Materials and Corrosion 2005, 56, No.

7 Superheater corrosion in a CFB waste boiler 449

Field test of superheater corrosion in a CFB

waste boiler: Part I – Metal loss characteristics
P. Andersson and M. Norell*

The corrosion was investigated on a superheater test coil in a crest. The nickel-based Alloy 65 was preferentially attacked di-
CFB waste boiler. The alloys ranged from ferritic steel T22 to ni- rectly under the crest of the deposit while the other alloys were pre-
ckel-based Alloy 65 and the metal temperatures were between 460 ferentially attacked at the edge. The corrosion rate increased with
and 540 8C. The thickness of the deposit was alloy and temperature temperature for X20, Alloy 304L, Alloy 310 and Alloy 825; de-
dependent. The low-alloyed steels developed thick deposits at all creased on Alloy 65; and was bell shaped on T22 and Alloy 28.
temperatures while the deposit thickness increased with the tem- Alloy 310 suffered from severe pitting corrosion in a line following
perature on the high-alloyed steels and the nickel-based alloy. the edge of the deposit crest. The best overall corrosion resistant
The corrosion attack was alloy dependent and related to the deposit alloy was Alloy 28.

1 Introduction Waste combustion is done primarily for two reasons; waste

disposal and energy conversion. Waste disposal by incinera-
The corrosive environment in biofuel combustion is gener- tion reduces its volume and thus the need for landfilling. For
ally worse than when fossil fuels are burnt. This is particularly long times it has been attractive to combine the disposal of
valid for the combustion of waste that is considered to be a waste with the recovery of its energy content as heat. Further-
biofuel, either it is in the form of Municipal Solid Waste more, as a biofuel, waste is not considered to give any net
(MSW) or Refuse Derived Fuel (RDF). In biofuel combustion emission of CO2 that would contribute to global warming. In-
large amounts of deposits may form on the superheater tubes creasing electricity cost is a further driving force to convert
and other surfaces. The deposit formation depends on e.g. the the produced heat into electricity. This conversion has created
flue gas and metal temperatures, as well as the flue gas com- a demand for increased steam temperatures and thus metal
position and flow. The deposit formed limits the efficiency of temperatures in the superheaters. Today the steam temperature
the boiler. First, the heat transport between the gas and the is limited by the corrosion attack [1, 9 – 10]. The corrosion
produced steam is limited. This effect is referred to as fouling. rate, burning MSW, has been observed to increase rapidly
Second, the deposits may induce rapid corrosion mechanisms at metal temperatures above 300 8C [9 – 10]. This has, for ex-
leading to tube failure. To avoid this, the steam temperatures ample, made Japan avoid steam temperatures above 300 8C
must be kept substantially lower than in combustion of fossil until 1995 [11]. Today, steam is produced at temperatures
fuels. up to about 450 8C in waste boilers around the world [1,
The operating temperature of the boiler depends among 12]. As a comparison, alloy development for steam turbines
other things on the kind of biomass that is burnt. Apart is now aiming at 650 8C or even 700 8C to improve the thermal
from waste, it may be e.g. peat, straw, woodchips, etc. The efficiency of turbines [13]. Ideally, the steam produced in
deposit forming elements in these fuels are mainly S, Cl, waste combustion should be equally hot.
O, K, Na and Ca. The danger of corrosion attack has been The boiler design is an important approach to compete
observed to correlate to the ratio between the chlorine and against the corrosion problems in waste combustion. How-
the sulphur contents in the fuel [1]. The higher this ratio is, ever, it must be combined with an appropriate choice, or de-
the more severe is the corrosion. Among the biofuels, waste velopment, of alloys. This calls for deep knowledge of the cor-
often has a high Cl/S ratio and it frequently gives a corrosive rosion in the boiler. The problem is complex. The composition
environment in the boiler. Waste is sometimes mixed with of the fuel, flue gases and deposits formed are examples of
other fuels to decrease the corrosivity, leading to decreased factors that both vary between different boilers and fluctuate
Cl/S ratio [2, 3]. Furthermore, MSW sometimes contains with time. Hence, research about corrosion in waste boilers
high levels of elements such as Pb, Zn and Sn that can be en- combines several approaches ranging from evaluation of field
riched in the deposit. Chlorides of these heavy metals may, experience, via field tests of various kinds to laboratory ex-
together with alkali chlorides, form eutectic chloride melts periments. The field tests, in their turn, range from separate
at low temperatures that will increase the corrosion rate probes that are inserted at available ports to large scale test
[4 – 8]. sections that are built in at ordinary positions. The previous
approach is flexible and less costly. It can be used e.g. to com-
pare the corrosion between different boilers. However, the
samples are usually limited to small rings that give limited
information about the corrosion and the probes are not tested
* M. Norell, P. Andersson,
Materials and Manufacturing Technology, at ordinary tube positions. Thus, the corrosion on a probe may
Chalmers University of Technology, not be the same as in the boiler. The larger test sections basi-
S-412 96 Göteborg (Sweden) cally have opposite characteristics on all these points. The
E-mail: mats.norell@me.chalmess.se samples are test tubes that can give extensive and representa-

F 2005 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/maco.200403852
450 Andersson and Norell Materials and Corrosion 2005, 56, No. 7

tive information about the corrosion in the boiler. Since they 2 Experimental
are tested at normal tube positions the flue gas flow pattern,
conditions for deposit build up and other factors are charac- 2.1 Field test
teristic for the boiler. Whatever test method, the reported cor-
rosion rates are often limited to the deepest thickness reduc- The field test was conducted for 738 h in a CFB-boiler at
tion. This can be sufficient to classify the alloys, but it only Korstaverket in Sundsvall, Sweden. The boiler consists of the
gives a moderate description of the corrosion attack. For in- fluidized bed, followed by a cyclone that separates the parti-
stance, the most severe corrosion on superheater tubes is nor- cles and reintroduces them into the bed. After the cyclone lies
mally expected at the edge of the deposit crests, but there are the upper draw with superheater tubes. The boiler was built in
only few systematic quantitative examinations whether this is 1984 and external systems were modified in 1995. In normal
the case [4, 14]. operation, the boiler is burning a mixture of 60% RDF and
Candidate alloys for waste boiler superheaters range from 40% woodchips. During the field test, the boiler was burning
ferritic and martensitic steels to austenitic steel and advanced RDF only. The boiler capacity is 5.5 ton RDF per hour pro-
nickel-based alloys. Extensive field tests of different alloys ducing 28 ton steam at 33 bar/ 285 8C. The fuel was shredded
have been done in Japan during the last years, especially and samples were collected throughout the test. These col-
by Kawahara and co-authors [15 – 17]. They have found lected samples were blended and 0.5 kg of the blended fuel
that the corrosion resistance increases with the sum of the sample was analysed. A detailed analysis of the flue gas com-
Cr, Ni and Mo contents. The effect of alloying was stronger position in the furnace is found in the paper about the corro-
the more corrosive was the environment. Probe tests in diffe- sion in the waterwall [19].
rent boilers showed that a CrþNi content of approximately The test coil was inserted in the upper draw at the top of the
80% gave the best corrosion resistance [15]. Latter field tests back pass, on candidate, but not ordinary, superheater posi-
gave the best performance for Ni-base alloys with high tions. A schematic view of the coil is shown in Fig. 1. The
amounts of Mo (19%) in the most corrosive locations (high superheater test coil constituted a separate steam circuit.
Cl-content in the deposits at 550 8C). However, besides being The coil consisted of two layers. The top layer served primar-
costly, such alloys are not a universal solution to the corrosion ily to heat up the steam to a higher temperature than in the
problems in waste combustion. For instance, Alloy 600 ordinary superheater of the boiler. This layer was also in-
(72Ni16Cr) may suffer from severe pitting [18]. In the present tended to shield the test tubes from the direct impact of
study, it will be shown that a Mo-containing Ni-base alloy cor- flue gases and particles. The test tubes were situated in the
roded more then less alloyed materials, especially at lower lower layer. Prior to the field test it was observed that the de-
temperatures. Many of the alloys studied here were tested posit formation was very limited in one quarter of the section
as composite tubes combining a ferritic core and a high-alloy where the coil should be inserted. This was probably related to
outer layer. This gives a possibility for favourable combina- the flue gas flow pattern in the boiler. Aiming to expose the
tion of creep strength, weldability and corrosion resistance, test tubes to as equal conditions as possible no test tubes were
etc. Field experience regarding composite tubes has been re- mounted in this region. A soot blower was situated in the
ported in the literature [4]. upper draw. However, it was mounted at least 1.5 m above
In this study, a field test was done in a Circulating Fluidised the test coil so it is not expected to have had any major influ-
Bed (CFB) burning RDF. The aim was to investigate the cor- ence on the deposit formation, nor on the corrosion.
rosion of different alloys in a realistic flue gas flow and in a The alloys were tested in a metal temperature gradient ran-
temperature range (460 – 540 8C) that was higher than normal ging from 460 8C to 540 8C. The metal temperatures of the
in waste boiler superheaters today, but relevant for the future. tubes were monitored by four thermocouples that were sol-
For this purpose, a test coil was inserted in the top of the back dered in the tube wall on the outer side, c.f. Fig. 1. Linear in-
pass. Alloys ranging from ferritic to nickel-base were tested ter- and extrapolations were used to estimate the metal tem-
for one month. Also, the corrosion in the waterwall was ex- peratures of individual test tubes. All alloys were tested at
amined in the field test. Results from that study have been
published previously [19].
The paper is divided in two parts. This paper, part 1, con-
centrates on the metal loss of all seven alloys tested. In order
to get a detailed picture of how the corrosion attack differed
between the alloys the thickness reduction was measured both
around and along the test tubes. The attack was correlated to
the position of the deposit crest and the tube temperature. The
test tubes were also long enough (380 mm) to get representa-
tive data not only from the thickness reduction but also from
the weight loss. Furthermore, the amount of deposit formed
was estimated to see how it related to the temperature or com-
position of the alloys as well as to the corrosion.
The second part of this paper will present an in-depth de-
scription of the corrosion of four selected alloys from the test
with significantly different corrosion behaviour. The deposits
and oxides formed on these alloys have been characterised.
That paper will attempt to analyse more in detail what factors
or reactions that determined the attack of the alloys.
Fig. 1. Schematic view of the superheater test coil in the upper
draw. The metal temperature and test tube positions are indicated.
Thermocouple locations are indicated T
Materials and Corrosion 2005, 56, No. 7 Superheater corrosion in a CFB waste boiler 451

three different temperatures except alloy T22 that was tested

at six temperatures.
The tested alloys ranged from ferritic Fe2Cr1Mo and mar-
tensitic Fe10Cr steels to the nickel-based Alloy 65. Though
the alloys T22 and X20 might not be considered candidate
alloys for these advanced conditions, they are commonly
used in many boilers and important as references. The com-
position of the alloys is shown in Tab. 1. The test tube length,
diameter and wall thickness were 380 mm, 44.5 – 54.0 mm
and 4.5 – 6.0 mm, respectively.
When the test was interrupted air was blown in the boiler Fig. 2. Schematic view of the split deposit for-
during cooling to avoid aqueous corrosion. After cooling the mation. The tube sectors were indexed from the
test tubes and their environment were photographed and highest deposit crest
roughly characterised. Deposit and ash samples were col-
lected and the test coil was removed. All tubes and samples
were stored in a sealed box containing silica gel to minimise
the hydration of the samples and avoid spallation of the de- two deposit crests were sometimes formed on the test tubes,
posit during storage. c.f. Fig. 2. Since the flue gas flow was not perpendicular to the
test coil one of the crests were higher than the other. The es-
timated direction of the highest crest was indexed as the sector
2.2 Evaluation at 08, with an uncertainty of
308. Ten thickness measure-
ments, about 10 mm apart, were done along a line in each sec-
Sections were prepared of all test tubes for subsequent ana- tor. The average from the ten points was then used to describe
lysis of the scale and deposit formation by using SEM and how the attack varied around the tubes as showed in circular
other methods. The deposits were first mechanically fixed diagrams below. The deepest sector average was used to com-
to the tubes with epoxy where the sections were cut. They pare the corrosion rate of the alloys. Additionally, the deepest
were taken a few centimetres away from the welds between (i.e. thinnest) point was determined on each tube. It was usual-
the test tubes to avoid the heat affected zone. The results of ly found in the deepest sector.
these analyses will be presented in part two of this paper.
The remaining tube, approximately 250 mm long, was used
to characterise the metal loss. All deposit and corrosion pro- 3 Results
ducts were removed by the use of a 600 bar water jet. After
cleaning, approximately 1 lm of oxide remained on the tubes. 3.1 Environment
High precision measurements of the weight and length of
the tubes were done before the test coil was built, and again The test was run for a month. However, the boiler was shut
after cleaning. The weight loss per length unit was converted down twice during that period for reasons external to the test.
to an equivalent average corrosion rate of the entire tube sur- The test coil was kept at temperature for 738 hours. Still, there
face. The accuracy of the measurement was equivalent to were some temperature fluctuations during this period. Some
about 90 lm/year. occasional temperature drops of at the most 120 8C occurred
The thickness reduction was measured around the tubes to lasting a few minutes. Once the temperature rose with 100 8C.
get a quantitative description of how the tubes had been at- The temperatures were constantly oscillating about
10 8C.
tacked. The remaining wall thickness was determined by After half the test period the thermocouples were broken. So
using a dial tube gauge and compared to that of unexposed far there was no indication that the metal temperature was de-
tubes. The accuracy in each measurement was about creasing due to the deposit formation. Throughout the test the
0.05 mm, equivalent to 0.6 mm/year. The thickness loss temperature of the outgoing steam was kept constant at
was measured in 12 sectors at 308 intervals around the tubes. 500 8C. The average test tube metal temperatures, as estimated
The top layer of the test coil splitted the fuel gas flow so that from the thermocouples, are indicated in Fig. 1.

Table 1. Chemical composition (wt-%) of the tested alloys according to melt analyses. Fe is balance.

Alloy C Cr Ni Mo Si Mn Others

T22 0.12 2.06 1.0 0.3 0.6

X20 (X20CrMoV121) 0.18 10.3 0.47 0.8 0.3 0.8 0.2Cu 0.3V
Alloy 304L 0.01 18.3 10.1 0.4 1.2
Alloy 310a 0.02 25.4 21.1 0.1 0.4 1.7 0.1Cu
Alloy 28a,b 0.01 26.7 30.4 3.3 0.4 1.6 0.9Cu
Alloy 825 0.01 22.4 39.5 3.2 0.4 0.2 2.0Cu 0.1Al
0.7Ti 0.5Co
Alloy 65a,b 0.01 20.7 61.3 8.3 0.4 0.4
[a] Composite tubes
[b] These alloy tradenames are Sanicro 28 and Sanicro 65, respectively
452 Andersson and Norell Materials and Corrosion 2005, 56, No. 7

Table 2. Content of some elements in the previously dried fuel sample. The sample humidity was 6.3 wt-%

Element wt-% Element wt-% Element wt-% Element wt-% Element mg/kg Element mg/kg
Ash 15.4 S 0.16 Ba 0.02 P 0.13 Pb 133 Cu 225
Al 1.35 Mn 0.03 Na 0.93 Cl 0.42 Cd 1.6 Ni 35
Si 3.70 Mg 0.25 K 0.43 F 0.01 Co 10 V 9
Fe 0.38 Ca 1.99 Ti 0.14 Br < 0.01 Cr 55 Zn 425

Refuse Derived Fuel is very inhomogeneous. Therefore, posit height varied between the test tubes from a few milli-
though the fuel was shredded and samples were collected metres to several centimetres. The changes were incremental
throughout, the analysis, c.f. Tab. 2, only gives an indication were the tubes were joined. Fig. 3 shows the height of the de-
of the fuel composition. posit crests as measured on tube sections. Though there is sub-
Collected samples of the deposits were analysed by means stantial scatter in the data, two trends can be observed. First,
of XRD. Independent of the alloy or the temperature the bulk the deposit thickness increased with the temperature, at least
of the deposits consisted mainly of alkali chlorides and calci- for the high-alloyed materials. Secondly, thicker deposits
um sulphate. This is in agreement with a previous analysis of were formed on the ferritic/martensitic steels than on the
these deposits [20]. The composition of the deposits will be high-alloyed materials.
discussed a bit more in detail in part two of this paper.
At the visual inspection of the test coil, it was strikingly
clear that the deposit formation was alloy dependent. The de- 3.2 Metal loss

3.2.1 General

The corrosion rates were very high for all alloys. The poor-
est tube would probably have been broken if the test had lasted
one or two weeks more. The rates ranged from 1.8 mm/year
for high-alloyed austenitic steel Alloy 28 at 462 8C to 34 mm/
year for the ferritic steel T22 at 516 8C. The corrosion rates are
summarised in Figs. 4 and 5. They are presented in mm/year
though this does not imply that the rate should be linear for so
long times. Fig. 4 shows the rates according to the thickness
reduction in the deepest sector and the deepest point as well as
the weight loss per length unit. For a better comparison of the
influence of the temperature Fig. 5 shows the rate in the dee-
pest point versus the metal temperature. The wall thickness of
unexposed tubes of X20 was uneven since the inner and outer
diameters were eccentric. This complicated the evaluation of
the thickness reduction. Therefore, the data for the tubes kept
at 478 8C and 534 8C is only indicative. Only the weight loss is
shown for the tube at 510 8C. The thickness data for Alloy
304L are somewhat uncertain for similar reasons.
The corrosion attack of the individual alloys will be pre-
Fig. 3. The height of the deposit crests on the test tubes sented below but first some general trends will be pointed

Fig. 4. The corrosion rates for all test

tubes according to the thickness reduction
in the deepest sector and the deepest point
as well as the weight loss per length unit
Materials and Corrosion 2005, 56, No. 7 Superheater corrosion in a CFB waste boiler 453

Fig. 5. The corrosion rate in the deepest point versus temperature

Fig. 7. The average thickness reduction around tubes of alloy T22.
The highest deposit crests were located at the top
out. The weight loss data gives a reliable measure of the av-
erage thickness loss on the entire tube surface. However, it is
The average corrosion depths in the 12 sectors around the
not very useful to estimate the tube lifetime. The occurrence of
tube periphery are shown in Fig. 7 for all test tubes of alloy
a few deep pits will not be reflected in the weight loss. Since
T22. The diagrams have been rotated so that the estimated
the weight loss also shows similar trends as the thickness re-
direction of the highest deposit crest is upwards for all tubes.
duction it will not be further discussed, except for alloy X20.
It is seen that the tube at 516 8C was attacked all around the
The difference between the deepest point and the deepest sec-
tube, but most at approximately
908 from the crest. The av-
tor is an indicator of the tube roughness. A large difference
erage corrosion of the tube at 502 8C was less severe, but in
may occur when a pitting attack took place on the tube, as
pits away from the deposits the thickness was still reduced
for e.g. the austenitic steel Alloy 310.
with up to 2 mm. For all other test tubes the pits were concen-
The temperature dependence of the ferritic steel T22 shows
trated to a band at the edge of the deposit, typically 30 – 608
a clear bell shaped form with a maximum attack at 515 8C.
from the crest. That was also the location of the deepest sec-
Despite that the decrease at high temperature was less pro-
tors.
nounced, also the corrosion rate of the Ni rich austenitic steel
Alloy 28 was slightly bell-shaped. For the Ni-based alloy Al-
loy 65 the highest corrosion rate actually occurred at the low-
3.2.3 The martensitic steel X20, full name X20CrMoV121
est temperature. All other alloys had rates that increased with
(Fe10Cr1Mo)
the temperature.
The corrosion of alloy X20 was quite severe not only under
the deposits and at the edges of them, but also on the sides of
3.2.2 The ferritic steel T22 (Fe2Cr1Mo)
the tubes. It was only on the leeward side that the corrosion
was slower and the pitting less severe. Both the overall corro-
The most striking with this alloy is the dramatic attack at
sion rate and the general pitting attack increased substantially
516 8C. The deepest point was 2.9 mm under the original me-
with the temperature, but there were a few deep pits formed
tal surface. Only 1.6 mm remained. Fig. 6 shows a photo of
already at 478 8C. The pits that were formed did not have the
this tube after water-jet cleaning. The deposit crest had
same shape as those on alloy T22. They were smaller, typi-
been about in the middle of this view. Large interconnected
cally 1 mm in size, with sharper edges between them (c.f.
pits are seen towards the sides. The appearance is that of
Fig. 8). Apart from the pits the surface was rougher or
hot corrosion attack.
wavy, especially at the highest temperature. It was noticed
that the oxide was darker and more adherent than on any other

Fig. 6. The water-jet cleaned surface of T22 at 516 8C. The deposit Fig. 8. The water-jet cleaned surface of X20 at 534 8C. The deposit
crest was located in the center crest was located in the centre
454 Andersson and Norell Materials and Corrosion 2005, 56, No. 7

Fig. 11. The average thickness reduction around tubes of Alloy

Fig. 9. The average thickness reduction around tubes of Alloy 310. The highest deposit crests were located at the top
304L. The highest deposit crests were located at the top

alloy. The oxide was not completely removed by the water-jet 3.2.5 The austenitic steel Alloy 310 (Fe25Cr21Ni)
cleaning. Since the unexposed tube had uneven wall thickness
no diagram with thickness reduction around the tubes is This alloy suffered from severe pitting in a manner different
shown. from all other alloys in the test. The pits were formed at all
temperatures following a narrow line about 308 from the de-
posit crest, c.f. Fig. 10. The number and size of the pits in-
3.2.4 The austenitic steel Alloy 304L (Fe18Cr10Ni) creased with increasing temperature. At the two higher tem-
peratures the pits actually penetrated the outer alloy down to
The attack of 304L was correlated to the deposit crest. This the ferritic core of this composite tube. The core was made in
can be clearly seen for the tube at 536 8C in Fig. 9. The thick- the ferritic steel T22. At the interface to the core tube the pits
ness reductions in the sectors at both 3008 and 30 – 908 are had flattened and spread a bit along the interface. In fact, it
deeper than at 3308. This corresponds to an attack that is dee- appears as if their growth in depth almost stopped at the inter-
per at the edges of the deposit than right under its crest. Also face. Few pits were actually deeper than the thickness of the
the tubes at 478 8C and 510 8C have corroded more on the edge outer alloy. This can also explain why the deepest points were
of the deposit than under the crest. Though this only occurred almost equally deep at 508 8C and 532 8C. The pitting attack
on one side at these temperatures, it should be remembered was observed mainly on one side of the deposit, a conse-
that the flue gas flow was not symmetric around the test tubes. quence of the top coil and the resulting difference in gas
The general corrosion of this alloy increased with the tem- flow, c.f. Fig. 11. The general corrosion rate increased with
perature. On the other hand, the pitting was less pronounced the temperature, but in view of the pitting attack this is of mi-
than on the ferritic/martensitic steels and also less temperature nor importance.
dependant.

Fig. 10. The water-jet cleaned surface of Alloy 310 at 508 8C. The
pits are located at 308 on the side of where the deposit crest was Fig. 12. The average thickness reduction around tubes of Alloy 28.
located The highest deposit crests were located at the top
Materials and Corrosion 2005, 56, No. 7 Superheater corrosion in a CFB waste boiler 455

Fig. 15. Detail of the water-jet cleaned surface of Alloy 65 at

474 8C. Half the tube diameter is shown. Pits are seen that occurred
on the side of the deposit crest
Fig. 13. The average thickness reduction around tubes of Alloy
825. The highest deposit crests were located at the top
3.2.7 The nickel-iron-based Alloy 825 (39Ni31Fe22Cr3Mo)

Also Alloy 825 corroded comparatively little in this test,

3.2.6 The high-alloyed austenitic steel Alloy 28 though the rates were substantial and increased with the tem-
(Fe27Cr30Ni3Mo) perature. It was the least corroded alloy at the middle tempera-
ture. Moreover, there were signs of pitting. As seen in Fig. 13
This was the least corroded alloy in the test both at the low- the attack was actually quite uniform, somewhat like for Alloy
est and highest test temperatures. The deepest point that was 28. The deepest sectors were found at the edges of the deposit
found after one month at 462 8C was about 0.14 mm deep. The crests, but they were not much deeper than under the crest.
corrosion was too moderate to establish its angular depend- Also, the leeward side of the tubes were attacked to almost
ence at this temperature. No signs of pitting were seen. similar extent.
The corrosion rate had a somewhat bell-shaped temperature
dependence and was considerably higher at 504 8C. The alloy
was slightly preferentially attacked 308 from the deposit crest 3.2.8 The nickel-based alloy Alloy 65 (61Ni21Cr9Fe 8Mo)
at both 504 8C and 528 8C, c.f. Fig. 12. The corroded surface
was also rougher in these sectors as compared to under the The attack of Alloy 65 was different from that on other al-
crest, but no actual pits were observed. loys in two ways. First, the corrosion rate decreased with the
temperature. At the lowest temperature the attack of Alloy 65
was substantially faster than that of the other high-alloyed ma-
terials in the test, Alloy 28 and Alloy 825, at their lowest tem-
peratures. At higher temperatures these three alloys performed
rather equal. Secondly, the most severe corrosion occurred
right under the deposit crest and not at the edge of it as
seen in Fig. 14. Furthermore, at the lowest temperature
some pitting was observed at the edge of the severely corroded
region, towards the side of the deposits, c.f. Fig. 15. Also un-
der the deposit crest the corroded surface was rough as from
pitting. Even though no distinct pits were observed on the high
temperature tubes, visual inspections indicated that similar
corrosion mechanisms took place at all temperatures.

4 Discussion

The high metal temperatures and the location of the coil

accelerated the corrosion rates in this field test. The corrosion
attacks therefore were very severe. On the most corroded tube
of the ferritic steel T22 the remaining goods in the deepest pit
was only 1.6 mm thick. Moreover, also some test tubes of the
high-alloyed materials had almost equally high corrosion
rates. Similar rates have been reported previously in waste
combustion under similar conditions [2]. The corrosion of
the different alloys will be compared and discussed below,
Fig. 14. The average thickness reduction around tubes of Alloy 65. but first some factors that might have caused this severe cor-
The highest deposit crests were located at the top rosion will be considered.
456 Andersson and Norell Materials and Corrosion 2005, 56, No. 7

First of all it should be remembered that the test was in- with increasing CrþNiþMo content from left to right. It is
tended to examine the corrosion at high temperatures for fu- clear that a major step in corrosion resistance prevails between
ture applications. Therefore, the outgoing steam temperature the austenitic steel Alloy 310 and Alloy 28. These two alloys
from the test coil was more than 200 8C higher than the steam are both austenitic with about equal Cr-contents but Alloy 28
produced during ordinary operation in this boiler. It is also contains about 10 wt-% more Ni (30.4 vs. 21.1) and 3.3 wt-%
known that the corrosion rate in waste combustion decreases Mo versus 0.1% for Alloy 310. The highest corrosion rates
with decreasing flue gas temperature [21, 22]. The test coil occurred on the ferritic and martensitic steels. Nevertheless,
was installed after the upper draw. In later boiler designs considering their higher alloying content the severe pitting of
the superheaters are placed further away from the cyclones Alloy 310 and the corrosion of Alloy 65 at the lowest tempera-
where the flue gas temperature is lower. Unfortunately, no ture were less expected.
thermocouple was available to measure the flue gas tempera- The corrosion rates generally increased with the tempera-
ture close to the test coil so the parameter is not known for this ture though in some cases the corrosion decreased with in-
test. Concerning the boiler, it should also be kept in mind that creasing temperature. For the ferritic steel T22 there was a
it usually burns a fuel with 60% RDF and not 100% as it was maximum in the corrosion rate at the two middle tempera-
during the test. The increased metal temperature and the fuel tures. The thickness reductions also differed. At the highest
composition were intentional choices to create a severely cor- and lowest temperatures the most severe attack occurred at
rosive environment for the test. the edge of the deposit crests, but at the middle temperatures
The severe corrosion in waste combustion is often attribu- the deepest sectors were found further away from the crests.
ted to the presence of Zn and Pb in the deposits [4 – 8]. These This suggests that a different corrosion mechanism has oc-
may contribute to the formation of eutectic chloride melts curred at the middle temperatures. When the deposit thick-
with low melting points. However, it is unlikely that such a ness, c.f. Fig. 3, is compared to the corrosion rates it is
mechanism played an important role in this test. Only minor seen that thinner deposits are formed on T22 at the middle
amounts of these elements were encountered in the deposits. temperatures were the highest corrosion rates occurred.
The analyses of the deposits and scales will be further treated This is an interesting observation since it is obvious that
in part two of this paper. the corrosion rates relate to the deposit thickness for all alloys.
The flue gas flow was not perpendicular to the test coil, A closer comparison of the ferritic steel T22 and the mar-
which resulted in the formation of two deposit crests on tensitic steel X20 shows that the corrosion rates were higher
many tubes. It can also be expected that the flow rate and con- for the martensitic steel at the higher and lower temperatures,
sequently the flue gas impact were more severe on one side of despite its higher Cr-content. However, also laboratory experi-
the tube. It has been shown that the corrosion rate increases ments simulating Cl-containing combustion gases have shown
with the flue gas flow rate [14]. Thus, the flue gas flow pattern that X20 can corrode faster than T22 at 450 8C [25]. Other
can explain that the preferential attacks were mainly observed experiments have shown that nickel free iron-chromium al-
on one side of the deposit crest. loys do not form protective Cr2O3 in waste combustion atmo-
Most alloys were preferentially attacked at the edges of the spheres but mixed oxides corresponding to the alloy composi-
deposit crest. The most obvious example was the austenitic tion [26].
steel Alloy 310 with pits restricted to a narrow band where In the literature it is often stated that nickel-based alloys
severe corrosion occurred. The reason for this is not known, with about 60 wt-% Ni and 8 – 18 wt-% Mo are the most sui-
but it is clear that a very specific combination of conditions table in waste boilers if the environment is severely corrosive
have triggered the mechanism. There were only two excep- [17, 27]. Although Alloy 65 has a composition within this ran-
tions. First, at the two middle temperatures the severely cor- ge, it did not perform well at the lowest temperature, 474 8C,
roded tubes of the ferritic steel T22 were preferentially at- where it actually had the highest corrosion rate of all alloys.
tacked further away from the deposit crest at
908. Secondly, However, it has been observed that the addition of nickel,
Alloy 65 was preferentially attacked right under the deposit chromium, and molybdenum can increase the corrosion resist-
crest. The preferential attack at the edge of the deposit has ance significantly in waste boilers at 550 8C but that alloying
previously been observed on tubes of Alloy 28 that had content may be less important at 450 8C [15]. Generally, over
been exposed for 5400 h at 340 8C in a waste boiler [4]. Similar the examined temperature range the corrosion resistance in-
observations have been reported for Alloy 310 [14]. Deposit creased with the alloying content. However, the corrosion be-
induced corrosion is a complex corrosion mechanism but it haviour of Alloy 65 was different from that of the other alloys.
involves several reactions between alkali chlorides, sulphates, Its corrosion rate decreased with the temperature and the
H2O, O2, Cl2, HCl, metal chlorides and oxides in different thickness reduction was highest directly under the deposit
layers of the scale and deposits [23]. The suggested mechan- crest. This indicates that the most severe corrosion of this al-
isms rely on the existence of these layers, different activities loy occurs under other conditions than for the other alloys.
for the reactants within them and the transport between them. It Deposit induced corrosion, where the deposit is providing
is therefore not surprising if rapid corrosion occurs where these chlorine gas to the metal surface, is still a possible mechanism.
layers exist but still are thin. Laboratory exposures with cou- However, directly under the deposit crest the oxygen activity
pons of Alloy 310 and Alloy 28 embedded in KCl-tablets have can be lower than under thinner deposits. Thus, protective oxi-
shown that the most severe attack can occur at a critical deposit des will form less easily.
thickness that was 1 – 2 mm in that study [24]. The high-alloyed austenitic steel Alloy 28 had a slightly
Comparing the corrosion rates, c.f. Figs. 4 and 5, it is seen decreased corrosion rate at the highest temperature. Such de-
that the alloys with the best overall corrosion resistance were creasing corrosion rates for high-alloyed austenitic steels and
the high-alloyed austenitic steel Alloy 28 and the nickel-based nickel-based alloys in waste incinerator systems have been
Alloy 825. At the lowest temperature the ferritic steel T22 ac- reported previously, though that was for the Alloys 310 and
tually performed quite well whereas the nickel-based Alloy 65 Alloy 825 [2]. The effect was attributed to the formation of
was about equally resistant as Alloy 28 and Alloy 825 at me- thicker and more protective oxides at higher temperatures.
dium and high temperatures. In Fig. 4 the alloys are ordered
Materials and Corrosion 2005, 56, No. 7 Superheater corrosion in a CFB waste boiler 457

The pitting attack observed on Alloy 310 was obstructed at l The deposit amount was alloy dependent. The low-alloyed
the inner tube where it spread out under the outer alloy. This steels formed thick deposits at all temperatures while this
implies that the composite structure has protected the tubes thickness increased with the temperature on the high-al-
from being penetrated by pitting corrosion. loyed austenitic steels and the nickel-based alloy.
The size of the test coil allows for a thorough evaluation of l The alloys, except Alloy 65, were preferentially attacked at
the corrosion of the test tubes. Nevertheless, due to the size of the edge of the deposit crest.
the coil it needs to be considered whether the corrosive envir- l Alloy 28 and Alloy 825 had the best overall corrosion re-
onment differed between the test tube positions. In practice, it sistance over the temperature range of all alloys in the test.
is barely impossible to answer that question. Instead, the dis- The attack on both alloys was uniform and no pitting was
tribution of the corrosion rates across the test coil can be as- observed.
sessed in relation to some striking results. The nickel-based l The corrosion rate on the ferritic steel T22 was bell-shaped
alloy Alloy 65 corroded more than any other alloy at the low- with the highest rate at 516 8C. At 460 8C was the corrosion
est temperatures. It could be argued that this was related to e.g. rate of the same order as for the more alloyed steels and
variations in the flue gas flow pattern. The tube was positioned lower than for the nickel-based alloy Alloy 65. The fastest
next to the quarter of the coil where the deposit formation was attack took place at
908 angle from the deposit crest.
limited. Nevertheless, the other alloys tested in the same area l The austenitic steel Alloy 310 was severely attacked by pit-
were less attacked. Another observation concerns the bell- ting in a narrow band along the edge of the deposits.
shaped corrosion of the steel T22. The two most corroded l The corrosion on the nickel-based alloy Alloy 65 differed
tubes of this alloy were also positioned in the vicinity of significantly from the other alloy. The attack was preferen-
the quarter with limited deposit formation. Yet in the same tial under the deposit crest. Furthermore, the corrosion rate
manner, positioned next to the most corroded tube, at decreased with increasing temperature.
516 8C, was the tube of the nickel-based Alloy 825 that
was the least corroded at this temperature. A final example
is that the corrosion rates of the austenitic steel Alloy 310
and the nickel-based Alloy 65 showed opposite temperature 6 Acknowledgements
dependencies. Still, these tubes were consistently positioned
next to each other so the result can hardly be related to differ- The Swedish High-Temperature Corrosion Centre at Chal-
ing corrosive conditions across the test coil. The possible in- mers University of Technology is acknowledged for financial
fluence of the flue gas flow pattern on the corrosion rate can support. Kvaerner Power and AB Sandvik Steel are acknowl-
never be entirely ruled out. However, it is concluded that the edged for a fruitful cooperation.
test tube temperature was the most important factor affecting
the corrosion rate of the test tubes.
In this part one of the paper the corrosion rates for waste 7 References
boiler candidate materials have been examined. The metal
loss have been characterised and related to the metal tempera- [1] K. Salmenoja, K. Mäkelä, CORROSION/2000, NACE Inter-
ture, the alloy composition and the deposit crest. In part two of national, Houston 2000, paper no. 238.
the paper a detailed study of the reaction products will be pre- [2] H. H. Krause, High Temperature Corrosion in Energy Sys-
sented. The focus will be on the metal oxides that were formed tems, The Metallurgical Society/AIME, Pennsylvania,
and the inner part of the deposits that were interacting with the USA, 1985, 83 – 102.
alloy and the oxide during the field test. The aim will be to [3] V. Zamansky, P. Maly, M. Klosky, Technical Report 755978,
study the corrosion mechanisms more in detail and identify Energy and Environmental Research Corporation, Irvine, CA
factors that determined the performance of the alloys. Four (US), 1998.
[4] A. Wilson, U. Forsberg, J. Noble, CORROSION/97, NACE
of the alloys that were tested in the field test have been se- International, Houston 1997, paper no. 153.
lected for this study, based primarily on their performance [5] K. Nakagawa, T. Isozaki, CORROSION/94, NACE Internatio-
in the field test. The ferritic steel T22 was partly chosen as nal, Houston 1994, paper no. 177.
a reference because it is a common alloy in many boilers. [6] K. Matsumoto, Y. Matsunaga, K. Nakagawa, Corr. Eng. 1997,
The austenitic steels Alloy 310 and Alloy 28 were chosen 46, 27.
for their difference in corrosion behaviour where Alloy 310 [7] Y. Kawahara, M. Kira, Corrosion 1997, 53, 241.
was severely attacked by pitting corrosion, unlike Alloy 28. [8] P. L. Daniel, L. D. Paul, J. Barna, Materials Performance
The nickel-based Alloy 65 was chosen since the corrosion at- 1988, 27, 22.
tack was significantly different from the other alloys. [9] H. H. Krause, I. G. Wright, V. K. Sethi, Heat Resistant Ma-
terials (Eds. K. Natesan, D. J. Tillack), ASM International,
Materials Park 1991, 623 – 631.
[10] O. Forsén, S. Yläsaari, Werkstoffe und Korrosion 1989, 40,
147.
5 Conclusions [11] K. Yukawa, Corrosion Engineering 1997, 46, 3.
[12] Y. Kawahara, Materials at High Temperatures 1997, 14, 191.
l The corrosion rates were accelerated due to increased metal [13] M. Staubli, K.-h. Mayer, W. Giselbrecht, J. Stief A. DiGian-
temperature and the location of the tube coil near the cy- francesco, T.-U. Kern, Materials for Advanced Power Engi-
clone. The highest corrosion rate on T22 was 34 mm/ neering, Part II (Eds. J. Lecomte-Beckers et.al.) Forschung-
year. Also some test tubes of high-alloyed materials had al- zentrum Jülich 2002, 1065 – 1080.
[14] M. Noguchi, H. Yakuwa, M. Miyasaka, M. Yokono, A. Mat-
most equally high corrosion rates. sumoto, K. Miyoshi, K. Kosaka, Y. Fukuda, Materials and
l The flow pattern of the gas highly influenced the deposit Corrosion 2000, 51, 774.
build up and the corrosion attack. However, local variations [15] Y. Kawahara, H. Hagiwara, M. Nakamura, E. Shibuya, K. Yu-
along the test coil seemed to be of less importance for the kawa, CORROSION/95, NACE International, Houston 1995,
corrosion than the metal temperature. paper no. 564.
458 Andersson and Norell Materials and Corrosion 2005, 56, No. 7

[16] Y. Kawahara, M. Nakamura, H. Tsuboi, K. Yukawa, Corrosion [22] A. J. B. Cutler, C. J. Grant, J. W. Laxton, D. D. Price, C. G.
1998, 54, 576. Stevens, Corrosion Resistant Materials for Coal Conversion
[17] Y. Kawahara, K. Takahashi, Y. Nakagawa, T. Hosoda, T. Mi- Systems, Applied Science Publishers, Barking, England, 1983,
zuko, CORROSION/2000, NACE International, Houston 159 – 177.
2000, paper no. 265. [23] Y. Kawahara, Corrosion Science 2002, 44, 223.
[18] E. Häggblom, L. Nylöf, Materials for Advanced Power Engi- [24] P. Andersson, M. Norell, R. Gautheron, Materials Science
neering, Part II (Eds. D. Coutsouradis et al), Kluwer Aca- Forum 2004, 461 – 464, 1015 – 1022.
demic Publishers, Dordrecht 1994, 1597 – 1606. [25] V. A. C. Haanappel, N. W. J. Haanappel, T. Fransen, H. D. van
[19] M. Norell, P. Andersson, CORROSION/2000, NACE Interna- Corbach, P. J. Gellings, Corrosion 1992, 48, 812.
tional, Houston 2000, paper no. 236. [26] A. Zahs, M. Spiegel, H. J. Grabke, Corr. Sci. 2000, 42, 1093.
[20] B.-M. Steenari, O. Lindqvist, B.-Å. Andersson, 15th Interna- [27] G. Sorell, Materials at high temperatures 1997, 14, 207.
tional Conference on Fluidized Bed Combustion, American
Society of Mechanical Engineers, New York, USA, 1999, 103. (Received: December 13, 2004) W 3852
[21] H. H. Krause, CORROSION/93, NACE International, Hou-
ston 1993, paper no. 200.