Professional Documents
Culture Documents
Sensitivity Analysis of Half-Bead and Alternative GTAW Techniques
Sensitivity Analysis of Half-Bead and Alternative GTAW Techniques
Sensitivity Analysis of Half-Bead and Alternative GTAW Techniques
BY P. J. ALBERRY
ABSTRACT. The ASME XI shielded metal tungsten arc welding (GTAW) procedure ture of the primary HAZ. Subsequently,
arc half-bead technique and an alterna- (Ref. 2). The third phase, a computer thermal cycles may cause further changes,
tive, mechanized, six-layer gas tungsten model sensitivity analysis of the likely vari- which can involve transformation and
arc technique have been assessed using a ability of the ASME XI half-bead technique tempering, depending on the peak tem-
computer model sensitivity analysis. These and the alternative GTAW technique, is peratures experienced.
techniques have been developed for re- reported here. The precise microstructure and hard-
pair welding of heavy-section steel com- ness at any given point in the HAZ will be
ponents in light water reactors without Computer M o d e l determined by its thermal history, grain
postweld heat treatment. The as-welded growth kinetics, phase transformation and
heat-affected zones are compared in General Description tempering behavior. The thermal history
terms of structure, grain size and hard- Details of the computer model for the will depend upon the welding conditions,
ness. The likely sensitivity of each tech- prediction of weld heat-affected zone joint geometry, spatial positioning of the
nique to systematic and random variations (HAZ) structures and hardnesses pro- weld beads, the weld bead shape and the
in welding parameters is assessed. duced by the SMA and GTAW processes resultant overlap of their HAZ's. Grain
have been given previously (Refs. 2-4). growth kinetics are a function of the orig-
Introduction inal structure and time at temperature.
Weld HAZ's in a given material consist
The phase transformation and tempering
of a range of structures that reflect the
In 1979, EPRI initiated a research con- behavior are determined by chemical
range of thermal cycles experienced. Ther-
tract EPRI-RP 1236 entitled "Repair Weld- composition, the prior structure, the aus-
mal cycles with peak temperatures close
ing of Heavy Section Steel Nozzles" for tenite grain growth kinetics, the peak aus-
to the melting point are experienced near
Babcock and Wilcox, Alliance, Ohio. This tenitizing temperature and the heating
the weld metal fusion boundary and usu-
project was designed to define and vali- and cooling rates during the thermal
ally result in the formation of coarse-
date an alternative to the ASME XI half- cycle(s).
grained structures. At greater distances
bead repair procedure for light water re- Thus, to define the structure and, hence,
from the fusion boundary, the thermal
actor nozzles. Both repair procedures are properties obtained in the HAZ of a given
cycle peak temperatures decrease pro-
intended for use without postweld heat material composition, it is necessary to
gressively and more fine-grained struc-
treatment. The half-bead procedure uses describe: 1) weld bead shape, 2) the ther-
tures are formed. Eventually, the thermal
a layer-by-layer refinement and temper- mal cycle as a function of welding proce-
cycle peak temperature decreases below
ing technique and is carried out manually. dure in all regions of the HAZ, 3) prior
the Aci and the base material structure is
Half of the first layer of weld metal is austenite grain growth as a function of
only tempered. This basic unit for a single
ground off manually to facilitate the re- prior structure and weld thermal cycle, 4)
weld bead is shown in Fig. 1.
finement and tempering action of subse- variations in transformation behavior as a
quent layers. The heavy dependence on In multipass welds, the HAZ formed by
function of thermal cycle and prior aus-
time-consuming manual activities was con- the deposition of a single weld bead is
tenite grain size, and 5) tempering effects
sidered to be a serious practical limitation modified by the superposition of succes-
as a function of thermal cycle experienced
in reactor environment due to the pro- sive adjacent weld bead deposits. The
for a given transformation product.
longed exposure to radioactivity. Conse- second thermal cycle modifies the struc-
quently, with the use of a computer Elements 3-5 for SA508 Class 2 (Table
model, an alternative GTAW technique 1) were characterized (Ref. 1) in detail in
capable of mechanization was designed. Phase 1.
The computer model work was carried KEY WORDS The thermal history can be calculated
out in three phases. The first phase was to from standard heat flow equations, but
Sensitivity Analysis
establish the basic welding metallurgy for the position of the fusion boundary can-
Half-Bead GTAW
SA508 Class 2 heat-affected zones (Ref. Alternative GTAW not be calculated accurately. To over-
1). The second phase was to use the Six-Layer GTA Weld come this, the weld bead dimensions are
computer model to develop a viable gas Computer Model established for a given range of welding
Repair Welding conditions, and the measured position of
Heavy Section Steel the fusion boundary is used as a reference
Light Water Reactor point for the heat flow calculations. The
GTAW Sensitivity HAZ isotherms are thus calculated relative
P. J. ALBERRY is with Marchwood Engineering Weld HAZ Comparisons
Laboratories, Central Electricity Generating to the known position of the fusion bound-
Board, Marchwood, Southhampton, England. ary. This ensures that the heat flow calcu-
Half-Bead
Repair and
Alternative
Six-Layer
GTAW
Repair Half-Bead Half-Bead f 'AC:, ' A d
(Refs. 1, 5) Repair Repair
Element Base Plate (Ref. 8) (Ref. 9)
lations are carried out in that region of the SMA Mean Bead Width = 0.00640 ally true. However, the model does not
HAZ where conduction heat transfer dom- (GHI) + 3.72; r = 0.90 (2) require highly accurate deterministic algo-
inates. This provides a better estimate of rithms. The calculated average values of
Mean Bead Height = 0.00058 (GHI) +
the spatial position of the fusion boundary the mean bead dimensions (MBD) have
1.41; r = 0.54 (3)
and associated HAZ relative to the plate variability, added by a randomizing pro-
surface than would be achieved using Mean Penetration = 0.00134 (GHI) + cess by the model to produce the mea-
heat flow calculations alone. 0.44; r = 0.82 (4) sured scatter in the empirical database.
The detailed methodology and algo- GTA Mean Bead Width = 0.00770 Hence, the randomized values for bead
rithms (Ref. 2) used in the model have (GHI) + 3.54; r = 0.95 (5) dimensions (RBD) are obtained from an
been validated against a number of equation of the type
Mean Penetration = -0.00020 (GHI) +
shielded metal arc (SMA) and gas tungsten
arc (GTAW) procedures (Ref. 5). How- 1.48; r = 0.13 (6) RBD = MBD + (V60 (2fn/5)
ever, the more important features rele- -0.5)o M B D (8)
vant to the sensitivity analysis are summa- where bead dimensions are in mm and r
where n is a random number generated in
rized below. is the correlation coefficient. the range 0 to 1, and CMBD is the measured
Since the GTAW process has a contin-
standard deviation of the mean bead
Mean Weld Bead Dimensions and Weld uous wire feed during welding, it is as-
dimension. The randomizing term \/6CT
Bead Dimension Variability sumed that all the wire is melted and de-
CS.fr/5) — 0.5 produces random variate
posited on the plate surface with little
The computer model used requires al- values with a mean value of zero and a
spatter and that the top surface of the
gorithms for the description of weld bead standard deviation of 1, which have an
weld bead can be represented by the arc
shape, which has an idealized geometry as approximately Gaussian form (Refs. 6, 7).
of a circle. Hence, it can be shown that the
shown in Fig. 1. This has the disadvantage The values of weld bead dimension
weld bead height can be obtained numer-
that the model is not generally applicable ically from: variability for the SMAW and GTAW pro-
u
but is only valid over the range of condi- cesses used in the investigation are given
tions used to obtain the weld bead di- 7t(WD)2WFS in Table 2. Since the bead width and bead
mension data. 4v , 360 V 2
The computer model algorithms for the 4ha
mean weld bead dimensions are ex- (7)
4h2 + a2 ) ) - & - * Table 2—Weld Bead Dimension Variabilities
pressed as a function of gross heat input,
which is defined as: where W D is the wire diameter (mm), tor the SMAW and GTAW Processes Used
WFS is the wire feed speed (mm • s~1), a is in the Sensitivity Analysis
Gross Heat Input (GHI) = V i v " 1 (1) the weld bead width (mm) (Fig. 1), and h
where V is welding voltage, I is the weld- is the weld bead height (mm) —Fig. 1. SMAW GTAW
Standard Standard
ing current, and v is the travel speed, Equations 2-6 are empirical and have
Deviation Deviation
m m - s - 1 (2.36 in./min). no known physical basis. The low corre-
Weld Bead Dimension (mm) (mm)
Empirical straight-line relationships valid lation coefficient for Equation 6 reflects
over the range of existing data (Refs. 2, 5) the fact that the measured values of pen- Bead width 1.1 0.63
obtained from the measured weld bead etration are largely independent of gross Bead height 0.33 0.11<a>
Penetration 0.33 0.37
dimensions using a least-squares regres- heat input over the range of welding con-
sion analysis are: ditions used. This is not likely to be gener- (a) Typical value.
300
< 250
0 0.50 1.00 1.50 2.00 2.50 0 0.50 1.00 1.50 2.00 2.50
Distance from fusion boundary, mm Distance from fusion boundary, mm
Fig. 2-Measured and calculated HAZ hardness values for the Fig. 3—Measured and calculated hardness values for the
ASME XI half-bead technique alternative six-layer GTA W technique
height are inversely correlated, a random- of weld variability, it allows for the exam- given in Fig. 2 and shows reasonable
ized value of bead height is obtained from ination of possible weld-to-weld varia- agreement. However, the base plate mea-
a numerical solution of Equation 7 using tions, which could be important in some sured hardness values show a wider scat-
the appropriate randomized value of weld nuclear repair applications. This may help ter than predicted. This is not surprising
bead width together with the appropriate to define the level of experimental valida- since a significant amount of base plate
welding conditions. tion required to establish a given level of segregation was reported (Ref. 5), as
The computer model now has the confidence that any given weld is ade- shown in Table 1.
capability to examine the effects of weld quately represented by a given set of test The alternative six-layer GTAW proce-
bead variability on a given weld. Since samples whose properties are well estab- dure (Ref. 5) (Table 4) uses the first three
weld bead variability is the inherent source lished. added layers to produce refinement and
the last three layers to carry out the bulk
Model Validation of the tempering. Computer model hard-
Table 3—ASME XI Half-Bead Technique ness predictions across the HAZ are com-
The computer model described above pared with corresponding measurements
Welding Procedure (Ref. 5)
has the capability to provide detailed in Fig. 3. Generally, there is reasonable
hardness distributions across SA508 Class agreement between the measured and
Welding Conditions^1
2 HAZ's for the SMAW and mechanized calculated hardness values, and the pre-
Welding Layer Layers 2
GTAW processes from a knowledge of dicted shape of the hardness traverse
Variables 1 Onward
the appropriate weld bead dimensions, band fits the measured data points for
Electrode type E8015-C3 E8015-C3 welding process variables and the basic maximum HAZ hardnesses to within 10%.
Current, A 95 140 HAZ metallurgy. The comparison of actual
Voltage, V 25 25 The following sections describe the use
hardness measurements across the HAZ
Travel speed, 3.39 3.60 of the computer model to provide a sen-
for the SMA half-bead and six-layer mech-
mm • s _ 1 sitivity analysis of the variation in HAZ re-
anized GTA welds (Ref. 5) and the corre-
Electrode 2.4 3.2 finement and peak HAZ hardness with
sponding computer model predictions is
diameter, mm weld parameters.
Weld bead 50 50 now summarized. The full details of the
overlap, % calculations have been described previ-
Preheat, °C 177 177 ously (Ref. 2), but sufficient information is Sensitivity Analysis
Gross heat 701 972 given here to provide the basis for the
input, J • m m - 1 following sensitivity analysis. Note that ASME XI Half-Bead Technique
calculated values of hardness are ob- HAZ Structure
(a) Half of the first layer of weld metal is manually ground off
before the deposition of Layers 2 o n w a r d .
tained from computer model calculations
based on algorithms that relate to Vickers The continuous cooling transformation
diamond pyramid hardness (Hv). Hard- data for SA 508 Class 2 HAZ's are available
ness measurements (Ref. 5) are Knoop di- (Ref. 1) for a full range of possible HAZ
Table 4—Welding Conditions for Six-Layer grain sizes. The bulk of the HAZ produced
amond hardness, which are equivalent to
Alternative GTAW Procedure (Ref. 5) by the ASME XI half-bead technique is re-
Vickers diamond pyramid hardness. An
approximate conversion to the Rockwell fined with an HAZ grain size of about
Welding Layer Number 0.015 mm and any unrefined areas will
C scale can be obtained by dividing the
Parameters 1 2 3-6 have an HAZ grain size of about 0.1 mm.
Knoop diamond hardness by a factor of
Current, A 180 200 220 10. The HAZ structures are predominantly
Voltage, V 11 11 11 martensite together with a small amount
Travel 3.60 2.96 2.54 In the ASME XI half-bead technique, the of lower bainite, as shown in Fig. 4. The
Speed, bulk of the refining action is carried out by probability of local unrefined areas can be
mm s 1 the second layer, and the bulk of the tem- estimated using the computer model.
Wire Feed 16.5 25.0 27.5 pering is carried out by further layers. De-
Speed, tails of the ASME XI half-bead weld proce-
HAZ Refinement
mm s _ 1 dure (Ref. 5) are given in Table 3. A
Wire 0.89 0.89 0.89 detailed comparison of calculated and The refining action produced by layer-
Diameter,
measured (Refs. 5, 8, 9) hardness values by-layer deposition is illustrated in Fig. 5
mm
for the HAZ's corresponding to the base where it can be seen that the refining zone
Preheat, °C 150 150 150
plate compositions given in Table 1 is of the second layer HAZ continuously
* 0.2
5 1.0
— •— Maximum coarse grained HAZ depth x
"— Minimum coarse grained HAZ depth =s 0.8
-1.5 -0.5 0 0.5 15
-10 -5 0 5 10
Grinding error, mm
D e v i a t i o n in v o l t a g e , volts
Fig. 6 — Effect of grinding error on the probability of obtaining full HAZ
refinement using the ASME XI half-bead technique
1.0 L
0.8 '.
446-s | N O V E M B E R 1 9 8 9
given in Table 6. The size of each effect — • — Maximum coarse grained HAZ depth
given determines its significance, and the —x—Minimum coarse grained HAZ depth
sign of each effect determines whether a
given input variable combination has a Over penetration perfect Under penetra •ion
c
positive or negative effect on the mea- First & second half bead
E 1 0
sured response, which in this case is the layer beads merge .- *~s^~* ^ Pract cal limit due to
maximum hardness value for a given >N, bead pile up
traverse across the HAZ. Six input variable S 0.8
combinations have effects that are signif-
M
< \
1
icant at the 95% confidence level. These 0.6 -
are grinding error (— 8.5), second layer "3 / \
travel speed (— 8.6), second layer weld Z 0.4 . / \ Fig. 8-Effect of
o weld bead overlap
bead overlap (21.5), grinding error and £0.2 - / on the probability of
first layer travel speed (11.8), grinding er- \ obtaining full HAZ
rt)
ror and first layer weld bead overlap refinement using the
(— 8.5), and second layer travel speed and
o o • ASME XI half-bead
a 10 30 50 70 90
weld bead overlap (8.5). These effects are Weld bead overlap % technique
complex but a simple interpretation can
be obtained in terms of the effect of each
parameter on the layer height and the
HAZ Refinement Fig . 5, usin 3. t w o ayers
I as a simple illus
tion. The principle of refinement and
subsequent effect of the change in layer
The six-layer mechanized GTAW tech- tempering from successive layers is the
height on the superposition of the HAZ
nique was designed using the computer same for both the half-bead and the
isotherms. For example, increased grind-
ing will reduce the thickness of the first model (Ref. 2) and validated (Ref. 5) by six-layer techniques, the major difference
layer. This will cause the isotherms from all depositing carefully controlled, thin layers being the number of layers involved.
subsequent layers to overpenetrate, of weld metal to cause multiple refine- Hence, the HAZ refinement sensitivity
thereby disturbing the balance of refine- ment and tempering from successive lay- analysis was approached in the same way.
ment and tempering and resulting in an ers. The layer thicknesses were calculated Some additional assumptions were re-
increase in the maximum hardness. so as to progressively shift five bands of quired because of the increased complex-
refining and tempering isotherms up ity caused by the use of six layers. It was
Decreased grinding will increase the through the first layer HAZ, as indicated in assumed that the variability of the ma-
thickness of the first layer. This will cause
the isotherms from all subsequent layers
to underpenetrate, thereby disturbing the Table 6—Details of Factorial Experiment Welding Conditions, Calculated Maximum HAZ
balance of refinement and tempering, and Hardness and Factorial Analysis of the Effects Produced
again resulting in an increase in the maxi-
mum hardness. Hence, both overgrinding Factor
and undergrinding have a negative influ-
Travel Travel
ence in that they increase HAZ hardness.
Sped Overlap Grinding Speed Overlap
In simplistic terms, overgrinding promotes Weld Treatment Layer 1 Layer 1 Error Layer 2 Layer 2
"overhardening" and undergrinding pro- No. Combination mm • s - ' mm mm • s _ 1 Max Hv Effect
motes "undersoftening," both of which
% %
are non-optimum conditions that increase I (1) 3.0 40 -0.5 3.2 40 397 —
HAZ hardness. Similarly, a decrease in the 2 A 3.8 40 -0.5 3.2 40 388 6.4
second layer travel speed increases the 3 B 3.0 60 -0.5 3.2 40 404 -0.1
4 AB 3.8 60 -0.5 3.2 40 400 4.1
second layer thickness, causing the tem- 3.0 40 0.5 3.2 40 402 -8.5
5 C
pering isotherms from the third layer to 6 AC 3.8 40 0.5 3.2 40 406 11.8
underpenetrate. Also, an increase in the 7 BC 3.0 60 0.5 3.2 40 398 -8.5
second layer weld bead overlap increases 8 ABC 3.8 60 0.5 3.2 40 409 1.0
the second layer thickness, causing the 9 D 3.0 40 -0.5 40 40 393 -8.6
refining isotherms from the third layer to 10 AD 3.8 40 -0.5 4.0 40 382 4.6
underpenetrate. Each of these effects, 1 1 BD 3.0 60 -0.5 4.0 40 398 -0.6
overpenetration and underpenetration, 12 ABC 3.8 60 -0.5 4.0 40 399 4.9
results in an increased maximum hardness 13 CD 3.0 40 0.5 4.0 40 358 -6.8
in the HAZ. 14 ACD 3.8 40 0.5 4.0 40 383 3.3
lr> BCD 3.0 60 0.5 4.0 40 364 -0.8
Id ABCD 3.8 60 0.5 4.0 40 390 3.5
17 E 3.0 40 -0.5 3.2 60 419 21.5
Six-Layer Mechanized GTAW Technique 18 AE 3.8 40 -0.5 3.2 60 411 1.0
HAZ Structure 19 BE 3.0 60 -0.5 3.2 60 425 -6.8
20 ABE 3.8 60 -0.5 3.2 60 419 1.0
The whole of the HAZ produced by the 40
21 CE 3.0 0.5 3.2 60 404 -2.1
six-layer mechanized GTAW technique 22 ACE 3.8 40 0.5 3.2 60 427 0.6
has a consistent, highly refined HAZ with 23 BCE 3.0 60 0.5 3.2 60 400 -4.9
a maximum HAZ grain size of 0.015 mm. 24 ABCE 3.8 60 0.5 3.2 60 403 2.1
There is only an extremely low probabil- 25 DE 3.0 40 -0.5 4.0 60 419 8.5
ity of occurrence of local unrefined areas, 26 ADE 3.8 40 -0.5 4.0 60 413 -0.3
which is estimated below. The HAZ struc- 27 BDE 3.0 60 -0.5 4.0 60 422 -2.8
tures produced are the same as those 28 ABDE 3.8 60 -0.5 4.0 60 422 4.8
produced by the half-bead technique, i.e., 29 CDE 3.0 40 0.5 4.0 60 421 6.1
predominantly martensitic together with a 30 ACDE 3.8 40 0.5 4.0 60 421 -0.9
31 BCDE 3.0 60 0.5 4.0 60 368 -2.1
small amount of lower bainite, as shown
32 ABCDE 3.8 60 0.5 4.0 60 421 5.1
in Fig. 4.
Publication of this report was sponsored by the Subcommittee on Dynamic Analysis of Pressure
Components of the Pressure Vessel Research C o m m i t t e e of the Welding Research Council. The price of
WRC Bulletin 336 is $20.00 per copy, plus $5.00 for postage and handling. Orders should be sent w i t h
payment to the Welding Research Council, Suite 1 3 0 1 , 345 E. 4 7 t h St., New York, NY 10017.
This Bulletin consists of four Sections t h a t present a preliminary evaluation of the current knowledge
of the elevated t e m p e r a t u r e behavior of a bolted flanged connection.
Section I—Introduction and Overview, by J. H. Bickford; Section II—Historical Review of a Problem
Heat Exchanger, by J. R. Winter; Section III—Development of a Simple Finite Element Model of an
Elevated Temperature Bolted Flanged Joint, by K. Hayashi and A. T. Chang; and Section IV—Discussion
of the ABACUS Finite Element Analysis Results Relative to In-the-Field Observations and Classical
Analysis, by J. R. Winter.
Publication of this report was sponsored by the S u b c o m m i t t e e on Bolted Flanged Connections of the
Pressure Vessel Research C o m m i t t e e of the Welding Research C o m m i t t e e . The price of WRC Bulletin
3 4 1 is $20.00 per copy, plus $5.00 for postage and handling. Orders should be sent with payment to the
Welding Research Council, Suite 1 3 0 1 , 345 E. 4 7 t h St., New York, NY 10017.