Filletweld test evaluation
Introduction
Abbreviations used
January2021  the new data and interpretations presented here come
from the series of four miniprojects, three in November 2020
and one in January 2021, abbreviated as follows:
Buttwelds  not part of this topic:

mp1
= 1st miniproject  "Uweldment in R.H.S.  fabricate, analyse, test"

mp2
= 2nd miniproject  "Weld test  weld at centre of simplysupported
beam"
For the fillet welds tests:

mp3
= 3rd miniproject  "Fillet welds tensile tested in beam test"

mp4
= 4th miniproject  "Tensiletest rig for beamconfiguration filletweld samples"
See Structures "index"
for context to the tests and this article.
Other abbreviations used:

bcfwtt = BeamConfiguration FilletWeld Tensile Test

RHS = Rectangular Hollow Section

FEA = Finite Element Analysis
Content
This page collates
 findings and data
 interpretation of these
from
 own tests
 own FEA
 other works where helpful to develop own analyses
in serving investigation and evaluation of filletweld
[definition]
performance.
At time of creating this page, late January2021, own content may
include

interpretation of tests performed November2020 to January2021

reference to fatigueresistant filletwelds programme Summer2011

possibly some filletweld tests looking at weld quality assurance 2014
Context
At time of writing, January2021, the major interests are

harnessing the filletweld "static" tensile break test method "as is"
to understanding filletweld "static" loading performance

developing the filletweld test method as a "dynamic" loading method
for investigating fatigueresistance / cyclicload performance of
fillet welds

applying these to
Proposal : research path to highstrength steel fatigueresistant
welded structures
utilising the 2011 observation that filletwelds can perform in
fatigue near the "mill" steel sections they are joining
Previous investigations and findings
Filletweld fatigueresistance investigation 2011
From 2011 
fatigueresistant fillet welds test programme
"index" page to the methods, tests and interpretations of a programme
investigating the possibilities for fatigueresistant fillet welds for
warships and bridges.
Also as
report
[PDF, 3.2MB]. Interesting result  FCAW Tfillet 1.4Million cycles
unbroken no cracks detectable, for 250thousand cycles expected to
break. Which could now be progressed if the "bcfwtt" could be made
cyclic loading for fatigue testing. Considered in
Proposal : research path to highstrength steel fatigueresistant welded structures
The beamconfiguration filletweld tensile test findings
These are
mp3 and
mp4
of the Nov2020toJan2021 test programme.
Necessary data from the beamconfiguration filletweld tensile test to
enable calculating the force on the testweld are

hydraulic piston force at moment the sample broke
(typically a "Force=Pressure*Area" estimate using
 hydraulic fluid pressure at moment of breakage
 the hydraulic cylinder's piston diameter
where pistonarea=pi.d^2/4)

The separation of the beam's bearers from the central plate ("L_ma")
(where 2*L_ma + centralplatethickness = separation of the two bearers)

the height of the beam section ("h")
Proceeding onwards : to then obtain the stress(es) in the test weld,
necessary data are

the size of the test weld(s)  typically the leglength ("z" in
standard welding terminology)

the length of the test welds
Proceeding further onwards : to evaluate the calculated values and arrive at
interpretations and conclusions, the following is among the information
needed

the welding conditions

the material specification of the weld(s)

the material specification of the two halfbeams
In the "bcfwtt", to obtain the force in the weld at breaking F_w
(which we want) from the force applied by the hydraulic piston F_p
(which we know) uses the formula
F_w=F_p*L_ma/2h
which is probably what one would intuitively guess.
Whose derivation is presented in
mp3
see section
Stresses in weld analysed
.
Data and its processing
"Fillet welds tensile tested in beam test" [mp3]
The measurements and analysis are presented in the webpage for
mp3
For the [ISO14341A] "G3Si1" weldmetal, the calculated breaking strength
is 567MPa  for expected typical value for "G3Si1" weldmetal of
560MPa.
"Tensiletest rig for beamconfiguration filletweld samples" [mp4]
The hydraulic pumpandcylinder used in
mp4
had no hydraulicfluid pressuregauge, so an accurate citable "F=P.A"
estimate of cylinder force applied to the test sample is not
available.
An estimate of force applied was made by a very inexact method
described. But where the resulting deduced weld breaking stress could
not be known at the time of the test. So there could be no human bias
in the applied force estimate.
The resulting filletweld breaking strength estimate of 572MPa, for
the expected 560MPa mean value for "G3Si1" GMAW/MIG weld metal, is
surprising for its apparent accuracy.
Evaluation and interpretation of tests
Writing Thursday 28January2021:

the tests, despite being done apparently rather approximately, present
estimates of weld strength in MPa  the stress they bear at breaking 
which are remarkably close to the accepted typical 560MPa
breakstrength of a "G3Si1" ("ER70s6") GMAW/MIG weld

the appearance of the broken welds are presented in
filletweld fracture pictures

the weld fracture surfaces in this "static" (ie slowly
increasing) overload are apparently ductile "microvoid coalescence"
fracture  judging by the very finely rough grey appearance

the observed fracture orientation of the test fillet welds accords
with the observation(s) of the practically experienced
Hicks (ref.)
in his book, see pg82, being along or close to the fillet leg between
weld and the plane of the longitudinal beam surface

the deduced value for weld fracture stress does match the established
expectation  but my derivation is different, being simpler and is
based on the observed mode (myself, Hicks
(ref.)) of fracture (developed further imminently)

in that the familiar tensile fracture strength of the weld (not
the shear strength, which would be half of the uniaxial tensile
strength) is output by dividing the force upon the weld at moment of
break by {weldleglength * weldlinearlength} (the area of the
fractured fillet "leg")
Added on Sunday 07February2021  FEA of the "bcfwtt":

the following two main findings are significantly made by this one
image  FEA output  von Mises stress
which can be identified in the referenced
webpage
on FEA of the "bcfwtt" (stress scale example  "1.001E+09" = 1.001GPa
/ 1001MPa)

the location and orientation of weld overload fracture
expected by FEA
concurs with the physical outcome of
own observations and
Hicks (ref.)

at the force known in the physical test to cause fracture FEA shows the intensity of stress in the weld appears enough to cause fracture
More detailed interpretation of filletweld tensile break measurements
The findings can now be considered in relation to established
knowledge and theory.
Concepts and terminology:
 as I visualise them, from
mp3
 as visualised by Hicks (ref.) on pg85
where Sigma=linearstress and Tau=shearstress.
Hicks (ref.) on pg85 presents the known
established formula which should connect
strength[MPa]<>force[N]
for a *doublesided* fillet weld
In this equation he presents as [6.3],
with symbols here given text name
P_perp=2.t.L/sqrt2 * (sigma_perp+tau_perp)
and introducing t=fillet throat thickness ("a" in common welding
terminology) and L=length of the weld
Taking [6.3] and [6.4]
P_perp=2.t.L/sqrt2 * (sigma_perp+tau_perp)
and
sigma_perp = tau_perp {for this geometry}
plus;
we recognise for a 454590 triangular cross section fillet that "t",
the weld throat, which is "a" in conventional weld terminology  that
sqrt2.a=z  the fillet leglength
so
sigma_perp=P/(2.sqrt2.t.L)=P/(2.z.L)
This is for a *doublesided* fillet weld.
Mine has two welds, but in series, not in parallel.
Each weld must take the full stress  so it is in stress terms a
singlesided fillet weld ("ssfw").
So in my case of applying "Hicks"
sigma_perp_ssfw=P/(z.L)
which is the same as what I found empirically and in forming
mathematics which matches the physical situation. The observation
that fracture is along the longitudinal leg (length "z") of the fillet
weld.
"My" derivation for fillet weld strength and the established
arithmetic expression for fillet weld strength give the same answer.
So, to that extent, my "finding" is "expected".
Whereas the mode of breakage differs, with my derivation expecting
fracture along the longitudinal fillet leg, and the established
expression implying fracture is expected across the weld throat.
Hicks qualitatively comments, on pg82, that fillet weld overload
fracture has the form which I found  close to the longitudinal fillet
leg. So he's derived and presented "the established" analysis in that
knowledge.
So it "doubly" affirms that the derivation applies.
I invite comment on these following things which I find surprising:

I know the shear yield is half the observed uniaxial tensile yield 
and you'd expect that to largely carryover into breakstress  so I
was kindof expecting a breaking force half of what's observed, given
the argument that this loading looks like pure shear and you'd expect
to see a lower break strength from more "efficient" loading on the
weld metal to induce overload failure ?

WRONG !!! > "why doesn't the "S355" substrate material with nominal
355MPa yield manifest as the strength, given the proximity of the
fracture around the weldmetalplatemetal boundary?"

The 355MPa is the yield stress on distributeddeformation bending  as seen when the beam bends  as observed in own buttweld tests
mp1
and
mp2

Is the break strength of S355 steel in local break constrained to no
distributed deformation 560MPa ?

Is the FEA simulation's
(
image
,
webpage
)
full span of weld leg but very narrow high stress zone  about 1.5mm
width at above 1GPa  such constraint against yielding ?

Would a "macro" of the break region (section and smooth to fine
abrasive (eg "800 grit") then etch with eg iodine
etchant and examine) show whether the break is through the weld metal
or "parent metal" of the RHS ?
(
example
of a "macro")
"Watch this space.." :)
References
John Hicks, Welded joint design [3rdedn.], Abington Publishing, 1999
(R. Smith, 28Jan2021, 29Jan2021 (mp4 str.calc. retn), 31Jan2021 (break tens),
03Feb2021 (560MPa 355MPa), 07Feb2021 (aname, FEA, FEA img preview), 08Feb2021 (aname fea bcfwtt))