142-s | APRIL 1999
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
ABSTRACT. An improved fume chamber
was constructed, and fume rates were
measured with unprecedented precision
for both steady- and pulsed-current
welding of mild steel using 92%
argon/8% CO
2
shielding gas. Compre-
hensive fume maps were constructed de-
picting fume rates over a wide range of
currents and voltages. Fume generation
was generally lower under pulsed-cur-
rent conditions. Theoretical arguments
explaining this difference are presented.
Introduction
Public agencies concerned with oc-
cupational safety and industrial hygiene
have recently pressed for more stringent
limits on metal-containing particles in
factory air. In many cases, these particles
originate as fume generated in welding
arcs. Those opposed to stricter limits
argue that such will require hundreds of
millions of dollars to be spent in capital,
maintenance and operation of ventila-
tion equipment while yielding negligible
gains in worker health. Proponents of
stiffer standards maintain that medical
fees, liability suits and lifestyle limita-
tions attributed to welding fume are
likely to cost much more.
Tougher proposed standards were re-
cently challenged in U.S. courts and re-
jected. Nevertheless, many observers ex-
pect tighter limits will eventually be
imposed. Meanwhile, fumes generated
in the welding of stainless steel and other
alloys have come under increased
scrutiny, and even tighter controls have
been proposed for them. Although im-
proved ventilation is the most common
way to clean shop air, other approaches
show promise. Some researchers claim,
for example, that fume can be reduced
60 to 90% by using power supplies that
deliver pulsed rather than steady current
(Refs. 1, 2).
Accurate fume-generation data and a
comprehensive fume formation model
are necessary for more sophisticated
fume control strategies. This paper in-
cludes precise fume generation data for
GMAW of mild steel using one shielding
gas under steady- and pulsed-current
conditions. A physical model introduced
by Gray, Hewitt and Dare (Ref. 4) is em-
ployed and amplified to explain our ob-
servations.
Although fume formation has been
studied by many scientists, results are dif-
ficult to reconcile from one researcher to
another. Limited accuracy of some results
is one problem, but interpretation and
correlation are complicated because of
the multitude of variables involved.
Many types of welding exist with and
without fluxes using a wide range of pos-
sible shielding gases. Numerous different
electrode and work materials or combi-
nations are possible. Much of the prior
work has been directed toward the solu-
tion of immediate problems in the work-
place. Often, fume generation studies in-
volve so many variables results are
almost impossible to use for theoretical
purposes. Our research was designed to
produce precise results for narrow con-
ditions. Although limited to GMAW of
mild steel with a single shielding gas, our
fume typography is typical of profiles one
should expect with other electrodes and
shielding gases. It is hoped that similar re-
sults for such systems will become avail-
able in the future.
Castner (Ref. 3) has provided what
may be the most comprehensive study of
a single system using the standard AWS
fume chamber. Although a comparison
of steady- and pulsed-current fume rates
was the main focus of his study, much
can be gained from examining the
steady-current data alone. Sets of data
obtained at fixed wire feed rate (essen-
tially constant current) and increasing
voltage were reported for a number of
different wire feed speeds. Selected re-
sults are illustrated in Fig. 1.
Curves in Fig. 1 were fitted to data
using a least-squares analysis. This is ap-
propriate if a function is smooth. On the
other hand, Gray, Hewitt and Dare (Ref.
4), who studied fume evolution in
GMAW of stainless steel, reported be-
havior that is discontinuous. Their mea-
sured rates, illustrated in Fig. 2, rise
through short-circuit to globular transfer,
then drop in the spray mode and rise
again in streaming transfer. Data from
other researchers and common experi-
ence tend to confirm this discontinuous
change in fume rate, dependent on
transfer mode.
An interesting fume profile emerges if
the cusped, rising/falling typography of
the Grey, Hewitt and Dare report is em-
ployed to fit Castner’s data. This is illus-
trated in Fig. 3 where two basic assump-
tions were applied. First was that of
continuity. That is, fume rates must vary
with current (from frame to frame of Figs.
1 and 3) in a continuous way. Second, the
effect of transfer mode as suggested by
Grey, Hewitt and Dare must be reflected.
Figure 3 represents our intuitive fit of the
data based on these assumptions. The
solid curves are ours. Lighter curves rep-
resent the original least-squares set from
Fig. 1.
The assumption that there must be a
continuous progression through Figs. 3A,
B, C, D and E requires a significant de-
parture from some of the data points, es-
pecially in C and D. Various problems
with the AWS standard fume chamber,
such as filter blanking and deposition of
particles on plate and chamber walls
(mentioned by Castner), could explain
such deviations. In fact, comparison of
two experiments with different power
sources but at conditions almost identi-
cal (Figs. 7 and 8 in Ref. 3) reveals dis-
crepancies as large as a factor of two at
some conditions. Fortunately, Castner’s
measurements cover a broad range of
Fume Formation Rates in
Gas Metal Arc Welding
BY B. J. QUIMBY AND G. D. ULRICH
A new fume chamber design improves the accuracy of fume generation data
KEY WORDS
Fumes
Gas Metal Arc Welding (GMAW)
Environment
Health
Fume Chamber
Industrial Hygiene
Mild Steel
A36
B.J. QUIMBY and G. D. ULRICH performed
this work at the Department of Chemical En-
gineering, University of New Hampshire,
Durham, N.H. Correspondence should be di-
rected to Dr. Ulrich, 34 Sheep Rd., Lee, NH
03824, or gdu@cisunix.unh.edu
currents and voltages, encouraging one
to interpolate and smooth the data. Fig-
ure 3F, a combination of the solid curves
from 3A through E, summarizes our spec-
ulated intuitive fit of these results.
Examination of Fig. 3F suggests that
fume generation rate might be presented
effectively by a three-dimensional plot.
In fact, an earlier publication by Willing-
ham and Hilton (Ref. 5) shows fume data
(Fig. 4) in a quasi-three-dimensional for-
mat. Figure 5 is our three-dimensional
model of data taken from Fig. 3F. Fume
rate is plotted on the vertical axis vs. a
horizontal plane defined by voltage and
current. Here, one sees the topography of
a rising “foothill” interrupted by a de-
pression or “valley” running parallel to
the current axis.
The valley in Fig. 5 represents spray
transfer conditions. The “ridge” at the left
corresponds to globular conditions; the
rising mountain at the right, to streaming
transfer. Fume rates rise with voltage as
one moves from short circuit (low volt-
age) to globular transfer (the ridge), then
drops into the valley during a shift toward
spray mode and finally rises again with
the onset of streaming (high voltage)
transfer. This figure illustrates the con-
nection between fume generation rate
and welding mode that some experi-
enced welders might claim is obvious.
The profile of Fig. 5 is consistent with
data from researchers who have plotted
fume rate in two dimensions — as a
function of either current or voltage, with
the other variable being held constant. It
agrees with results reported by Willing-
ham and Hilton for a current-voltage
path defined by the “feel” of an experi-
enced welder. Two-dimensional graphs
can be obtained by cutting the three-di-
mensional model (Fig. 5) with vertical
planes — one parallel to either the cur-
rent or voltage axis or one along the volt-
age-current path chosen by an experi-
enced welder.
If pulsing the current extends the
length and breadth of the valley, a dra-
matic reduction in fume could be ex-
plained as a shift from globular or steam-
ing (high fume) conditions at steady
current to spray (low fume) transfer under
pulsed conditions.
Experimental Procedure
This research was designed to docu-
ment the three-dimensional fume profile
with a precision and reproducibility un-
approached in past studies. Most stan-
dard welding fume chamber designs are
inflexible, erratic and imprecise,
plagued by variable purge-air rates, short
welding times, filter blanking, plate over-
heating and loss of fume through deposi-
tion on chamber
walls. A thorough re-
view of past fume
chambers and their
problems can be
found in Ref. 6.
Another considera-
tion, in this laboratory,
was our role in a New
England Welding Re-
search Consortium
project to study the
physiological effects
of inhaled fume. We
needed a device that
could not only yield
accurate, reliable
fume rates but could
also deliver fume-laden air to laboratory
animals for long periods at a steady state.
The fume chamber designed and con-
structed to address these constraints is il-
lustrated in Fig. 6. This cutaway sketch
shows a rotating pipe workpiece on
which a weld bead is deposited auto-
matically with an indexed gas metal arc
(GMA) welding gun. The pipe is enclosed
by a chamber maintained under slight
pressure. Compressed air is metered and
added at a rate that simulates air flow in
WELDING RESEARCH SUPPLEMENT | 143-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 1 — Selected fume formation rates for GMAW of carbon steel at steady current with 85%
argon/15% CO
2
shielding gas (Ref. 3).
Fig. 2 — Fume generation data reported by Gray, Hewitt and Dare
(Ref. 4).
a shop setting. Positive air-
flow eliminates changes in
the purge rate as fume
builds up on the filter. (In
designs dependent on vac-
uum draw, airflow
changes with time.)
The cylindrical work-
piece allows longer peri-
ods of steady-state opera-
tion, and it can be cooled
internally by either air or
water. This avoids an in-
crease of workpiece tem-
perature with time charac-
teristic of rotating disc
designs. Photographs of
our chamber are shown in
Figs. 7 and 8. To confirm
that results from this de-
vice are compatible with
accepted standards, tests were con-
ducted using 100% CO
2
shielding gas at
the two conditions specified in the Amer-
ican Welding Society standard fume test
(Ref. 7). Measured fume rates were well
within ± 10% of specified values as re-
quired.
Nine variables were held constant
and three were changed. Those parame-
ters held constant are listed below along
with justifications for holding them so.
More extensive explanations and a thor-
ough review of prior research can be
found in Ref. 6.
Constant Parameters
Workpiece and Electrode Material
Prior research confirms that virtually
all GMAW fume comes from the elec-
trode rather than the work unless the
workpiece is coated with oil, paint or
some other substance. Since carbon steel
(A36) is the most common metal welded
today, it was chosen for this study. Carbon
steel electrode welding wire ER70S-3 was
used. Each pipe was cleaned with a sand-
ing wheel on an electric grinder before
use as recommended by the AWS proce-
dure (Ref. 7).
Wire Type
Flux-cored welding wire generates
much more fume than solid wire because
of flux evolution, and each flux behaves
differently. To keep parameters manage-
able, only solid-core wire was used.
Wire Diameter
Electrode diameter has a modest ef-
fect on fume rate because of differences
in voltage, current and (possibly) welding
mode. A wire diameter of 1.2 mm (0.045
in.) is specified in the AWS standard. It is
also the median common standard in
most GMAW and was used in all experi-
ments here.
Polarity
Other researchers have found that po-
larity can affect fume rate, but since most
GMAW is done with a positively charged
electrode, such was used here.
Nozzle-to-Work Distance
Prior research has shown that nozzle-
to-work distance has a minor influence on
fume rate. The “typical” setting of 19 mm
(0.75 in.) was used for all our profiles ex-
cept one. This one (steady-current, 174
mm/s [410 in./min] wire feed rate) was
conducted with a nozzle-to-work dis-
144-s | APRIL 1999
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 3 — Intuitive fit of Castner data from Fig. 1. Continuity from frame to frame and curve shapes
consistent with Gray, Hewitt and Dare were assumed. Lighter curves are from Fig. 1. Dark curves
are ours. F is a composite of A through E.
Fig. 4 — Quasi-three-dimensional graph of fume rate vs. cur-
rent and voltage (Ref. 5).
tance of 12 mm (0.5 in.). Emission profiles
were similar for both separation distances
over the voltage operating range, but ab-
solute fume rates ranged about 20%
higher at 12 mm.
Electrode Angle
Others have found that variations in
electrode angle have only a slight effect on
fume rate (within reasonable operating
limits). A 10-deg drag angle was used here
as recommended by the AWS standard
procedure (Ref. 7).
Welding Speed
Changing the torch travel speed by a
factor of two reportedly increases fume
rate by about 5% (Refs. 8, 9). A value of
6 mm/s (14 in./min) as recommended by
AWS standard procedure (Ref. 7) was
used throughout this study.
Shield Gas Composition
It is widely known the type and com-
position of shielding gas profoundly af-
fects fume generation rate. The gas rec-
ommended for our power supply is 92%
argon/8% CO
2
, a common choice in in-
dustry. Therefore, this shielding gas mix-
ture was chosen for all experiments (ex-
cept calibration) conducted in this
research.
Shield Gas Flow Rate
Others have found shielding gas flow
rate to affect fume rate. Presumably,
shielding gas rates must be high enough to
protect the weld zone from oxygen in the
air but low enough to minimize turbulent
mixing. A value of
16.5 L/min (35 ft
3
/h),
in the midrange of
values recommended
by the manufacturer
of our welding equip-
ment, was used here.
Variable Parameters
In GMAW, three
variables — voltage,
current and wire feed
speed — are interde-
pendent. In this re-
search, voltage and
wire feed speed were
dictated by the opera-
tor while current,
which is approxi-
mately constant at a
given wire feed
speed, was controlled
by the power supply.
Wire Feed Speed
Five different wire feed rates were
chosen: 76, 102, 127, 148 and 174 mm/s
(180, 240, 300, 350 and 410 in./min),
similar to values selected by Castner (Ref.
3). These wire feed speeds encompass
normal welding modes and represent
common practice in GMAW of mild
steel.
Voltage
Voltages ranged from 18 to 34 V in
these experiments. (Not all voltages
could be employed at all wire feed
speeds. Excessive voltage at low wire rate
melts through the workpiece. Low volt-
age and high wire feed rate creates a vis-
ibly unsatisfactory weld.)
Steady/Pulsed Current
Fume rates were measured under
both steady- and pulsed-current condi-
tions. At steady-current conditions, wire
feed speed and voltage were set by the
operator and current was controlled by
the power supply. In pulsed-current ex-
periments, wire feed speed, pulse width
and frequency were set by the operator.
The power supply controlled voltage and
average current at steady levels during a
weld. These parameters were read and
recorded by the operator.
Sampling Procedure
Samples were collected on filters of
WELDING RESEARCH SUPPLEMENT | 145-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 5 — Three-dimensional visual model based on an intuitive
fit of Castner’s steady-current fume data (Ref. 3).
Fig. 6 — Cutaway sketch of the University of New Hampshire (UNH) fume
chamber.
Fig. 7 — Photograph of the UNH fume chamber. Slide gate
mounted at the top of the chamber is for filters and sampling.
Torch and self-darkening lens are mounted at the one o’clock po-
sition.
various types mounted at the chamber
exhaust port. (Fume quantity was the
major concern in this study. In a related
project, particle size distributions were
measured using an Electrical Aerosol An-
alyzer. Results of that work will be re-
ported elsewhere.)
To create a data map, bead-on-pipe
welding was conducted for 20 s at a fixed
condition. Then, the chamber was purged
by sweep-air for an additional 2 min to en-
sure complete collection of fume. The fil-
ter was removed and weighed. A new,
preweighed, predried filter was installed,
and the process was repeated at a differ-
ent experimental condition.
Numerous experiments were con-
ducted with welding durations of 30 and
60 s. Measured fume generation rates
were independent of welding time. To
prolong workpiece life, a duration of 20 s
was used for most experiments.
Results
The equipment described above was
used to define two separate fume profiles.
Both apply to GMAW of mild steel with
92% argon/8% CO
2
shielding gas. One
is for steady-current welding, the other for
pulsed current. Type ER70-S mild steel
electrode welding wire of 1.2-mm diam-
eter was used to weld a continuous bead
on a 10-in. Schedule 40 mild steel (A36)
pipe.
Equipment was calibrated using the
AWS standard procedure with mild steel
and 100% CO
2
shielding gas. Calibration
fume rates were in excellent agreement
with AWS standard values. Repro-
ducibility was checked several times dur-
ing the collection of approximately 150
data points. Standard deviation was
found to be within ± 5% for this system
compared with values in the range of ±15
to 20 % typical of other recent studies.
Individual templates were developed
by measuring fume rate vs. voltage at a
fixed wire feed speed. Data for steady-
current operation are illustrated in Fig. 9.
Each frame is for a given constant wire
feed speed (or current). These results
were used to construct the three-dimen-
sional typography shown in Fig. 10.
Pulsed-current results are presented
in Fig. 11 for individual wire feed rates.
A three-dimensional representation of
these data is shown in Fig. 12. Pulsed-
current operation introduces other para-
meters: pulse width, frequency and
waveform. The representation in Fig. 12
is for data gathered at a constant pulse
width of 2.5 ms with waveform parame-
ters, as delivered by the Hobart Arc-Mas-
ter 500 power supply. To generate each
curve in Fig. 11, wire feed speed was set
and frequency was increased (from about
146-s | APRIL 1999
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 8 — Photograph of UNH fume chamber, side view.
Fig. 9 — Fume formation rates for GMAW of carbon steel at steady current with 92% argon/8%
CO
2
shield gas. Wire feed speeds and average currents vary from frame to frame as specified.
100 to 250 Hz), raising power level. Volt-
age and average current were controlled
automatically by the power supply and
recorded by the operator.
The effect of using pulse width rather
than frequency as a primary variable is il-
lustrated in Fig. 13. Here, wire feed speed
and frequency were held constant, and
pulse width was varied from 2.0 to 4.0 ms.
Current and voltage were controlled by
the power supply to deliver constant
power of 4600 W (from 2.1 to 2.7 ms).
Then, power rose linearly from 4600 to
6400 W at 3.9 ms. (At a pulse width of
6.25 ms, one would have steady current.
At this hypothetical extrapolated condi-
tion, we estimate the potential would have
been approximately 34 V. This is useful as
the basis for another point in Fig. 13, that
is, a fume value of 1.1 g/min at 6.25 ms as
obtained by extrapolating steady-current
data from Fig. 9C to 34 V.)
Another useful correlation is illus-
trated in Fig. 14. This shows fume rate vs.
frequency at the same wire feed speed as
in Fig. 13, but here power was held con-
stant by allowing frequency to change
with pulse width. The upper curve is for
a power level of 4600 W, which corre-
sponds to the globular peak fume rate in
Fig. 11C. The lower curve is for 5900 W,
the spray valley minimum in Fig. 11C.
Filter Studies
In prior presentations (Ref. 10) and in-
formal discussions, we have questioned
the efficiency of the fiberglass filter used
in the AWS standard test. This medium is
not designed for filtration but for aircraft
insulation. It is recommended in the AWS
standard (Ref. 7) because that fume
chamber depends on a blower to exhaust
fume, and finer high-efficiency filters
clog or “blank” before a test can be con-
cluded. (Arguments that a second AWS
filter shows no added fume pickup are in-
conclusive because ultra-fine dust that
might escape the first filter pad would
also pass through the second.)
To resolve doubts about the filter, we
performed tests using both the AWS rec-
ommended medium- and high-efficiency
(HEPA) filters. Each medium was tested
separately. Further tests were made using a
HEPA filter behind the AWS fiberglass pad.
Fume mass was basically the same at
identical welding conditions indepen-
dent of primary filter type. In fact, even in
this positive-pressure system, blanking of
a HEPA primary filter was serious enough
to cause leakage of fume from the cham-
ber seals, limiting test time. When the
HEPA filter was used as a backup to the
AWS pad, fume was visibly evident on
the HEPA filter, but its weight was too
small to register.
We conclude that even
though some fume does
pass through the AWS
medium, it is too small in
mass to affect results.
Thus, the recommended
fiberglass pad is suitable
for measuring welding
fume rate on a mass basis.
Indeed, it was used for
most of our experiments.
If one is concerned about
ultra-fine fume and parti-
cle number populations,
on the other hand, this
medium might be inap-
propriate.
Discussion
Design and construc-
tion of an improved fume
chamber was one goal of
this research. Its use to
document three-dimensional fume maps
for a single, common GMAW material
and shield gas under steady- and pulsed-
current conditions was another goal. The
chamber illustrated in Figs. 6–8 was built
and used to measure fume rates for solid
mild steel (ER70S-3) electrode welding
wire 1.2 mm in diameter, and a 92%
argon/8% carbon dioxide shielding gas at
steady- and pulsed-current conditions
(pulse width of 2.5 ms).
Results are illustrated in Figs. 10 and
12. Both maps display a typography sim-
ilar to that foreshadowed in Fig. 5. That
is, fume rate rises gradually as current,
voltage and wire feed speed increase and
the welding mode migrates from short-
circuit to globular. Globular transfer cre-
ates peak fume rates, as illustrated in Figs.
10 and 12 by ridges running parallel to
the wire-feed-speed axes. Voltage along
this ridge is almost constant (26.5 V with
steady current and 23.5 V with pulsed
WELDING RESEARCH SUPPLEMENT | 147-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 10 — Three-dimensional representation of steady-current
fume data from Fig. 9. (GMAW of carbon steel at steady current
with 92% argon/8% CO
2
shielding gas.)
Fig. 11 — Fume formation rates for GMAW of carbon steel, pulsed current, with 92% argon/8%
CO
2
shielding gas. Wire feed speeds vary from frame to frame as specified.
current). At higher voltages, fume rates
decline dramatically as the welding
mode shifts toward spray transfer. At 127
mm/s wire feed speed with steady cur-
rent, the peak-to-valley fume rate differs
by a factor of two. At the same wire feed
speed with pulsed current, the drop in
going from globular to spray mode is
even more dramatic, a factor of nine. At
even higher voltages, fume rates rise
again sharply and for both current types
as transfer shifts from spray mode to
streaming.
Several important points should be
made in comparing Figs. 10
and 12. For example, fume
rates in the spray valley are
consistently lower with
pulsed current, less than
half those observed with
steady current. Also, the
voltage range for spray
transfer (i.e., the width of
the valley) with pulsed cur-
rent is about double that
found with steady current.
In general, claims that fume
rates are lower with pulsed-
current welding are con-
firmed by these data. There
is one notable interesting
exception. At a wire feed
speed of 127 mm/s and at
globular conditions, the
maximum fume rate with
pulsed current (the “peak”
in Fig. 12) is higher than that with steady
current.
The influence of other pulsed-current
welding parameters is illustrated in Fig.
13. Here, one sees a fume rate of 0.65
g/min at “machine standard” conditions
(160 Hz, 2.5 ms pulse width, a point just
to the right of the globular peak in Fig.
11C). With an increase in pulse width to
3.5 ms, fume rate drops by a factor of three
to 0.2 g/min, a level typical of spray trans-
fer. Similar variations are evident from Fig.
14. Here, frequency was changed while
the power supply automatically adjusted
pulse width to maintain constant power.
Again, we see dramatic variations in fume
rate with frequency. (It should be noted
that a family of typographies similar to Fig.
12 could be generated under pulsed-cur-
rent conditions, one for each pulse width.
Other variations are possible depending
on the waveform delivered by the power
supply.)
For a given welding situation, fume
generation rate is essentially dictated by
welding mode. According to our re-
search, one should, if given a choice, op-
erate under spray transfer conditions to
minimize fume evolution in GMAW of
carbon steel with solid electrodes. Fume
can be cut further by a shift from steady
to pulsed current if the latter is also tuned
to the spray-mode minimum. Welding at
globular and streaming conditions in-
creases fume rate dramatically with both
steady and pulsed current.
Modeling Considerations
How does one explain the dramatic
difference in fume rate between globular
and spray transfer conditions? A detailed
answer to this question is the basis of
work to be published later, but a brief
discussion is appropriate here. Our ex-
planation is based on the model intro-
duced by Gray, Hewitt and Dare (Ref. 4).
They list seven potential sources of fume.
Of these, evaporation and explosive
droplet detachment are relevant to this
discussion. We assume that these two
148-s | APRIL 1999
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 12 — Three-dimensional representation of pulsed-current
fume data from Fig. 11. (GMAW of carbon steel at pulsed-cur-
rent with 92% argon /8% CO
2
shielding gas.)
Fig. 13 — Fume formation rates vs. pulse width in pulsed-current
GMAW of mild steel. Wire feed speed was fixed at 127 mm/s and
pulse frequency at 160 Hz. Voltage and current were automatically
controlled by the power supply at values shown. Extreme data point
at 6.25 ms extrapolated from steady-current data of Fig. 9C.
Fig. 14 — Fume formation rate vs. frequency in pulsed-current
GMAW of mild steel. Wire feed speed was fixed at 127 mm/s. Volt-
age, current and pulse width were automatically controlled by the
power supply to maintain power constant (4.6 kW in top curve and
5.9 kW in the bottom curve). High fume conditions correspond to
globular transfer at this wire feed speed. Low fume corresponds to
the spray valley minimum ( see Fig. 11C).
mechanisms are both active during glob-
ular transfer. That is, as a droplet forms
and grows at the tip of the electrode, its
surface heats up and electrode metal
evaporates to diffuse into the gas stream
where it later oxidizes and condenses to
form fume. Then, when a droplet de-
taches from the electrode, the area of the
“neck” decreases to a point where enor-
mous heat is released because of the ris-
ing resistance and high current. This gen-
erates metal vapors at explosive rates,
ejecting micro-droplets into the gas
phase. Some droplets, or “spatter,” are
too large to remain in the atmosphere,
but smaller droplets, or “sputter,” persist
as fume. Thus, under globular condi-
tions, evaporation and explosive detach-
ment both contribute to the fume.
Why do fume rates drop during spray
transfer? In spray mode, the arc spot is no
longer focused at the bottom of the grow-
ing drop, but the arc moves around de-
taching droplets and up the side of the
melting electrode. (See recent high-speed
photographs of Jones, Eagar and Lang,
Ref. 11, for illustration.) Thus, explosive
ejection is no longer a major contributor
to fume. In fact, there may be even less
droplet superheating with the expanded
arc contact area, creating even less evap-
oration than in the globular mode.
Why are fume rates generally lower
with pulsed current? If the droplet de-
tachment frequency matches pulse fre-
quency, detachment can occur during
the low background current phase of the
cycle and there will be less explosive
fume ejection. Also, it is likely that puls-
ing, by promoting early droplet detach-
ment, results in even less superheating of
molten electrode droplets and less metal
evaporation.
Why is the spray valley wider with
pulsed current than with steady current?
Pulsing evidently promotes droplet de-
tachment under spray conditions over a
wider voltage range. This causes spray
transfer under what might be globular
conditions with steady current.
Why, under special conditions (the
spike in Fig. 12), does pulsed-current
globular transfer create more fume than
steady-current globular? If droplet and
current frequencies are quite different,
detachment may occur not during the
background current cycle but during
peak current, which is considerably
greater than the steady-current value at
similar power levels. This high peak cur-
rent passing through the detaching
droplet neck will cause even more ex-
plosive fume expulsion than what occurs
at steady-current globular conditions.
Summary
Welding fume will undoubtedly per-
sist as a subject of litigation and legisla-
tive debate. Its role in workplace health
and safety will command more scrutiny
as time passes. This research was con-
ducted to improve the accuracy of fume
formation data and to correlate fume rate
with welding mode.
Fume formation rates measured in the
past have lacked precision. Heile and
Hill’s pioneering work (Ref. 9) reported a
standard deviation of ± 20%, but some sta-
tistical variations were as great as ± 40%.
Other researchers report standard devia-
tions of ± 15% to 20% using established
U.S. and European fume chamber designs.
The fume chamber of new design in-
troduced in this paper was operated
through a comprehensive range of
steady- and pulsed-current conditions
that produce acceptable welds. With
more than 150 test results, many of them
replicates, the standard deviation was ±
5%. Thus, excessively broad limits of ac-
curacy need no longer be a hindrance to
data interpretation.
Each electrode type/shielding gas
combination has its unique fume profile.
A different profile results if pulsed current
is used. In fact, each pulse width exhibits
its own fume typography.
Conclusions
A fume chamber of new design was
used to measure fume generation rates
for gas-shielded metal arc welding of
mild steel under both steady- and
pulsed-current conditions using a single
common shield gas. A continuous three-
dimensional typography is revealed
when fume rate is plotted above a plane
defined by wire feed speed and voltage.
In general, fume rates rise as power is in-
creased through the short circuit mode.
Rates peak under globular transfer con-
ditions and then drop dramatically as the
mode shifts to spray transfer. Fumes in-
crease again as the mode shifts to
streaming transfer.
Profiles are similar for both steady-
and pulsed-current welding, except the
spray-transfer “valley” is wider and lower
with pulsed current.
The fume chamber design used here
is recommended for its high accuracy
and flexibility. This particular chamber
has been moved to another New England
Welding Research Consortium location
(the Harvard School of Public Health)
where it is being used for MIT-Harvard
projects involving stainless steel fume
and the effects inhaled welding fumes of
various kinds exert on animals.
Acknowledgments
We offer thanks to the Edison Welding
Institute, which provided major funding
for this work. Additional help from the
University of New Hampshire, ESAB
Welding & Cutting Products, Hobart
Brothers Co., Hornell Speedglass, Inc.,
John Deere and Co., Miller Electric Co.
and the Torit Div. of the Donaldson Co.
was necessary to the success of this pro-
ject. In particular, Todd Holverson of
Miller Electric and Leo Wildenthaler of
Hobart provided valuable personal at-
tention to the project during equipment
design and shake-down stages. Appreci-
ation is also expressed to colleagues in
the New England Welding Research Con-
sortium under the direction of Professors
Joseph Brain of the Harvard School of
Public Health and Thomas Eagar of the
Massachusetts Institute of Technology,
who provided valuable guidance and as-
sistance.
References
1. Press-Time News. 1991. Rectangular
wave pulsing reduces weld fume generation.
Welding Journal 70(11): 5.
2. Irving, R. 1992. Inverter power sources
check fume emissions in GMAW. Welding
Journal 71(2): 53–57.
3. Castner, H. R. 1995. Gas metal arc weld-
ing fume generation using pulsed current.
Welding Journal 74(2): 59-s to 68-s.
4. Gray, C. N., Hewitt, P. J., and Dare, P. R.
M. 1982. New approach would help control
welding fumes at source, part two: MIG fumes.
Welding and Metal Fabrication, Oct., pp.
393–397.
5. Willingham, D. C., and Hilton, D. E.
1986. Some aspects of fume emissions from
MIG welding stainless steel. Welding and
Metal Fabrication, July, pp. 226–229.
6. Quimby, B. J. 1998. Fume formation
rates in gas metal arc welding. Master’s thesis,
University of New Hampshire Department of
Chemical Engineering. Also obtainable from
Edison Welding Institute, Columbus, Ohio.
7. ANSI/AWS F1.1, Method for Sampling
Airborne Particulate Generated by Welding
and Allied Processes. 1985. American Weld-
ing Society, Miami, Fla.
8. Albert, R. V. 1996. Fume generation in
gas metal arc welding. Ph.D. thesis, University
of New Hampshire Department of Chemical
Engineering.
9. Heile, R. F. and Hill, D. C. 1975. Par-
ticulate fume generation in arc welding
processes. Welding Journal 54(7): 201-s to
210-s.
10. Ulrich, G. D., Eagar, T. W., and Brain,
J. D. 1993. Fume physics, chemistry and phys-
iology. Oral presentation, Edison Welding In-
stitute, Ninth Annual Welding Research Con-
ference, Columbus, Ohio.
11. Jones, L. A., Eagar, T. W., and Lang, J.
H. 1998. Images of a Steel electrode in Ar-2%
O
2
shielding during constant current gas metal
arc welding. Welding Journal 76(4):135-s to
141-s.
WELDING RESEARCH SUPPLEMENT | 149-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT