A Laboratory Evaluation of Precoat Filtration Parameters for the Solvent Refined Coal Process Sidney Katz’ and Billy R. Rodgers Chemical Technology Division. Oak Ridge National Laboratory, Oak Ridge, Tennessee 37830
The effect of temperature, pressure, filter cake thickness, and filter cake compressibility upon the filtration fate of unfiltered oil from the Solvent Refined Coal liquefaction process was determined using a laboratory scale precoat filtration system.
Introduction In a recent evaluation of developments in coal liquefaction technology by EPRI (1974), the separation of the mineral matter and undissolved organic matter was identified as an unanswered critical problem area. The removal of inorganic constituents during coal liquefaction, a necessity for ecological, health, and operation reasons, has proven to be difficult because of the product viscosity, small size of the suspended inorganic particulates, and the small density difference between the liquid and solid phases. Although similar separations have been examined using hydroclones and centrifuges by Phinney (1975), only precoat filtration was reported by Scotti et al. (1975) and Schmid (1975) to reduce the sulfur and ash content to acceptable levels. Unfortunately, filtration rates achieved by precoat filtration have been low (approximately 10 gal h-’ ft-*). This paper describes a laboratory study of precoat filtration, the results of which define some important filtration variables and provide guidance in developing filtration systems. Materials Coal-Derived Liquids. The unfiltered oil was produced by the Solvent Refined Coal pilot plant a t Wilsonville, Ala., which is operated by Catalytic Inc. Illinois No. 6 coal was used as feed to the process, and it produced an unfiltered oil containing 3.19% ash. Filtered oil was the same liquid after precoat filtration at this laboratory and contained less than 0.15% ash. Precoat. Johns-Manville “Fibre-Flo” was used for all filtrations. Apparatus The principal component of the precoat filtration test system is the filtration assembly shown in Figure 1;auxiliary to it is a regulated nitrogen supply and a continuous weighing device. The constant temperature bath portion of the filtration assembly is constructed of standard 1.5-in. diameter pipe. It contains an end cap with a Ih-in. welded compression fitting through which passes the filter exit tube. Silicone oil is used as a heat exchange medium to permit high-temperature filtration. The filter is an 18-in. stainless steel tube of 0.173-in.* cross-sectional area. The exit tube is sealed to the bath during each filtration with a heat-resistant plastic ferrule. Compression fittings permit rapid assembly and disassembly for cleaning and operation. The diameter of the filter is large enough for convenient transfer of viscous fluids and to minimize wall effects during filtration, but small enough to limit filtration time to a few hours. The filter volume is several times the volume of a normal charge (10.5 cm3) to allow for head space. A wire screen (approximately 40 mesh) is cut to fit
snugly in the supporting wall of the reducing fitting and forms the precoat support. The precoat is deposited by introducing onto the support screen 800 mg of the material as a suspension in 15 cm3 of filtered oil. The compacting force is provided by a cylinder of nitrogen gas.
Results The Effect of Temperature, Pressure, and Filter Cake Thickness on Flow of Filtered Oil through Stationary Beds of Precoat and Coal Derived Solids. Three series of measurements were made by flowing filtered oil through precoat and through precoat plus filter cakes of various thicknesses. The first series used only precoat deposited as previously described and compacted by a pressure differential of 10 psi. For the second and third series, increments of filter cake were added to the same precoat bed by filtration of 3.5 and 4.7 g of unfiltered oil, respectively. The flow rate was taken as the slope of a plot of collected filtrate volume vs. time. About 10 min of flow was allowed a t each point to ensure that stable flow had been established. Results for the three series are shown in Figures 2 to 5 and demonstrate considerably higher flow rates for the precoat bed as compared to precoat plus coal-derived solids, Figure 3 indicates that the order of measurements has no significant effect on the results. The reproducibility of the measurements is indicated by the following pairs of points: (13, 181, (19,24),and (25,301. The importance of thin filter cakes in achieving high filtration rates is illustrated in Figure 5 . Average Bed Compressibility from Flow of Unfiltered Oil through Accumulating Coal-Derived Solids. Eight filtrations were made of unfiltered oil through precoat and the accumulating filter cake of coal-derived solids. In the previous series, filtrate vs. time results indicated a constant flow rate throughout the test period. In this series, the filtrate flow rate continuously decreased as filter cake accumulated. The curves of Figures 6 and 7 demonstrate that the effects of temperature, pressure, and filter cake thickness were similar to those discussed in the previous section. These curves may also be used to estimate bed compressibility, which must be evaluated to optimize operating conditions. The estimate of average compressibility was made in three steps. The data contained in Figures 6 and 7 were used to obtain plots in Figure 8 of an integrated form of Poiseuille’s equation for flow through parallel capillaries --0 - p t a o V p r 1P (VIA) 21P A Next, the slopes and intercepts were tabulated (Table I), and average cake resistance and resistance of the medium were calculated using viscosity and density measurements. Finally, the average compressibility was determined as the slope of a plot of the equation
-(-)
+-
Ind. Eng. Chem., Process Des. Dev., Vol. 15, No. 3, 1976
407
U t G/)S SUP,PL*
112" COHFf?5SYON
FITTIK.
7200'C
F I L T E R SUPPORT IN COMPRESSON FITTING
PI ASTIC FERRULE IN COMPRESSION FITTING
40
0
EXIT TUBE
Figure 1. Laboratory scale filtering assembly.
20
30 40 PRESSURE (psi1
50
60
Figure 3. Flow of previously filtered solvent refined oil through precoat plus a 0.Il-in.filter cake.
The numbers indicate the order In which the experiments were run
A 250 ' C
/ 150'C
z o o y
l 5 0 y
"
0
40
20 30 40 PRESSURE I P S I )
50
Figure 4. Flow of previously filtered solvent refined oil through precoat plus a 0.26-in. filter cake. 0 0
IO
20
30
40
50
60
70
PRESSURE ( p s i )
Figure 2. Flow of previously filtered solvent refined oil through precoat only. In a = s In AP
+ In a'
as shown in Figure 9. The average compressibility,s , was established as 0.49 from the slope, and the particle size constant was determined as 2.5 h ft4/gal in. CPlb1I2from the intercept.
Conclusions A laboratory-scale apparatus was developed which permitted evaluation of the precoat filtration parameters for the 408
Ind. Eng. Chem., Process Des. Dev., Vol. 15, No. 3, 1976
Solvent Refined Coal process. The flows through precoat alone are much larger than those through precoat plus filter cake; this demonstrates the small contribution to flow resistance from the precoat and provides estimates of the maximum flow rates possible with zero filter cake thickness. Filter cakes of the order of 0.1 in. thick or less should provide significant improvements in flow rates; only where very thin filter cakes are used does the precoat contribute materially to the flow resistance. Scale-up to large filtration equipment for Solvent Refined Coal liquids from Illinois No. 6 coal is possible by using the data in Figures 3 and 4. For other coal-derived liquids, the parameters, a , d,r , and s as determined in this study permit estimates of filtration rates by use of the filtration equation
120
I
I
I
I
I
1
1
100
125
150
175
I
110
I
I
25
50
1
75 V/A
x
1021ga1 ff2,
Figure 8. Effect of temperature and pressure on precoat filtrations. Figure 5. Effect of cake thickness on flow rate. Precoat
filtmtlon
at ISO0C
33
31
' 0 '0
26
0
' 0$ ' 2
2 3l 30
1
32
I
I
34
36
I
I
38
40
I
42
I
44 45
(On A P )
Figure 9. Determination of compressibility. TlMElminl
Figure 6. Filtration as a function of temperature at 50 psi.
Table I. Evaluation of Filtration Equation Parametersn
-e . . . ( ! ) + I . VIA
2AP A
AP
Condition
0
II
-
30
45
€0
?5
90
TIMElminl
Figure 7. Effect of pressure on filtration at 150 "C
105
P, Temp,
Pawl
psi
"C
2 9
cy
50 50 50
100 150 200 150
0.77 0.37 0.20 0.52 0.38 0.33 0.27 0.24
27.4 19.5 13.6 10.9 16.0 17.4 17.0 20.2
20 40 50 60 80
150
150 150
150
p1AP
r
0.50
3.84 1.59 0.58 1.27 1.45 2.04 2.72 1.27
0.14 0.04 0.28 0.16 0.18 0.20 0.07
' Calculations based on:
p (100 "C) = 1.097, experimental measurement; p (150, 200 "C) = 1.097, assumption; o = 0.432 lblgal, experimental measurement; p (100 "C) = 6.5 cP; p (150 "C) = 4.4 cP; p (200 "C) = 3.4 CP.
(eq 1). The compressibility of the filter cake, which was determined to be 0.49, indicates t h a t increasing the filtration pressure will only moderately increase the resistance of the filter cake t o flow. Ind. Eng. Chem., Process Des. Dev., Vol. 15,No. 3, 1976
409
Nomenclature V = volume of filtrate collected A = filter cake surface area r = resistance of medium (all items except the filter cake) s = compressibility % = time AP = pressure drop across cake, precoat and support p = viscosity of filtrate o = mass of accumulated solids (dry cake basis) per unit volume of filtrate ct = average specific cake resistance cy' = particle size constant
L i t e r a t u r e Cited Electric Power Research Institute Report 206-0-0, Parts II & 111, "Evaluation of Coal Conversion Processes to Provide Clean Fuels", Palo Alto, California, Feb. 1974. Phinney, J . A., Chern. Eng. Prog., 71,65 (1975). Schmid. B. D.. Chem. Eng. Prog., 71,75 (1975). Scotti. L. J.. Jones, J. F., Ford, L., McMunn, B. D., Chem. Eng. Prog., 71, 61 ( 19 75).
Received for review October 17, 1975 Accepted January 26, 1976
This research was sponsored by the Research and Development Administration under contract with the Union Carbide Corporation.
Hot Spot Simulation in Commercial Hydrogenation Processes Stephen B. Jaffe Mobil Research and Development Corporation, Research Department, Paulsboro, New Jersey 08066
The occurrence of steady-state hot spots in a commercial hydrogenation process unit has been explained in terms of limited regions of low flow. A mathematical model has been developed which accounts for the temperature rise with rapid reaction of the fluid in the affected low flow region and for temperature drop with the eventual mixing of cooler fluid from the surrounding region. By matching the commercial profiles, an estimate can be made of the velocity of the low flow region and its lateral extent.
Introduction The axial temperature profiles shown in Figure 1 were observed in a commercial hydrogenation process unit. They represent steady-state profiles taken at the same time through parallel thermocouple wells. These hot spots can potentially cause damage to both the catalyst and reactor vessel as well as degradation of product quality. I t is of great importance to understand the nature and origin of these hot spots and estimate their size. Careful inspection of the profiles shows that the precipitous temperature rises and falls along the length of the reactor are independent of the positions of the interstage cooling. Temperature rises in such processes are not unusual; what is extraordinary are the temperature falls. Since the reactor is adiabatic and under commercial hydrogenation pressure, endothermic reactions are negligible, we are led to the conclusion that the profiles represent disturbances which are local. The marked differences in the two profiles support this conclusion. I t is proposed that limited regions of low flow may be responsible for the hot spots. Reactants entering the affected low flow region convert at a rate greater than those in the surrounding region. The heat liberated by the increased rate results in a higher temperature. As the reactants pass out of the low flow region, they eventually mix with the cooler reactants from the rest of the bed and the heat is dissipated. The cause of such a low flow region, such as catalyst fines or physical obstructions, has been the subject of much work a t our laboratory and will not be discussed here. Rather we recognize that a limited region of low flow can indeed account for the hot spot phenomena and describe the system mathematically with the aim of estimating the lateral extent of the disturbed region. 410
Ind. Eng. Chem., Process Des. Dev., Vol. 15, No. 3, 1976
This paper provides a synthesis of the method of modeling heat release in petroleum hydrogen process of Jaffe (1974) and the method of modeling flow and mixing in packed beds of Deans and Lapidus (1960). The treatment, while obviously approximate, does serve to simulate the observed phenomena adequately. Heat Generation Heat is generated in petroleum hydrogenation processes by the net consumption of hydrogen (Jaffe, 1974). When hydrogen is consumed by a paraffin or naphthene through cracking or ring opening, a u C-C bond is broken and 7-10 kcal/mol of Hz is released. When hydrogen is consumed by saturating an aromatic, a TT C-C bond is broken and 14-16 kcal/mol of H2 is released. When an olefin saturates, a x C-C bond is destroyed and 27-30 kcal/mol is released. The heat released per mole of hydrogen consumed for these three types is remarkably constant over a wide range of hydrocarbon reactions with hydrogen. Conventionally, the petroleum hydrogenation reactions are modeled by identifying lumped species such as boiling point cuts in the petroleum mixture (Stangeland, 1974; Qader and Hill, 1969; Zhorov et al., 1971). A kinetic scheme is thereby devised accounting for the conversion of reactant lumped species into product lumped species. Rate constants for the kinetic scheme are determined by fitting experimental data. Once heats of reaction are assigned an adiabatic reactor simulation may be made. This approach, while generally useful at ordinary operating conditions for predicting conversion and selectivity, cannot be extrapolated to severe conditions present in the hot spot. What generally happens a t high temperatures to a model constructed in this way is that reactants become exhausted and the predicted reaction and
