Paper No.

COIIRI?OSIONc)L,
The

391

NACE 1Mernational

Annual Conference

and Expositiorl

MORE EXPERIENCES V/ITH CORROSION AND FOULING IN A REFINERY AMINE SYSTEM

Michael J, I,itschewski Sun Company, Inc. 1819 Woodville Road Toledo, Ohio 43616
ABsrlucT

This paper describes a roller coaster ride of corrosior[ and subsequent fouling in a Refinery MDEA system. The accelerated corrosion was first initiated by addition of caustic and the following up and down corrosion rate was a result of operating conditions imposed by increased sour crude charge, fouling and mis-application of MDEA. System vari:.bles that were controlled during this period included equipment metallurgy, the addition of caustic to neutralize heat stable salts (HSS), ion exchange to remove HSS and sodium, amine circulal ion rate, reboiler steam rate and the injection of corrosion inhibitor, PROCESS DE:SCRIPTION The amine plant, shown in Figure 1, processes refinery sour wet gas from the Fluid Catalytic Crackin, Unit (FCC), and crude units and dry gas from other tefinery streams. Sour propane - propylene is treated in a liquid amine absorber. The amine regenerator or stripper has a horizontal kettle type reboi using nominal 60 psig steam. Filtration is provided by 5-10 micron bag type and a clay filter. The ricl amine flash drum operates at 5 psig The Regenerator and Dry Gas Absorber have 21 and 20 trays while the Wet Gas Absorber and the Liquid Treater have packed beds. Trays, associated ~ttachment hardware and support rings are 304 stainless steel. Packing is 316 stainless. The Lean ~ mine Cooler and Regenerator Overhead Conden: are SB-33 8-2 Titanium. Of the two Lean/Rich exchangers, one is SA-214 carbon steel, the other 31 stainless, System volume is about 23,000 gallons clf 33-35/0 MDEA. Normal acid gas loadings are 0,20 mol acid gas/mol rich MDEA and ,0014 mol acid gas/mol lean MDEA.
Copylight 01996 by NACE International. Requests for permission to publish this manuscript in any form, in part or in whole must be made in writing to NACE International, Conferences Division, P.O. Box 218340, Houston, Tex=s 77218-8340. The material presented and the views expressed in this paper are solely those of the author(s) and are not necessarily endorssc by the Association. Printed in the U.S.A.

Background The system used MEA for 10 years but was switchetl to a proprietary MDEA in 1986. The amine change was to reduce energy consumption and corrosion which allowed for elimination of an expensive corrosion inhibitor program, Corrosion was a major concern while using MEA, The regenerator was replaced in 1985 after 9 years on MEA and reboiler bundles were replaced on an almost yearly basis, In a effort to improve bundle life several metal lurgies were tried (on MEA) inclL ding 304L, 316, and Sandvik 3RE60. These failed, often in a very short time. Several factors contributed to failure, including excessive vibration from disengaging gases, non-flooded top tubes and corrcls on. Reboiler life remained our primary corrosion concer I after switching to MDEA although some corrosion had been observed in the bottom of the regenerator, Routine amine analysis indicated that the amine was well stripped but poor sampling techniqulx contributed to low Acid Gas Loading (AGL) results. Acid Gas Loading results of a sample colle{:ted with a small water cooler indicated periods cf higher AGL in the lean amine. Higher AGL COUICcontribute to corrosion in the regenerator and reboiler, Corrosion monitoring consisted only of a carbon st:~ 1coupon in the liquid outlet piping of the reboiler. This coupon was mounted on a Type 316L stainless holder and inserted into the stream via a valve and packing. Normal historical corrosion rates were less than 8 mpy. The same holder and coupon metal Iurgy was used through the entire reported history. In October 1989, a newly designed bundle was placed in service after a failure with MDEA as the solvent, The bundle had an X pattern of omitted tubes to allow for gas disengaging, The carbon steel tubes had a heavier wall (O.120) than previous bundles (0.065). November 1991, several tubes were plugged anc the bundle was eventually replaced in March 1992 Bundle life had increased from 10-12 months to 18-;0 months. Inspection of the bundle showed severt pitting and weak acid or acid gas attack, Failure WM in the top 4-5 rows of tubes where general OD thinning had reduced tube diameter sufficiently to al ow tubes to vibrate in the baffles and wear through by fretting.
In

Corrosion coupons of several metal lurgies were inst~ lled on the March 1992 bundle and another set on the new bundle installed in September 1993. The coupons were installed on the back baffle as close to the heat transfer surface as possible. The results are given in Table 1. Sour crude charge was gradually increased such thal average weight percent sulfur rose from 0,34 wt% to 0.46/0 in 1994, Prior to 1989, unit shutdcwns were taken every 2.5 to 3 years, at the time of this incident the unit was 3 years into an eventual 5 ~ear run.
in 1985

DISC1J3SION Caustic and Corrosion The MDEA solution contained about 2.6 wt% HSS ESMDEA at the time of the March 1992 bundle replacement. The amine was clean even though the :]recoat filter was out of service. The clay filter WM

keeping the particulate concentrations

acceptably 10W.

After viewing the reboiler bundle, the amine supplie - recommended neutralization of the HSS with caustic. Neutralization of HSS with caustic has beer reported in literature for MEA and DEA but little had been published on neutralization of HSS with caustic in MDEA solutions in 19921Z, The basic concept is to react a strong base with the acids that accumulate in the amine, This should release the amine bound by the acids to resume acid gas absorption. The acidic and the sodium ions remain in the amine solution. In April of 1992, two additions of Rayon grade (high purity) caustic, totaling 200 gallons, were made to the MDEA solution, About 30 days after caustic addition , the corrosion coupon was analyzed and showed a slightly elevated rate. Although the rate was within the historical ral:~e, the nature of the metal loss was more localized Because the system has no purge, sodium had increa;ed to 2.3 wtOAand ash to 0.88% A second coupon showed an elevated corrosion rate of 18 milsiyear (mpy). Coupon frequency was increased to about 30 days because of the localized nature of the ;Lttack. Coupon corrosion rates are shown in Table 2. Corrosion of the coupons was most severe at the Ieac ing edge of the coupon and the coupon holder, a type of crevice attack. Identical corrosion was observed on all 5 coupons after caustic addition but prior to sodium reduction to less than 150 ppm by ion exchange. An alternate theory that the attack was under deposit corrosion was not supported by system conditions as shown in Table 3. The amine solution remained clear and iron levels did not incr:z se until nearly seven months after caustic addition.

Crevice corrosion, once initiated, is very difficult to ]top due to the stagnant conditions within the crevice. Stainless steels like other alloys forming protective oxide layers are susceptible to this type of attack, A 3 16L stainless steel thermowell on the ret oiler and the 304 stainless steel packing support ring attachment hardware (Wet Gas Absorber) failed The failure of the support ring attachment hardware resulted in collapse of the bed. Additional evidence that crevice corrosion was init ated by the caustic addition can be seen in Table 3. Chromium content in the amine began to rise immediately following caustic addition from historically less than 1 ppm to a peak of 12 ppm. It remained ligh, despite reduced corrosion rates following ion exchange and dilution from higher than normal fresh amine additions. Three separate cleaning methods, a complete solvent change out and the initi ition of a corrosion inhibitor program occurred before chromium content returned to near normal co lcentrations. The coupons attached to the reboiler bundle during tile period of caustic addition and high sodium levels showed heavy pitting as compared to those installed after sodium removal by ion exchange. Corrosion rates and descriptions of the coupons attached to the reboiler bundle are listed in Table 1. Formate Concentrations The formate ion concentrations in the MDEA solvcmt are historically high as listed in Table 3. Formic acid is a primary contributor to HSS and high conce~itrations can contribute to increased corrosion activity. Figure 2 identifies the corrosion rates of tht; reboiler outlet coupon and the formate level corresponding to the exposure period.
391,3

The corrosion rates appear to reflect formate ion concentrations but closer examination reveals that the corrosion rate increase is more closely linked to the : odium concentration (Table 3). Coupon corrosion rates had historically been less than 12 mpy with fcr]nate concentrations up to about 8000 ppm, After caustic addition, corrosion rates increased rapidly to 44.4 mpy while formate ion concentrations increased to 11,900 ppm. The increase in formate l:~el certainly contributed to the accelerated corrosion, however the rapid increase in corrosion ar d formate concentrations both occurred after caustic addition. An unanswered question is the possible impact of cal]stic on the formate ion formation. Kim, Palmer and Millimans discuss one method of formate ion formation in the presence of carbon monoxide that indicate MDEA has a slow carbon monoxide absor~tion rate as compared to potassium carbonate. The method they discuss may imply that caustic addition could increase the carbon monoxide absorption ra:e into the solvent and thus increase formate ion form~tion, The reported mechanism requires no oxygen. Our investigation into the increased formate ion concentration revealed that the crude diet was essentially unchanged. All wet gas streams are watel washed three times prior to reaching the amine unit. Routine analysis of these water streams showed no increase in organic acids, including formic. Retlnerv MDEA Use The system operated with MEA and a corrosion inhibitor program prior to 1986. The only challenge: experienced were cracking in non-stress relieved pip; welds, corrosion and replacement of the regenerator and failure of the amine reboiler bundle. The regenerator corrosion was attributed to galvanic action between collapsed 304 stainless tr~~s and the regenerators carbon steel walls. The reboiler problem was attributed to bundle design ar d was corrected in March 1992 after switching to MDEA, No corrosion problems were noted and the :;ystem was clean at the end of each 2-3 year run. The regenerator was periodically water washed and the effluent was usually black. In 1986, the solvent was changed to MDEA and the corrosion inhibitor program stopped. About 2.5 years later, some operational problems were experi:rced, but were not clearly documented, After 3 years of service on MDEA, the system was brought down for routine turnaround. During the inspections, much of the equipment inspected, showed moderate to heavy buildup of sludge. The regenerator trays below the rich amine feed nozzle h,~d 2-3 inches of material accumulated. This material, when analyzed was primarily iron sulflde After turnaround, the unit began the five year run vd addition. Following the February 1993 Ion Exchang: low levels, They remained low only about 3 months author believes these higher corrosion rates were the our system, as given below. ich included the previously discussed caustic treatment, the corrosion rates returned to normal before returning to a 20-30 roil/year range. The result of inherent weaknesses of using MDEA in

First inherent weakness of MDEA in our system is tie iron sulfide corrosion product fouling of equipment and the limited means to remove acids ud corrosion products in our MDEA system. system, fouling while using MEA was less signific;lrt due to the amine reclaimer which removes and acids and the use of an inhibitor to disperse solics and protect metal surfaces, MDEA cannot easily or economically reclaimed continuously, Filtration removes solids and foaming agents but filtration had been inadequate to remove corrosion p -oducts.
39f4

In our solid:; be Lean

MDEA is sold as non-corrosive and to take advantage of that we eliminated an expensive corrosion inhibitor program, There is ample evidence that cl(;an MDEA is less corrosive than DEA or MEA but most of the literature deals with carbon dioxide systtms and does not compare the amines at normal refinery operating concentrations or conditions, Pea ce and Brown reported corrosion rates of 20-26 mpy in a 50/0 MDEA solution saturated with hydro!,en sulfide.5 Increasing MDEA concentration increased corrosicn rates as shown in Figure 3 and reported in literature.f7 All high corrosion rates from the caustic episode were eliminated from the graph. Only those corrosion rates generated with similar operatic, ~ conditions at different amine concentrations were used, A second weakness of MDEA in our system is the high MDEA volubility in propane at high amine concentrations. Because of the systems liquid treater and the high volubility in propane, MDEA was limited to about 35% concentration. This is comparable to an MEA concentration of about 18 wtOAor a molar basis and is considerably less than the 45-50% recommended operating range for MDEA. Exceeding 35% resulted in high amine losses, and o her adverse downstream effects. The 3 5A concentration worked well until the syste~n fouled and the refinery was attempting to maximize sour crude production, The first noticeabl a result of the fouling was the lean amine temperature. Normally, lean amine cooler outlet ten- perature target was 120 F, but following the ion exchange, amine temperature gradually increased to 150 F. The hotter lean amine and low concentration severely restricted acid gas removal capacity. To compensate for this low concentration, the unit would run with higher circulation rates and then overstrip in the regenerator. As shown by Figure 4, ]lormal amine circulation rate prior to the initial accelerated corrosion cycle were 450-460 BPH. They gradually increased to 550-600 BPH as the fouling progressed. Overstripping was measured by steam rate to the reboiler. This increased from a normal of 10-12 Mlbs/hr to a routine of 17-18 Mlbs/ n- with peaks as high as 22 Mlbs/hr. By overstripping, the capacity of the amine was rnaximi:red and the higher circulation rates increased the relative contact time, in effect increasing the relati~c: amine concentration. The combination of increased circulation rate and reljoiler steam rate was not enough to prevent hydrog-en sulfide breakthroughs and the refinery was forced to cut sour crude charge rates, It appears that these operating conditions only served to incre~se corrosion and promote additional fouling, Table 2. Overstripped amine may contain too little hydrog m sulfide to maintain a protective iron sulfide sea e in the reboiler and hot lean amine piping, allowing fc)r more aggressive corrosion.
In September solvent 1994, the system was brought down for an acid cleaning of selected equipment and the

was changed out, A filming type corrosion nhibitor program was initiated with the intent of stabilizing the iron sulfide protective scale and remo.ling solids. Since the startup, the amine solution remained clear, filter bag changes were reduced from every 4 hours to once a day or less, and corrosiot~ rates began to return to normal. Additionally, the rcf nery was able to resume sour crude production targets and meet H~S specifications.

It should be pointed out that a corrosion inhibitor is, like MDEA, not a panacea. Certain conditions must exist for it to work properly. This was made c Iear when the corrosion rate jumped from 3.7 mpy to 20 mpy just prior to unit shutdown. Investigation revealed that changes in system operating
391/5

conditions had set up circumstances that limited the ilbilities of the filming amine type inhibitor The 3.7 mpy coupon was removed O 1/06/95 and ths 20 mpy coupon installed at that time, As with previous low corrosion rate coupons, the 3,7 mpy cc upon had a good black, protective film formed on the surface. At about the same time the second coup m was installed, amine concentration dropped from about 37/0 to 33/0 and the unit was having difficulty meeting HZS specification. To compensate, circulation rate was increased -1 4/0, from 550 to 625 BPH and to overstrip the amine, steam rate was also increased 20/0. This condition existed during the entire 33 day exposure period of the second coupon. The coupon, when removed, was white as no iron sulfide film had formed, The operating conditions of over circulation and overstripping had ,;imply prevented the formation of a film for the inhibitor to stabilize. The author believes that the inhibitor was still benefi ;ial during this incident, The corrosion rate was lower than those measured during similar operating conditions, the solvent remained clean and there was no increase in the frequency of filter bag chan~e outs. Proper corrective action to restore amine concentrati(m and reduce operating severity were initiated, but the unit was shut down shortly after and the effect of these changes on corrosion were not measured. CONCLUSION As Refineries push toward longer runs and economics favor running more sour crude, maintaining cleanliness in these important environmental units w 11become more critical. Minimizing corrosion is key to maintaining cleanliness. Three recommendations have been made to improve the long term reliability of this system. Corrosion coupon results were used to change the reboiler metallurgy to type 304 stainless steel This should extend the bundle life to well beyond thf current 18 months.
1.

2, A filming type corrosion inhibitor program was itlitiated to minimize corrosion and maintain system cleanliness. Fouling is reduced by the inhibitor disx:rsing solids until they can be properly removed by filtration, 3. Investigate replacement of MDEA solvent with Z.solvent with higher hydrogen sulfide capacity and lower propane volubility at operating concentrations.
ln

addition to these recommendations,

three conclusi ms were formulated

1, Caustic addition has the potential to initiate cre~ it:e corrosion in stainless steel as determined by the failed stainless steel equipment, increased chromium concentrations in the amine and increased corrosion rates of the stainless steel coupons attachec to the reboiler. 2. Caustic addition may promote the formation of Fcrmate ions 3. Over stripping MDEA results in too low a hydrogen sulfide concentration to form a protective iron sulfide scale in the regenerator bottom section, reboi er and hot lean amine piping.

391,6

REFERENCES 1, Butwell, K. F,, Kubek, D,J, and Sigmund, P. W,, ~dkanolamine Treating, Hydrocarbon March 1982, 2. Ball, H. T., Designa ndOperationo fAmineU nits, Petroenergy, October 23-27, 1989, Processing,

3. Kim, C. J., Palmer, A.M. and Milliman, G. E., Abs~rption of Carbon Monoxide into Aqueous Solutions of K2C03, Methyldiethanolam ine and D [ethylethanolam ine, American Chemical Society, 1988, 4. Sales brochure 5. DuPart, M.S, Bacon, T, R,, and Edwards, D. J., Understanding Corrosion in Alkanolamine Treating Plants, Hydrocarbon Processing, April 1993
6. Pearce, R.L, and Brownlie, T.J, Selective Hydrc)gen Sulfide Removal, Gas Conditioning

Gas

Conference, Norman, Ok, March 8-10, 1976 7, Keller, A.E and Mecum, S.M., Heat-Stable Salt F.emoval From Amines By the HSSX Process Using Ion Exchange, The Laurence Reid Gas Conditionin~~ Conference, March 2, 1992.
T,% BLE IN-SITU 1

CORROSICIN COUPON (rates are in mpy)

RESULTS

Days Exposure

508 days

:i:.:.:::..ft$.f:!:.:.:+:~:. . yJJ$J\$]i.:::?c*$;J*&#$ ::~:::~:~J:::~:J:::~:::~:::::~:::~:~:::J:::m~:~. :::::::::::::::::::::::::$W ::::::::::::::::::::~::::::::::w.:::w::~ ...................................:::#$::..:::~.j$ ... :.:!.i}:y:.:.:.:..:.:.:.:.: ...+ ....... .,.. 347 days ........... ............:=.:.:+:.!.:...:.:.:.:.:.:.:.:.:.:.: :.::::::::::::::::::::::::::::::t.:.:.:.:.:.:.:.:. ~<@%#8%XW&W:; %i&i@gT:r:T:::

Material 5 Cr-. O5 moly*

9-93 Cor Rate

CommetIt!;
Badly Att~ ked Uniform 17inning Uniform Tl inning Heavy Pit~lg Pitting
Like New Like New

9-94 Cor Rate

Comments
Uniform Thinning Some Attack Uniform Thinning Like New Like New
Like New Like New

12,6
13

0,08
15,9

Ss

1.1 1.2

316SS 317LSS
AL6XN 904L

o o
0 0

o
o

2205 SAF2304 Titanium * A182F5a **A182F9 **CIO1O

o 0 o o

Stained

Like New Like New Like New

o 0 0 0
in mpy)

Like New Like New Like New Like New

3917

HISTORICAL
Date Installed 04/17/90 05/02/90 05/21/90 06/ 12/90 07/17/90 08/0 1/90 10/02/90 05/14/91 05/14/92 Days 15 19 22 35 15 62 224 366 109

TAIB .E 2 CARBON STEEL CORROSION

RATES

NaOH 30 days prior

08131/92 10109192 12/04192

39 25 32
30 28

37,9 44.4 38.2

01/05/93
02/ I 5/93

~)n

Exchange

03/15/93 04/28/93 06/04/93 06/29/93 08/02/93 09/03/93 10/06/93 1 l/12/93 12/15/93 02/ I4/94 04/14/94 10/20/94 11/21/94 01/06/95

44 37 25 34 32 14 37 33 60 59 136 32 46 33

4.1 2.1 19.9 23.7 26.2 27.8 11,2 46,8 gone 7.6 30 8.4 3,7
20

~ heroical Cleaning ~;id cleaning & solvent change out reinstallation finit down for Turn around

391(8

CONDITION
Date 01/90 07/90 ()()/g 1 01/92 04/92 05/92 06/92 11/92 12/92 ol/93 02/93? 03/93 04/93 05/93 06/93 07/93 ox/93 oy/y~ 10/93 I l/93 01/94 05/94 06/94 07/94 og/gJ 10/94 11/94 02/95 *All units in ppm 61 17 25 21 58 138 310 ~y 20 100 120 120 105 120 135 Iron* 1 1 2 10 4 5 6 20 350 455 Sodium 45 40 50 60 1,400 2,300 2,1OO 2,400 2,300 2>()()0

TAIBI.E 3 MIIEA SOLUTION

496 16 35 38 -1 1 2 4 75 80 82 22 40 58
5x

. 2 1 1 4 2 I
<1

Grcy 7.666 10.25O 10.650 500 850 2,640

Yellow Dark Amber Dark Amber Dark Amber Dark Amber

*1- Caustic added *2 - Ion Exchange *3 - Chemical Cleaning

\_

-/

39110

FIGURE 2 FORMATE AND CORROSION

RATE

COR RATE MPY

14 12 10
u co :

] 50

8 6 4 2 0

30 20

1
10 0
BAR = FORMATE; LINE = CORROSION RATE

CORROSIONRATES AS A FUNCTION MPY 30

SQUARED= 0.86

E
/

,/ ,/

y./~

20

,/

ES

,, / ..

//

,.// //-

10
,/

,/ /4ii
// /

//

El

0
3

I
34

,/

/ //
I

,/

x
36

M
I 1 I

38

40

42

44

AMINE

CONCENTRATION

WT%

Z6AN

Z6nr

(363(3

0 m

(f)

0 0 co
391/13