Professional Documents
Culture Documents
Design, Fabrication, and Testing of A Composite Side Door For A Mid Suv
Design, Fabrication, and Testing of A Composite Side Door For A Mid Suv
Mid-Size SUV
John N. Owens
GM Research and Development
Materials and Processes Lab
Abstract
As part of a mass-savings initiative, a composite intensive side door project was started
at GM R&D. In order to allow more innovation in the design, two normally limiting
constraints were eliminated. Firstly, the Class A requirement for the outer surface was
relaxed and secondly, labor intensive handcrafting was allowed for the purpose of
prototyping. The composite door was constrained to fit the existing door opening and to
use carry-over internal hardware.
Using stiffness criteria, finite element analysis was used to develop a minimum mass
design using composite sandwich structures. Preforms were handcrafted from molded
foam cores wrapped with a combination of woven and stitch-bonded unidirectional glass
fibers. Prototype doors were molded using vacuum assisted resin transfer molding. The
resulting composite door met the stiffness targets and yielded a 35% mass reduction
compared to its steel counterpart.
Background
Structural applications of polymer composite materials are expected to provide significant mass
reductions over the steel components that they replace. With an aim towards more fuel-efficient
vehicles, mass reduction has become an increasingly important strategy. While wholesale
substitution of polymer composite materials for steel could reduce mass dramatically, a less
aggressive approach of targeting specific parts is more likely to succeed.
Closures are often targeted for mass reduction because they have minimal influence on the
structural integrity of the vehicle as a whole. Prototype closures can often be substituted onto existing
vehicles with little or no changes to the rest of the vehicle. This facilitates testing and reduces the
development costs significantly. In this work, attention is focused strictly on the front side door. Due
to vehicle symmetry, any mass reduction achieved in a side door is doubled on the vehicle (for two
door models). An additional advantage of developing a low mass side door is that it can be easily
commonized for high volume. Thus, a successful development of a low mass side door could have
wide applicability.
F o r th is p a rticu la r sid e d o o r p ro je ct, G M s m id -sized SUV (Envoy, Trailblazer) was chosen as a
target vehicle for mass reduction. The project objectives were to replace the production steel door
with a polymer composite door that was a minimum of 30% lower mass while keeping projected cost
increases to a minimum. The composite door had to meet or exceed all of the stiffness requirements
of the steel door and would be expected to meet the federally mandated side impact requirements
(FMVSS-214) as well. In this investigation, all internal hardware was required to be carry-over from
the steel design.
A key element in achieving the mass reduction was to design intrusion protection into the
composite structure rather than including a separate beam. While composites have been shown to
absorb energy well in crush[1-2], side impact requires energy absorbtion in a bending mode. With the
right architecture, composites can be successfully designed for this application [3-6]. Previous inhouse development on various composite door materials had shown that a glass fiber fabric oriented
at 45 provided a progressive failure mechanism that absorbed a large amount of energy. With the
side impact requirements in mind, it was decided to take advantage of this particular glass fiber
orientation in the composite door concept. After looking at several possible constructions, initial
estimates suggested that the 30% mass savings target would require use of a sandwich construction.
Without the mass efficiency of a sandwich structure, a 30% mass savings was not thought to be
possible using glass fibers. Once the decisions were made to use a glass fiber fabric reinforcement
around a core material, design optimization procedures could begin.
+45 Layer
-45 Layer
+45 Layer
0 Layer
Core
+45 Layer
-45 Layer
+45 Layer
Class A
The thickness of each ply within the general structure was left as a variable for the structural
optimization. The optimization goal was to determine the combination of ply thicknesses that would
yield a minimum mass structure that still met or exceeded the structural stiffness criteria.
Structural stiffness criteria were established based on standard door design guidelines. These
guidelines established the maximum allowable deflection for application of a fixed load at a specific
location on the door. Seven stiffness criteria were used: fore window header inboard, fore window
header outboard, aft window header inboard, aft window header outboard, inner beltline outboard,
outer beltline inboard, lower torsional, upper torsional, and vertical rigidity. These tests will be
discussed later in the structural validation section.
Initial optimization analyses were conducted without constraints on the ply thicknesses within each
region. This resulted in mass of approximately 11 kg or a 42% mass reduction from the 18.90 kg
steel door. Manufacturing constraints were progressively imposed on the model through several
optimization iterations. The first manufacturing constraint imposed was to link layers from several
adjacent regions to form mats of constant thickness (to reduce the number of fabric pieces). Fabric
thicknesses were then constrained to be an integral multiple of 0.65 mm (single ply thickness). A 0.30
mm minimum thickness was then imposed on all unidirectional layers. Minimum core thickness for
specific areas was also imposed for fabrication and handling considerations. As constraints were
added, the resultant mass of the door also increased. After many iterations, a balance between
manufacturability and mass savings was achieved that resulted in a design mass of 13 kg ( a 30%
mass reduction).
In the final optimized design, unidirectional carbon fiber was used to reinforce the inner belt beam
instead of unidirectional glass fiber. This change was made because with glass fiber, the model
called for a 8 mm thick wall which was deemed excessive.
tool halves. With the top and bottom mold halves floating; any pressure applied to the tool is
transferred to the resin and preform within the tool cavity. The concept was to infuse resin into the tool
with the tool halves slightly separated (approximately .125 to .250 mm) and then rely on air pressure
in the autoclave to compress the tool halves together while the resin cures. If the mold contained
insufficient resin, the tool halves would over-close, yielding a composite door that was thinner than
designed and if the mold contained an excess of resin, the composite door would be too thick. With a
repeatable preform, determination of the desired quantity of resin was expected to take only a few
trials. With this approach, sealing of the tool was of critical importance. Pressure on the resin in the
mold could only be maintained if no resin was allowed to leak out of the mold and if no air from the
autoclave was allowed to leak into the mold.
Complete filling of the mold was also considered a technical challenge due to the complexity of the
composite door geometry and the variation of fiber orientation and skin thickness throughout the part.
Although computer simulations of resin infusion processes have improved dramatically, it was
concluded that the composite door was too complex to obtain reliable flow predictions. Instead, it was
decided to specify a tool with many vacuum ports (vents) to reduce the chances of entrapped air
causing dry spots.
With resin infusion processes, flow along the edges of a tool (race-tracking) is usually the biggest
problem in obtaining repeatable flow patterns. In order to avoid this unpredictable flow phenomena, a
perimeter runner feed system was implemented by extending the flange region of the door tool by
approximately 20 mm. This excess flange area would be trimmed from the door in a post mold
operation. Preforms were designed to fill only 12.5 mm of the excess flange and the remainder was
designed to serve as a resin runner. With this approach, the resin inlets would feed the perimeter
runner and the mold would fill from the outside edges towards the middle versus the more
conventional system of filling from an interior point toward the outer edges. This long flow front should
result in a faster fill time and a more repeatable filling pattern. The disadvantage of this approach is
that the vacuum ports (vents) must be on the part surface.
A vacuum vent on the appearance surface (Class A side) of a door is undesirable. In order to
avoid venting the appearance side, the concept of using vents through the foam core of the sandwich
structure was developed. Instead of venting trapped air out through the mold wall, a hole was drilled
in the foam core allowing air to escape through the core to the non-appearance side of the composite
door. In order for this concept to work, the resin flow front on the appearance side of the foam core
must reach the hole in the foam core ahead of the resin flow front on the non-appearance side.
Fortunately, the surface veil used on the appearance side of the door dramatically increased the
permeability of the outer preform layers virtually insuring that the flow front on the appearance side
would reach the vent location first.
Tooling Design
Once the tooling concept was decided, attention was focused on seal design required to make the
molding process work. A two seal system was chosen; a primary resin seal, and a secondary
vacuum seal. The role of the primary seal was to prevent any resin from leaking out from between
the two mold halves. Because the two mold halves were designed to be floating (no hard mold to
mold contact points), the primary seal needed to fully seal at mold gaps ranging from 0.50 mm above
to 0.50 mm below the designed mold gap. In addition to holding resin pressure of up to 0.27 MPa,
the primary seal also had to be vacuum tight. With these requirements, a silicone rubber extrusion
with a hollow bulb was chosen.
The purpose of the secondary seal was to facilitate closing the two mold halves and keeping them
together. With the preform in the tool, it was expected that some initial compression force would be
required to close the two mold halves to the desired mold gap. Capability to use vacuum to pull the
mold halves together when they were 6 mm apart was needed. Thus the secondary seal needed to
engage long before the primary seal in order to pull the mold halves closer together at least to the
point that the primary seal would engage. For this purpose, a silicone rubber V-seal was chosen.
Figure 3 shows a cross sectional view of the final sealing system. The primary seal was located in the
lower tool half where it served as a visual guide in locating the preform. The gap between the primary
seal and the preform was designed to serve as a resin channel or runner to distribute the resin quickly
around the perimeter of the part.
Fiberglass Preform
Lower Mold Half
Alignment
Pads
Vacuum
Vents
Underlying
Ribs
Figure 4. Illustration of lower composite door tool w/ alignment blocks and vents
In the above figure, the green tool surface is part of the door while the cyan tool surface is off the part.
In order to facilitate preform placement and final trimming of the composite door, the trim lines were
scribed into tool during the CAD-CAM machining operation. The scribed trim line in combination with
the primary seal (not pictured) provided good visual aides when laying in the preforms.
The top half of the tool also used the same construction as the lower half, but was of a much simpler
shape. The shape of the top tool half was basically the class A outer surface of the production door
outer with the flange areas extended. In the corners, square receptacles were added to mate to the
alignment blocks of the lower half.
The laminated tool was instrumented with eight dielectric flow sensors (locations shown above).
These custom built sensors were composed of a copper wire (12AWG) running along the central axis
of a inch OD copper tube. Epoxy potting compound was cast into the tube to electrically insulate
the wire from the copper tube. The sensors were embedded in the tool with the end flush to the mold
surface. By placing a sinusoidal voltage on the central wire and using the copper tube as an antenna,
flow arrival could be detected by a sudden increase in signal when the liquid resin provides a
conductive path from wire to tube. Resin cure could also be detected with these sensors because the
resin conductivity decreases dramatically during cure. A calibration test was performed in which resin
was introduced into a glass fiber fabric in contact with a sensor placed in a constant temperature
oven. Frequency of the sinusoidal input voltage was a used as a parameter and the antenna signal
was measured as a function of time. The results of the calibration test are plotted in Figure 5.
100
0.1 kHz
0.2 kHz
1.8
0.5 kHz
90
1.0 kHz
1.6
5.0 kHz
80
10.0 kHz
1.4
50.0 kHz
Signal
70
Gel point
T (C)
60
0.8
Temperature (C)
100. kHz
1.2
50
0.6
40
0.4
30
0.2
0
0
20
40
60
80
100
120
140
160
180
20
200
Time (min)
Preform Development
The first step in generating the fiber preform patterns was to sub-divide the inner and outer surfaces
of the composite door into regions corresponding to the contiguous fiber pieces. Using the CAD file
from the designer, the task of cutting and sewing the existing surfaces was performed. Initially, the
surfaces were sub-divided into seven main sections based on knowledge of the composite skin
thickness transitions. These main sections were:
1. Main inner
2. Main outer
3. A-pillar inner
4. A-pillar outer
5. B-pillar inner
5. B-pillar outer
7. belt beam
These seven surfaces form all of the inner and outer surfaces of the composite door. The break lines
were chosen with an eye towards creating surfaces that could be draped with a single piece of fabric
A-pillar
inner
A-pillar
outer
B-pillar
inner
Belt
Beam
B-pillar
outer
Main
inner
View from
Outside vehicle
Main
outer
View from
Outside vehicle
View from
Inside vehicle
View from
Inside vehicle
Once the 3-D surfaces were draped with fabric, a flat cutting pattern was produced corresponding to
the exact dimensions of fabric necessary to form the surface. These flat patterns were exported from
PATRAN back into Unigraphics via the DXF format. Any outside edges of the patterns were then
identified and extended outwards to create the extra 12mm run-off that was designed into the tooling.
This extra 12mm of material was to be molded into the part and then trimmed off in a post-mold
machining operation. Figure 8 presents the 3-D surface of the A-pillar inner and the corresponding
flat cutting pattern.
3-D surface
For draping
True end
Of Part
Extended pattern
For run off
Flat pattern
For cutting
Flange
3.0 mm
Uni-directional
Fiberglass Layer
Window
Seal Flange
Braid +
Fabric Ply
Foam Core
Figure 9: Cross-section of B-pillar showing foam core wrapped with unidirectional glass
fiber
Although the FEA analysis applied the 3.0 mm uni-directional reinforcement over the entire length of
the B-pillar, engineering intuition suggested that the 3.0 mm thickness should be tapered with a series
of ply drops moving up towards the top of the header. A cross-section of the B-pillar is shown in
Figure 9.
The plan for performing the B-pillar was to build the targeted 3.0 mm thickness of unidirectional fibers
by wrapping 10 individual plies of 0.3 mm around the foam core. Unigraphics was used to measure
the perimeter of the foam core (the blue lines) and the perimeter of the braided layer (yellow lines) at a
series of points between the belt line and the upper corner. The perimeter measurements were used
to determine the width of the unidirectional patterns. Because of the desire to taper the unidirectional
fiber thickness, a set of five heights were chosen to create a succession of ply drops that take place
as the section rises toward the top corner of the window header.
The inner belt beam also proved to be a design challenge. Although initially targeted to be reinforced
with uni-directional glass fiber, the belt beam reinforcement had to be upgraded to carbon fiber. With
glass fiber reinforcement, FEA analysis was driving the skin thickness of the belt beam beyond 7 mm.
The original intent of the belt beam design was to use a constant section beam thus enabling the
application of a balsa wood core with uniform thickness. Unfortunately, a tapered design evolved that
necessitated a needlessly complicated ply lay-up and a machined balsa core. The final construction
of the inner belt beam is depicted in Figure 10. It consists of a machined balsa wood core with an
angle cut top and rounded bottom. Four plies of uni-directional carbon fiber were applied to either
side of the balsa core and the entire beam was wrapped with a fabric layer on the outside. One of the
carbon fiber plies was cut extra tall to extend up into the window flange region for stiffening of the
flange.
Belt Beam
Sections
Fiberglass
Fabric
Uni-carbon fiber
Plies (0.8mm)
Balsa
Core
Typical
Section
10
the patterns for this project were hand positioned to facilitate the ease of operation instead. Figure 11
illustrates the cutting table jobs that were developed for this project.
Veil
Inner
Outer
11
Figure 12. Photograph of foam core moldings for the composite door.
12
PREFORM ASSEMBLY
After complete sets of fiber reinforcement patterns and foam cores were fabricated, the task of
preform assembly was begun. The basic strategy was to start assembling the headers and work
down to the belt beam and finally to the main body of the door. The innermost layers were laid directly
into the tool using the end-of-part scribe lines and the resin seal as visual guides. The glass fiber
braids were pulled over the header foam cores. The braided tubes were then pulled taut and stapled
directly to the core.
The figure below shows the lay-up of the fabric into the door header. The cast epoxy tooling
compound models of the foam core were used to help shape the fabric into the header of the tool.
After the initial shaping, the cast models were replaced with the actual foam core headers with their
braided overwrapping.
Resin Seal
B-pillar
inner
1.
2.
A-pillar inner
Uni-carbon on
Balsa beam
Foam core
Tooling model
3.
4.
Figure 13: Preform assembly of the door header.
The inner belt beam was laid into the tool next. An outer layer of glass fiber fabric was placed in the
tool first followed by the four layers of uni-directional carbon fiber. The balsa core was then added
and the final four plies of uni-directional carbon fiber were laid on top. After completing the inner belt
beam, the aluminum tooling insert was positioned into the tool and tightened into position with three
bolts coming from the bottom side of the tool.
13
Although most of the fabric patterns could be formed directly in the molding tool, the main door inner
with its high, U-shaped wall needed a special performing tool. A single-sided tool with hand clamps
along the edges was used (see Figure 14). The fabric pattern was positioned on the forming tool and
a thermosetting binder (with blue dye for visibility) was sprayed into regions of sharp bends. A sheet
of nylon vacuum bag film was then draped over the forming tool and sealed to the edges with yellow
rubber sealing tape. The hand clamps were used to hold the fabric down at several strategic points
along the edges. A vacuum was then drawn to pull the film down tightly over the fabric and press it
against the form. With the vaccum holding the fabric to the form, an infra-red lamp was lowered over
the performing tool to cure the thermoset binder and lock the fabric into shape.
14
distributed around the bottom of the tool. The degassed resin was then poured into plastic separatory
funnels (with stopcock on the bottom) and connected to the three resin inlet positions via a short
length of flexible tubing. Finally, the resin inlet valves were opened to evacuate the flexible tubing
leading to the funnel stopcocks.
At this point, the data acquisition program was started thereby defining zero time. A few seconds
later, the funnel stopcocks were opened and the resin drawn into the tool by the vacuum. After the
resin flow was started, the vacuum vents on the underside of the tool were constantly monitored for
resin. A 30 cm length of transparent flexible tubing was attached to the vent ports to facilitate
observation. After resin was observed in a vent line, the vent line was sealed off with a pinch clamp
to prevent more than approximately 10 ml of resin from exiting the tool. The vents in the header
region of the door were sealed off in rapid succession between 3 and 5 minutes after the start of fill.
The remainder of the vents were sealed off later at times ranging up to 15 minutes.
The progress of the resin infusion could be partly assessed by observing the signals from the in-mold
dielectric sensors as well as direct observation of resin in the vent lines. Upon the resin arrival, the
apparent dielectric loss signal from the sensor would jump from around 0.5 up to a value greater than
0.9 . The dielectric loss signals from the 8 in-mold sensors are plotted in Figure 15. As expected, the
sensor in the door header detected resin arrival first, approximately 2 minutes after the start of fill.
Resin arrival at the remaining locations was detected later depending on location but usually within 5
to 10 minutes of the starting time.
Sensor #
1.7
1
2
3
4
5
6
7
8
1.5
Loss Factor
1.3
Transition Line
from dry to wet sensor
1.1
0.9
0.7
0.5
0.3
0
10
Time (minutes)
Figure 15: Plot of loss factor vs. time for in-mold sensors showing resin arrival
After all of the vacuum vent lines were sealed with the pinch clamps, the resin filled tool was rolled
back into the autoclave for cure. Vacuum was maintained on the trapped volume between the
primary seal and the resin seal in order to keep the tool closed. After sealing the autoclave door, air
pressure was allowed to build in the chamber effectively squeezing the two mold halves together
while heating the tool up to the desired cure temperature.
T h e m o ld g a p s a t th e fo u r co rn e rs w e re m e a su re d b y th e L V D T s. F ig u re 1 6 p re se n ts a p lo t o f th e
mold gaps versus time for a typical resin infusion molding. During the resin infusion stage, the
15
Autoclave Pressure
Applied at 38 minutes
0.1
Gap (mils)
0.08
0.06
Tool initially open
more than .070"
0.04
Target mold gap
0.02
0
0
10
20
30
40
50
60
70
80
90
100
-0.02
Time (minutes)
Figure 16: Mold gaps at the corners during composite cure cycle
Due to the combination of a thick composite tool and poor heat transfer within the autoclave, the cure
cycle lasted several hours. To insure complete cure, a minimum of two hours at a mold temperature
greater than 80C was desired. The temperature profiles inside the autoclave and at two distinct
locations within the tool are plotted against time in Figure 17. The autoclave heaters were turned off
after approximately 10 hours and the mold was allowed to slowly cool within the autoclave until the
following morning.
95
90
Tool #1
Tool #2
Air #1
Air #2
Air #3
85
Temperature (C)
80
75
70
65
60
55
50
0
120
240
360
480
600
Time (minutes)
16
720
840
960
1080
1200
Figure 17: Temperatures during the cure cycle of the composite door molding
The major difference in the three composite doors molded was the mass of resin that was allowed to
b e d ra w n in to th e m o ld . C a lcu la tio n s b a se d o n th e p re fo rm s m a ss a n d th e ta rg e t fib e r vo lu m e
fraction gave an ideal target of 2550 g of resin, however this target did not include resin needed to fill
the perimeter resin runner or the resin flow into the vacuum lines. Probably the biggest uncertainty in
the resin calculations was the volumetric shrinkage of the foam cores. Although the foam core tools
were originally fabricated from a model core based on precise composite skin thickness, the molded
foam cores exhibited significant cure shrinkage. For this reason, a large excess of resin was
prepared for the first molding trial. With the first molding, the objective was to mold a composite door
with no dry spots, therefore additional resin was added until all of the vacuum ports had been filled.
For the first trial, a total of 5400 g of resin was infused into the tool. The overall appearance of the
resulting part was excellent (see Figure 18), however the door was approximately 1.7 kg overweight.
17
were still dry (no resin) at the culmination of the infusion process. Complete preform wet-out was
dependent on obtaining additional resin flow by squeezing the tool halves together in the autoclave
with air pressure. Fortunately, sufficient additional resin flow was generated and a completely wet-out
part was molded.
After trimming, the mass of the second composite door was 10.0 kg. Although this mass still
exceeded the design target of 9.2 kg, it represented a 0.9 kg mass reduction from the first composite
door. Because the second composite door was still completely wet-out, it was decided to further
reduce the mass of resin for the third molding trial.
For the third molding trial, a total of 4000 g of resin was used. As expected, even more of the vacuum
vents were still dry at the end of the resin infusion step. After rolling the tool into the autoclave and
applying air pressure, the tool halves were once again squeezed together forcing additional resin flow.
Although all regions of the third door appeared to be wet-out, the class A outer surface appeared to
b e re sin sta rve d in se ve ra l lo ca l re g io n s ( S e e F ig u re 1 9 ). D u e to th e a p p e a ra n ce o f th e th ird d o o rs
outer surface, it was decided that 4000 g was the lower limit for resin infusion. The mass of the
trimmed third composite door was 9.8 kg which is 0.6 kg greater than the design target but within
acceptable limits.
Figure 19: Third molded composite door showing resin starvation on outer skin
18
Header Rigidity
For the all of the header rigidity testing, the latch side flange was clamped at three locations to one
test stanchion and the door hinges were fixed to the opposite test stanchion. Load was introduced at
two different locations on the window frame header, the front location located along the A-pillar and
the rear location located near the top of the B-pillar. Inboard and outboard load deflection curves
were measured at each test location. A maximum load of 360 kN was applied at the front location
while a maximum load of 540 kN was applied to the rear location. Photos illustrating the test set-up
are provided in Figure 20. Maximum deflections and permanent set values for each test were taken
from the load vs. deflection data collected.
a.
b.
Figure 20: Photos of front (a) and rear (b) loading points for upper frame rigidity tests
For comparison to the composite door, test data from the steel door were obtained. The maximum
deflection and permanent set data from both the production steel and the composite front side doors
are given in Table I below.
Table I: Header Rigidity Test Result Comparison
Rear Out
Rear In
Front Out
Front In
Deflections
SSTS max
Steel Door
19 mm
15.3 mm
19 mm
16.0 mm
19 mm
7.7 mm
19 mm
8.2 mm
Composite
15.7 mm
19.1 mm
11.0 mm
11.5 mm
19
Permanent Set
SSTS max
Steel Door
1.5 mm
3.1 mm
1.5 mm
2.3 mm
1.5 mm
0.4 mm
1.5 mm
0.4 mm
Composite
0.1 mm
1.6 mm
0.4 mm
0.7 mm
Although the composite door was more compliant than its steel counterpart, the measured deflections
were either at or below the designated maximum allowables. The permanent set measurements at
the rear header location were much lower for the composite door than for the steel door. In fact the
steel door permanent set values exceed the target.
Beltline Stiffness
GM standard procedures were followed for measuring the inner and outer beltline stiffness of the
composite door. The composite door was fixed to stanchions by its latch and hinges only. No clamps
on the latch side flange were used. A hemi-spherical load nose was used to introduce the load at the
midpoint of the beltline. The load was directed outboard on the inner belt beam and inboard on the
outer belt beam. For these tests, deflection is defined by the closing of gap between inner and outer
window flanges. Figure 21 provides a photo of the experimental set-up.
Figure 21: Beltline Rigidity test set-up for inner belt beam of composite door
The measured deflections and permanent set values for both the composite door and its steel
counterpart are given in Table II,
Test
Beltline Outer
Beltline Inner
SSTS
7.0 mm
3.0 mm
Across the beltline, the composite door is significantly stiffer than the steel baseline part, especially on
the inner belt. Whereas, the composite door easily meets the target for deflection on the inner belt,
20
the steel door deflects twice the allowable distance. Permanent set values for the composite door
were also very low-- less than half of those of the steel door.
Unfortunately, the FEA predictions for deflection were again not very accurate. For the beltline outer,
the structural analysis calculated 6.5 mm vs the measured value of 3.1 mm. For the beltline inner, the
calculated value was 3.6 mm as compared to the test value of 1.7 mm. The structural analysis
appears to have overestimated the beltline deflections by more than a factor of two. A review of the
NASTRAN analysis deck did not identify the source of the overestimation. Possibly, the deflection
values were taken directly from the distance traveled by a single node rather than the desired
difference between opposing nodes on either side of the window pocket. Without this overestimation
of beltline deflection, the design could have used thinner composite skins in the inner beltbeam, or
perhaps even used glass fiber instead of carbon fiber and still met the design targets.
Torsional Rigidity
GM standard procedure was followed for measuring the upper and lower torsional rigidity of the
composite door. For these tests, the composite door was fixed by the latch and hinges to stanchions
and the load was introduced via a hemi-spherical load nose. Deflection was measured at a point
directly opposite the load nose in line with the applied load. Figure 22 provides a photo of the
experimental set-up.
21
Although the composite door was significantly more compliant than the steel door in torsion, it came
close to meeting the design requirements. The deflection during the lower torsional test exceeded
the target by only 0.6 mm. The structural analysis had predicted that the composite door would
deflect 6.1 mm and 6.9 mm in the upper and lower torsional tests respectively. Actual deflections
were a bit higher than predicted but the predictions were within a reasonable tolerance.
Vertical Rigidity
GM standard procedure was followed for measuring the vertical rigidity of the composite door on a
test fixture. In this test, the door is supported only by its hinges to a fixed test stanchion and a 900 N
downward load is applied at the latch. Deflection and permanent set are measured at the bottom of
the door directly beneath the latch assembly. A photo of the test set-up is provided in Figure 23.
22
while the composite door met the target. In this particular loading case, the single piece design of the
composite door with the hinges adhesively bonded to a thick composite sandwich section proved to
be very effective.
References
1. P . T ho rto n, R . Jeryan, C rash E nergy M anagem ent in C o m po site A uto m o tive S tructures, Int. J. Im pact E ngng, 1988, & ,
167-180
2. M . P hilipps,L . P atberg, R . D ittm ann, H . A dam , S tructural Analysis and Testing of Composites in Automotive
C rashw o rthiness A pplicatio n, S A E paper # 981140, S A E Internatio nal C o ngress and E xpo sitio n, 1998
3. E . C hirw a, L . O L arey, S . M uhtaseb, S ide Im pact P ro tectio n: the U se o f 3 -D Reinforced Composite P anels,
International Journal of Crash, 1996,1(3), 315-330
4. L . P atberg, M . P hilipps, R . D ittm ann, F ibre-reinforced composites in the car side structure, Proc Instn Mech Engrs ,
1999, vol.213 part D, 417-423
5. S . C heo n, D . L ee, K . Jeo ng, Composite side-door impact beams fo r passenger cars, Composite Structures, 1997 38(14), 229-239
6. H. Adam, L. Patberg, M. Phillips, R. D ittm ann, T esting o f N ew C o m po site S ide D o o r C o ncepts, S A E P aper # 980859,
SAE International Congrss and Exposition, 1998
23