C

MI

a

ARRA

1

tmtwp

ts(mst

ooothdcswm

0d

Nuclear Engineering and Design 246 (2012) 123– 127

Contents lists available at ScienceDirect

Nuclear Engineering and Design

j ourna l ho me page: www.elsev ier .com/ locate /nucengdes

rack analysis in thermite welding of cathodic protection

arjan Suban ∗, Simon Bozic, Andrej Zajec, Robert Cvelbar, Borut Bundaranstitute of Metal Constructions, Mencingerjeva 7, SI-1000 Ljubljana, Slovenia

r t i c l e i n f o

rticle history:eceived 7 April 2011eceived in revised form 19 July 2011

a b s t r a c t

For protection of steel pipes that are exposed to corrosion, cathodic protection is commonly used. Jointbetween steel surface and copper conductor is done by thermite welding. During the welding process

ccepted 2 August 2011and due to the nature of it, steel surface in the solid state comes in contact with liquid copper. Contactof steel with liquid metal (copper, zinc) in some cases cause phenomenon known as the liquid–metalembrittlement or LME. Phenomenon was previously studied in cases such as soldering or hot-dip galva-nizing but for thermite welding no records were found in accessible literature. The purpose of this paperis to draw attention to some irregularities and consequences arising from it, which in this type of weldingcan occur. At the end of article some measures to reduce these risks are given.

. Introduction

To prevent corrosion of the material several methods by whichhe material may be protected are used in practice. One of these

ethods is the cathodic protection, which is commonly used to pro-ect steel, water or fuel pipelines and storage tanks, steel pier piles,ater-based vessels including yachts and powerboats, offshore oillatforms and onshore oil well casings.

There are two types of cathodic protection: passive protec-ion with active (sacrificial) anode and protection with an externalource of direct electrical current. Cathodic protection with activesacrificial) anode is more simple and requires the use of an anode

aterial (Mg or Zn anode), which protects the cathode (in our caseteel). One of the methods to connect cathode with the anode ishermite welding of the connecting copper wires to the cathode.

In order to determine the potential impact of thermite weldingf Cu wires on the steel, we performed a series of tests of this typef welding. Test data collected during a study of welding proceduren nuclear power plant construction projects (pipeline of SW sys-em) are presented and evaluated. These data lead to an importantypothesis that solely visual inspection is not sufficient because itoes not enable identification of some crucial procedural deficien-ies. Therefore, inspection has to be extended by implementation ofome other methods like macro- and micro-scopic examination ofeld cross-section that enable detection and identification of someaterial defects, which can lead to failure.

∗ Corresponding author. Tel.: +386 1 2802 120; fax: +386 1 2802 151.E-mail address: marjan.suban@imk.si (M. Suban).

029-5493/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.nucengdes.2011.08.001

© 2011 Elsevier B.V. All rights reserved.

2. Thermite welding

Thermite welding, also known as exothermic welding, of Cuto steel is a welding process for joining these two materials thatemploys superheated copper to permanently join the weldingparts. The process takes advantage of an exothermic reaction ofa copper thermite composition to heat the copper, and requiresno external source of heat or current. The chemical reaction thatproduces the heat is a reaction between aluminium powder anda mixture of copper oxides: copper(II) oxide and copper(I) oxide.Reaction is described with the following chemical formula:

3CuO + 2Al → 3Cu + Al2O3 + Heat (1)

This chemical reaction reaches a temperature above 1400 ◦C.This type of welding is especially useful for joining dissimilar met-als, e.g. Cu and steel for creating electric joints, like in our casefor cathodic protection and can be found on market under variouscommercial names such as Cadweld or Techweld. Thermite weld-ing process itself is shown in Fig. 1a–d and includes the followingsteps:

• grinding, cleaning and drying base material surface and installa-tion of Cu wire (Fig. 1a);

• installation of the welding mould and filling it with the thermitemixture (Fig. 1b);

• ignition and combustion of the thermite mixture (welding)(Fig. 1c);

• removing the welding mould and sealants (Fig. 1d).

If the welding process is optimal then formed weld has goodmechanical strength and excellent corrosion resistance. It is alsohighly stable when subjected to repeated short-circuit pulses, and

124 M. Suban et al. / Nuclear Engineering and Design 246 (2012) 123– 127

mite w

do

osciomoofmbp

3

tSewp6

hp

••

•

4

4

o

••••

ibipm

Fig. 1. Steps of ther

oes not suffer from increased electrical resistance over the lifetimef the installation.

However, the cost of the process is relative high comparing tother welding processes, requires supply of replaceable moulds,uffers from lack of repeatability, and can be impeded by wetonditions or bad weather when performed outdoors. In our exper-mental work presented in continuation we observed that qualityf the weld is influenced by the surface preparation and thaticrostructural changes and microcracks in steel appear. Because

f the good electrical conductivity and high stability in the casef short-circuit pulses, this type of welds are one of the optionsor grounding conductors and bonding jumpers. It is the preferred

ethod of bonding, and indeed it is the only acceptable means ofonding copper to steel in the case of cathodic protection of theipeline.

. Experimental method

The welding setup shown in Fig. 1 was used for the laboratoryests. Copper conductor NYY 1 mm × 16 mm was welded on 24 in.,CH 40 steel pipe made of ASTM A106 Gr. B (P255G1TH). Cadweldxothermic system was used for welding. Tests were performedith new and multiple-times used (old) moulds. Preheating tem-erature of steel pipe was set to: room temperature (15 ◦C), 40 ◦C,0 ◦C and 80 ◦C.

While some engineering specifications require just visual or per-aps even X-ray examination of completed exothermic weld, weerformed the following examinations in addition:

macroscopic examinations of the welds;microscopic investigation of joints (a review of materialmicrostructure, the depth of penetration of copper into steel, theoccurrence of cracks);measurements of micro-hardness at the respective places of theweld cross-section (microhardness tester is calibrated to a 500 gstandard).

. Results

.1. Macroscopic examinations

In the macroscopic investigations the following characteristicsf welded joints were observed:

lack of joint with steel base;shape of the weld;porosity of the weld;size of the heat affected zone (HAZ) in the base metal.

Fig. 2a (side view) shows the appearance of the weld withnsufficient area of completed joint between copper and steel

ase material which lead to weld breakdown. Cause for that is

n improper welding surface preparation. Uneven weld shape andorosity of the weld can be attributed due to the over-used weldingould (Fig. 2a; cross-sectional view). From the weld macro-section

elding procedure.

it is easy to measure size of HAZ in steel base material as shown onFig. 2a for bad weld and on Fig. 2c (cross-sectional view) for weldof good quality.

4.2. Microscopic examinations

In the microscopic examination the cross sections of all samplesprocessed under different conditions were examined with an opti-cal microscope. From microscopic examination of joints followingremarks can be made.

• The copper and steel create intermediate layer, which containsthe penetrated copper. The thickness of this layer is from 10 to20 �m.

• In some places, just below the steel surface, martensitemicrostructure occurs due to rapid cooling (see Figs. 3 and 4).

• Just below the surface of the base material an increased grain sizecan be found (see Fig. 4b).

• In the martensite microstructure microcracks were observed.Their length was up to 0.12 mm. The microcracks were filled withcopper. In the samples without martensite microstructure micro-cracks were not detected. Fig. 4 shows the cross-section opticalmicrographs of joints with presence of microcracks filled withcopper.

When examining the microstructure of the base material, wenoticed that just below the welds cracking occurs in which theliquid copper was penetrated (Magnabosco et al., 2006; Li andLin, 2001). This phenomenon (liquid–metal embrittlement, LME)occurs due to contact of liquid metal (Cu) with a solid metal (steel),as copper and steel have different melting temperatures (differ-ence is approximately 450 ◦C). At certain location the formationof hard martensite microstructure appears, which also representthe start for the cracks. Steels are severely embrittled by liquidcopper, and extremely high rates of crack growth up to 100 mm/scan occur during LME so that cracks are essentially instantaneous(Lynch, 2003; Clegg, 2001). In this case, fracture is facilitated byadsorption-induced weakening of interatomic bonds at crack tips,with transport of liquid metal (copper) to crack tips occurringrapidly by capillary flow. Diffusion of atoms along grain bound-aries ahead of cracks is not involved, although this can occur insome circumstances (specific couples of materials). Cases of LMEare well documented in literature for steel soldering using high-temperature copper alloys (Godec and Grdun, 2001; Semenov,1999, 2001) or hot dip-galvanizing for steel–liquid zinc interac-tion (Feldmann et al., 2010), but articles describing LME in thermitewelding cannot be found in the available literature.

For comparison between liquid copper induced and liquid zincinduced cracks, Fig. 5 shows cracking of steel under the corrosionattack by liquid zinc from the zinc bath. The crack is filled with com-ponents of the zinc melt and it is intercrystalline (Feldmann et al.,

2010). Authors of the report did not found formation of martensitemicrostructures in steel.

In our case, it must also be pointed out that the martensitemicrostructure, that is present in the surface layer of steel, is

M. Suban et al. / Nuclear Engineering and Design 246 (2012) 123– 127 125

Fig. 2. Some examples of defects revealed by macroscopic examination.

Fig. 3. Measurements of thickness of Cu–Fe fusion zone (intermediate layer).

ocrac

sihc

F

Fig. 4. Two examples of micr

ubjected also to diffusion of hydrogen. Literature indicates thatn the case of cathodic protection there are cases where atomicydrogen can be absorbed in the martensite layer, causing soalled hydrogen embrittlement (Hörnlund et al., 2007).

ig. 5. Surface crack in steel–liquid zinc interaction (2000×) (Feldmann et al., 2010).

ks in steel filled with copper.

4.3. Measurements of micro-hardness

Micro-hardness measurements just below the weld edgeshowed that the values in the martensite can be as high as367.15 HV (Fig. 6 left-side). Such result was obtained in the caseof over-used welding mould and at a temperature of steel pipeapproximately 15 ◦C. If a new mould was used at a temperatureof 15 ◦C hardness was still high (350 HV) but lower as in a case ofold mould.

When we use the new welding mould and the base materialis only slightly preheated (dried) to a temperature of approxi-mately 40 ◦C, the values of the hardness of martensite immediatelydecrease to values from 210 to 250 HV. If the preheating tempera-ture is increased to 60 ◦C, the hardness value just below the weldedge is safely decreased (below 200 HV). Preheating temperaturesat 80 ◦C did not lead to lower results in micro-hardness measure-

ments.

Fig. 7 shows that by using of new mould and at minimum pre-heating, we can achieve reduction in hardness just below the weldedge on steel side. Reduced local hardness is result of absence

126 M. Suban et al. / Nuclear Engineering and Design 246 (2012) 123– 127

Fig. 6. Measurements of micro-hardness at weld edge for various preheating temperatures.

100

150

200

250

300

350

400

Mic

ro-h

ardn

ess [

HV

]

Temp. 15 °C

Preheat. temp. 40 °C

Preheat. temp. 60 °C

of

5

tPwc

F

Base mat.HAZWeld edge

Fig. 7. Micro-hardness results in weld cross-section.

f hard martensite microstructures. Also microcracking cannot beound in these cases.

. Reduction of material strength due to microcracks

Calculation of reduction of static strength of steel pipe due to

he presence of filled crack can be done using equations derived byanasyuk et al. (2005). Relation between static strength of materialith empty crack pc and material with filled crack pfc for a size of

rack shown in Fig. 8 is:

ig. 8. Schematic representation of tension in cracked plate (Panasyuk et al., 2005).

Fig. 9. Crack length versus cycle behaviour of the specimens filled with nickel (EN5)and copper (EC3) at 0.95 Kmax crack prop-opening load (Song et al., 2001).

pc

pfc= 1 − ˛

1 + 2˛ˇ= � (2)

where the coefficients ̨ and ̌ are calculated using Eq. (3):

˛ = Efc

Eand ̌ = l

c(0 < ̨ ≤ 1; 1 < ̌ < ∞) (3)

The coefficient � varies from 0 to 1. The case � = 0 corresponds tothe strength of the material with completely “healed” crack (filledwith material of equal strength) and � = 1 corresponds to the casewhere the crack is not “healed”. In our case for size of crack presentsin Fig. 4b, where we use in calculations Young modulus for steel (E)and copper (Efc), the coefficient � is 0.01. This means that just alittle reduction of static strength can be expected.

In this research crack propagation due to fatigue was notinvestigated, but some previous research articles can be foundin literature. Song et al. (2001) investigated in article mechanicalproperties and in-crack distribution of filled crack. Results of thisresearch show in Fig. 9, assume that cracks in steel filled with othermetal (Cu or Ni) lead to retardation of crack propagation.

6. Conclusions

Because of presence of liquid–solid metal contact in weldingarea larger crystal grains, martensite microstructure and conse-quently, the microcracks appear. Those microcracks observed in

our research work are filled with liquid copper.

Formation of hard and brittle martensite microstructure justbelow the surface can be observed, because of rapid cooling rateof weld, which may be initials for the propagation of cracks. To

ering a

rtBcmihp

trrtelC

R

C

Semenov, V.N., 2001. Nature of initiation and propagation of cracks in precipitation-

M. Suban et al. / Nuclear Engine

educe this unfavourable microstructure, it is necessary to reducehe rate of cooling, which can be achieved by preheating the steel.y preheating to a temperature of at least 40 ◦C, which is espe-ially important at low surrounding temperatures, reduction ofartensite microstructure can be achieved, hardness at weld edge

s lowered and initiation of microcracks is suppressed. This pre-eating also reduces moisture, which may appear on the surface ofipe and causes some other welding defects.

In conclusion, some literature review was made to check howhese microcracks influence the static and fatigue strength of mate-ial. Calculation made for static strength showed that only minoreduction of strength can be expected. In the case of dynamic load,he crack propagation of filled crack is slower that in the case ofmpty crack. However, filled crack can also lead to material col-apse. Some further experimental investigation an review of ASMEode Cases in this field still need to be done.

eferences

legg, R.E., 2001. A fluid flow based model to predict liquid metal induced embrit-tlement crack propagation rates. Eng. Fract. Mech. 16, 1777–1790.

nd Design 246 (2012) 123– 127 127

Feldmann, M., Pinger, T., Schäfer, D., Pope, R., Smith, W., Sedlacek, G., 2010. Hot-dipZinc Coating of Prefabricated Structural Steel Components, JRC Scientific andtechnical Reports, EUR 24286 EN-2010.

Godec, B., Grdun, V., 2001. Krhkost nizkoogljicnega jekla zaradi stika s tekoco kovino.Mater. Tehnol. (3/4), 181–186.

Hörnlund, E., Fossen, J.K.T., Hauger, S., Haugen, C., Havn, T., Hemmingsen, T., 2007.Hydrogen diffusivities and concentrations in 520 M carbon steel under cathodicprotection in 0.5 M NaCl and the effect of added sulphite, dithionite, thiosul-phate, and sulphide. Int. J. Electrochem. Sci. 2, 82–92.

Li, J.H., Lin, R.Y., 2001. Joint zone evolution in infrared bonded steels with copperfiller. Metall. Mater. Trans. B 6, 1177–1183.

Lynch, S.P., 2003. Failures of engineering components due to environmentallyassisted cracking. J. Fail. Anal. Prev. 5, 33–42.

Magnabosco, I., Ferro, P., Bonollo, F., Arnberg, L., 2006. An investigation of fusion zonemicrostructures in electron beam welding of copper-stainless steel. Mater. Sci.Eng. A 424, 163–173.

Panasyuk, V.V., Sylovanyuk, V.P., Marukha, V.I., 2005. Strength of cracked structuralelements healed by injection methods. Mater. Sci. 6, 777–783.

Semenov, V.N., 1999. Effect of copper-silver solder melt on the properties of high-strength and high-temperature alloys and steels. Met. Sci. Heat Treat. 10,426–433.

hardening alloy soldered under the action of copper–silver molten pool. Met. Sci.Heat Treat. (11/12), 473–475.

Song, P.S., Sheu, B.C., Chou, H.H., 2001. Deposition of plating metals to improve crackgrowth life. Int. J. Fatigue 3, 259–270.