolatilization During Coking of Athabasca Residues," Ind. Eng. Chem. Res. 43, 5438-5445 (2004).

Oh, M. S., W. A. Peters and J. B. Howard, "An Experimental and Modeling Study of Softening Coal Pyrolysis," AIChE J. 35, 775-792 (1989).

Olmstead, W. and H. Freund, "Thermal Conversion Kinetics of Petroleum Residua," AIChE Spring Meeting, New Orleans, LA (1998), paper 34b.

Murray R. Gray (1) *, Tuyet Le (1) and Xin A. Wu (2)

(1.) Department of Chemical and Materials Engineering, University of Alberta, Edmonton, AB, Canada T6G 2G6

(2.) Syncrude Canada Ltd., 9421-17 Avenue, Edmonton, AB, Canada T6N 1H4

* Author to whom correspondence may be addressed. E-mail address: murray.gray@ualberta.ca

Table 1. Properties of Athabasca vacuum residue

Toluene insolubles, wt.% 1.8
MCR, wt.% 27.8
Sulphur, wt.% 5.7
Nitrogen, wppm 7206
Density, kg/m3 @ 20[degrees]C 1086.8
Boiling fractions, wt.%
524[degrees]C- 10
524-650[degrees]C 40
650[degrees]C + 50