Steam Purging andFoundation Repair of anAtmospheric Ammonia

Storage Tank

Liquid and vaporous ammonia can be safely and quickly cleanedfrom an atmospheric ammonia tank using the direct injection of lowpressure steam.

K.A. Vick, J.B. Witthaus, H.C. MayoFarmland Ind., Inc., Lawrence, Ks.

In May of 1978, a 6,800 metric ton atmosphericammonia storage tank at Cooperative FarmChemicals Association in Lawrence, Kansas,was cleared of NH3 for the purpose of cleaningand inspection. A major portion of the resid-ual liquid ammonia was purged from the tankby direct injection of steam. The tank isone of two identical tanks at Lawrence whichwere built in 1959. Neither of the tanks hadbeen entered since being commissioned. Thetanks were designed without sumps in thefloor and the main cold product pumps losesuction when there is still 30 centimeters(\\h inches) of liquid in the tank. Further-more, the only connections flush with thefloor of the tanks are three 2.5 centimeter(1 inch) nozzles. Each tank contains a 3.8centimeter (1% inch) heating coil around theinside circumference of the tank floor. Thisheating coil was designed to be used with acirculating glycol solution to vaporize theremaining ammonia liquid. However, calcu-lations indicated a minimum of 16 days wouldbe required to boil off the remaining liquid.Having one tank out of service for an ex-tended period of time severely restrictsoperating flexibility at the Lawrence complex.Therefore, the development of a safe, yetquick way of evacuating the tank vas needed.

Ammonia-Steam Equilibrium Calculations

A proposal was made that we purge theliquid ammonia from the tank by direct in-jection of steam to vaporize the liquid. Toinsure the performance and safety of the0149-3701-80-3910 $01.00 © 1980American Institute of Chemical Engineers

steam injection method, calculations weremade to model the system.

The first area of investigation waswhether there was any chance that a vacuumcould be pulled on the tank by the condensingof ammonia vapors. To study this possibility,calculations were made as to the effect ofinjecting 450 kg. (1,000 pounds) of 1,136 kPa(150 psig) steam into three different volumesof saturated ammonia vapor at -33°r. (-28°F).By performing an energy balance on the steam,initial ammonia vapor, and the resultant liq-uid and vapor, the final vapor volume can bedetermined. The total enthalpy of the systemafter the injection of steam can be deter-mined by:

H system = hstm x (Steam Injected)

x (Initial Ammonia Vapor)

If any liquid is formed due to condensationof vapor, it must be in equilibrium withthat vapor. The equilibrium relationship andenthalpies of the aqua solutions can be de-termined from the table of aqua ammoniaproperties shown in reference (1). Withthese relationships, the following equationapplies to the tank after injection of steam.

H system = h[_ L + hv V

! Jhere :

h|_ = enthalpy of aqua ammonia liquid

54

h = enthalpy of vapor inv equilibrium with liquid

L = liquid mass

V = vapor mass

Let T = L + V

Then by rearranging:

H system = h,_ T + (hy - hL) V (1)

Now by assuming an ammonia concentration ofthe liquid, the enthalpies can be obtainedfrom the aqua tables and the kilograms ofvapor remaining can be determined directlywith the Equation (1). With the new vapormass determined, the liquid concentration canbe calculated and a check on the assumedconcentration is made. This interative pro-cess is repeated until the assumed and cal-culated compositions are equal. From theenthalpy of the vapor, the temperature anddensity can be calculated and the final vaporvolume can then be determined. If this finalvolume is equal to or greater than the initialvolume of vapor, then the tank will not pulla vacuum.

A summary of the cases can be seen inTable 1. In Case 1, the heat liberated by thesteam into the small volume is large enough toprevent any condensation. In cases 2 and 3,some ammonia is condensed, but the equilibriumtemperature has increased enough to cause theremaining vapor to expand beyond the initialvolume of all the ammonia vapor. When thereis liquid ammonia in the tank initially, thesafety of the steam injection is even furtherenhanced because additional ammonia vapor isformed.

Once the safety of the steam injectionmethod was established, calculations were madeto determine the amount of ammonia vapor thatwould be generated and to what degree it couldbe recovered. This was done by making anenergy balance around the tank after every4,540 kg. (10,000 pounds) of steam added. Byusing Equation (1) and interating on theliquid ammonia concentration, the vapor gener-ated was determined. The data points generatedby this technique were then connected with aline to produce a plot of ammonia vapor over-head as a function of the steam injected, asshown in Figure I. This curve is based on 113metric tons (125 short tons) of liquid ammoniaand 9 metric tons (10 short tons) of ammoniavapor in the tank before steam injection.

Initially, the vapor rate is over 2.5 timesthe steam injection rate, but this rate dropsoff as the solution becomes more dilute.After about 45,000 kilograms (100,000 pounds)of steam have been injected, the vaporizationrate becomes equivalent to the steam rate.When this curve is extended out, the trendwill continue with progressively less ammoniavapor being recovered per kilogram of steaminjected. Figure II shows the weight percentammonia in the residual liquid as a functionof the steam injection. This indicates thatat a liquid concentration of approximately 30percent ammonia by weight, the one to onevapor-steam relationship is reached.

It was proposed that this ammonia vaporwould be recovered with the existing refrig-eration compressors and then be dumped backinto the second storage tank. This was to bedone by adjusting the steam rate to hold thedesired tank pressure and running all thecompressors at full rate. The limiting factoron how much vapor could be recovered with thecompressors was the water content of the vapor.Small amounts of water in the ammonia productare acceptable, but as the liquid concentra-tion decreased towards 30 percent ammonia,the water vapor going overhead starts to in-crease rapidly. In Figure III, the equilib-rium water vapor concentration is plottedagainst the temperature in the liquid. Oneother variable to consider when steam inject-ing was the volume of liquid remaining in thetank after equilibrium is reached. Thisliquid volume is plotted against steam injec-tion in Figure IV. If the volume of the aqua-ammonia solution to be disposed is more of adetermining factor than the concentration ofthis solution, then the steam injection shouldbe stopped when the minimum liquid volume isattained. This minimum volume is reached atan ammonia concentration of approximately 30percent.

In order to monitor the concentration ofthe aqua solution while injecting steam, thetank temperatures were observed. Figure Vshows the equilibrium temperature as a func-tion of the liquid ammonia concentration.Several existing temperature probes locatedin the vapor space at different heights in thetank were used for these measurements. Theseprobes were also used as a check on steambypassing. If the steam did not condense inthe vapor space, it would cause the vaportemperature to increase above the calculatedvalue and the temperature indicators wouldshow this problem. To minimize the risk ofsteam bypassing, the steam should be injected

55

as far from the vapor outlet nozzle aspossible. No bypassing was indicated duringthe injection of steam, but monitoring theoverhead vapor temperature is a good pre-caution. The overhead vapor line is a goodlocation for a temperature indicator if con-nections are not available in the side of thetan k i

Modifications to Piping

Preparations for purging the tank werecomplicated by the fact that the majority ofthe tank piping was common to the secondstorage tank. Figure VI is a simplifiedpiping diagram of the facility. In order tofacilitate continued operation of one tankwhile the other tank was out of service,piping modifications were required. Due tothe fact that both tanks shared one vapor lineto the refrigeration compressors and the flare,a bypass line was installed as shown in FigureVI. This allowed continued refrigeration ofone tank while the other tank was venting tothe flare. The bypass line was installedahead of time without interrupting the opera-tions of either tank. As previously mentioned,the only connections flush with the floor ofthe tank are three 2.5 centimeter (1 inch)nozzles directly across the tank from theproduct pump suction line. These nozzles arealso on the opposite side of the tank fromthe vapor outlet nozzle, which helps toprevent bypassing of the steam. Therefore,these nozzles were used for both steam in-jection and liquid pumpout. The temporarysteam injection and liquid pumpout piping isdetailed in Figure VII. This piping was alsoinstalled ahead-of time, and for the most part,consisted of screwed pipe and flexible ammonialoading hose. With completion of thesemodifications, the tank was reading forpurging.

Ammonia Recovery and Purging

Ammonia was removed from the tank in thenormal manner until the product pumps lostsuction. At this point, there were approxi-mately 30 centimeters (11% inches) of liquidremaining, which corresponds to 113 metrictons of ammonia. At an average vapor temper-ature of 4°C (40°F) there was an additional 9metric tons of ammonia vapor in the tank. Allthe available refrigeration compressors werethen placed in service and steam injectionwas initiated. The steam injection rate wasadjusted to maintain the tank pressure at995 Pa to 1120 Pa gauge (5 inches hLO gauge).The tank pressure was very responsive to

adjustments of the steam rate, and the in-jection rate leveled out at 680 kilograms(1500 pounds) per hour. At this point, anattempt was made to pump the remaining liquidammonia into the second storage tank using agear type pump. However, the pump performancewas hampered by the presence of viscous oilin the suction lines. Therefore, this effortwas abandoned. Steam injection into the tankwas maintained at 680 kilograms per hour for75 hours. The tank temperature and pressurewere closely monitored during the time.Figure VIII represents the tank temperatureas a function of steam injection, with thetank temperature being the temperature 4.5meters (15 feet) up on the sidewall of thetank. Due to the fact that the tank neverreached equilibrium conditions and heat waslost through the tank walls and floor, theammonia recovered was less than predicted byFigure I. An estimated 91 metric tons (100short tons) of NH^ were recovered from thetank by the refrigeration compressors duringthis time leaving a 30 percent aqua-ammoniasolution in the tank. At this point,attempting to recover additional ammonia usingthe refrigeration compressors was uneconomicaldue to the increasing amount of water in thetank. Therefore, the decision was made toopen the bypass line to the flare and blockin the line to the compressors. With thetank open to the flare, steam injection wasincreased to 1590 to 1850 kilograms (3,500to 4,000 pounds) per hour. The injectionrate was maintained at this level for 20 hoursresulting in the removal of approximately 13.6metric tons (15 short tons) of NHo and anaverage tank temperature of 34°C (93°F).Approximately 95 cubic meters (25,000 gallons)of 10% NH3 remained in the tank. Steaminjection was then reduced to 227 kilograms(500 pounds) per hour to maintain pressure inthe tank, and removal of the aqua-ammoniasolution was initiated using a gear typepump. Forty-five cubic meters (12,000gallons) of the solution were pumped into atank truck, steam stripped, and dumped intoan oil separation pond. The concentrationof the solution dumped into the pond averaged5% NHß. At this point, before the tank wasopened to the atmosphere, the block valves inthe bypass line to the flare were closed as aprecaution against a flare initiated explo-sion. The top and bottom manyways werecracked open and two fire hoses were droppedthrough the top manway. In order to knockdown the remaining ammonia vapors in the tank,water was sprayed into the top of the tankfor twelve hours. The resultant solution inthe tank was only 5% ammonia. A high volume

56

sump pump was then dropped into the tankthrough the bottom manway and the solutionwas pumped directly to the separation pond.The tank was flushed an additional twelvehours using the fire hoses and a sump pump.An air blower was then placed in the bottommanway. Four hours after the air blower wasinstalled and 7 days after the first steamwas injected, the tank was entered. Norespiratory equipment was necessary and onlya slight smell of NHß remained in the tank.Visual inspection revealed the tank to beclean except for a thin film of oil on thefloor and the walls and the usual weldingscrap. The oil and NH3 smell were removed byblasting the walls and floor with firehoses.

Tank Inspection

A wide variety of inspection techniqueswere used to determine the condition of thetank. Roughly one-third of the floor weldswere vacuum tested with no leaks discovered.All the nozzle welds around the bottom of thetank were dye checked, and the wall to footerplate weld was inspected using the magneticpartical method. Again, no cracks were found.In addition, random X-rays of the sidewallseams revealed some weld porosity and slaginclusion; however, the weld flaws were notserious enough to warrant repair. Visualinspection of the tank floor revealed auniform settlement of the floor plate alongthe periphery of the tank, starting about 1meter (3 feet) out from the wall. Therefore,a detailed survey of the tank floor was made.The results of the survey are shown in FigureIX. The tank was originally constructed witha uniform slope in the floor of 12.7 centi-meters (5 inches) from the center of the tankto the walls. The floor profiles generallyfollow such a slope with the exception of theoutside 1 meter. Figure X shows a detail ofthe floor profile of the periphery of thetank. The average drop in the floor is 12.0centimeters (4 3/4 inches) in the outsid^ 1.meter. Due to the sharp bending of the floorplates in this area, a consultant was broughtin to perform a structural analysis. Theconsultant, Weidlinger and Associates, 110East 59th Street, New York, NY 10022,determined that the floor plate had gonethrough the yield condition; however, due tothe high quality of steel used, no crackinghad occurred and the floor was judged safe.The structural point of greatest concern wasthe floor plate to footer plate weld. There-fore, 30 meters (100 feet) of this weld wererandomly X-rayed. The X-rays failed to revealany problems in the weld.

Modifications to Tank Foundation

The two Lawrence tanks have a designpressure of 3.23 kPa gauge (13 inches H20gauge), sit on a gravel base and do not bearon a ring wall. In addition to this, theundertank heating has not been used forseveral years and the ground below is com-pletely frozen. There is a hard, non-porousshale layer located between 2 and 3 metersbelow the surface. Expansion due to freezingof this shale layer is small. Calculationswere made on the expansion of the gravellayer from the surface down to the shale layerwhen all the void space was filled with waterand frozen. The maximum predicted growth ofthis layer was less than 3.8 centimeters(1.5 inches).

The probable cause of circumferentialsettlement was the high loads imposed by thetank wall on the 15. 2 centimeters (6 inches)wide footer plate. This caused large shearstresses in the soils immediately below thefooter plate, resulting in settlement. Dueto a lack of foundation under the tank wall,there was a possibility of future settlementaround the circumference. This, combined withan uplifting of the tank in the center whenthe ground refroze, lead to the decision toinstall a concrete ring wall under the tank.The purpose of the ring wall is to provideuniform bearing for the tank wall load. Across section of the new foundation is shownin Figure XI. The ring wall was installed in15 segments, with a maximum of three segmentsexcavated at one time. Eight monitoringprobes have been welded to the bottom of thetank wall for use in monitoring future tanksettlement. The results of the latest surveyare shown in Table II. These indicate a netrise of the foundation since tank refillingof 1.1 centimeters (0.43 inches).

Conclusions and Recommendations

Based upon our experience with the tankat Lawrence and a 13,600 metric ton tank atHastings, Nebraska, the conclusion can bemade that liquid and vaporous ammonia can besafely and quickly cleaned from an atmosphericammonia tank using the direct injection oflow pressure steam. The time and moneyrequired to complete such a project iscontingent upon the tank size, the tankdesign, the amount of NH3 recovered, and therefrigeration capacity available. Neverthe-less, it is reasonable to expect that themajority of atmospheric ammonia tanks in theUnited States which have a readily available

57

source of steam, can be cleared, cleaned,inspected, and returned to service in lessthan two weeks at a minimum expense.

LITERATURE CITED

1. Jennings, Burgess H., and Shannon, FrancisP., "Thermodynamics of AbsorptionRefrigeration," Part 1. RefrigeratingEngineering. May, 1938.

VICK,K.A. wrrraAus,j.B. MAYO, B.C.

DISCUSSION

WALT LEHLE, Gulf Central: Do 1 understand that thetank after repair still doesn't have any bottom heater?

KEVAN VICK: No, under tank heaters are still not inservice.LEHLE: And you haven't gotten that small rise?VICK: Pardon?LEHLE: The small rise that you indicated on the side -that's all that the tank bottom rose? Is that what you aresaying?VICK: After it was returned to service, it rose 1.1centimeters. This was done to the refreezing of the ground.LEHLE: Did you find the same condition in the other tank?VICK: At Hastings?LEHLE: YesVICK: Those tanks are of a different design and they haveunder tank heaters which are in service all the time. We didnot fine the same condition in those tanks.LEHLE: Did you make any attempt to try to unplug thosedrains.VICK: The 1" nozzles?LEHLE: YesVICK: Yes, they were unplugged by injection of steam.

LEHLE: But you didn't use them to drain the ammonia?VICK: We attempted to use a gear pump to pump theammonia from that tank to the second tank. However, theammonia vaporized. There simply was not enough suctionhead, and we weren't successful in doing that.LEHLE: We have done this a number of times, using agear pump to reduce the amount of liquid down to amanageable amount. However, we don't have any bottomoutlets such as you described. What we have are diptodeswhich go down to within an inch of the bottom of the tank, onthe ring wall. So, we are able to remove almost all of theliquid ammonia.JON BLANKEN, UKF, Holland: I would like tocompliment you on the method you found and the way youpresented it. We have inspected four tanks, three of themon a pile foundation, and we emptied them by putting spaceheaters in-between the ground and the foundation, whichtook a long time. We emptied the fourth tank by injectingaqueous ammonia which can be done, as has been provenbut, which will take more time than the method you haveused? I would like to add that the four tanks we haveinspected we never found anything wrong with them. Thatmeans no stress corrosion cracking or anything.

58

TABLE l

EFFECT OF INITIAL STEAM INJECTION OFAMMONIA VAPORS

BASIS: 450kg. OFII36kPa STEAM INJECTEDINITIAL AMMONIA VAPOR TEMP.-33°C.

INITIALAMMONIA LIQUID FINAL

CASE

I2

3

VAPORVOL.(m3)

85

1160

10,000*

FORMED(kg)

NONE

80 at 10%CONC.469 at 35%CONC.

VAPOR TEMPVOLJtm3) (°C)

31,268

2,272

12,363

9570

25

* ESTIMATED VOLUME OF TANK.

BASIS:

113 METRIC TONS OFPURE AMMONIA LIQUIDIN TANK INITIALLY AT-33.3°C AND ATM.PRESSURE.

0 10 20 30 40 50 601000 KILOGRAMS OF 1136 kPa STEAM INJECTED

Figure 1. Ammonia vapor generated vs. steam injectioninto an atmospheric ammonia tank.

TABLE IT

PROBE

12345678

TANK SETTLEMENT SURVEY

INITIAL MEASUREMENTON 7/12/78 (METERS)

30.19130.1974

30.199030.205430.186330.197430.1831

NEW MEASUREMENTON 3/30/79 (METERS)

30.200630.2054

30.210130.214830.194330.203830.1879

CHANGE(cm.)

0.950.80

Ml0.950.8O0.640.48

59

HluOCEof 100H5 90

2

80

70

eo8 50

<

40

30

20

10

BASIS:113 METRIC TONS OF PUREAMMONIA LIQUID IN TANKINITIALLY AT -33.3°C. ANDATMOSPHERIC PRESSURE.

w iS O 10 20 30 40 50 601000 KILOGRAMS OF IB6kPa STEAM INJECTED

Figure 2. Residue liquid concentration vs. steam injectioninto an atmospheric ammonia tank.

.

oQ.

Szcc1zUJo(gUJCL

1-

Oai

Sz>(Tm_jZ)

S

10

9

8

7

6

5

4

3

2

1

0

BASIS:ATMOSPHERIC PRESSURE

-30 -20 -10 0 10 20 30 40 50TEMPERATURE OF AQUA SOLUTION (°C)

Figure 3. Water in overhead vapor as a function of liquidtemperature.

60

ISO

I-100

§90S2 80

§70

*60

50ÜJ

I 40

l30

§20

10

BASIS:113 METRIC TONS OFLIQUID AMMONIA INTANK INITIALLY AT-333°C.AND ATMOSPHERICPRESSURE.

£ O 10 20 30 40 50 601000 KILOGRAMS OF H36kPa STEAM INJECTED.

Figure 4. Residual liquid volumes vs. steam injection intoan ammonia storage tank.

80

70

60

50

o

a:

QC

10

E-1go

g-20S

-30

BASIS:ATMOSPHERIC PRESSURE

0 10 20 30 40 50 60 70 80 90 DOWT PERCENT AMMONIA IN AQUA SOLUTION

Figure 5. Equilibrium temperature vs. aqua ammoniaconcentration.

MM»

-&U&T.-NBW

I

TAMK

flPUMPS

TO MOLDIKJ6 COMPR-BSSOeS

Figure 6. Simplified piping diagram.

V"7"/ ih-<s«>

64.VA. ,- T J- MM" l »

Vv~A e &A-VA.&6«!PUMP

BE VA.

*ex^*

J

5

-rS

l'"lm=0 &AI^K TRV

STEAMdPAIEGEft

XONen

[I5O PS1& 6TBAMI (TO S6COWP STORA66 TANK |

Figure 7. Temporary steam injection and liquid pumpoutpiping.

61

•MU),

IS 30 45 80

Steam - M Kgs75

Figure 8. Tank temperature vs. steam injected.

0 .a 3.05 6.10 ai4 12.19 15.24 I2J9 914 6.10 105 310 *

-12.7

N-24J -156 -10£ -SS -7.0 -3.2 0 +3.2 -1.6 -1.3 -3.8 -10.5 -225

NORTH-SOUTH PROFILE

-12.7-7

-22.7 -S8 -67 -3.2 -.8 0 +3.5 +.3

-̂-12.7

\W-6.7 -7.0 -79 -23.5

EAST-WEST PROFILE

»•DISTANCE FROM TANK WALL IN METERS.NOTES: PROFILE *'s ARE MEASURED IN CENTIMETERS RELATIVE TO THE TANK t.

BROKEN RED LINES REPRESENT THE DESIGN PROFILE.

Figure 9. Tank floor profiles

BOTTOM t PROFILE(EAST-WEST)

BOTTOM t PROFILE(NORTH-SOUTH)

aVERTXcm.)o

HORIZIcm)

IBin(Oö

to oif) rn

1.1I

_£_ I00

EXISTINGRING WALL

u)-SAND LEVELING PAD. ~ (121.9cm.))-4"( 10.2cm) FOAM GLASS INSULATION.

©-POLYETHYLENE SHEET.© - 5-10" LAYERS OF CRUSHED ROCK.

Figure 11. Ring wall foundation.

Figure 12. Closeup of new tank foundation.

Figure 10. Bottom plate profile at periphery of tank. Figure 13. Installation of new tank foundation.

62