Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Preferential weld corrosion: Effects of weldment microstructure and composition (April 2005)

   

C-M Lee, S Bond and P Woollin

Paper presented at NACE 2005 Houston, Texas, 3-7 April 2005

Abstract

Preferential weldment corrosion (PWC) of carbon and low alloy steels used for pipelines and process piping systems in CO 2 -containing media has been observed increasingly in recent years. In particular, this has been on weldments made by the manual metal arc (MMA) process using electrodes containing Ni or Ni plus Cu. This paper presents the results of a joint industry research programme which was conducted collaboratively by three research organisations to investigate this corrosion mechanism and to seek practical solutions.

The effect of composition and microstructure on PWC in CO 2 -containing media was investigated on 12 weldments produced in X52 and X65 grade pipe materials using TIG and MMA processes. Corrosion tests were conducted in a re-circulating vessel on segmented weld electrodes in CO 2 -containing media, with two levels of chloride content. The addition of increased amounts of nickel and silicon was detrimental, whilst additions of molybdenum and chromium (of up to 0.7wt%) did not give improvements in PWC behaviour. Autogenous weldments, made without filler additions, and weldments made with matching composition consumables gave the best PWC resistance. It is also shown that empirical relationships exist between PWC and hardness levels and microstructure, with unrefined microstructures, having high hardness, being detrimental. The implications of the data for design of welding procedures to minimise PWC are considered.

Introduction

Preferential weldment corrosion (PWC) of carbon and low alloy steels used for pipelines and process piping systems in CO 2 -containing media has been observed increasingly in recent years. Attempts to control PWC have previously involved making minor additions of noble metals (e.g. Ni, Cr, Mo, Cu) to the weld consumable in order to make the weld metal cathodic with respect to the adjacent parent pipe and HAZ. This approach has proved successful in preventing preferential weld attack in seawater injection systems where additions of either Ni or Ni and Cu can be used to prevent preferential corrosion providing care is taken to avoid over-alloying, which can induce attack of the HAZ [1] .

The use of nickel-containing weld consumables has also been widely adopted in production systems. However, while some companies have satisfactory experience of using nickel-containing consumables, there have been examples of severe preferential attack of the weld metal in sweet environments [2] .

Studies have shown that PWC in CO 2 -containing media is influenced by a complex interaction of several parameters including the environment, flow conditions, scaling effects, parent steel composition and welding procedure [3-5] . No general agreement exists on the role of alloying elements and microstructure in preferential corrosion of welds [3] .

A recently completed joint industry project investigated the PWC of ferritic steels in CO 2 -containing environments. This paper reports the findings on the effects of weld microstructure and composition on PWC from the research programme. Two other papers, published in this conference, will focus on corrosion flow loop studies of PWC and its inhibition in CO 2 environments [6] and PWC of 1%Ni welds as a function of solution conductivity and inhibition [7] . Based on the findings of the project a set of guidelines for the prediction, control and monitoring of PWC of ferritic steels has recently been published [8] .

Experimental method

Welding

Three parent pipe materials were used on the project. These were:

  • API 5L X52/ASTM A333 grade 6, 356mm diameter, 25mm wall thickness.
  • API 5L X65, 508mm diameter, 10.3mm wall thickness.
  • API 5L X60 (with 0.5%Cr addition), 273mm diameter, 12mm wall thickness.

The chemical analyses of these pipes are given in Table 1. Welding was carried out by two processes, namely manual metal arc (MMA) and automated cold wire tungsten inert gas (TIG) welding to produce 12 different weld compositions and microstructures. Four elements, Ni, Cr, Mo and Si, were introduced into the weld metal in small amounts using commercially available consumables. The two welding processes were used to achieve low and high levels of dilution in welds containing Ni and Cr. Welds made with nominally matching consumables and autogenous welds were also included for comparison. Details of the welding consumables and significant alloying elements are given in Table 2.

Table 1 - Chemical analysis of the parent pipe materials

Pipe GradeElement, wt%
CSiMnPSCrMoNiAlCuNbV
5L X52 0.15 0.28 1.21 0.013 <0.002 0.088 0.012 0.057 0.039 0.092 0.034 0.037
5L X60 0.09 0.27 0.89 0.012 0.004 0.59 0.027 0.010 0.027 0.004 0.029 0.047
5L X65 0.07 0.20 1.17 0.019 0.005 0.033 0.003 0.019 0.031 0.021 0.051 0.004

As <0.004, except X52 for which As=0.024; Co ≤ 0.011; Pb, Zr <0.005; Sn ≤ 0.010; Ti ≤ 0.003, except X65, for which Ti = 0.042; W <0.01; Ce <0.02; Sb <0.002; B, Ca <0.0003.

Table 2 - weld details

Weld numberWeld type/(Target Dilution)*Parent steel gradeWelding processAWS consumable designation (closest)Significant alloying elements% Alloying element in weld metal
W41 Matching filler X52/A333-6 MMA E7018 None -
W21 Matching filler X52/A333-6 TIG ER70S-6 None -
W39 Matching filler X65 TIG ER70S-6 None -
W25 High Si (20%) X52/A333-6 TIG ER70S-6 1.05%Si 0.55(Si)
W20 High Ni (30%) X52/A333-6 MMA E8018-C3 1%Ni 0.58(Ni)
W26 Low Ni (50%) X52/A333-6 TIG ER80S-Nil 1%Ni 0.41(Ni)
W29 High Cr (20%) X60 (0.5%Cr) MMA E8010-G 0.6%Cr 0.67(Cr)
W31 Low Cr (50%) X52/A333-6 TIG ER80S-G 0.6-0.7%Cr 0.34(Cr)
W43 High Cr (20%) X52/A333-6 MMA E8010-G 0.6%Cr 0.57(Cr)
W33 Mo addition (20%) X52/A333-6 MMA E7015-A1 0.5%Mo 0.32(Mo)
W35 Autogenous X52/A333-6 TIG - - -
W37A Autogenous X65 TIG - - -
- Not applicable
* Matching filler implies matching composition

Corrosion Tests

Test Vessel. The design of the test vessel is shown in Figure 1; the cylindrical body of the main vessel was made from titanium. A paddle was driven from a motor mounted beneath the vessel to provide circulation. The test vessel was connected to a glass reservoir to provide additional solution volume and pumps were used to circulate the solution through the oxygen monitoring equipment.

spcmlapr2005f1.gif

Fig.1. Schematic diagram of test vessel

 

Sample Preparation. A section of each weld was removed and machined to 90mm x 15mm x 7mm. The segments were electrically isolated by cutting the fusion boundary and separating the HAZ from parent steel with a slitting wheel (<1mm thick) through the thickness of the sample. Electrical connections were made to the five segments at the back of the sample, which was mounted in resin together with an alloy 316 stainless steel counter electrode. The complete sample had a small hole drilled at the centre for inserting a salt bridge ( Figure 2).
spcmlapr2005f2.gif

Fig.2. Detail of corrosion test sample

 

Test Parameters. Details of test parameters are given in Table 3. Two chloride levels were employed: 35 and 0.35g/l NaCl, both at 60°C and acidified with 1 bara CO 2 .

Table 3 - Corrosion test details

 High ChlorideLow Chloride
Temperature 60°C 60°C
Paddle rotation 44rpm 44rpm
(Peripheral speed) (0.5 m/s) (0.5 m/s)
Chloride (g/l NaCl) 35 0.35
CO 2 1 bara 1 bara
O 2 Level <10ppb <10ppb
pH 5-6 5-6
Duration 30 days 30 days
Test Procedure. The test samples were placed in the test vessel with the exposed surfaces facing inwards. Prior to initiation of the tests the solution was pre-purged with N2 for 24hrs followed by CO 2 gas purge for a further 2hrs. The pre-purged test solution was pumped into the reservoir and also the test vessel, and the paddle was started. The peripheral speed of the paddle was set to 0.5 m/s. The CO 2 gas purge at 1 bara pressure and solution circulation between the test vessel and reservoir was maintained and the test allowed to proceed for 30 days at a constant temperature of 60°C.

The oxygen level of the test solution was monitored using a dissolved oxygen meter and maintained <10ppb. The pH and Fe 2+ ion concentration were also monitored on samples taken periodically from the test solution. The Fe 2+ were measured using atomic adsorption spectrometry. On completion of the test after 30 days, the specimens were removed, and rinsed with distilled water and ethanol.

The electrochemical measurements were carried out using a computer controlled potentiostat with a weld test facility on each of the five segments (i.e. weld metal, 2 x HAZ, 2 x parent). Linear Polarisation Resistance (LPR)measurements was used for calculating corrosion rates in the high conductivity solution and Electrochemical Impedance Spectroscopy (EIS) was used in the low conductivity solution. Details of the corrosion rate calculation methods are given in one of the associated papers [7] .

Results and discussion

Weld Characterisation

Compositionally, the 12 welds produced can be divided into 6 main categories; welds containing Ni, Cr, Si and Mo alloying additions, welds made with nominally matching filler and autogenous welds. Table 4 shows results of hardness and grain size measurements of the 12 welds tested in the programme.

Table 4 - characterisation of weld metal microstructures

Weld No.Parent steelProcessWeld typeMeasured grain size (µm)Mean weld root hardness (HV5)Microstructure
RA
W41
W21
W39
X52
X52
X65
MMA
TIG
TIG
Matching
Matching
Matching
5.6
11.7
6.3
185
229
195
3
1
3
0
3
0
W20
W26
X52
X52
MMA
TIG
High Ni
Low Ni
6.8
4.6
195
218
3
3
0
1
W29
W43
W31
X60
X52
X52
MMA
MMA
TIG
0.5%Cr
High Cr
Low Cr
3.5
5.7
4.7
230
235
221
0
3
3
1
1
1
W25 X52 TIG Si 9.6 224 1 3
W33 X52 MMA Mo 5.5 220 2 1
W35
W37A
X52
X65
TIG
TIG
Autogenous
Autogenous
7.5
7.9
209
186
2
3
1
0
R = Degree of refinement to an equiaxed ferrite and pearlite microstructure.
A = Amount of ferrite with aligned second phase.

Table 5 - comparison of weld metal corrosion rates

Weld No.Parent steelWelding processWeld typeCalculated Average 5-10 days+ Weld Metal Corrosion Rate (mm/yr)Calculated Peak Weld Metal Corrosion Rate (mm/yr)
High Chloride Solution (**)Low Chloride Solution (*)High Chloride Solution (**)Low Chloride Solution (*)
W41
W21
W39
X52
X52
X65
MMA
TIG
TIG
Matching
Matching
Matching
5.4
3.4
2.5
6.2
6.8
2.4
7.2
8.5
6.8
7.9
7.8
4.8
W20
W26
X52
X52
MMA
TIG
High Ni
Low Ni
3.5
0.6
14.0
11.4
7.9
5.9
18.2
17.0
W29
W43
W31
X60
X52
X52
MMA
MMA
TIG
0.5%Cr
High Cr
Low Cr
3.2
0.5
1.9
6.2
6.5
5.2
7.0
8.7
5.5
6.2
8.3
7.6
W25 X52 TIG Si 5.2 10.1 7.0 12.6
W33 X52 MMA Mo 3.8 6.0 6.9 9.1
W35
W37A
X52
X65
TIG
TIG
Autogenous
Autogenous
4.8
4.8
6.1
2.7
8.0
6.6
6.9
5.1
+ 5-10 days was chosen, as it was the period of highest sustained corrosion rate prior to surface film formation
*   Calculated from EIS data
** Calculated from LPR data
The deposition of welds with the two welding processes and a range of welding consumables produced weld root beads with a range of microstructures. Most of the welds had root microstructures that had been largely refined to equiaxedferrite and pearlite. However two, W21 and W25, consisted predominately of ferrite with aligned 2 nd phase. The exception to the pattern was weld W29 which had a microstructure consisting principally of grain boundary ferrite and acicular ferrite.

The microstructures were each assigned ratings in two categories; firstly the degree of refinement of the microstructure towards one of equiaxed ferrite and pearlite, labelled R, and secondly the extent of ferrite with alignedsecond phase, labelled A. A value of 0 to 3 was assigned in each category, to represent the visually assessed extent of refinement and aligned second phase present. R=0 showed no refinement to equiaxed ferrite and pearlite, R=3represented a fully refined structure of equiaxed ferrite and pearlite. A=0 indicated little or no ferrite with aligned second phase, whilst A=3 indicated the highest level of aligned second phase observed. Table 4 also shows the assigned ratings of each of the 12 weldments.

Corrosion Tests

Current Measurements vs Time. The galvanic current measurements from selected samples are shown in Figures 3 and 4. Note that the y-axis scales for the graphs are not the same for low and high chloride tests due to a large difference in the magnitude of the measured currents. In general, the magnitude of the current was the highest for welds containing alloying elements. When reduced corrosion rates arose from film formation, towards the end of the test, the galvanic currents followed a similar falling trend.

spcmlapr2005f3.gif

Fig.3. An example of current vs. time plot of weldment showing cathodic weld metal behaviour: W20, X52, High Ni, MMA, tests conducted in low chloride solution at 60°C, 1 bara CO 2

spcmlapr2005f4.gif

Fig.4. An example of current vs. time plot of a weldment showing anodic weld metal behaviour: W33, X52, with Mo addition, MMA, tests conducted in low chloride solution at 60°C, 1 bara CO 2

In Ni-containing weld samples (W20 and W26), the weld metal showed a consistent cathodic behaviour with respect to the other segments, in low and high chloride tests, Figure 3 . Two samples, X52 matching-composition MMA weld (W41) and X52 Mo-containing MMA weld (W33), showed weld metal that exhibits consistently anodic behaviour, as compared to the other segments, in low and high chloride tests, Figure 4. In the Cr-containing welds, only the X60 0.5%Cr pipe MMA weld (W29) showed a consistent cathodic behaviour. The other two welds with Cr additions showed both anodic and cathodic behaviour in the different tests. TheX65 autogenous weld (W37A) showed a consistent cathodic behaviour in all tests, but the X52 autogenous weld (W35) showed both anodic and cathodic behaviour in different tests.

Corrosion Rates vs Time. Figures 5 and 6 show calculated corrosion rates from selected samples from low chloride (Test 4) and high chloride (Test 3) tests. Typically, the weld metal shows preferential corrosion in the early part of the test and then falls to alevel nearer to that of the parent steel. After the initial peak, the falling corrosion rates are attributed to changes in the metal surface condition. In the period from 5 to 10 days, the metal surfaces are relatively free from asurface film, so they corrode relatively freely in the solution. After this period, as Fe 2+ ions build up in the solution, some segments of the weldment samples begin to form a protective scale on the surface and corrosion rates begin to drop dramatically. This effect is typically seen first in theweld metal. This is followed by the preferential corrosion of the HAZ which shows the next highest corrosion rates and these segments subsequently form a protective scale and also typically show reduced corrosion rate towards the endof the test. This result explains the occurrence of preferential HAZ attack rather than preferential weld metal attack, as observed in some field failures.

spcmlapr2005f5.gif

Fig.5. An example of calculated corrosion rate vs time results for tests conducted in low chloride solution at 60°C, 1 bara CO 2 (W20, X52, High Ni, MMA)

spcmlapr2005f6.gif

Fig.6. An example of calculated corrosion rate vs time results for tests conducted in high chloride solution at 60°C, 1 bara CO 2 (W21, X52, matching composition, TIG)

 

Not all samples and segments showed a reduction in corrosion rate during the test. This indicates that a build-up of protective scale did not occur in these segments. This is confirmed by visual examination of samples after the test, where certain segments were either partially covered with the protective film or the film formed was loosely adherent to the surface and so unable to provide complete protection.

Effects of Chloride Concentration. Figure 7 shows the average calculated EIS corrosion rates between 5-10 days from the low chloride test and Figure 8 shows the peak LPR corrosion rates from the high chloride test. These values were chosen to be representative of the freely corroding period before film formation and after the initial transient.

spcmlapr2005f7.gif

Fig.7. Average EIS calculated corrosion rate during the period 5-10 days; tests conducted in low chloride solution at 60°C, 1 bara CO 2

spcmlapr2005f8.gif

Fig.8. Maximum calculated LPR corrosion rate; tests conducted in high chloride solution at 60°C, 1 bara CO 2

With the exception of the Ni and Si-containing weld metal segments in the low chloride test, the corrosion rates of all parts of the weldment are generally greater in the high chloride test. However, the difference between the corrosion rates of the weld metal and parent steel, i.e. the degree of preferential corrosion, is greater for the low chloride solution test. For all weldments, it was noted that preferential weld metal/HAZ corrosion was lower in the high chloride environment compared to the low chloride.

Effects of Composition. Figures 7 and 8 show that the weld metals made with matching composition consumables or without consumable addition gave lower corrosion rates and better resistance to preferential weld metal attack than those made with alloyed consumables, with Ni being particularly detrimental. The data indicate that silicon has a detrimental effect on preferential weld metal corrosion also but this effect was less than that for nickel. It is noted that in another study on parent material, silicon was shown to have a beneficial effect on CO 2 corrosion resistance at temperatures greater than 93°C 9 . Molybdenum addition at the level examined had little or no effect on preferential weld metal corrosion although other workers have found some beneficial effects [2,10] . The data for chromium-containing weld metals were inconclusive. However the data did not rule out the possibility that there might be some beneficial effect of higher levels of Cr, as would be expected intuitively. Overall the best performance with regard to reducing preferential weld metal corrosion was obtained with matching composition or autogenous weld metals. However, preferential weld metal corrosion was not avoided for any of the weld metals examined. Figure 9 shows an example of a 0.6wt%Ni-containing weld metal that was particularly prone to PWC, where compositional effects dominate over microstructure.

spcmlapr2005f9.jpg

Fig.9. Weld metal with 0.6wt%Ni, a fairly fine ferrite grain size and little aligned second phase. Weld was deposited in X52 pipe by the MMA process, using E8018-C3 coated electrodes, W20. The composition promotes PWC but the microstructure does not

 

Nevertheless chromium is known to be beneficial, in the parent material, if present in sufficient levels [10,11] and one potential solution to PWC would be to examine higher chromium levels in the weld metal, with a view to reducing preferential weld metal corrosion rates further. It should be recognised however that this might tend to increase preferential HAZ attack if the corrosion resistance of the weld metal is increased substantially.

Influence of Microstructure. The root weld metal microstructure classification is given in Table 4. It was found that there was a reasonable correlation between microstructure and preferential weld metal corrosion for weld metals with no significant deliberate alloying (i.e. welds made with matching compositionconsumables and autogenous welds). Weld metals having largely unrefined ferrite with aligned second phase root microstructures (low R, high A), tended to give preferential weld metal corrosion. Refined weld metals with less alignedsecond phase (high R, low A) tended to have similar corrosion rates to the parent steel, except where a compositional effect, in particular nickel addition, had an overriding detrimental influence, in particular in the low chloridesolution.

Further correlations were developed, relating the peak corrosion rate of the weld metal in each sample to the root weld metal hardness and grain size. Figures 10-13 show plots of these correlations. The plots show a modest trend of increasing corrosion rate with increasing hardness and increasing grain size for weld metal without significant alloying additions, consistent with the visual microstructure assessments. Once again, the addition of Ni, and to a lesser extent Si, tends to give higher corrosion rate for a given hardness level or grain size in the low chloride tests. The effects of Cr and Mowere small but not consistently beneficial or detrimental. Figure 14 shows an example of a weld metal microstructure that promotes PWC where there is no over-riding composition effect.

spcmlapr2005f10.gif

Fig.10. Variation of peak weld metal corrosion rate with mean Vickers hardness, low chloride solution test

spcmlapr2005f11.gif

Fig.11. Variation of peak weld metal corrosion rate with mean Vickers hardness, high chloride solution test

spcmlapr2005f12.gif

Fig.12. Variation of peak weld metal corrosion rate with grain size, low chloride solution test

spcmlapr2005f13.gif

Fig.13. Variation of peak weld metal corrosion rate with grain size, high chloride solution test

spcmlapr2005f14.jpg

Fig.14. Unalloyed weld metal with a fairly coarse prior austenite grain size and a high level of aligned second phase. Weld was deposited in X52 pipe with the TIG process and an ER70S-6 wire, W21. The microstructure promotes PWC but the composition does not

In some cases preferential HAZ corrosion increased with increasing HAZ hardness, consistent with the weld metal data. This suggests that pipe with low carbon equivalent therefore might be preferred. However, the trend was not observed consistently. This probably reflects interaction with the weld metal, i.e. preferential attack of weld metal may tend to discourage preferential HAZ attack and vice versa.

Conclusions

A strong link between preferential weld metal corrosion and composition was found, particularly in a low chloride solution, although preferential weld metal corrosion cannot be prevented solely by the use of an alternative weld metal composition selected from those studied. Greatest resistance to preferential weld metal corrosion was obtained for autogenous root deposits or for welds deposited using consumables without significant alloying addition. Addition of 1% nickel was detrimental, as was 1% silicon. Addition of 0.5% molybdenum or 0.6-0.7% chromium to the weld metal had no consistent beneficial effect with respect to preferential weld metal corrosion.

Preferential weld metal corrosion increased with increasing hardness, increasing grain size, an increasing level of aligned second phase and a decreasing level of microstructure refinement of the root by the subsequent passes.

Preferential HAZ corrosion occurred during test when film formation occurred on weld metal preferentially. Some evidence suggested that high HAZ hardness might encourage preferential HAZ corrosion but this was not conclusive.

Welds made by TIG and MMA processes were compared but this was primarily to achieve different weld metal compositions rather than to allow a direct comparison of the welding processes. No distinct difference between the performance of welds made by the two processes, with respect to preferential weld corrosion, was found.

Acknowledgement

This work was carried out in the Joint Industry Project 'Risk of Preferential Weldment Corrosion of Ferritic Steels in CO 2 -Containing Environments', conducted 2000-2003 by TWI, CAPCIS, and Institute for Energy Technology (IFE). The project was sponsored by BP, ENI SpA, Health & Safety Executive (UK), Petrobras, Saudi Aramco, Shell UK Ltd, and Total Fina Elf.

References

  1. 'Preferential Corrosion of Carbon Steel Weldments in Oil and Gas Production Systems', Joint Industry Project, CAPCIS/TWI, 1989.
  2. S. Olsen, B. Sundfær, J. Enerhaug, 'Weld corrosion in C-steel pipelines in CO 2 environments - Comparison between field and laboratory data', CORROSION/97, Paper no. 43, NACE International (Houston, TX 1997).
  3. D.A. López, T. Pérez, S.N. Simison, 'The influence of microstructure and chemical composition of carbon and low alloy steels in CO 2 corrosion. A state-of-the-art appraisal', Material and Design, 24 (2003).
  4. M.W. Joosten, J. Kolts, P.G. Humble, 'Aspects of selective weld and HAZ attack in CO 2 containing production environments', CORROSION/96, Paper no. 79, NACE International (Houston, TX 1996).
  5. F.A. Martins, A.P.C. Rodrigues, L.Y. Reznik, D.S. de Freitas, L. Bott, 'Temperature effect on API Grade steels in CO 2 environments', Eurocorr 2004, Nice, 2004.
  6. E. Gulbrandsen, A. Dugstad, 'Corrosion loop studies of preferential weld corrosion and its inhibition in CO 2 environments', CORROSION/2005, paper no. 05276, (Houston, TX: NACE International, 2005).
  7. S. Turgoose, J.W. Palmer, G.E. Dicken, 'Preferential Weld Corrosion of 1% Ni Welds: Effects of Solution Conductivity and Corrosion Inhibitors", CORROSION/2005, paper no. 05275, (Houston, TX: NACE International, 2005).
  8. D. Queen, C-M. Lee, J. Palmer, E. Gulbrandsen, 'Guidelines for the Prediction, Control and Monitoring of Preferential Weld Corrosion of Ferritic Steels in Wet Hydrocarbon Production Systems Containing CO 2 ', 1st Int. Symposium on Oilfield Corrosion, Paper no. SPE 87552, Aberdeen (Society of Petroleum Engineers, Richardson, TX 2004).
  9. D.W. Stegmann, R.H. Hausler, C.I. Cruz, H. Sutanto, 'Laboratory studies on flow induced localized corrosion in CO 2 /H 2 S environments, I. Development of test methodology', CORROSION/90, paper no. 5, (Houston, TX: NACE International, 1990).
  10. M. Kimura, Y. Saito, Y Nakano, 'Effects of alloying elements on corrosion resistance of high strength linepipe steels in wet CO 2 environment', CORROSION/94, paper no. 18, (Houston, TX: NACE International, 1994).
  11. H. Takabe, M. Ueda, 'The formation behavior of corrosion protective films of low Cr bearing steels in CO 2 environments', CORROSION/2001 paper no. 01066, (Houston, TX: NACE International, 2001).

For more information please email:


contactus@twi.co.uk