1

Improving the FatigueResistance of ThermiteRailroad Rail Weldments

F. V. LawrenceY-R. ChenJ. P. Cyre

2

Outline

! Fatigue problems with thermite welds

! Improving the rail head

! Improving the rail web and base

3

Metallic Fatigue

A.M. Zarembski – Bulletin 673, 1979, Volume 80 of AREA proceedings

ACELA

4

Rolling contact fatigue

Thermite Weld

Rail

B

H

W

Railroad car wheelmoving over railcauses fatigue tooccur in both the railhead and base.

5

Fatigue crack initiation sites

Internal Fatigue Crack

Rail Head

Rail Web

Web-to-base Fillet

Fatigue Crack at Weld Toe in Fillet

Rail Base

Fatigue Crack at Weld Toe in Base

! ≈ 40% of all servicefailures are due tothermite field welds.

! ≈ 10% of allderailments are dueto broken fieldwelds.

6

Fatigue crack in rail head

Internal fatigue crack initiation in rail head

7

Fatigue crack in rail base

Cold Lap

Site of crackinitiation

Limit of fatiguecrack growth

8

Thermite weld service failures

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

Base Web-basefillet

Web Head-webfillet

Head ?

Location

Record of 244 service failures on a Class 1 railroadinvolving thermite field welds.

9

Service failures or “markouts”?

! Most field-weld service failuresoriginate at web or base.

! But defects detected and removedfrom the rail head before a servicefailure can occur (“markouts”)exceed service failures by 2:1!

10

Implications

! Fatigue cracks in the web and base areless frequent but are the principalcause of service failures since they aredifficult to detect. Crack initiationoccurs at external stress concentration.

! Fatigue cracks in head are morefrequent but are generally removed.Crack initiation occurs at internal stressconcentration.

11

Outline

! Fatigue problems with thermite welds

! Improving the rail head

! Improving the rail web and base

12

Porosity types

Shrinkage Gas (Spherical)

Thermite welds studied contain about 1.5% shrinkage porosity.

13

Porosity initiates fatigue

Formation of shells in tangent trackat interdendritic shrinkage porosity.

Running surface.

Odario 1992

14

Interdendritic shrinkage porosity

15

Possible solution?

! Eliminate weld metal! (?)

! Developed a modified thermitewelding process called “SqueezeWelding” in which ends of joint forcedtogether to expel most of the thermiteweld metal.

16

Squeeze welding

Force Force

Rail Cross-section

Expelled Imputities

FinalWeldThickness

Rail EndsMoved TogetherWhile WeldMetal StillMolten

17

Weld longitudinal-sections

Standard Squeezed

WM

HAZ

BM

Fry 1992

18

Laboratory test results

10

100

1000

104 105 106 107 108

Withee - Squeezed

Liu - Squeezed/Vibrated

Liu - SqueezedLiu - Vibrated

Liu - Standard

Withee - Vibrated

Withee - Standard

Max

imum

Str

ess,

S max

(MP

a)

Fatigue Life, Nf(cycles)

Fatigue behavior of small specimens taken from head ofweld shows some improvement.

Withee 1998

87 mm

19 mm

9.27 mm

R 208 mm19 mm

6.35 mm

19

But distribution unchanged!

0.0

0.2

0.4

0.6

0.8

1.0

100 101 102 103 104 105

Standard Weld

Squeezed Weld

Vibrated Weld

Cu

mu

lativ

eP

rob

ab

ility

Pore Size, area (µm2)

size range of poresinitiating failureimputed fromSEM images

Pore size distribution unchanged!Withee 1998

20

Largest pore size controls!

1.0

10.0

100.0

104 105 106 107

Standard Weld (C)Squeezed Weld (B)Vibrated Weld (D)Regression Analysis

Initi

alS

tres

sIn

tens

ityF

acto

r,K o

(MP

a*m

1/2)

Fatigue Life, Nf(cycles)

B1

C1C5

D5B4

B3

B2

13

Withee 1998

Single relation for all treatments depending only on pore size(and applied stress).

87 mm

19 mm

9.27 mm

R 208 mm19 mm

6.35 mm

21

Implications

! Reducing the size of the largestpores and/or the volume of weldmetal should increase in the(average) fatigue life.

! Largest pore per unit volume(porosity) and the volume of weldmetal jointly determine the fatiguestrength.

22

Theoretical study

p

Σ(t)

(t)

Fry 1995

23

Stress history experienced

Str

ess

MP

a)

24

Fatigue occurs at critical depth

Fatiguedamageparameter

Worst depth

Worstlocations onpore

Dep

thbe

low

runn

ing

surf

ace,

Y(m

m)

No residual stress Considering residual stress

Fry 1995

25

Effects of pore shape?

2

0

0.5

1

1.5

2

2.5

3

3.5

0 0.5 1 1.5 2 2.5

2

4

6

8

10

152025

3040

50

60

72

λY

/λZ

Ratio of pore's longitudinal and transverse axes,λX

/ λZ

FBY

PCV

FBX

SpherePCHFBZ

PCT

Detail Fracture

Vertical Split Head

Shelling

RAHELS Predictions

FavorsDetailfracture

Favors verticalsplit-head

Favorsshelling

Fry 1995

26

Model predictions

! Critical depth for fatigue crackinitiation (≈ 15mm) determined bywheel-contact-induced residualstresses.

! Model predicted that shelling, verticalsplit heads and detail fracture couldall initiate at shrinkage poresdepending upon the pore shape.

27

Central portion of weldmentmachined and ground flat to12.7 mm thickness. Stepped penetrameter.

specimen film

New measurement technique

Chen 2000

28

L1

Typical radiograph

!?!

Difference incontrast due tomicro-porosity(shrinkageporosity.

Porosity notuniformlydistributed!

Chen 2000

29

Radiographs of field welds

F1 F2 F3

30

0.52% 1.72%interface

Optical determination of porosity

31

25000

27000

29000

31000

33000

35000

37000

39000

0 500 1000 1500 2000 2500 3000 3500 4000

Distance

Measuredchanges ingrey scale inphotoshop.

Penetrameter with 0.11 mm steps indicateat least 1% sensitivity

F1

BM WM

0% 0.9% porosity

Radiographic image density

32

0

0.2

0.4

0.6

0.8

1

1.2

1.4

B-3 B-4 A-6 A-7 A-8 A-1 L-1 B-1 A-2 A-3

Po

rosi

ty(%

)

Porosity in 10 thermite welds

Average porosity in 10 “markouts” varies considerably!

Chen 2000

33

Developing detail fracture

Detail fracturein head ofrail appears tobe developingin associationwith an areawith a highconcentrationof shrinkageporosity?

34

Conclusions

! Large variation in porosity from weld toweld. Porosity not uniformlydistributed.

! Porosity clusters at weld centerlinefrequently seen. Fatigue cracks inhead often associated with associatedwith porosity clusters.

35

Why?

A-1 A-2 A-3 A-4 A-8

A-9 A-10 B-1 L-short L-long

100mm

Apparentlythere are largevariations inthermalconditionsduring thermitewelding.Observedvariations inmelt-back (weldprofile) onradiographs.

36

Outline

! Fatigue problems with thermite welds

! Improving the rail head

! Improving the rail web and base

37

Thermite weld service failures

0.0%

5.0%

10.0%

15.0%

20.0%

25.0%

30.0%

35.0%

Base Web-basefillet

Web Head-webfillet

Head ?

Location

Record of 244 service failures on a Class 1 railroadinvolving thermite welds.

38

Web-to-base fillet !?!

! Why does this happen ?????

! Answer:! Residual stresses!

! Weld toe geometry!! Flank angle.! Cold laps.

39

Webster et al.

Residual stresses

Neutral Axis

Critical locations:• Web-to base fillet• Rail base

Compression Tension

40

Weld toe flank angle

≈ 85̊ Flank Angle

41

Weld toe geometry

Toe Radius (r)

Roughness (R)

Flank Angle

Weld Metal Base Metal

(θ)

Fatigue Severity = 1+0.27 tanθ0.25 t

r

1+0.1054Su R( )−1

Improve by:

• Flank angle↓

•��������� ↑

•���������↓

42

Current Orgo-thermit mold profiles

A-A

B-B

C-C D-D

E-E

4530

A-A

AA

BB

CC

DD

MoldRail and weld

Measured profiles ofOrgo-thermit molds

43

Modified Orgo-thermit mold profiles

Suggestedmodifications toOrgothermitmolds

AA

BB

CC

DD

Modified Current

44

Nature of critical defects

0.0% 5.0% 10.0% 15.0% 20.0% 25.0% 30.0% 35.0%

Cold Lap

Slag

Hot Tear

Porosity

Lack of Fusion

Columnar Grains in Head

Grind Burn

Hot Pull-apart

Inclusion in Head

Sand burn in

?

Analysis of 244 service failures on a Class 1 railroadinvolving thermite welds.

45

Cold laps - Dimitrakis

Cold lap No cold lap

Cold laps greatly reduce the fatigue life of a weldment

46

Cold laps at thermite weld toe

Cold Lap

Base Metal

47

Weld toe cold laps

LoadingDirection

D

Weld Metal

Base MetalHeat Affected Zone

Weld Toe LocationWithout Cold-Lap Defect

Curved PathVertical Path

θ

r

φ

48

Effect of cold laps

Condition Percentage of Fatigue Life

Flank angle (θ) = 30Þ 100%

Flank angle (θ) = 45Þ 56%

Flank angle (θ) = 60Þ 44%

Cold lap depth (D) = 0 100%

Cold lap depth (D) = 1mm 20%

Cold lap depth (D) = 2mm 15%

49

Causes of cold laps

! Gap between mold and rail in thecritical web-to-base fillet area.

! Inadequate melt back causingincomplete fusion at the weld toe?

50

Variations in melt-back

A-1 A-2 A-3 A-4 A-8

A-9 A-10 B-1 L-short L-long

100mm

Melt backvariesconsiderablyin thelocation ofthe web-tobase fillet

51

Melt back dimensions

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

6

5

4

3

2

1

0

Hei

gh

to

fR

ail(

in.)

Length of Melt Back (in.)

Weld Sample #2

Weld Sample # 3

Weld Sample #1

52

Melt back at web-to-base fillet

0

10

20

30

40

50

26-30 31-35 36-40 41-45 46-50 51-55 56-60 61-65 66-70

Melt-back width (mm)

Collar width defined by mold

Bad! Good!

53

UIUC Experimental Program

UIUC modified 1”thermiteweld molds are usedwith a 1.4” rail gap.Mold sealed at weld toewith refractory paste.

And: Reduced flankangle!

Standard 1” thermiteweld: Large flank angleand cold laps.

Rail Rail

Weld MetalMelt back

54

UIUC Experimental Program

Luttingpaste fromRailtech.

Sealing pastefrom Railtechw/ BrazingFlux.

55

UIUC Experimental Program

Uni Ram Blurefractory paste.

Leecote moldwash and UniRam Blurefractory paste.

56

Weld Fabrication

1

3

4

65

2

57

Fatigue testing

Standard 4-point bending test.

58

Modified Weld Specimen #28

Crack initiationpoints

Limit of fatigue crack growth

59

Effect of modifications

Cold lapformationbeyondsealingpaste

60

Standard weldments

100

1,000

10,000

1,000 10,000 100,000 1,000,000 10,000,000

Cycles to Failure, N f

Process A

Process B

Process C

Process D

TAMU

1

3

61

100

1,000

10,000

1,000 10,000 100,000 1,000,000 10,000,000

Cycles to Failure, N f

Process A

Process B

Process C

Process D

Modified UIUC

TAMU

UIUC experimental welds

1

3

62

100

1,000

10,000

1,000 10,000 100,000 1,000,000 10,000,000

Cycles to Failure, N f

Process A

Process B

Process C

Process D

Modified UIUC

TAMU

UIUC experimental welds

1

3

63

Summary - Head failures

! Head failures caused by internaldefects notably porosity and highconcentration areas of porosity.

! Thermal conditions duringsolidification may cause oneweldment to be good and another tobe bad?

64

Summary - Web-base failures

! Web and base failures aggravated bysevere external geometry and coldlaps.

! Thermal conditions duringsolidification play a role in web-basefatigue problems?

! Fatigue life can be increased bymodifications of external weldgeometry.