mechanical integrity of primary reformer hot outlet headers...steam-methane reformers have either...

2011 [83] AMMONIA TECHNICAL MANUAL

Mechanical Integrity of Primary

Reformer Hot Outlet Headers

This paper discusses mechanical integrity issues for primary reformer hot outlet headers. These are

typically welded assemblies of wrought Alloy 800H or 800HT or cast 20Cr-32Ni-Nb alloy. The typical

design and operating conditions for these headers are presented. Problems and failures that have been

reported at past Ammonia Plant Safety Symposiums are reviewed and summarized. With this

background information, recent failures of hot outlet headers are described along with the identified

causes of these failures. The most common failure mechanism is high-temperature creep but the

location and causes of creep damage vary. Creep failures have occurred both at weld joints and within

base metal. The sources of these problems are discussed and suggestions for mitigation are presented.

Based on the findings of the failure and material properties investigations, stress analysis and material

damage modeling is used to estimate remaining life.

Carl E. Jaske

Det Norske Veritas (USA), Inc., Dublin, OH USA

Greg T. Quickel

Det Norske Veritas (USA), Inc., Dublin, OH USA

Introduction

he purpose of this paper is to review and

discuss the primary mechanical integrity

issues that are associated with the hot out-

let headers of primary reformers. A number of pa-

pers presented at past Ammonia Plant Safety

Symposiums have addressed various issues with

these hot outlet headers. To provide background

for recent issues with similar headers, the past

work is reviewed and the various types of prob-

lems that have been encountered in the past are

summarized. Relevant issues on similar headers

in hydrogen and methanol reformers are also dis-

cussed. In all cases, the hot outlet headers operate

in the high-temperature creep range for their ma-

terial of construction, which is either a wrought

Alloy 800H or 800HT or a cast 20Cr-32Ni-Nb al-

loy. Because of the operating temperature range,

failures typically occur by high-temperature creep

or creep-fatigue, with the fatigue damage induced

by thermal cycling. Thus, creep strength is an im-

portant parameter addressed in the review discus-

sion.

Following the background review, examples of

recent problems with hot outlet headers are pre-

T

832011 AMMONIA TECHNICAL MANUAL


sented and discussed. The causes of these prob-

lems and their relation to past problems are eva-

luated. The purpose of this evaluation is to help

operators of primary reformers minimize the po-

tential of similar problems in future operation of

their units and to provide guidance for estimating

the safe and useful life of headers that are current-

ly in service. Such guidance will help avoid un-

planned shutdowns and schedule repairs or re-

placements in a timely, cost-effective manner.

Past Hot Outlet Header Issues

Steam-methane reformers have either hot or cold

outlet headers to collect the process gas from the

outlet of the catalyst tubes. Cold headers are made

of carbon steel and insulated so that they operate

at temperatures well below the creep range, while

hot headers are not insulated and operate at high

temperatures close to the reformer outlet gas tem-

perature. The focus of this paper is on hot outlet

headers that operate in the high-temperature creep

range of the materials from which they are con-

structed. Kawai, et al. [1] point out that there are

two basic types of hot header system designs: (1) a

hot collector where the catalyst tubes are attached

directly to the manifold by means of fittings and

the hot gases exit from the manifold via riser

tubes and (2) a hot subheader with a cold transfer

line where the catalyst tubes are attached to the

subheader via outlet pigtails. In either case, the

hot header or subheader is cylindrical vessel with

numerous welded attachments at holes that allow

the hot process gas to flow into the header and

central tee section where the hot process gas flows

out of the header.

Because the headers operate at temperatures near

or greater than 800 °C, they are typically made

from either wrought Alloy 800Hor 800HT or from

cast 20Cr-32Ni-Nb alloy. The fittings and tees are

also made using these same alloys. Welding of the

headers, tees, and fittings is done with a high-

nickel material or a “matching” material. Com-

monly used welding electrodes include Inco-Weld

A, Inconel 182, Inconel 112, and Inconel 117.

Commonly used filler metals include Inconel 82,

Alloy 625, and Alloy 617. The common “match-

ing” filler metals are Thermanit 21/33, Thermanit

21/33 So, and UTP A 2133 Mn. The final selec-

tion of welding material depends on the design

and the welding technique that is used.

The primary load on outlet headers is internal

pressure but bending loads are often induced by

movement of attachments and/or restraints at sup-

ports. These system bending loads are related to

thermal expansion and complex interactions with-

in the manifold system, making stress analysis to

compute their magnitude difficult and adding sig-

nificant uncertainty to the results of the calcula-

tions. Furthermore, the bending loads are often

displacement controlled so that stresses induced in

the high-temperature creep regime can relax and

redistribute with time at operating temperature. In

contrast, internal pressure is usually maintained at

a relatively constant level and is a sustained load

so that stresses induced by internal pressure are

relatively straight forward to compute.

This section of the paper reviews hot outlet header

issues that have been discussed in previous publi-

cations. Issues with primary ammonia reformers

as well as issues with similar components in hy-

drogen and methanol reformers are addressed. In

comparing this information, keep in mind that

these different types of steam-methane reformers

operate in different ranges of temperature and

pressure as illustrated in Figure 1. Ammonia re-

formers operate at higher pressures and lower

temperatures than hydrogen or methanol refor-

mers. This difference in operating conditions af-

fects the hot outlet header design. Thus, the outlet

headers in ammonia reformers usually operate at

higher stresses and lower temperatures than those

in hydrogen and methanol reformers.

Performance of Header Materials

In 1963, Francis and Glass [2] reported the failure

of outlet piping and described high-temperature

stress-rupture failures in Type 304 stainless steel

84 2011AMMONIA TECHNICAL MANUAL


piping. Avery and Valentine [3] reviewed the de-

sired characteristics of materials for outlet mani-

fold systems, while Harnby [4] pointed out that an

outlet manifold system made of Alloy 800 had

better creep resistance than one made of the HU

alloy. Later, Blackburn [5] pointed out that mani-

folds made of Alloy 800 were less prone to crack-

ing than those made of traditional cast alloys, such

as HU, HK, and HT, because of its superior duc-

tility. He proceeded to note that the cast 20Cr-

32Ni-Nb alloy was an improvement to Alloy 800

for hot outlet header applications. These papers

provide historical background on alloys for use in

hot outlet headers.

0

2

3

4

700 800 900 1000

NH3

H2

CH3OH

Limit of Tube

Metal Use

1

Gas Temperature, °C

Internal Pressure, MPa

0

2

3

4

700 800 900 1000

NH3

H2

CH3OH

Limit of Tube

Metal Use

1

0

2

3

4

700 800 900 1000

NH3

H2

CH3OH

Limit of Tube

Metal Use

1

Gas Temperature, °C

Internal Pressure, MPa

Figure 1. Typical operating temperature and pres-

sure ranges for ammonia (NH3), hydrogen (H2), and

methanol (CH3OH) reformers, from Kawai, et al. [1].

Shibisaki, et al. [6] found that Si, Nb, and C addi-

tions had a negative effect on the ductility of

20Cr-32Ni-Nb cast alloys while Mn had a positive

effect on their ductility. Parks and Schillmoller [7]

indicated that Alloy 800H has a low tendency for

embrittlement but noted that the 20Cr-32Ni-Nb

cast alloy has higher creep-rupture strength than

Alloy 800H. Thomas, et al [8] pointed out a se-

rious embrittlement issue with the 20Cr-32Ni-1Nb

cast material and suggested that niobium-free

grades should be developed in order to avoid such

in-service problems. Hoffman [9] evaluated the

aging of 20Cr-32Ni-Nb castings after 21 months

of service at 852 °C and found both M23C6 carbide

and inter-metallic Nb-Ni silicide precipitates

present in the aged material. The silicides were

mostly G phase plus some η’ phase. They are ex-

pected to decrease creep resistance and promote

liquation cracking during welding.

Hoffman and Gapinski [10] evaluated samples of

a 20Cr-32Ni-Nb cast outlet manifold that had de-

veloped fatigue cracks after it had been in service

for approximately 192 months at an outlet gas

temperature of 871 °C. In addition to carbide and

silicide precipitation, they observed carbide coa-

lescence. The observed aging significantly de-

graded the tensile strength, Charpy V-notch im-

pact resistance, stress-rupture strength and

weldability of the material. The carbides and sili-

cides were solutionized by annealing at 1093 °C.

However, complete solutionizing of secondary

precipitates and spheroidization of the primary

carbides required a minimum annealing tempera-

ture of 1149 °C. Better stress-rupture properties

were achieved by solution annealing at 1204 °C

than by solution annealing at 1093 or 1149 °C.

The cracked region of the manifold was success-

fully repaired after in-situ solution annealing at

1149 °C for three hours and operated successfully

for four more years before the furnace was rebuilt.

The authors recommended keeping Si content as

low as possible and maintaining a stoichiometric

Nb/C ratio of 7.7.

Welding and Repairs

Roach and VanEcho [11, 12] found that weld-

ments generally have lower stress-rupture strength

than comparable base metals. Inconel 182 weld-

ments had the lowest stress-rupture strength when

compared with Inconel 82, Inconel 112, and Inco-

Weld A weldments. Several reformer operators

[13-15] have experienced accelerated creep dam-

age, embrittlement, and failures with Inconel 182



weld metal. Failures have been attributed to a loss

of rupture strength caused by embrittlement. Inco-

nel 182 is now known to exhibit a loss of rupture

strength and ductility at elevated temperatures.

Commonly used materials for joining Alloy 800

are the Inco-Weld A electrode and Inconel 82 fil-

ler metal. Avery and Valentine [5] specifically

recommended filler metal Inconel 82 for Alloy

800 welds. Roach and VanEcho [11, 12] indicated

that Inconel 82 weldments have higher stress-

rupture strength than Inco-Weld A and Inconel

182 weldments, but lower stress-rupture strength

than Inconel 112 weldments. McMillan [13]

found success in using Inco-Weld A material in

place of failing Inconel 182 filler metal.

Shibisaki, et al. [16] recommended the use of a

high-nickel alloy metal (Inconel 82) for welding

20Cr-32Ni-Nb metals. Kobrin [15] obtained supe-

rior results using Inconel 82, Alloy 625 filler met-

al, and Inconel 112 in place of Inconel 182. Or-

bons [17] suggested using Inconel 112 in place of

Inco-Weld A for applications at temperatures

above 816 °C because of its superior resistance to

creep damage. Roach and VanEcho[11, 12] also

found that Inconel 112 had superior creep resis-

tance to those of Inco-Weld A, Inconel 182, and

Inconel 82 weldments. van Wortel [18] noted that

filler metals 82 and 617 were found to be suscept-

ible to relaxation cracking. For applications that

require the highest strength and corrosion resis-

tance, Inconel 117 electrode and Alloy 617 filler

metal are recommended [19] for use at tempera-

tures above 788 °C.

Al-Jubeihi [20] reported making successful repair

welds by grinding out cracks, depositing Inco-

Weld A stringer beads, and grinding out welds

and base metal defects one bead at a time while

using dye penetrant tests and ultrasonic examina-

tions to ensure quality repairs. Another successful

repair method [21] involves grinding into through-

wall cracks to remove 60% of the cross-section

and grinding back the surface area by 5 mm to

prevent re-cracking associated with embrittlement.

The beveled area is then covered using Inconel

182 electrodes, and the previously ground grooves

are filled up by welding following a stringer bead

technique. The shortcoming of this repair method

is that Inconel 182 welds have poor high-

temperature creep strength.

Pande and Swenson [22] successfully completed a

preliminary repair by welding a reinforcement

cone onto sound material in the tees to extend op-

erating life. Blackburn [5] developed a method to

improve weld efficiency using parent metal as fil-

ler after completing the root weld with Inconel 82

wire.

Shi and Lippold [23] found that repair practices

that lead to failures often involve elevated temper-

atures throughout welding repairs leading to em-

brittlement, which in turn gives rise to a ductility

dip. When the elevated temperature restraint ex-

ceeds the ductility minimum, failure occurs.

Another common problem is the mechanism of

hydrogen induced crack formation which leads to

weld failures. Pattabathula, et al. [24] noted this

can be mitigated by increasing the temperature at

the weld above dew point, by adding outside sur-

face insulation, and by installing windshields on

sides of the furnace.

Another problem with welding operations, pointed

out by Shibisaki, et al. [25], involves differences

between weld and base metal. The difference in

thermal expansion coefficients can lead to elastic-

plastic strain of the weld material. Orbons [17]

suggested high-temperature preheating around the

weld to prevent cracking of the base metal caused

by weld shrinkage. Lobley, et al. [26] found that it

was necessary to bore machine and solution an-

neal a 20Cr-32Ni-Nb cast header prior to repair

welding because of its severe embrittlement in the

aged condition, especially near the inner surface.

Anderson and Bruno [27] reviewed weldability is-

sues that they encountered during the welding of a

new cone piece to an existing tee (aged material)

in a hydrogen reformer outlet manifold. Both



pieces were 20Cr-32Ni-Nb castings from the same

foundry. Metallurgical analysis of samples of the

previously installed, aged cone piece on which an

Electrode 117 weld bead had been deposited re-

vealed both heat-affected zone (HAZ) liquation

cracking (LC) and reheat cracking. As shown in

Figure 2, the liquation cracking occurred near the

fusion zone (FZ) and was believed to be caused by

low melting point silicate precipitates that had

formed during service. The reheat cracking was

believed to have been caused by a combination of

mismatched mechanical properties and high inclu-

sion content. A full solution annealing of the aged

material was required to weld it successfully.

Figure 2. Liquation cracking in heat-affected zone

(HAZ) of aged 20Cr-32Ni-Nb casting, from Anderson

and Bruno [27].

Penso and Mead [28] discussed cracking that they

found on existing 20Cr-32Ni-Nb statically cast tee

pieces when a new centrifugally cast 20Cr-32Ni-

Nb reducer cones were welded to them. The loca-

tion of this cracking is shown in Figure 3. It was

found to be associated with intermetallic silicate

precipitates in the existing (aged) material. The

tees were cut out and solution annealed at

1150 °C, followed by machining out of all of the

cracks. Welding was done using a GTAW process

with Alloy 617 filler wire, a 50 °C preheat, and a

maximum interpass temperature of 150 °C. A but-

tering layer was first applied using low amperage.

The joints were then tack welded, and a root pass

was applied. DP inspection was performed fol-

lowing every second fill pass and after blend

grinding the cap pass.

Figure 3. Location of cracking in aged 20Cr-32Ni-

Nb cast tee piece after welding a new cone piece to it,

from Penso and Mead [28].

Jack [29] reported cracking in a 21/33 weld be-

tween a tee and a subheader in a cast 20Cr-32Ni-

Nb outlet manifold after 14 years of service at ap-

proximately 793 °C. This cracking is shown in

Figure 4. Attempts to grind out and repair the

cracking were not successful, so tapered sleeves of

Alloy 800HT were designed and welded over the

cracked weld and similar welds in the outlet mani-

fold.

Figure 4. Cracking at subheader to tee weld; both

are 20Cr-32Ni-Nb castings, from Jack [29].



Dejaeger, et al. [30] discussed cracks in an Alloy

800H outlet header. In this case, the header was

manufactured using a solution annealed Alloy

800H plate material that was cold formed into a

cylinder and then longitudinally welded with Ni-

base filler. No heat treatment was applied after the

manufacturing process. After two years, a leak

was detected. Examination revealed the cause to

be high residual stresses from fabrication (cold-

working/welding) or resulting from cyclic defor-

mation and accompanying relaxation processes.

Stress relaxation cracking was controlled in the

replacement header by a stabilization heat treat-

ment at 950 °C performed before welding. After

welding, a second heat treatment at 850 °C was

performed to decrease weld residual stresses. This

method successfully controlled the stress relaxa-

tion cracking in the Alloy 800H material. Howev-

er, this stabilization heat treatment temperature

was less than the recommended value [18] of

982 °C, and the post-weld heat treatment (PWHT)

temperature was less than the recommended val-

ue [31] of 900 °C.

Brongers and Jaske [32] evaluated an Alloy 800H

outlet manifold that had welds made using weld-

ing Electrode 112. It was removed from service

after normal operation at approximately 870 °C

for 64,000 hours because nondestructive testing

indicated cracks in the circumferential butt welds.

Creep-rupture tests were performed at stress levels

comparable to the service stress and temperatures

above the design temperature were selected to ac-

celerate the testing. The results showed that the

base metal did not suffer significant creep damage

from the prior service exposure and that the weld

metal had a very low resistance to creep damage.

Metallographic analysis of a weld sample taken

from a comparable in-service manifold supported

this conclusion, and data for new material re-

ported in the literature support the finding of rela-

tively low high-temperature creep strength for

welds made using welding Electrode 112. The au-

thors recommended using welding Electrode 117

for repairs.

Recent Hot Outlet Header Issues

To highlight some recent issues with hot outlet

headers, three case histories are reviewed. The

first two cases cover outlet headers from methanol

reformers, while the third case is for an outlet

header from a hydrogen reformer.

Case 1. Metallurgical Analysis of a Header

to Bull Tee Weld from a Methanol

Reformer

A metallurgical analysis was performed on a bull

tee from a methanol reformer outlet header. The

tee was connected to two header tubes and an in-

creaser cone. The tee was removed when leaks

were discovered at the girth welds after approx-

imately 50,000 hours of service.

The header, tee, and increaser cone were made of

a cast alloy 20Cr-32Ni-Nb alloy. The welds used

to join these components were made using an In-

conel 82 root pass, Inconel 82 or Inconel 182 rein-

forcement, and Inconel 182 for the remainder of

the weld. Thus, the majority of the weld filler

metal in the joints was Inconel 182.

Approach. The following steps were performed

for this analysis. The bull tee was visually in-

spected and photographed in the as-received con-

dition. Liquid dye penetrant testing was performed

on the three large diameter girth welds of the tee.

Cross-sections were removed from cracks in the

welds, mounted, polished, and etched. Light pho-

tomicrographs were taken to document the mor-

phology of cracking and microstructures. Samples

for chemical analysis were removed from the re-

ducer cone, header tube, and tee and weld metal

between the increaser cone and tee to determine

the chemical composition.

Results and Discussion. Figure 5 is a photograph

of the bull tee. Pigtails extended vertically from

fittings attached to the tee and indicate the top of

the tee. Girth welds attached the tee the increaser

cone and tee to two header tubes. The outside di-

ameter (OD) of both header tubes was 345 mm.



The wall thicknesses of the of the header tubes

were measured at the 12:00, 3:00, 6:00, and 9:00

o’clock orientations. The average wall thickness

of the header tube was 21.4 mm.

Figure 5. Photograph of bull tee following liquid dye

penetrant testing of the girth welds.

Figure 5 also shows the tee after performing liquid

dye penetrant testing of the three welds. The red

dye shows the locations of indications. Four loca-

tions that contain indications were identified. The

indications were located at: 1) both header to tee

girth welds, 2) a pigtail to header tube girth weld,

and 3) the tee to increaser cone girth weld.

Figure 6 is a photograph of the crack-like indica-

tion located at the toe of a header to tee girth

weld. The indication is approximately 122 mm

long and present from the 1:15 to 2:30 orientation.

The liquid dye penetrant testing clearly showed

the locations of the crack indications at the girth

welds.

Figure 7 is a photomicrograph of the mounted

cross-section (Mount W3) that was removed a

header to tee girth weld (GW 3). The dashed line

in Figure 6 indicates the location where the cross-

section was removed and the arrow indicates the

polishing direction. Cracks are present in the HAZ

of the tee, HAZ of the header tube, and in the

weld metal. Although difficult to see, a columnar

microstructure is also present in the header. Fig-

ure 8 is a photomicrograph of Mount W3 showing

the HAZ of the tee and the weld metal. Creep

cracks are present at the grain boundaries in the

HAZ of the tee. Creep voids and aligned creep vo-

ids are present in the weld metal. Figure 9 is a

photomicrograph of Mount W3 showing the HAZ

of the header tube and weld metal. Relatively

wide, and also narrow, creep cracks are present in

the HAZ of the header tube. This photomicro-

graph was taken just below the outside surface

where the crack-like indication shown in Figure 6

was located.

Figure 6. Photograph of a crack located at girth

weld between a header and the tee; area indicated in

Figure 5.

Figure 7. Stereo light photomicrograph of a

mounted cross-section removed from the girth weld

shown in Figure 6 (Glyceriga Etchant).

Figure 6

Flow

Flow

Flow

Pigtail

Reducer

cone

Header Header

Flow

Crack Header

Polishing

direction

for cross-

section

Header

tube side Tee side

Figure 8

Figure 9



Figure 8. Light photomicrograph of the mounted

cross-section showing the HAZ and weld metal near

the tee side; mirror image of area indicated in Figure

7 (Glyceriga Etchant).

Figure 9. Light photomicrograph of the mounted

cross-section showing the HAZ and weld metal near

the header tube side; mirror image of area indicated

in Figure 7.

Figure 10 is a photomicrograph of the header base

metal microstructure. The observed microstructure

is consistent with that expected for the 20Cr-

32Ni-Nb alloy. The figure shows primary carbides

located at the grain boundaries and smaller sec-

ondary carbides dispersed in the grains. The finer,

secondary carbides are indicative of high tempera-

ture exposure. Although it is difficult to be certain

since the appearance of the pre-service micro-

structure is not known, it appears that the network

of primary carbides was reduced and that they are

slightly blocky, which is also indicative of high-

temperature exposure.

The most severe creep damage was on the header

tube side rather than the tee side. The intergranu-

lar crack path is consistent with a high-

temperature creep mechanism. The finer, second-

ary carbides and blocky primary carbides are in-

dicative of high-temperature exposure.

Figure 10. Light photomicrograph of the header

base metal microstructure (Glyceriga Etchant).

The results of the chemical composition analysis

conducted on a sample removed from the base

metal of the header are shown in Table 1. The re-

sults of the analyses are consistent with the no-

minal chemical composition specifications for the

alloy at the time of its manufacture.

The results of the chemical composition analysis

conducted on a sample removed from the weld

metal between the increaser cone and tee are

shown in Table 2. Except for the nickel plus co-

balt (Ni + Co), manganese (Mn), iron (Fe), and

chromium (Cr) content, the results of the analysis

are consistent with specified values for Inconel 82

and 182. The Mn and Cr content are consistent

with the Inconel 82 specification and not with In-

conel 182 specification. The Ni + Co and Fe con-

tent are consistent with the Inconel 182 specifica-

tion and not the Inconel 82 specification.

HAZ Weld metal

HAZ

Weld metal

Primary

carbide

Secondary

carbide



Table 1. Results of chemical analysis of samples re-

moved from the tee, header tube, and increaser cone

compared with the nominal chemical composition for

the 20Cr-32Ni-Nb alloy.

Element Tee

Wt. %

Header

Wt. %

Increaser

Cone

Wt. %

Nominal1,

Wt. %

Carbon 0.12 0.15 0.14 0.10 – 0.15

Manganese 1.05 1.01 0.77 1.50 (max)

Silicon 0.63 0.50 0.65 1.50 (max)

Chromium 20.62 19.91 20.01 19 – 25

Nickel 33.03 32.61 33.02 31 – 33

Niobium 1.17 1.08 1.05 0.50 – 1.50

Phosphorus 0.019 0.016 0.017 0.050

(max)

Iron Balance Balance Balance Balance

Table 2. Results of chemical analysis of a sample

removed from the weld between the tee and increaser

cone compared with the nominal chemical composi-

tion for Inconel 82 and 182.

Element

Weld

Wt. %

Inconel Filler

Metal 82 2

Wt. %

Inconel Welding

Electrode 182 2

Wt. %

Nickel + Cobalt 65.9 67.0 (min) 59.0 (min)

Carbon 0.037 0.10 (max) 0.10 (max)

Manganese 2.74 2.5 – 3.5 5.0 – 9.5

Iron 8.52 3.0 max 10.0 max

Sulfur 0.006 0.015 (max) 0.015 (max)

Silicon 0.17 0.50 (max) 1.0 (max)

Chromium 20.05 18.0 – 22.0 13.0 – 17.0

Niobium + Tantalum 2.42 2.0 – 3.0 1.0 – 2.5

Phosphorus 0.003 0.030 (max) 0.030 (max)

The fractional increase in Ni content would not be

expected to have a substantial effect on creep re-

sistance. The differences in the Mn, Fe, and Cr

content are likely related to dilution or enrichment

that can occur with the various passes of the Inco-

nel 82 and 182 welding materials.

Summary. A summary of the key observations

follows:

• Creep voids were present in the heat af-

fected zone (HAZ) and weld metal of the

header to tee girth weld but not in the base

metal.

• Creep cracks were present in the HAZ and

weld metal at one girth weld.

• Intergranular creep cracks penetrated the

outer surface in the HAZ and weld metal,

near the toe of the girth welds.

• The fine, dispersed secondary carbides are

indicative of high-temperature exposure.

• The microstructures of the base metals

were consistent with those expected for

20Cr-32Ni-Nb cast alloy.

• Except for Ni compositions for the tee and

reducer cone of 0.03 and 0.02 weight %

greater than specified, the base metal

composition was consistent with specified

values for the alloy.

• The composition of the girth weld was

consistent with that of Inconel 82 or 182,

and in some cases, with both specifica-

tions.

The cracking is consistent with a high temperature

creep mechanism associated with the low creep

strength of the filler material. This type of failure

can be mitigated by using a filler metal with im-

proved creep strength in the weld joint such as a

“matching” filler or and Inconel 617 filler.

Case 2. Metallurgical Analysis of Outlet

Header from Methanol Reformer

A metallurgical analysis was performed on two

sections from a methanol outlet header that leaked

in-service. One of the sections contained the leak

as shown in Figure 11 and the other (non-failed)

was located away from the leak in the same head-

er. The header samples were removed from ser-

vice after approximately the minimum design life

of 100,000 hours. The header tube sections were

centrifugally castings made of a 20Cr-32Ni-Nb al-

loy.



Approach.Approach.Approach.Approach. The following steps were performed

for the analysis. The header tube sections were vi-

sually inspected and photographed. A ring section

was removed from the through-wall portion of the

header tube section; the other header section was

already in the form of a ring. The two ring sec-

tions were polished and light photomicrographs

were taken to document the observed creep dam-

age. The ring sections were then etched in Adlers

reagent to reveal their macrostructures and photo-

graphs were taken. A small portion of each ring

was then removed, mounted, polished, and etched

with glyceregia for detailed examination. Photo-

micrographs were taken to document the micro-

structure and maximum service temperatures were

estimated.

Using elastic finite-element stress analysis, the

maximum tensile hoop stress in the region of the

header failure was calculated to be 10.2 MPa.

Failure stresses for average stress-rupture lives of

100,000 hours (11.4 years) were computed for

header operation at various temperatures using the

manufacture’s Larson-Miller parameter relation-

ship for the 20Cr-32Ni-Nb alloy. A temperature of

963 °C was calculated to cause stress-rupture fail-

ure in 100,000 hours. This was 73 °C above the

reported maximum header operating temperature

of 890 °C. Based on this result, it is likely that the

material was somewhat weaker than indicated by

the manufacture’s data and that it was overheated

to some degree at the failure location.

Results and Discussion. Figure 11 is a photo-

graph of the as-received header tube section

(Header Section 1) that failed. The figure shows

the through-wall portion of the failure. Secondary

cracks are also visible in the vicinity. The cracks

were present from the right side of the tube sec-

tion up to approximately 174 mm from the right

side. Also visible is a bulge in the tube where the

cracks are present. The maximum diameters of the

bulged and non-bulged regions of Section 1 were

174.0 mm and 171.0 mm, respectively.

Figure 11. Photograph of the header tube section

(Header Section 1), as received.

Figure 12 is a photograph of the ring from Section

1, following polishing and etching. The location

of ring Section 1 is indicated in Figure 11. A co-

lumnar dendritic macrostructure was visible on

both ring sections.

Figure 12. Photograph of Ring Section 1 from

Header Section 1; dashed lines in Figure 11 indicate

the location of the ring (Adlers reagent).

Measured dimensions of the tube sections are

listed in Table 3 along with reported nominal di-

mensions of the tube when it was new. The no-

Through-wall

Creep

fissures

Through-wall



minal inside diameter (ID) of the new header tube

was not reported, so it was calculated based on the

reported minimum outside diameter (OD) of a

new header tube minus twice the minimum sound

wall thickness. The strain values of the header

tube sections were based on calculated minimum

ID values. There was more strain at the bulge in

the header tube than in Header Section 2, which

was to be expected. Since baseline values of orig-

inal dimensions are not available for each tube,

the strains are only approximate because of inhe-

rent variations in the dimensions of new tubes.

Table 3. Measured dimensions of header tube sam-

ples compared to minimum values reported by client.

Description Outside

Diameter

(OD)

Wall Thick-

ness (WT)

Inside

Diameter

(ID)

Strain*

mm in mm in %

Minimum

new tube 168.3 6.625 15 0.591 —

Bulge on

Header Sec-

tion 1

174.0 6.850 15.0 0.592 4.10

Header Sec-

tion 2 172.0 6.77 15.4 0.607 2.08

* Strain based on ID minimum and measured values. ID

calculated based on OD and WT minimums and measured

values.

Figure 13 is a photomicrograph of Ring Section 1

adjacent to the OD surface. Relatively wide and

narrow creep cracks that appeared to follow the

dendritic columnar structure were present for the

full wall thickness. A low density of creep voids

was found in the cross-section. The density of vo-

ids was not high adjacent to the ID or OD surface.

The absence of creep voids near the cracks is evi-

dence of short-term overheating. Note that evi-

dence of austenite grains and twinning was appar-

ent at the ID where the through-wall cracks were

present. This suggests that the ID was cold

worked during fabrication. It is not known if this

contributed to the failure.

Figure 14 is a photomicrograph of Ring Section 2

adjacent to the OD surface. Aligned creep voids

(not linked) were present adjacent to the OD and

ID and appeared to be more prevalent adjacent to

the OD. The residual life of this tube section is

approximately 70% based on metallurgical analy-

sis and guidelines of past studies [33] of reformer

tubes. For 11 years of prior operation, this result

implies approximately 26 years remaining life, as-

suming that future operating conditions are the

same as those of the past.

Figure 13. Light photomicrograph of Ring Section 1

adjacent to the OD surface, near the through-wall lo-

cation.

Figure 14. Light photomicrograph of Ring Section 2

adjacent to the OD surface.

Figures 15 and 16 are photomicrographs of the

microstructures of Ring Section 1, near to and

away from a crack, adjacent to the OD. The pho-

tomicrographs were taken at the same magnifica-

tion. The network (cored structure) of primary

Aligned

creep voids



carbides does not appear to be intact near the

crack. A cored, lacelike structure was present

away from the crack, indicating that lower tem-

peratures were present away from the crack. Also

evident near the crack was the presence of sec-

ondary carbides and nitrides. The nitrides were not

evident away from the crack, which is to be sus-

pected. Nitrides are known to form adjacent to

fracture surfaces. Figure 17 is a photomicrograph

of the microstructure from Ring Section 2. The

cored, lacelike structure and secondary carbides

are present as was shown away from the crack in

Ring Section 1.

Figure 15. Light photomicrograph of the microstruc-

ture of Ring Section 1 near a crack and OD surface

(glyceregia etchant).

Based on the microstructures observed in the

header samples and the reference material, their

maximum service temperatures were estimated to

be:

• Ring Section 1: 870°C

• Ring Section 2: 840°C

The reference material had four times as much

carbon as the header samples, which made it diffi-

cult to estimate these values. Thus, there is uncer-

tainty regarding them.

The amount of niobium (1.91 wt% compared with

the specified range of 0.90 to 1.35 wt%) in the

header section was greater than the maximum val-

ue of the specified range. The amounts of the oth-

er elements met the manufacturer’s specification

for 20Cr-32Ni-Nb. Higher than desired niobium

content may have reduced the creep strength of

the 20Cr-32Ni-Nb alloy. This possibility was

pointed out in the previous review of published

investigations.


ture of Ring Section 1 away from a crack and near

the OD surface (glyceregia etchant).


ture of Ring Section 2 (glyceregia etchant).

Summary. A summary of the key observations

follows:

• The metallographic evidence (wide creep

cracks, low density of voids) suggests that

Header Section 1 failed by short-term over-

heating.



• Ring Section 2 showed metallographic evi-

dence of creep damage. The remaining life of

this section was estimated to be 70%, while

the maximum service temperature was esti-

mated to be 840°C.

• Using finite-element stress analysis, the max-

imum tensile hoop stress in the region of the

header failure was calculated to be 10.2 MPa.

For this stress, a temperature of 963 °C was

calculated to cause stress-rupture failure in

100,000 hours, indicating that the material

was either overheated, weaker than expected,

or a combination of both.

Case 3. Metallurgical Analysis of Outlet

Header a Hydrogen Reformer

During an inspection, a significant increase in di-

ameter was noted in one of two headers in a hy-

drogen reformer. The headers were made of the

cast 20Cr-32Ni-Nb alloy. The portion of the re-

former that fed gas to this header was known to

have been operating at higher than desired tem-

peratures. Further inspections revealed the pres-

ence of cracks in the header, especially in the

around the holes where fittings for the outlet pig-

tails were attached. As a result, the operator de-

cided to replace the header and cut out two ring

sections for metallurgical failure analysis. Ring

Section 1 was from the central portion of the

header tube where the increase in diameter was

the largest and the most material damage was ex-

pected to be found. Ring Section 2 was from near

the end of the header tube where the increase in

diameter was the smallest and the least material

damage was expected to be found.

Approach. Ring Sections 1 and 2 were visually

inspected and photographed and then were po-

lished and photographs were taken to document

the observed creep damage. Two portions from

Ring Section 1 and two portions from Ring Sec-

tion 2 were removed, mounted, polished, and

etched with glyceregia. Photomicrographs were

taken to document the creep voids prior to etching

and to document the microstructure after etching.

Results and Discussion. Figure 18 is a photo-

graph of Ring Section 1 following polishing. The

ring section was approximately 220 mm in diame-

ter by14.5-mm thick by 64 mm wide. Ring Sec-

tion 1 contained an inlet port for a pigtail, as indi-

cated in the figure. The port was approximately

31.8 mm in diameter and the wall was thicker at

this location, approximately 15.9-mm thick.

Figure 18. Photograph of Ring Section 1 after po-

lishing.

Indicated and visible on the polished surface is an

OD surface breaking crack that is within the limits

of the previously mentioned port. Although not

visible in the figure, a crack of similar size is lo-

cated on the ID surface of the port, below the visi-

ble crack.

Ring Section 2 was approximately 217 mm in di-

ameter by 14.3 mm thick. The location of interest

on the ring had a red marking.

Metallographic Examination. Figures 19 and 20

are photomicrographs of Met Section 1a (removed

from Ring Section 1) taken adjacent to the OD

and ID surface, respectively. Met Section 1a was

removed near the port of the pigtail inlet. The re-

sidual life of the tube section that Ring Section 1

was removed from is approximately 20% based on

the metallurgical analysis and guidelines of past

studies [33]. Based on the smaller size of the indi-

vidual voids in the figures compared to the thicker

width of the cracks indicated in Figure 19, it ap-

Port for

pigtail inlet

Cracks

Polished

surface



pears that the material tore after the creep voids

aligned. This tearing of the material suggests, but

does not indicate, that high-temperature creep was

involved. A high density of voids, but no cracking

was observed near the ID surface. The aligning

and linking up of the voids appears to follow the

dendritic columnar structure adjacent to the OD

surface. The location of the voids adjacent to the

ID surface suggests an equiaxed fine-grained

structure. The density of the voids is higher adja-

cent to the ID surface.

Figure 19. Light photomicrograph of Met Section

1a, taken adjacent to the OD surface and near the

pigtail inlet (as polished).


1a, taken adjacent to the ID surface and near the pig-

tail inlet (as polished).

Met Section 1b was removed from Ring Section

1opposite the port of the pigtail inlet. The density

of the voids throughout Met Section 1b was low.

Minor aligning of the creep voids was visible ad-

jacent to the OD surface.

Figure 21 is photomicrograph of Met Section 2a

(removed from Ring Section 2) taken adjacent to

the OD. The residual life of the tube section that

Ring Section 2 was removed from is approximate-

ly 60% based on the metallurgical analysis and

guidelines of past studies [33]. Met Section 2b

was removed opposite from the location marked

in red on the OD surface. The density of voids in

on both met sections, throughout the wall thick-

ness, is uniform and high. The voids adjacent to

the OD surface are aligned (as shown in Fig-

ure 21) and appear to follow a dendritic columnar

structure. The location of the voids adjacent to the

ID surface suggests an equiaxed fine-grained

structure. Thus, even though Ring Section 2 had

less creep damage than Ring Section 1, it still had

a significant amount of creep damage.



red marking (as polished).

Figures 22 and 23 are photomicrographs of the

microstructure of Met Section 1a (removed from

Ring Section 1) near the pigtail inlet. The figures

show primary carbides located at the grain boun-

daries and secondary carbides dispersed in the

grains. The network of primary carbides is more

Cracks/aligned

creep voids

Aligned

creep voids



intact near the ID surface than near the OD surface

which indicates that the OD surface was subjected

to higher temperatures than the ID surface.



pigtail inlet (glyceregia etchant).


1a, taken adjacent to the ID surface and near the pig-

tail inlet (glyceregia etchant).

Figure 24 is a photomicrograph of the microstruc-

ture of Met Section 1a (removed from Ring Sec-

tion 1) adjacent to the OD surface. A demarcation

is visible in the figure. There is no evidence of a

network of carbides above the demarcation and

there is a network below the demarcation. The re-

gion where the demarcation is present was likely

overheated since there is a severe reduction in the

amount of primary and secondary carbides there.



pigtail inlet showing the decarburized region (glyce-

regia).

For the microstructure of Met Section 1a (re-

moved from Ring Section 1), opposite the port of

the pigtail inlet, the network of primary carbides is

less intact and more secondary carbides are dis-

persed in the matrix adjacent to the OD surface as

compared to the ID surface.

For the microstructure of Met Section 2a (re-

moved from Ring Section 2), at locations 180°

from each other, the network of primary carbides

adjacent to the OD surface at both locations is less

intact than at the ID surface. Again, the structure

of the primary carbides indicates that the OD sur-

face was subjected to higher temperatures than the

ID surface. The quantity of secondary carbides

appears to be similar when comparing the loca-

tions adjacent to the OD and ID surface.

Summary. A summary of the results follows:

• For Ring Section 1, the width of the creep

cracks near the port for the pigtail inlet suggests

a high temperature creep mechanism. A decar-

burized region near the port indicates that the

material was locally overheated. The creep

Demarcation



cracks and decarburized region were both lo-

cated adjacent to the OD surface. Aligned creep

voids, but no cracking, were identified in Ring

Section 2.

• The remaining lives of Ring Section 1 and 2

were estimated to be 20% and 60%, respective-

ly.

Creep-Rupture Strength

Jaske [34] has previously reviewed the creep-

rupture strength of weldments for hot outlet head-

ers and compared their stress-rupture behavior

with that of the cast 20Cr-32Ni-Nb alloy. Howev-

er, whereas as test data were available for the

weldments, only manufacturers’ curves (no data

points) were available for the cast base metal. The

lack of such published data makes it difficult to

assess the cause of the recent base metal failures

of hot outlet headers. Some operators and refor-

mer design engineers are of the opinion that the

creep-strength of the 20Cr-32Ni-Nb alloy is ac-

tually well below the published curves. For this

reason, they apply large design margins in estab-

lishing stress and temperature limits. Develop-

ment of reliable creep and stress-rupture data for

the 20Cr-32Ni-Nb alloy would be highly desira-

ble.

Conclusions

Hot outlet headers for steam-methane headers

have historically been subject to failure issues.

Early failures were related to the use of low-

strength alloys until the use of Alloy 800H or

800HT or 20Cr-32Ni-Nb became the construction

materials of choice. Then, more failure problems

became associated with welded joints and weld

metals with low creep strength. The strength and

quality of weld joints has increased, but they are

still prone to problems, especially when un-

planned for bending loads are induced in the

header. More recently, several failures have been

associated with the cast 20Cr-32Ni-Nb base metal

bringing into question its creep-rupture strength.

References

1. Kawai, T., Takemura, K., and Zaghloul,

M. B., “Design Features, Material Selection

and Performance of Steam Reforming Fur-

nace,” Proceedings of the International Plant

Engineering Conference (INPEC), Bombay,

November 1984.

2. Francis, M. F., and Glass, J. A., “Failure of

Primary Reformer Outlet Piping,” Proceedings

of Ammonia Plants & Related Facilities Sym-

posium, Vol. 4, American Institute of Chemi-

cal Engineers, New York, 1963, pp. 1-3.

3. Avery, R. E., and Valentine, H. L., “High-

Temperature Piping Systems for Petrochemi-

cal Processing,” Proceedings of Ammonia

Plants & Related Facilities Symposium, Vol.

8, American Institute of Chemical Engineers,

New York, 1967, pp. 20-24.

4. Harnby, G., “Brittle Fracture of Centrifugally

Cast Stainless Steels,” Proceedings of Ammo-

nia Plants & Related Facilities Symposium,

Vol. 8, American Institute of Chemical Engi-

neers, New York, 1967, pp. 16-19.

5. Blackburn, J., “Cast Reformer Headers and

Manifolds,” Proceedings of Ammonia Plants

& Related Facilities Symposium, Vol. 16,

American Institute of Chemical Engineers,

New York, 1976, pp. 16-20.

6. Shibasaki, T., Takemura, K., Kawai, T., and

Mohri, T., “Experiences of Niobium-

Containing Alloys for Steam Reformers,”

Proceedings of Ammonia Plants & Related

Facilities Symposium, Vol. 27, American In-

stitute of Chemical Engineers, New York,

1986, pp. 56-68.

7. Parks, S. B., and Schillmoller, C. M., “Update

in Alloy Selection for Ammonia Furnaces,”

Proceedings of Ammonia Plants & Related

Facilities Symposium, Vol. 37, American In-

stitute of Chemical Engineers, New York,

1996, pp. 220-227.



8. Thomas, C. W., Knowles, D. M., and Keen,

D. J., “Embrittlement in Cast Reformer Outlet

Manifold Components,” Proceedings of Am-

monia Plants & Related Facilities Symposium,

Vol. 44, American Institute of Chemical En-

gineers, New York, 2003, pp. 136-143.

9. Hoffman, J. J., “High Temperature Aging

Characteristics of 20Cr32Ni1Nb Castings,”

Paper No. 00512, CORROSION/2000, NACE

International, Houston, Texas, 2000.

10. Hoffman, J. J., and Gapinski, G. E., “Proper-

ties and Microstructures of Outlet Manifold

Components,” Ammonia Technical Manual

(2001/2002), American Institute of Chemical

Engineers, New York, 2002, pp. 10-21.

11. Roach, D. B., and VanEcho, J. A., “Creep-

Rupture Properties of HK-40 and Alloy 800

Weldments,” Materials Performance, 20 (11),

1981, pp. 43-49.

12. Roach, D. B., and VanEcho, J. A., “Investiga-

tion of Mechanical, Physical, and Creep-

Rupture Properties of Reformer Materials,”

Technical Report on Materials for Steam Re-

former Furnaces, Battelle Memorial Institute,

Columbus, Ohio, 1973.

13. McMillan, C. L., “Service Behavior of Re-

former Outlet Manifolds,” Proceedings of

Ammonia Plants & Related Facilities Sympo-

sium, Vol. 16, American Institute of Chemical


14. Keen, D. J., Firth, D. M., and Thomas, C. W.,

“Learnings from Completing Risk Based As-

sessments on 8 Ammonia Plants and Down-

stream Urea, Nitric Acid and Fertilizer

Plants,” Proceedings of Ammonia Plants &

Related Facilities Symposium, Vol. 46, Amer-

ican Institute of Chemical Engineers, New

York, 2005, pp. 255-273.

15. Kobrin, G., “Metallurgical Evaluation of a

Primary Reformer,” Proceedings of Ammonia



New York, 1977, pp. 103-110.

16. Shibasaki, T., Mohri, T., and Takemura, K.,

“Experience with Cast Material for Steam Re-

former Surfaces,” Proceedings of Ammonia



New York, 1993, pp. 166-176.

17. Orbons, H. G., “Weld Cracking in Reformer

Outlet Parts after 12 Years in Service,” Pro-

ceedings of Ammonia Plants & Related Facili-

ties Symposium, Vol. 28, American Institute

of Chemical Engineers, New York, 1987, pp.

45-52.

18. van Wortel, H. “Relaxation cracking in auste-

nitic welded joints – an underestimated prob-

lem,” Stainless Steel Europe, June 1995.

19. Anon., “INCOLOY® alloy 800H & 800HT,”

Publication Number SMC-047, Special Metals

Corporation, September 2004.

20. Al-Jubeihi, H. E., “Improved Service Factor

by Primary Reformer Modifications,” Pro-



of Chemical Engineers, New York, 1982.

21. Thielsch, T., and Cone, F. M., “Failures, Inte-

grity and Repairs of Petrochemical Plant Re-

formers,” Proceedings of Ammonia Plants &



York, 1992, pp. 30-52.

22. Pande, J. O., and Swensen, B., “Ammonia

Plant Nil: Creep Damages in Outlet Manifolds

of Primary Reformer,” Proceedings of Am-

monia Plants & Related Facilities Symposium,

Vol. 35, American Institute of Chemical En-

gineers, New York, 1994, pp. 175-183.

23. Shi, S. and Lippold, J., “Repair Weldability of

Service-Aged Heat Resisting Stainless

Steels,” Proceedings of ASM 7th International

Conference on Trends in Welding Research,

ASM International, Materials Park, OH, 2005.

24. Pattabathula, V., Rani, B., and Timbres, D. H.,

“The AIChE Ammonia Safety Symposium 50

Years of Shared Experiences,” Proceedings of



Ammonia Plants & Related Facilities Sympo-

sium, Vol. 46, American Institute of Chemical


25. Shibasaki, T., Takemura, K., Mohri, T., and

Hashimoto, H., “Experience with Metal Dust-

ing in H2/CO/CO2/H2O Atmosphere,” Pro-



of Chemical Engineers, New York, 1995, pp.

165-179.

26. Lobley, G. R., Kermad, A., Al-Buraiki, I. A.,

and Al-Nuaim, T. A., 2003, “Assessment of

Aged Steam Reformer Furnace,” Proceedings

of Cape 2003, Cape Town, Republic of South

Africa, April 7-11.

27. Anderson, P., and Bruno, T., “Weldability Is-

sues in Fe-20Cr-32Ni-1Nb Castings,” API

Roundtable Discussion: Issues with Alloys in

Hydrogen Furnaces, API Spring Refining and

Equipment Standards Meeting, Dallas, May

2006.

28. Penso, J., and Mead, H., “Repair Case Histo-

ries and Mitigation of Cast 20-32 Nb Reduc-

ers,” API Roundtable Discussion: Issues with

Alloys in Hydrogen Furnaces, API Spring Re-

fining and Equipment Standards Meeting,

Dallas, May 2006.

29. Jack, B., “Subheader to Tee Weld Failures,”

API Roundtable Discussion: Issues with Al-

loys in Hydrogen Furnaces, API Spring Refin-

ing and Equipment Standards Meeting, Dallas,

May 2006.

30. Dejaeger, J., Guns, L., and Korkhaus, J., “Ex-

periences with Alloy 800H in Ammonia

Plants,” Proceedings of Ammonia Plants &



York, 1994, pp. 134-144.

31. Anon., “2007 ASME Boiler & Pressure Ves-

sel Code, Section VIII, Division 1, Rules for

Construction of Pressure Vessels,” The Amer-

ican Society of Mechanical Engineers, New

York, 2007.

32. Brongers, M. P., and Jaske, C. E., “Creep-

Rupture of Service-Exposed Base Metal and

Weldments of Alloy 800H,” PVP-Vol. 409,

Aging Management, Component and Piping

Analysis, Nondestructive Engineering, Moni-

toring and Diagnostics — 2000, The Ameri-

can Society of Mechanical Engineers, New

York, 2000, pp. 143-153.

33. Dillenger, L., and Roach, D., "Metallurgy and

Metallography of Reformer Alloys," Report to

Materials for Steam Reformer Furnaces

Group, Battelle, Columbus, Ohio, May 1,

1975.

34. Jaske, C. E., “Creep Strength of Welds in Hy-

drogen Reformer Outlet Manifolds,” Paper

PVP2009-77091, Proceedings of the 2009

ASME Pressure Vessels and Piping Confe-

rence, The American Society of Mechanical

Engineers, New York, July 2009.


mechanical integrity of primary reformer hot outlet headers...steam-methane reformers have either...

Documents