Vous êtes sur la page 1sur 1

Moving Towards Virtual Testing

J Ankersen, GAO Davies, and D Hitchings

INTRODUCTION
Destructive testing is a major cost contributor in qualification of new designs in the Virtual testing has particular relevance to UAVs since: materials have a high strength to weight ratio compared with e.g. aluminium. Through
aircraft industry. Both time and money can be saved by using virtual testing, thickness strength is, however, several times lower than the in-plane strength.
•Production runs may be short with frequent spec changes
particularly when exploring unconventional designs. Virtual testing aims at reducing: Delamination failure is therefore of particular interest here.
•Unconventional shapes are likely
1. Non-recurring costs by reducing the need for testing at component level resulting An interface element is used for modelling delamination in FE77. This element has zero
in a faster design process •Data sheets may be non-existent thickness and is inserted between conventional elements allowing these to separate
once the critical strain energy release rate (SERR) is reached, see diagram below.
2. Recurring costs by means of more accurate design tools The current solution to a design tool consists of a MSC.Patran user interface
coupled to our in-house Finite Element (FE) package FE77. Both in-plane failure The performance of the FE77 interface element has been tested successfully on
The current work is part of the Material and Structures package under the FLAVIIR
and delamination can be predicted. An explicit solver is used for coping with severe several benchmarks with a known solution. For further validation it was used for
research programme which looks at technologies for a future Unmanned Air Vehicle
non-linearities which are inevitable when modelling failure. predicting delamination failure in a more realistic structural detail. Experimental testing
(UAV). The programme is funded jointly by BAE Systems and EPSRC and has ten
has now commenced and results can be compared with the FE predictions.
university partners. Only laminated composite materials are considered here. Carbon fibre reinforced

Failure prediction in skin-stiffener detail – Tee-pull test

V
Green 45° ply
Orange 0° ply
Light blue noodle
Dark blue steel
‘dummy’ skin

Interface

Specimen description Finite element model Test


The stiffener is formed by two C-sections put back to back with a •For the tension load case a slice model was set up for Finite Element Analysis •Specimen is shown mounted in test machine for tension loading. Dummy
capping strip at the top and bottom resulting in an I-section, shown in metal skin is fully clamped at ends.
•Lower part of stiffener meshed finely
yellow. Overall dimensions of the stiffener are 60mm height, 60mm
•Four type A specimens tested, nine repetitions of each type is planned
width, and 78mm length. Four variants of the test specimen were •Fully clamped ends of metal dummy skin
specified with thickness of the C-sections and fillet radius as variables. •Of particular interest is the initiation and growth of delaminations in the
•Each RTM ply modelled with a layer of special shell elements
Only results for the variant with thick C-section (1.9mm) and large fillet ‘noodle’ region where web and lower flange meet
radius (5.4mm) - type A - are shown here as testing is still ongoing. •Bolted joints not modelled. Fully connected mesh in bolt region with interface
•Load and displacement monitored on test machine and LVDT
The stiffener is bolted to a dummy metal skin. Loading brackets for between bolt lines.
tension load and shear load are also bolted to the stiffener. The •High speed filming used at 40000/67000 frames per second. Resolution is
•Interface elements were inserted at anticipated fracture planes. Stiffness
laminate consists of a IM7 4 Harness Satin Weave/PR500 RTM Cloth. restricted to 125x256 and 125x125 respectively
factors of 0.1 or 1.0 were used

Two stage failure


Results and comparison Initial failure where a
crack forms in the
noodle and web
25 initial failure
Two distinct failure patterns emerge from the four specimens associated with a
tested: final failure small reduction in
20
stiffness.
1) Initial failure occurs at around 13kN with a crack forming
Load [kN]

unstably in the web and noodle. Further loading 15


(displacement) leads to final failure with cracks forming also
unstably in the noodle/C-section interface and flange. An 10
interface stiffness factor of 0.1 predicts this stage-wise Final failure is here defined as failure
failure (specimen 2A and 5A). 5 extending into the flange. The stiffener
still carries load but at a much reduced
2) Sudden failure occurs without prior warning at 18kN with 0 stiffness. All crack fronts have been
cracks forming in the web, noodle/C-section interface and 1A 2A 5A 8A FE FE stopped by bolt lines. The noodle crack
flange in a single stage. In the FE prediction this behaviour tf=0.1 tf=1.0 closes again.
is seen with an interface stiffness factor of 1.0 (specimen 1A Specimen
and 8A).

20
test 5A LVDT
test 8A LVDT
Non-linear softening was observed in all tests, the cause of
15 FE soft load tf=0.1 Single stage failure
which is not fully understood. Initial stiffness is in agreement
between test and prediction which remains linear up to FE soft load tf=1.0
Load [N]

failure. Stage-wise failure starting in the noodle is seen with 10 Failure is abrupt with a large drop in
low interface stiffness (tf=0.1). This is also the case in e.g. stiffness. Cracks form in an unstable
test 5A. A high Interface stiffness (tf=1.0) matches the manner in the web, noodle/C-section
response of Test 8A. Failure in the FE prediction with soft 5 interface and extends into the flange and is
loading is seen as a jump in displacement while hard loading again arrested by the bolts. Cracks also
used in the test results in a load (force) drop. 0 form in the fillet but not in the noodle.
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Displacement [mm]

CONCLUSIONS
Material law for interface element in FE77
•Delaminations in the tests only occurred within interfaces which had been identified as critical during the
The tangent stiffness drops off gradually to near zero as the critical energy •Interface stiffness is related to stiffness of associated element FE prediction stage
(Gc) is reached. The strength (Rc) is also reached as this point. Hereafter,
•User specified stiffness factor gives additional control over actual
no load is carried. This law works well with higher order elements.
initial interface stiffness
•Predicted failure loads are overall in good agreement with the test results seen so far. It should be noted
that the noodle toughness and strength values are likely to vary considerably in the actual specimens
•Resin rich layer thickness and interface stiffness can be related by the
shown expression •Hard loading in FE predictions should be completed. This may yield slightly different results where failure
Force
is abrupt under soft loading
Response at each  E re sin
 •There may be some correlation between variation in failure loads in test and FE – and the variation in FE
sampling point u t re sin
Rc interface stiffness i.e. modelled resin rich layer thickness

For a given interface law with a stiffness factor of 0.1 and a neat resin
Area g c modulus of 5GPa, the resin rich layer thickness is 0.08mm which is a
Slope ko realistic value. If the stiffness factor is increased, the corresponding
resin rich layer thickness decreases.
r0 rfail
Displacement

Vous aimerez peut-être aussi