Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979
535
Coal Liquefaction and Deashing Studies. 1. Consol Synthetic Fuel Process Joseph A. Klelnpeter,' Donald C. Jones, Philip J. Dudt, and Francls P. Burke Conoco Coal Development Company, Research Division, Library, Pennsylvania 15 729
Cresap Pilot Plant CSF Process operations were simulated in cycling operations of a 10 Ib of coal/h continuous unit. Mass spectral and H NMR analyses indicated essentially steady-state operation in the third of the four completed cycles. Integrated gravity settling produced 0.22 wt % ash extract. Using an antisolvent did not improve deashing.
Introduction and Summary Pittsburgh Seam 8 coal was liquefied by the Consol Synthetic Fuel (CSF) donor solvent process and deashed via gravity settling at 600 O F with and without an antisolvent. These continuous 10 lb coal/h tests were done to provide guidance for the 20 T P D Cresap CSF Pilot Plant. Highlights of these experiments follow. Mass spectral and NMR analyses showed that the CSF solvent closely approached steady state in three cycles of integrated operation and t h a t tetralins, hydrophenanthrenes, and hydropyrenes were the predominant hydrogen donors. Extract from steady-state operations contained 0.22 wt 70 ash, which slightly exceeds the EPA specification of 0.20%. The low settler upflow velocity (0.3 in./min) used translates to about 34 40-ft diameter settlers for a commercial plant processing 25 000 T P D coal. Using paraffinic anti-solvent to precipitate a fraction of the coal extract to enhance settling gave little improvement in solids removal. Massive agglomerates formed at a moderate anti-solvent rate and forced shutdown of the rake-equipped settler. Settler performance improved as the liquefaction solvent approached steady-state composition and with increasing liquefaction solvent/coal ratio. About 80% of the extract was recovered in the settling step. A second stage unit would increase overall recovery to near 95%. Earlier CSF liquefaction plus gravity settling deashing studies done in our laboratories with the same coal achieved about four times the ash removal rate and a somewhat lower ash content (Gorin et al., 1977). The current steady-state study was more extensive and is more likely to correspond to large-scale operations. The better separation in the previous work may have resulted from the use of liquefaction solvent which was further from steady state than that used here. The lower boiling of two petroleum-derived aromatic liquids was selected as the Cresap start-up solvent because of superior solids separation. The work reported here was the first phase of a two-part engineering study on deashing coal liquefaction products via gravity settling. It was done under subcontract to Fluor Engineers and Constructors as part of their prime contract with DOE to renovate and operate the Cresap, W. Va., CSF Pilot Plant. The development of the CSF process and the theory of solvent deashing have been presented by Gorin et al. (1977). The second part of this study involved anti-solvent promoted gravity settling of PAMCO SRC-I product and is discussed in the following paper of Kleinpeter et al. In both cases, the coal liquefaction step
simulated as closely as possible the steady-state operation of 20-50 TPD pilot plants and the gravity settling unit was operated in tandem with the liquefaction step. Experimental Section Equipment. The 10 stage reactor (4.23 gal) shown in Figure 1was stirred at 400 rpm. The coal slurry (30 lb/h) was heated from about 150 O F to 675-700 O F in the first stage of this reactor. By the fifth stage, the temperature was the same as the outlet temperature (750 OF). The annuli between the stirrer shaft and the partition between each stage are 3 / 8 in. except for the fifth stage, where the annuli are in. We suspect that a certain degree of backmixing between stages occurs in spite of the small flow area between the stages. The contactor vessel (1.3 gal) was mixed at 335 rpm and was used to stabilize the liquefaction product temperature a t 600 O F . When anti-solvent was used, it was added to this vessel to ensure good mixing with the liquefaction reactor effluent prior to gravity settling. The essential design features of the gravity settler are shown in Figure 2. The rake operated at 2-3 rpm to facilitate removal of the settler underflow. Definitions. All product yields are expressed as a percent of the moisture and ash free (MAF) coal. Yield and quality definitions are given below. Extract is the liquefaction product which boils above 554 OF a t 1 mmHg in a Vigreux column and which is soluble in a hot mix of m,p-cresol. Nondistillable liquefaction solvent polymer and coal-derived material are included in this fraction. Distillate yield is the overhead from the above vacuum distillation minus the liquefaction solvent and anti-solvent, if used. The atmospheric equivalent of this fraction is 995 OF-. Coal conversion is defined as (MF coal - cresol insolubles) 100 x MAF coal Extract ash is the weight percent of mineral matter in the extract as determined by ashing a sample at 1472 O F . Start-up Solvent for Cresap In the first task of the program, the lower boiling of two petroleum derived aromatic solvents was selected as start-up solvent for the Cresap Pilot Plant because it gave superior deashing in integrated liquefaction-gravity settling operation of the process development unit (PDU) shown in Figure 1. Table I compares performance of the two solvents. A 4-in. gravity settler (Figure 2) was used at 600 O F without anti-solvent for solids removal. The overflow rate was 185 lb/h ft2 and the upflow velocity was
0019-7882/79/1118-0535$01.00/0@ 1979 American Chemical Society
536
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 Table 11. Analyses of Ireland Mine Coal
COIL-OIL SLURRY
Proximate Analysis, as Received, Wt % moisture 1.01 volatile matter 42.21 fixed carbon 46.91 oxidized ash 9.87
INTI
Ultimate Analysis, MF, Wt % hydrogen 5.07 carbon 72.61 nitrogen 1.11 oxygen (difference) 6.33 sulfur (total) 4.91 sulfur (pyritic) 1.86 0.01 sulfur (sulfate) sulfur (organic) 3.04 ash 9.97
savEm
Figure 1. Hydrogen donor extraction unit with settler.
Table 111. Comparison of Solvent Boiling Range Analyses wt%
Drive
403X 464X 4 6 4 ° F 5 1 8 ° F 518"F+
403°FNeville solvent Sure Sol 180 cycle 4 feed solvent Cresaprecycle solvent Cresap hydro oil Cresaprecyclel hydro (3/1)
88XSun sol I80
UNDERFLOW
Figure 2. Dimensions for 4, 6, and 8-in. diameter settlers. Table I. Liquefaction and Solids Separation Results with Neville and Sure Sol 180 Solvents
yields, wt % MAF conversion extract, 9 9 5 O F + distillate, 9 9 5 "Foverflow extract ash, wt %
Neville LX 745 405X 8 7 8 ° F
Sure Sol 180 405 X 6 9 1 "F
66.0 76.2 -13.6 6.2
69.7 62.2 2.8 1.9
1.9 in./min. The liquefaction reactor was operated at 750 and 400-500 psig using a 1.78 solvent/coal ratio and a slurry rate of 30 lb/h. The analysis of the -65 mesh Ireland Mine Pittsburgh Seam 8 used is presented in Table 11. Boiling range analyses for the Neville and Sure Sol 180 petroleum-based solvents are given in Table I11 with data for coal-derived solvents discussed later. The lower boiling Sure Sol 180 is a poorer solvent for the high molecular weight aromatic components of coal extract than is Neville solvent. Thus, more extract probably precipitated when the liquefaction reactor product was cooled from 750 O F to the 600 O F settler temperature for the Sure Sol 180 than for the higher boiling solvent. The extract precipitated cannot be accurately determined, so this effect cannot be quantified. The precipitated extract agglomerated the coal ash and residue particles, which are in the micron and submicron size range. Without this agglomeration, little clarification would have resulted. Sure Sol 180 gave 1.9% ash extract vs. 6.2% for the Neville solvent. This large relative difference was the basis O F
4
750'F
0.2 2.3 0.0
1.2 12.9 1.4
26.3 71.0 49.4
72.3 13.8 49.2
12.9
18.0
60.0
9.1
23.0 15.4
31.7 21.4
34.7 53.7
10.6 9.5
H
550 prig D Ibs S l u r r y l h r I 7 8 Sol /MF Cwl
1750 Ibs Extract 430 Ibr Solvent 045%Arh m Extracl
SETTLING
MOT
XC PSIQ u
$
~
20% Eafract
~
~
~
~
~
s
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979 537 Table IV. Stage I and I1 Liquefaction and Solids Separation Results yields, wt % MAF coal run no. 10A 10B 11A 11B 12 av cycle 1 cycle 2 cycle sb cycle 4
GUIDE SHAFT
wt%
converdistil- overflow sion extract late ash, wt 5% Stage Ia 84.4 73.0 6.2 0.32 82.9 74.2 5.7 0.36 0.46 83.9 70.8 9.9 0.35 82.4 72.2 6.2 0.43 83.2 74.2 4.1 0.38 83.4 72.9 6.4
extract to underflow 20.6 19.7 21.3 21.1 19.8 20.5
78.2 79.3
Stage I1 71.4 1.8 73.0 0.3
_-
__
0.28 0.21
22.1 20.8
81.5
76.1
0.6
0.21
20.2
_-
__
88% Tetralin/l2% Sure Sol 180.
__
No results; see
text.
TFICTOR
TES ( 4 1
which could be hydrocracked to donor solvent. The stage I program depicted in Figure 3 produced 1750 lb of extract with an average ash content of 0.45 wt % . Table IV gives results of five runs made at the stage I conditions given in Figure 3. The overflow rate from the 8-in. settler was 70 lb/h ft2 and the upflow velocity was 0.3 in./min. % Tetralin. The liquefaction solvent composition of 88% Sure Sol 180 and 12% tetralin was chosen for stage I because it gave the coal conversion planned for the Cresap Pilot Plant operations. Several concentrations of tetralin in Sure Sol 180 were used before achieving the desired conversion. Results are given below. wt % tetralin
conversion
0 12 20 30
69.7 83.3 85.3 89.1
Solvent/Coal. Following stage I, a brief study was made of the effect of liquefaction solvent/coal ratio on solids removal, with the results indicated below. solvent/MF coal
extract ash, wt %
1.78 2.03
0.4 0.3
The lower settler feed viscosity a t the higher ratio produced the better separation.
Extract Hydrocracking A spinning basket experimental reactor was developed and used to hydrocrack the 1750 lb of low ash extract produced in stage I. The catalyst basket assembly is shown in Figure 4. Each basket was 1in. in diameter by 50 in. long. The assembly was rotated at 160 rpm through the liquid to provide good contact between catalyst and extract. Details of this system and a comparison of the results with those previously obtained in a fluidized bed bench-scale reactor have been presented by Dudt and Kleinpeter (1977). The hydrocracking step produced 660 lb of 446 X 878 O F donor solvent which was rich in hydroaromatics. This donor solvent was very complex; mass spectral analysis indicated more than 100 component masses. By contrast the Sure Sol 180/tetralin used in stage I had only 30 major components. The 403 X 878 O F solvent recovered from stage I product was essentially the same as the feed solvent except for the conversion of tetralin to naphthalene during the hydrogen transfer reactions in the liquefaction step.
DRIVE SHAFT FOR SPINNING ASSEMBLY
Figure 4. Spinning catalyst basket assembly. STAGE I RECYCLE
DONOR SOLVENT
DISTILLATION
EXTRPCT
OISTILLPTIW
EXTRPCT
I
O I S F L L A T I O N E ~X T R P C T
I I
-- I
DISIILLPTKW
1
EXTRPCT
RECYCLE SOLVENT
Figure 5. Stage I1 CSF program.
Hydrocracking coal extract produced coal derived donor and other solvent species. Stage I1 Liquefaction a n d Deashing Results The donor solvent made by hydrocracking stage I extract was used in the four-cycle stage I1 program diagramed in Figure 5. The liquefaction solvent in each cycle contained 30% of the donor solvent and 70% recycle solvent from the previous cycle. For cycle 1, the recycle solvent was from the stage I program. Liquefaction and deashing conditions were the same as in stage I (Figure 3). The stage I1 program was designed to study liquefaction and deashing in the CSF process as steady-state conditions were approached. Theoretically, the fraction of stage I recycle solvent replaced in the four-cycle program of Figure 5 would be (1G0.74) = 0.76. This rapid enrichment of the liquefaction solvent with donor liquids is a unique feature of the CSF process. At steady state, the liquefaction solvent is composed predominantly of hydrogen donors and donor precursors. This means that solvent balance problems and solvent quality deficiencies can never occur in CSF. Solvent balance and quality are problems in the SRC process (Kleinpeter et al., following paper). A more rigorous cycling program would have hydrocracked extract from each cycle to produce donor solvent for the next cycle, but such a program would have required more time than was available. We chose the approach of Figure 5 because CSF extract quality does not vary greatly with liquefaction solvent as long as the solvent is high in
538
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979
hydrogen donors. Extract analyses from stage I1 discussed later will further verify this aspect of CSF extraction. Table IV compares stage I and I1 liquefaction and deashing results. Mass spectral, NMR, and other analyses discussed later show that this system had made a good approach to steady state by the end of cycle 3. The primary conclusions from stage I1 concerning deashing of the CSF product are the following. (a) Deashing efficiency improved as steady-state conditions were approached. (b) Use of a paraffinic anti-solvent in cycles 3 and 4 gave no improvement in deashing. (c) Excess anti-solvent caused operating problems even with a rake-equipped settler (cycle 3). (d) The EPA specification of 0.1 lb/MM Btu (-0.2 wt % ash on extract) was slightly exceeded in all cases which would be long term operable. Specification extract could be made by lowering the upflow velocity, but the value used (0.3 in./min) corresponds to 34 40-ft diameter settlers for a commerical plant processing 25 000 T P D coal. Cycles 1 and 2. The conversion of the Pittsburgh Seam 8 coal averaged 78% in cycle 1, compared with 83% in stage I using 88% Sure Sol 180/12% tetralin, so the 30% donor solvent in cycle 1was not equivalent to 12% tetralin in stage I. Pure Sure Sol 180 gave about 70% conversion. Therefore, cycle 1 solvent did give a significant increase in conversion over the start-up solvent. The average percent ash on extract was 0.28 wt 70vs. 0.32-0.40% in the best periods of stage I. Recycle solvent from cycle 1was recovered and mixed 70/30 with donor solvent to provide the liquefaction solvent for cycle 2. Average conversion increased one point over cycle 1 and there was a decrease in extract ash from 0.28 to 0.21 wt % . Compared with stage I and cycle 1, the cycle 2 deashing results show that deashing efficiency improved as start-up solvent was replaced with coal derived recycle solvent. Anti-Solvent in Cycle 3. Using anti-solvent in cycle 3 produced deposits that would prohibit long-term settler operation. At an anti-solvent/liquefactionsolvent (AS/LS) ratio of 0.29 by weight the deposits were 1-4 in. diameter and forced termination of the run after six h. Eleven pounds of solids were found in the settler after shutdown. The test was repeated using 0.09 AS/LS. Although operable for this 18-h test, the settler deposits which formed would preclude long-term operability. No data had been taken when the higher anti-solvent rate caused abortion of the first attempt of cycle 3. In the second attempt (cycle 3B), the extract ash was 0.16 wt % in the first 3.5 h material balance, but it increased to 0.21% in the second balance period. The increased ash content in the second period may be due to solids which accumulated during the run. Material balance calculations indicated that about 3 lb of material accumulated in the settler during the run. Accurate conversion and yield data could not be calculated because of the solids accumulation, so none are presented in Table IV. The anti-solvent used in cycle 3 was Soltrol 130, a 365 X 405 O F paraffinic naphtha produced by Phillips Petroleum Co. The earlier work in our laboratories (Gorin et al., 1977) used normal decane as anti-solvent. Soltrol 130 is nearly as effective an anti-solvent as normal decane and would be more practical than the pure paraffin for a commercial plant. Cycle 4. The final cycle of stage I1 used 0.05 AS/LS. The average ash content of the extract was 0.21%, essentially the same as in cycle 2 without anti-solvent. The average conversion in cycle 4 was 81.5% vs. 79.3 in cycle 2, indicating a slight improvement in liquefaction per-
Table V. Ash Size Distribution of Extract Made with and without Solids Separation wt % extract ash
stage 11, cycle 4 stage I, run E-5
settler + 5 p m yes 0.14 no
10.96
1.2 X 5 pm
0.45X - 0 . 4 5 1 . 2 pm pm
0.01
0.02
0.12
total 0.29
0.03
0.02
0.26
11.27
~
~
~~
formance as start-up solvent was displaced. No deposits formed in the settler during cycle 4 but the low anti-solvent rate gave no better deashing than cycle 2 without anti-solvent. Sufficient CSF extract precipitated on cooling to 600 O F that deashing was not improved by adding small quantites of Soltrol 130 to produce further extract precipitation. An aromatic anti-solvent such as process derived light oil might be better for the CSF process than an aliphatic liquid like Soltrol 130. Light oil is a poorer anti-solvent than Soltrol 130 so operation would not be as sensitive to anti-solvent ratio. Fine Ash Limits Clarity. Millipore filtrations of cycle 4 extract showed that nearly half of the ash passed through a 0.45-pm filter. This “fine ash” probably was very small particles plus soluble organometallics. Since this fine ash cannot be removed by filtration, it sets a lower limit on the ash content of the coal extract. This limit is about 0.10 to 0.15 wt % and is significant when compared to the EPA fuel oil specification of 0.1 lb of ash/MM Btu, which is about 0.2 wt 9’0 extract ash. Table V gives the Millipore filtration results for extract from cycle 4 and from a stage I run where the settler was bypassed. Gravity settling removed most of the > 5 pm material but much of the ash smaller than 5 pm was not removed. These ash analyses were done at 1112 O F instead of 1472 OF as used for all analyses previously discussed. The 1112 OF method gave 0.28% ash for a cycle 4 extract sample. This compares very well with the 0.29% obtained by summing the ashing results from each of the filtrations. If the 1472 OF method had been used on the Millipore filtrates, the ash values may have been slightly different, but the relative contribution of each size range would probably have been similar. That is, about half of the total cycle 4 ash would have been less than 0.45 pm. Extract Recovery. Table IV shows that the extract lost to the settler underflow was about 20%. Optimizing the underflow rate would probably have given about 85% extract recovery. About 95% extract recovery would result if two settling stages were used, as planned for the Cresap Pilot Plant. Steady-State Evidence. The liquefaction solvent recovered by distillation of the cycles 3 and 4 product had similar mass spectrometric and ‘H and 13CNMR analyses. The extract from these cycles had nearly the same solubility characterizations. Solvent from cycles 1 and 2 differed significantly from cycles 3 and 4 solvent. This evidence implies that a reasonably good approach to steady state was achieved by cycle 3. Table VI gives the mass spectral 2-number analyses of the output solvents from stages I and 11. Oxygenated molecules (phenols and dibenzofurans) and several nonbenzoid aromatics (indan, indene, acenaphthene, acenaphthylene, fluorene) have been grouped separately. The 2 number is derived from the generalized empirical formula CnH2n+z; for example, naphthalene (C,,H,) and its alkylated homologues have a 2 number of -12. Other examples are given in Table VI which also lists the average molecular weight for each solvent as calculated from the mass spectral data. From a comparison of the output
Ind. Eng. Chern. Process Des. Dev., Vol. 18, No. 3, 1979 539 Table VI. Mass Spectral Analyses of CSF Coal Liquefaction Solvents (wt % ) -2
stage I feed
stage I recycle
donor solv.
6 8 10 12 14
4.2 16.8 0 74.7 1.2
3.5 6.7 0 88.9 0.1
1.4 15.8 4.0 21.9 14.0
16
0.1
0.0
18 20 22 24 oxygenated molecules nonbenzoid av mol wt
0.1 0.0 0.0 0.0 2.0
0.9 148.0
cycle 1 output
cycle 2 output
cycle 3 output
cycle 4 output
1.6 5.4 0.3 81.6 2.6
2.3 5.2 0.6 69.8 5.6
1.o 4.9 0.6 58.0 8.3
1.3 5.2 0.6 53.9 9.0
12.6
1.8
4.0
6.4
6.9
0.0 0.0 0.0 0.0 0.8
5.2 3.6 3.5 0.8 11.0
1.1 0.3 0.3 0.0 3.6
1.7 1.5 0.7 0.3 5.9
4.6 2.4 1.8 0.4 8.4
4.8 3.2 1.9 0.3 11.0
0.0 145.3
6.1 197.8
1.3 153.7
2.2 162.6
3.2 172.0
3.9 172.7
Table VII. H and 13C N M R Analyses of CSF Coal Liquefaction Solvents
stage I feed solvent stage I recycle solvent donor solvent output solvent, cycle 1 output solvent, cycle 2 output solvent, cycle 3 output solvent, cycle 4
CH,
Table VIII. Solubility Characterizations of Clarified Extract solubility characterization, wt %
H distribution normalized to 100% CH,
possible identifications benzenes tetralins, indans octahydrophenanthrenes naphthalenes biphenyls, tetrahydrophenanthrenes acenaphthalenes, fluorenes, dihydrophenanthrenes phenanthrenes dihydropyrenes pyrenes, fluoranthenes benzanthracenes, chrysenes phenols, dibenzofurans
aro- C aromaticity WH matic ( b y I3C NMR)
0.5
8.3
39.8
51.4
81.1
0.9
5.8
35.5
57.8
84.0
8.2 2.5
24.7 8.7
35.8 38.8
31.3 50.0
60.8 79.7
3.4
12.2
35.4
49.0
74.9
4.1
16.5
36.7
42.7
70.3
6.0
17.9
36.7
40.4
71.7
solvent analyses, and the mean molecular weights, we conclude that the solvent had closely approached steady state by cycle 3. The average deviations between cycles calculated from Table VI are: cycle 1-cycle 2, 46.8%; cycle 2-cycle 3, 39.5%; cycle 3-cycle 4, 12.0%. The mean molecular weights of the cycle 3 and cycle 4 output solvents are essentially identical. The most prevalent compound class is the naphthalenes, a t 2 = -12. Potential hydrogen donors include the tetralins, the hydrophenanthrenes, and the hydropyrenes. The 'H NMR data (Table VII) show a steady decrease in aromatic hydrogen as the liquefaction solvent approached steady state. Aromatic protons are those attached to an aromatic ring. There was a corresponding increase in the aliphatic character of the hydrogen, as evidenced by the CH3 and CH2 proton increases in the output solvents. CH2 protons are located on methylene carbons, @ or further from an aromatic ring. CH3 protons are terminal methyl protons y or further from an aromatic ring. This increase in aliphatic character suggests longer side chains and/or an increase in hydroaromatics. The a hydrogens, which include hydroaromatic and benzilic protons on carbons immediately adjacent to an aromatic ring, did not vary significantly in the recycle solvent. The carbon aromaticity by 13CNMR followed the same trend as the aromatic hydrogen. That is, it indicated a decrease in aromatic character of the solvent as steady state was approached. The increase in aliphatic character of the output solvents may account for the improved deashing performance but poorer coal conversion in the stage I1 work as opposed to stage I. That is, if we surmise that the more aliphatic
stage I1 cycle no.
oils
asphaltenes
preasphaltenes
1 2 3 4
16.5 18.1 17.6 19.5
36.1 38.0 42.0 41.5
47.4 43.9 40.4 39.0
material is a poorer physical solvent, we would expect it to dissolve less coal in the extraction step and to precipitate more extract upon cooling of the extractor effluent prior to its introduction to the settler. This would result in an improved solids separation efficiency. The donor solvent was the most aliphatic of the solvents. This is probably the result of hydrocracking one ring of various polynuclear aromatic compounds in the coal extract. The fragments of the cracked ring form the aliphatic side chains of the remaining nucleus. The quality of the 995 OF+ extract as determined by solubility characterization increased during the first three cycles of stage I1 but was then essentially constant, as shown in Table VIII. However, the extract quality variation was not great during stage 11. This justifies using donor solvent from stage I extract throughout stage 11, as discussed earlier. Oils are cyclohexane soluble, asphaltenes are benzene soluble-cyclohexane insoluble, and preasphaltenes are benzene insoluble-cresol soluble. The quality rating, of course, is oils > asphaltenes > preasphaltenes. The close similarity of the products in cycles 3 and 4 is further evidence that the system had made a close approach to steady state by cycle 3. Earlier Work More Optimistic Lower Upflow Velocity. A much lower upflow velocity was required to produce a similar extract ash level in the stage I1 program than in earlier work done at our laboratories by Gorin et al. (1977). CSF extract from Ireland Mine coal was deashed in both cases, using essentially the same experimental system. A comparison of the conditions and results of the previous and current programs is given in Table IX. The earlier results are much more favorable than the current data. Since the settling area is inversely proportional to the upflow velocity, the previous data imply about one-sixth the total area for a scaled-up design than would be projected from this work. While the previous work provided an excellent base for the current program, we feel that stages I and I1 were
540
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 3, 1979
Table IX. Extraction Conditions and Deashing Results from Previous and Current CSF Studies extraction conditions upflow wt% % donor solvent- velocity, extract program solvent to-coal in/min ash cycle 3 cycle 4
30 30
Current 1.8 1.8
0.3 0.3
0.19 0.20
0.9 0.9
0.22 3.30 0.16 0.16
Previous run run run run
A
B C
D
25 100 25 100
1.5 1.5 2.0 2.0
2.0
2.0
rigorous simulations of the start-up and operation to steady state of the Cresap Pilot Plant and that the pilot plant performance will be similar to that achieved in the work reported here. Nonequilibrium Solvent. The earlier work used Cresap Pilot Plant recycle solvent and hydro oil (donor solvent from the extract hydrocracking section of the plant). Good operation of the pilot plant extraction &ion was achieved during the 1968-1970 program but hydrocracker operation was limited. Therefore, limited donor solvent was available to the extraction system and the recycle solvent probably contained a significant portion of the start-up solvent, which was a low boiling petroleum derived aromatic. Table I11 shows that the Cresap solvents used in the work of Gorin et al. (1977) were much lower boiling than cycle 4 feed solvent. We infer from the boiling point data and from gas chromatographic analyses of the earlier solvents that both the “Recycle Oil” and the “Hydro Oil” were significantly lower in molecular weight than the steady-state cycle 4 solvent of this work. The mass spectral analyses of the stage I1 solvents (Table VI) show that molecular weight increased as steady state was approached. This suggests that the earlier solvents were not nearly as close to equilibrium as cycle 3 and 4 solvent of the stage I1 program of this work. We think that the large difference in settling rates between this work and the previous program was due mainly to the differences in the liquefaction solvents. The work with Neville solvent and Sure Sol 180 reported in this paper clearly demonstrated that different solvents can give vastly different settling rates. This is why such great efforts were made in this work to generate a steady-state liquefaction solvent. Recommended F u t u r e Work The major focus of the seven-month experimental program reported here was simulation of the start-up and
operation to steady state of the Cresap Pilot Plant, with particular emphasis on solids separation. Many of the deashing system variables were set at values known to produce satisfactory operation. Further work should be done in the following areas in order to optimize the deashing process. The removal of part of the liquefaction solvent prior to settling would reduce the operating pressure of the settler and permit the use of a higher boiling anti-solvent. Operating at less than the 600 OF used in this work may improve solids separation by precipitating more extract. However, too low a temperature could produce excessive precipitation and cause operability problems. The settler withdrawal rate used corresponded to about 35% solids in the underflow. The extract lost to underflow was about 20%. More extract could be recovered by reducing the settler underflow to correspond to 40-45% solids. A CSF process-derived aromatic light oil might give good solvent deashing without causing the operability problems encountered with Soltrol 130. A process derived liquid would be a poorer anti-solvent than Soltrol 130 so system operability would not be as sensitive. The effects of mixing rate, residence time, and temperature in the contactor should be studied. Acknowledgment Permission from the Department of Energy and Fluor Engineers & Constructors, Inc. to publish this work is gratefully acknowledged. This work was funded through Fluor Subcontract No. 448434-9-0002 as part of DOE Contract No. EX-76-C-01-1517. Dr. Everett Gorin, retired Director of Process Research for Conoco Coal Development Company, made many helpful suggestions during the planning and execution of this program. Literature Cited Curran, G.P., Struck, R. T., Gorin, E., I d . Eng. Chem. Process Des. Dev., 6, 166 (1967). Dudt, P. J., Kleinpeter, J. A., “Hydrocracking of Coal Extract from the CSF Liquefaction Process in a Spinning Basket ExperimentalReactor”,Preprints, Second Pacific Chemical Engineering Congress, Denver, Aug 1977. Gorin, E., Kulik, C. J., Lebowitz, H. E., Ind. Eng. Chem. Process Des. Dev., 16,95 (1977). Kleinpeter, J. A,, Jones, D. C., Dudt, P. J., Burke, F. P., I d . Eng. Chem. Recess Des. Dev., following paper in this issue (1979). Phinney, J. A,, Chem. Eng. Prog., 71(4), 65 (1975). U S . Department of the Interior, Office of Coal Research, “Project Gasoline Pre-Pilot Plant Phase-I Research on CSF Process”, R & D Report No. 39, Vol. 11, July 11, 1968-Dec 31, 1970.
Received for review August 16, 1978 Accepted February 6 , 1979 This paper was presented at the 85th National Meeting, American Institute of Chemical Engineers, Philadelphia, Pa., June 4-8,1978.
