Operation of adsorbers for purifying ammonia synthesis make-up gas

Apr 1, 1993 - Operation of adsorbers for purifying ammonia synthesis make-up gas. Gregory R. Schoofs. Ind. Eng. ... ACS Legacy Archive. Note: In lieu ...
0 downloads 0 Views 2MB Size
Ind. Eng. Chem. Res. 1993,32, 613-619

613

SEPARATIONS Operation of Adsorbers for Purifying Ammonia Synthesis Make-up Gas Gregory

R. Schoofst

Department of Chemical Engineering & Applied Chemistry, Columbia University, New York,New York 10027

The operation of adsorbers which purify ammonia synthesis make-up gas was assessed theoretically and empirically. A mathematical model predicts that nitrogen adsorption in 13X molecular sieve decreases the effluent flow rate and effluent molecular weight, and increases the effluent temperature at the start of the adsorption step. These interrelated, transient, process changes can greatly disrupt downstream unit operations. Data and observations from commercial ammonia synthesis gas adsorbers agree well with the effects predicted by the model. The effluent temperature rise can be used to monitor the condition of the adsorbent, the effectiveness of the previous thermal regeneration step, and flow distribution in the adsorber. A combination of activated alumina followed by 13X molecular sieve should reduce the transient process changes due to nitrogen adsorption, increase the capacity of the adsorber, and provide greater operating flexibility relative to 13X molecular sieve as the sole adsorbent. Introduction Many processes have been proposed and constructed to improve the yields and energy efficiency of ammonia manufacture (Slack, 1977). An important trend has been to remove as many contaminants as possible from the make-up synthesis gas (hereafter referred to as synthesis gas), which maintains the activity of the iron catalyst and improves the economics of ammonia production (Slack, 1977; Satterfield, 1980). The Braun Purifier Process (Grotz, 1967)was among the first to use adsorbers to reduce the total concentration of carbon dioxide and water in the synthesis gas to less than 1 ppm. The Braun Purifier Process also has a cryogenic purifier between the adsorbers and the synthesis loop which removes methane, argon, and carbon monoxide from the synthesis gas. Adsorbers which treat the synthesis gas have been retrofitted into many existing ammonia plants (Le Blanc, 1986), in part because of the technological and commercial success of the Braun Purifier Process (Slack, 1977; Setoyama et al., 1979; Verduijn, 1979). Adsorbers for treating ammonia synthesis gas typically contain 13X molecular sieve and operate in a cycle which generally consists of the following steps: adsorb, depressurize, thermally regenerate, cool, and repressurize. Two adsorbers are used. One adsorber is on stream adsorbing the contaminants, while the other is being reactivated by sequencing through the other four steps. It has been reported that the synthesis gas compressor downstream of the adsorbers sometimes behaves erratically and must be shut down after a freshlyreactivated adsorber is brought on line. This can subsequently force the entire ammonia plant to shut down. In this paper we analyze the performance of the repressurization and adsorption steps of adsorbers used to purify ammonia synthesis gas. We show that nitrogen adsorption in freshly reactivated 13X molecular sieve can change the flow rate, molecular weight, and temperature of the synthesisgas enough to severely disrupt the synthesis gas compressor and other downstream process operations. + Present

address: Schoofs Incorporated, 1675 School Street,

P.O. Box 61, Moraga, CA 94556.

We also describe practices which minimize the effects of nitrogen adsorption and the deleteriouseffects it can cause. Preferential nitrogen adsorption in 13X molecular sieve has been recognized previously and exploited in rapid cycle, pressure swing adsorption (PSA)processes (Ruthven, 1984; Yang, 1987). This paper shows that preferential nitrogen adsorption in 13X molecular sieve also occurs when a thermally regenerated adsorber is repressurized and brought on stream, and that the unwanted sorption of nitrogen can cause significant operational problems downstream of a thermally regenerated adsorber. Theory

A. Mat hematical Model. A one-dimensional equilibrium model was used to model the repressurization and adsorption steps of adsorbers for treating ammonia synthesis gas. The assumptions include the following: 1. The adsorber operates adiabatically. Large industrial adsorber vessels are usually well-insulated, and adiabatic operation is closely approached. 2. The synthesis gas is a binary mixture containing 65 mol % hydrogen and 35 mol % nitrogen. Other trace components are neglected. 3. The gas phase behaves ideally. At the highest pressure and lowest temperature considered herein, the compressibility factor of the synthesis gas is approximately 1.01. This value was calculated using the correlation of Lee and Kesler (1975), and is very close to unity which describes an ideal gas. 4. The gas phase moves via plug flow during the repressurization and adsorption steps. Axial dispersion is formally neglected in the sense that the mathematical model contains no second-order derivatives. 5. Equilibrium exists between the gas and adsorbed phases at all points within the adsorber. Nitrogen diffuses very rapidly through the pores of 13X molecular sieve (Ruthven, 1984), which eliminates the need to include intraparticle mass transfer in the model. The time constant for thermal diffusion through the adsorbent particle is roughly 5 s, which is much less than the time needed to repressurize the adsorber. Interparticle trans-

0888-5885/93/2632-0613$04.00/00 1993 American Chemical Society

614 Ind. Eng. Chem. Res., Vol. 32, No. 4, 1993

port resistances are approximated by treating the mass and energy balance equations as an equilibrium stage model, as discussed below. 6. The axial pressure drop is negligible. In ammonia synthesis gas adsorbers the repressurization rate is small enough so that the axial pressure drop can be neglected, based on the criteria developed by Sundaram and Wankat (1988). During the adsorption step the pressure drop is usually a few pounds per square inch, which is a very small fraction of the total pressure. Hence the pressure drop during the adsorption step can be neglected also. The mass balance for each component may be written

Table I. Boundary and Initial Conditions. boundary and initial conditions for process variable repressurization step adsorption step space velocity a constant LLO(t, z=L)IL 0.0 s-1 composition at end of y ~ ~ ( t = O , z ) nearly 1 repressurization step 0.35 YN2(t,Z=O) 0.35 pressure at end of P(t,z) P ( t ) given by the repressurization rate repressurization step, and initial pressure a constant temperature at end of T(t=O,z) 277.5 K repressurization step 277.5 K T(t,z=O) 277.5 K

where

Some process conditions are proprietary and cannot be divulged. It should be apparent, however, that specifying the process variables listed in the left column will permit the mass and energy balances to be solved numerically.

ci = yiP/RT

(2)

and (3) The component mass balances may be summed to yield the total mass balance

The energy balance may be written

where n

n

hf

cf(T - T,,,)

(7)

and

C = PIRT (8) Additionally, the energy balance should include Joule heating during the repressurization step, a thermodynamic phenomenon ignored in previous analyses of adsorption processes. The case of repressurizing an empty vessel by increasing the moles of gas in the vessel while keeping the vessel volume constant has been described previously (Dodge, 1944). We have extended this analysis by assuming that the vessel contains adsorbent and support media, and assuming that the gas, adsorbent, and support media are in thermal equilibrium. If the feed gas, adsorbent, and support media are at the same temperature the increase in temperature with pressure initially (TI), throughout the adsorber due to Joule heating is given by

aT - b P - 2ac - P(b2- 4ac)'/' -ap ~ U b2 P -( 4ac)'I2 where a = (m,c, + mbcb)(k - l)/vv b =P

+ Pl(k - 1)- aT1 c = -PkTl

(10) (11)

(12) Table I lists the boundary conditions which accompany the heat and mass balances for the repressurization and adsorption steps. The adsorber was cooled with nearly

pure nitrogen gas to a uniform temperature prior to the repressurization step. Synthesis gas containing about 65 mol % hydrogen, 35 mol 5% nitrogen, and trace amounts of contaminants enters the adsorber during both the repressurization and adsorption steps. B. Nitrogen and Hydrogen Adsorption in 13X Molecular Sieve. Physical properties of 13X molecular sieve were obtained from data sheets furnished by the Davison Chemical Division of W. R. Grace & Co. (1986). Nitrogen adsorption equilibria in 13Xmolecular sieve have been measured previously over the appropriate temperature and pressure ranges (Yang et al., 1982). Isosteric enthalpies of adsorption as a function of the amount of adsorbed nitrogen were calculated by applying the Clausius-claperyon equation to these data. The heat capacity of nitrogen adsorbed in 13X (Na-X) molecular sieve has not yet been measured. The heat capacity of nitrogen adsorbed in type X molecular sieves should be approximately 5R,corresponding to an adsorbed state with five vibrational modes (Barrer, 1978). The data of Barrer and Stuart (1959)indicate that the heat capacity of nitrogen adsorbed in K-X moelcular sieve approaches 5R as the temperature increases to 219 K and the amount of adsorbed nitrogen increases to 1.38 mol of Nz per kg of K-X molecular sieve. Pending better data, we used 5R for the heat capacity of adsorbed nitrogen in type 13X molecular sieve. No data exist in the open literature at present for hydrogen adsorption in 13X molecular sieve. Pressure swing adsorption processes for separating mixtures of hydrogen and hydrocarbons are known to produce extremely high purity (ashigh as 99.9999 % ) hydrogen, which implies that virtually no hydrogen adsorbs in 13X molecular sieve at temperatures near ambient (Cassidy, 1980). We assume that none does. C. Algorithm. The numerical integration of eq 1-12 employed backward finite differences in space and a fourth-order Runge-Kutta algorithm over time (Carnahan et al., 1969). All terms in these equations were retained, but the amount of adsorbed hydrogen was set equal to zero. Forty equilibrium stages were used in the simulation. This first-order backward difference discretization is equivalent to considering the adsorber as a series of wellmixed, stirred tanks, and the approximate numerical solution obtained this way is mathematically equivalent to incorporating the expected external mass-transfer resistances and dispersion into the model (Wen and Fan, 1975). The equilibrium model described here provides an upper limit to the nitrogen adsorption phenomenon and the problems it can cause. The model does not explicitly

Ind. Eng. Cham. Res., Vol. 32, No. 4, 1993 615 100.0,

WET SYNTHESIS GAS

DIMENSDNLESS AXIAL POSITION: \

1 0

0 00

0 05

.

I

.

.

........

,

0 50.0 6

,

I

I

.

.

....

,

~:~

40.0

30.0

*,

20.0 NITROGEN

10.0

0.0

13X MOLECULAR SIEVE

0.0

0.68

1 .oo

DRY C02-FREE SYNTHESIS GAS

Figure I. Schematic diagram of the admrber from which the data were acquired. The wet synthesis gas flows downward during the repressurizationandadsorptionsteps. Theadnorbent, 13Xmolecular sieve, is sandwiched between two layera of non-adsorbing ceramic balla. Corresponding dimensionlessaxial positions are listed to the right of the adsorber. The dimensionless axial position varies in proportion to the volume of adsorbent and ceramic balls.

include a variety of factors which may diminish the predicted extent of nitrogen adsorption and hydrogen enrichment, includingmass- and heat-transfer resistances, deviationsfrom plug flow, and adsorption of species other than nitrogen in the 13X molecular sieve. Experimental Section Data were collected from an industrial adsorber used to purify the synthesis gas to a modem ammonia synthesis converter. Figure 1 shows a schematic diagram of the adsorher and identifies the corresponding dimensionless axialpositions. Wetsynthesisgasflows downward through the adsorber during both the repressurization and adsorptionsteps. Non-adsorbingceramicballs placed at the bottom of the vessel support the adsorbent, type 13X molecular sieve. Another layer of non-adsorbing ceramic balls at the top of the bed helps distribute the gas flow evenly and protects the 13X molecular sieve from thermal and mechanical shocks. After accounting for the nonadsorbing ceramic balls, the model outlined in the previous section mathematically describes the operation of this adsorber. In this paper, the ‘bed” is defined to include the two layers of ceramic balls and the 13X molecular sieve. As depicted in Figure 1, the dimensionless axial position of 0.0 corresponds to the top of the upper layer of ceramic halls and the inlet end of the bed. The 13X molecular sieve lies between dimensionless axial positions of 0.05 and 0.68. The dimensionless axial position of 1.0 corresponds to the bottom of the lower layer of ceramic balls and the effluent end of the bed. The dimensionless axial position varies in proportion to the volume of adsorbent and ceramic balls.

0.1

0.2

0.3 0.4 0.5 0.6 0.7 0.8 DIMENSIONLESS AXIAL POSITION

0.9

1.0

Figure 2. Predicted axial hydrogen and nitrogen compositions in the ammonia synthesis gas adsorber after the repreasurization step. The adsorber contains nearly pure nitrogen cooling gaa at the start of the repressurization step. The adsorber is repressurized with synthesisgaa containing 65 mol % hydrogen and 35 mol % nitrogen.

Sensors placed in the adsorher entrance and exit pipes monitored the temperature and pressure during the adsorption step. The adsorber inlet and outlet gas compositions were also measured simultaneously at the start of the adsorption step. Sample bombs were purged with the respective gas before the sample was trapped. The samples were subsequently analyzed by gas chromatography. A continuous gas densitometer would provide more complete, time-resolved data, but unfortunately no such instrument was available.

Results Nitrogen adsorption in the freshly reactivated 13X molecular sieve during the repressurization step greatly enriches the hydrogen content of the synthesis gas inside the adsorber. Figure 2 shows the predicted gas-phase nitrogen and hydrogen compositions along the length of the adsorber after the repressurization step. The dimensionless axialpositions correspond to thcae shown in Figure 1 and described in the Experimental Section. At the inlet end of the bed, the nitrogen and hydrogen compositions are at the feed levels of 35 and 65 mol % , respectively. Nitrogen adsorption in the 13X molecular sieve depresses the nitrogen mole fraction in the middle of the bed. Near the effluent end of the bed, the residual gas remaining at the end of the cooling step gets compressed over the nonadsorbing ceramic balls, and the gas-phase composition approaches the cooling gas composition of nearly pure nitrogen. Figure 3 shows the predictad axial temperature profiles after the repressurization step. The dimensionless axial positions correspond to those shown in Figure 1 and described in the Experimental Section. Joule heating produces a uniform temperature rise of 4.2 K over the entire length of the adsorber. The temperature maxima located near the inlet end of the adsorber results from the large amount of nitrogen adsorption in this region. The temperature decreases between dimensionless axial positions of 0.13 and 0.40 as the amount of nitrogen adsorption decreases. In this range, very little nitrogen remains in the synthesis gas which entered the adsorber. The temperature slowly rises befweendimensionless axial positions of 0.40 and 0.68 because of increased nitrogen adsorptionfromtheresidualcoolinggasasitiscompressed toward the effluent end of the bed. The temperature profile reaches a plateau over the non-adsorbing ceramic

616 Ind. Eng. Chem. Res., Vol. 32,No. 4,1993

1.4 295.0

1.3

290.0

5 1.1 1.2

285.0

10.9

Y

Bn W-

E

7 7

pc

1.0

2 I + W

280.0 0

0.0

0.1

0.2

0.3 0.4 0.5 0.6 0.7 0.8 DIMENSIONLESS AXIAL POSITION

0.9

1.0

Figure 3. Predicted axial temperature profiles in the ammonia synthesis gas adsorber after the repressurization step. The adsorber is at a uniform temperature at the start of the repressurization step. I

0.6

300.0

g 20.0 0

Y

295.0

pc

2 15.0

W' pc

Y0

$ 290.0

0

= 10.0

E a

'

k

I

6

2

285.0

5.0

W LL

0.01' 0.0

"

'

I

1 .o

"

"

I

"

"

2.0 DIMENSIONLESS TIME

"

3.0

2.0 DIMENSIONLESS TIME

4.0

3.0

Figure 5. Predicted effluent flow rate for the ammonia synthesis gas adsorber at the beginning of the adsorption step. The dimensionless time equals the number of bed volumes of synthesis gas that have passed through the bed, based on an empty vessel. The dimensionless effluent flow rate is the ratio of the superficial effluent gas velocity to the superficial feed gas velocity.

325.0

z.

1 .o

0.0

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

,

1

1

1

1

,

,

,

,

,

,

,

,

,

,

1

i

' ' 1

4.0

Figure 4. Predicted composition breakthrough curve for the ammonia synthesis gas adsorber. The dimensionless time equals the number of bed volumes of synthesis gas that have passed through the bed, based on an empty vessel.

balls at dimensionless axial positions greater than 0.74;in this region the temperature increases because of Joule heating. Figure 4 illustrates the predicted composition breakthrough curve as the adsorption step commences. The dimensionless time equals the number of bed volumes of gas which have passed through the bed, based on an empty vessel (uo,2t/L).The nitrogen and hydrogen breakthrough curves have been combined to illustrate how the average molecular weight of the adsorber effluent gas changes with time. The enormous decrease in the effluent molecular weight results from nitrogen adsorption in the freshly reactivated 13X molecular sieve and the large difference in the molecular weights of nitrogen (28 g/mol) and hydrogen (2g/mol). So much nitrogen adsorbs in the 13X molecular sieve that the flow rate of the effluent gas decreases to approximately 78 % of the synthesis gas feed rate, as shown in Figure 5. Here the dimensionless effluent flow rate is the ratio of the superficial effluent gas velocity to the superficial feed gas velocity (uo,~/uo,I). Figures 4 and 5 indicate that the effluent composition and flow rate return to those of the inlet synthesis gas after about 1 bed volume of synthesis gas has passed through the adsorber. The composition front qoves down the column at a rate approaching the convective velocity because nitrogen is present in high concentrations and has a small enthalpy of adsorption. Gas samples collected from the adsorber are consistent with these predictions. Sample bombs at the adsorber

275.0 " 0.0

' g l l l l l l l l l l l l l l

10.0

20.0

1

1

1

1

1

30.0 40.0 50.0 DIMENSIONLESS TIME

1

1

60.0

1

1

1

1

70.0

1

1

1

(

~

80.0

Figure 6. Experimental and theoretical temperature breakthrough curves for the ammonia synthesis gas adsorber. The dimensionless time equals the number of bed volumes of synthesis gas that have passed through the bed, based on an empty vessel.

inlet and outlet were simultaneously purged with the respective gases for about three dimensionless time units before the samples were trapped. Subsequent analysis by gas chromatography indicated that the sample collected at the adsorber inlet had an average molecular weight of 11.2g/mol; the theoretical value equals 11.1 g/mol. The sample collected at the adsorber outlet had a lower average molecular weight of 9.4 g/mol. This value agrees closely with the predictedvalue of 9.3g/molobtained from Figure 4 by averaging the effluent molecular weight over three dimensionless time units. Figure 6 displays the close agreement between the experimental and predicted temperature breakthrough curves during the adsorption step. The dimensionlesstime equals the number of bed volumes of gas which have passed through the bed, based on an empty vessel. Approximately 20 dimensionless time units elapse before the effluent temperature reaches ita maximum value. This lag occurs because the heat released by nitrogen adsorption in the 13X molecular sieve must be transported past the thermal reservoir of adsorbent and ceramic balls located toward the effluent end of the bed. The temperature then falls for dimensionless times greater than 20 as the 13X molecular sieve and the ceramic balls are cooled by the incoming synthesis gas. The large heat capacity of the adsorbent and ceramic balls relative to the synthesis gas requires that roughly 50 bed volumes of synthesis gas must

~

Ind. Eng. Chem. Res., Vol. 32, No. 4, 1993 617

pass through the adsorber before the effluent temperature approaches the feed temperature. In the commercial unit studied, the bed was cooled to within 1 K of the feed temperature prior to the start of the repressurization step. Thus, the results presented here indicate that the temperature exotherm observed during the adsorption step is caused primarily by nitrogen adsorption in the freshly reactivated 13X molecular sieve, with a small contribution from Joule heating. The results clearly demonstrate that this temperature exotherm does not result simply from repressurization (Le., solely Joule heating), nor is it due to an incomplete cooling step because the bed was at a uniform temperature and composition at the start of the repressurization step. Water and carbon dioxide adsorption do not contribute much to the temperature exotherm observed during the adsorption step. Based on energy balances across the respective mass-transfer zones (White, 1988), we predict that water and carbon dioxide adsorption produces atotal temperature rise of less than 1K. It is important to recognize that the phenomenon of transient, preferential nitrogen adsorption and the associated downstream consequences occurs whenever the pressure is increased in an ammonia synthesis feed gas adsorber containing activated or freshly reactivated 13X molecular sieves. This occurs after each thermal regeneration, and also during the initial startup of the adsorber and during subsequent startups followingevery shutdown. Discussion

A. Effects of Nitrogen Adsorption Downstream of the Adsorber. Nitrogen adsorption in activated or freshly reactivated 13X molecular sieve during the repressurization step and the beginning of the adsorption step changes the effluent gas flow rate, molecular weight, and temperature. These process changes can produce dramatic transient effects downstream of the adsorber. If the freshly reactivated adsorber is quickly switched on line while the other adsorber is simultaneously taken off line, the multistage synthesis gas compressor downstream of the adsorbers will behave wildly as it responds to each of the large process changes resulting from nitrogen adsorption. The flow rate at the compressor inlet will decrease immediately after the freshly reactivated adsorber is brought on line. Lower molecular weight gas subsequently enters the compressor after it travels through the piping from the adsorber to the compressor. After the molecular weight of the effluent gas returns to the design conditions, warm gas emerging from the adsorber enters the compressor. Large, sudden changes in the flow rate and molecular weight can cause the compressor to surge, which in turn can damage the compressor and force the entire ammonia plant to shut down (Hile, 1967; Fromm and Rall, 1987). This is the first paper to attribute these incidents to transient, preferential nitrogen adsorption in activated 13X molecular sieves, however. Transient, preferential nitrogen adsorption may also affect the performance of the synthesis loop. Increasing the concentration of hydrogen relative to nitrogen impairs the ammonia synthesis reaction because dissociative adsorption of nitrogen on the iron catalyst is the ratelimiting step (Stoltze and Norskov, 1985). Thus a lower partial pressure of nitrogen in the synthesis loop decreases the rate of ammonia production. B. Adsorber Operation. Ammonia plant operators have judiciously selected operating practices which minimize the deleterious effects of nitrogen adsorption and permit smoother process operation. First, operators do

not quickly switch the freshly reactivated adsorber on line and simultaneously take the other adsorber off line. Instead, a small fraction of the process flow is diverted through the freshly reactivated adsorber and subsequently bled into the bulk of the process flow which has passed through the other, on-line adsorber. Mixing the parallel process streams after they have passed through the adsorbers dilutes the flow rate, molecular weight, and temperature changes due to nitrogen adsorption in the freshly regenerated 13X molecular sieve and minimizes the deleterious effects downstream of the adsorbers. Second,the process flow directions and sequences tend to dilute the hydrogen enrichment. The synthesisgas flows down (in the same direction that gravity acts) during both the repressurization and adsorption steps, and a long standby period can be used between the repressurization and adsorption steps. Nitrogen adsorption in the 13X molecular sieve during the repressurization step produces a composition gradient with hydrogen-enriched, lowdensity gas in the middle of the adsorber (Figure 2). Although the temperature gradient weakly counteracts the composition gradient, the criterion for quiescence (Rosner, 1986)

Gr'/4S~'/2