Sie sind auf Seite 1von 14

Failure Analysis of

Automotive Suspension Coil


Springs
This article is available online at www.aist.org for 30 days following publication.

F ollowing the trend of the auto industry to


continuously achieve weight reduction,
coil springs are not exempt. A consequence
(FEA) of stress distributions around typical
failure initiation sites are also presented.

of the weight reduction effort is that spring Introduction


materials are subject to significantly larger A mechanical spring is defined as an elastic
stresses compared to similar designs decades body that has the primary function to deflect
ago. Utilizing a higher strength of steel pos-
sesses both advantages and disadvantages. A
disadvantage of subjecting materials to higher
levels of stress is that a failure can be more This paper discusses automotive suspension coil
catastrophic. As the stress level is increased, springs, their fundamental stress distribution, material
material and manufacturing quality become
more critical. Material cleanliness and decar- characteristics, manufacturing and common failures.
burization that were not major issues decades An in-depth discussion on the parameters influencing
ago now become significant.
the quality of coil springs is also presented.
To assure coil spring integrity, failure analy-
sis of broken coil springs is valuable both
for the short- and long-term agendas of car
manufacturers and parts suppliers. This paper or distort under load, and to return to its
discusses several case studies of suspension original shape when the load is removed. As
spring failures. The failures discussed concern to the design of a coil spring, a common but
insufficient load-carrying capacity, raw mate- simple approach is referenced here.1 The so-
rial defects such as excessive inclusion levels, called un-wound spring, as shown in Figure
and manufacturing defects such as delayed 1, is commonly used. The assumption is that
quench cracking. Also discussed are failures an element of an axially loaded helical spring
due to complex stress conditions and chemi- behaves as a straight bar in pure torsion. The
cally induced failures. Finite element analysis following notations are typically used:

Figure 1

Wound and un-wound coil springs.

Authors
Y. Prawoto, M. Ikeda and S.K. Manville, NHK International Corp., Wixom, Mich. (yunan.prawoto@gmail.com, maiko.ikeda@nhk-intl.
com, sonia.manville@nhk-intl.com); and A. Nishikawa, NHK Spring Co. Ltd., Yokohama, Japan

September 2008   ✦   35
P = applied load, 64 PR 3n 8PD 3n
α = pitch angle, δ = Rϕ = =
τ = shear stress, Gd 4 Gd 4
R = coil radius and (Eq. 3)
d = wire diameter.
The spring rate therefore becomes:
The torsional load component is then cal-
culated as PR cos α, the bending moment as P Gd 4
PR sin α, the shear force as P cos α, and the
k= =
δ 8nD 3
compression force as P sin α. Traditionally,
(Eq. 4)
when the pitch angle is less than 10°, both the
bending stresses and the compression stresses Equation 4 is still commonly used by sus-
are neglected. pension designers to estimate the spring rate.
Assuming that the shear stress distribution As opposed to the uncorrected shear stress in
is linear across the wire cross-section, and PR Equation 1, Wahl2 proposed a corrected shear
cos α = PR, the following should be valid: stress. The uncorrected shear stress neglects
a great many factors which modify the stress
16 PR
τ= distribution in actual helical springs. The cor-
π ⋅ d3 rected shear stress, τa, is obtained by multiply-
(Eq. 1) ing the uncorrected stress with a correction
factor K, which depends on the spring index,
The shear stress here is usually called uncor- c = D/d. Figure 2 shows the corrected shear
rected shear stress. The total length, l, is 2πRn, stress distribution.
where n is the number of active coils. Using Furthermore, by taking x as the distance
the fact that γ = τ/G, it can be rewritten as from the point where the shear stress is
16PR/(π · d3G), and the total angular torsion zero, Wahl proved that the following equation
ϕ becomes: holds:
2 πRn
2γ 32 PR 64 PR 2 n
ϕ= ∫ dx = 4 dx =
πd G Gd 4 τa =
32 xPR 2
( )
d
0
π ⋅ d 4 R − d 2 / 16 R − x
(Eq. 2)
(Eq. 5)
where G is the modulus of rigidity. The total
deflection caused by the angular torsion is: With the introduction of the spring index c,
the maximum shear stress at the inner side of
the coil, where x = d/2 – d2/16R, becomes:
16 PR 4c − 1
Figure 2 τ a1 =
π ⋅ d 3 4c − 4
(Eq. 6)
Given by the neutral surface of a cantilever
of circular cross-section loaded by a force P,
the term 4.92 P/πd2 should be added to obtain
maximum shear stress:
16 PR  4c − 1 0.615 
τ max = +
π ⋅ d 3  4c − 4 c 
(Eq. 7)
Likewise, the minimum shear stress:
16 PR  4c + 1 0.615 
τ min = −
π ⋅ d 3  4c + 4 c 
(Eq. 8)
Equations 7 and 8 are used by design engi-
neers for coil springs when neglecting the
curvature. Further theory can be found in
Reference 2. Also, given the simplified deri-
vation of the equations, the larger the pitch
Uncorrected shear stress vs. corrected shear stress distribution. angle, the more will be the resulting error. In
reality, coil spring makers today use analysis

36   ✦  Iron & Steel Technology


Figure 3

Materials used for coil springs.

procedures that are often confidential, and


therefore cannot be discussed here. These Figure 4
analysis procedures require the design engi-
neer to input the coil diameter, design height,
design load, spring rate, etc. The procedures
determine the optimum shape and dimension
of the coil. FEA is often used to check the
accuracy of the stress distribution design.

Automotive Coil Spring


Materials
Although the history of the gasoline-powered
automobile can be traced back to 1870, when
the first car was made in Austria, the mass
production of cars did not start until the early
1900s both in Germany and in the United
States. The first automotive coil spring was on
the Model-T Ford in 1910, where the suspen-
sion combined the leaf spring and the coil
spring. The earliest coil spring material had
approximately a 500 MPa design stress level.
Coil spring materials have developed to the
point where today it is common to have a coil
spring with a design stress of around 1,200
MPa.
Figure 3 shows the approximate timeline
vs. the design stress level. Design stress lev-
els have increased significantly over the last
couple decades. This increase is expected to
continue as weight-reduction trends also con-
tinue. However, having a higher stress level
increases the sensitivity of defects. A small
defect will consequently result in a more sig- Finite element model.
nificant effect.

September 2008   ✦   37
Figure 5

FEA results can be obtained by creating 3-D


parts of a coil spring and its seats, followed by
meshing the parts with a 3-D solid element.
Finer meshing with higher-order elements will
produce more accurate results.
However, with a higher number of ele-
ments and a nonlinearity due to the contact
between a coil spring and its seats, or between
the coil and itself, each analysis could take
hours to run. While the accuracy of the result
is important, the computational time must be
reasonable to incorporate FEA into coil spring
design. To resolve lengthy computational time
in a solid model, a 3-D beam element is often
Example of compressed coil spring. selected to model a coil spring and seats.3
Since the deformation of a seat under com-
pression is very minimal and can be ignored,
the material properties of seats are set very
Designing Coil Spring Using high to act as a rigid body. Contact between
Finite Element Analysis a coil and seat, or between the coil and itself,
A primary reason to use FEA in coil design is detected by gap elements. A typical FEA
is the ability to reduce error caused by the model is shown in Figure 4.
simplification of equations, mainly concern- In-house software was developed to model
ing the pitch angle. An FEA-based design coil/seats (pre-processor) and to display the
begins with the selection of the element type, analysis results (post-processor). The simula-
how the model should be constructed, how tion is performed by commercial FEM soft-
accurate the results should be, and how fast ware, MARC or ABAQUS.6 In general, a new
the model should be run. The most accurate FEA model is created each time a coil/seat
profile is modified, and the simulation will be
repeated until all requirements are satisfied.
Figure 6 Sometimes it may require numerous simula-
tions to achieve a desired coil spring design;
therefore, a faster simulation is necessary.
This FEA model may not generate accu-
rate analysis results due to model simplifica-
tion and assumptions, but will generate the
approximated results and significantly reduce
the amount of design time. The determina-
tion of whether or not the coil design using
FEA is good enough is an important issue.
The determination is made by comparison
with past analysis results or with actual experi-
mental results. FEA is also used in the evalua-
tion of the actual product and re-designing in
case of a testing failure.
Figure 5 shows a condition of the simplified
modeling both in terms of the element type
and in its condition, static in the modeling vs.
dynamic in reality.
The finite element modeling of a coil spring
is performed after a simple calculation using
Wahl’s equation and confirmation after both
manufacturing and testing, as shown in Figure
6. In this paper, the FEA modeling is also used
to simulate how various defects affect local
stress distributions.

Major Imperfections in Coil


Manufacturing
Raw material selection is always the most
Simple flowchart of coil spring design. important decision in obtaining the best qual-
ity of any product, including coil springs. The

38   ✦  Iron & Steel Technology


Figure 7

(a) (b) (c)

Typical defects in raw materials: inclusion (a), inappropriate microstructure (b) and decarburization (c).

selection of the raw material usually includes performed. Heat treatment–related defects
the enforcement of cleanliness, microstruc- are another major cause of a coil failing early.
ture, and decarburization inspection. Figure 7a These defects include, but are not limited to,
shows a typical raw material defect in the form quench cracking, insufficient tempering and
of an inclusion; also shown is a microstructure over-tempering.
matrix defect (b) and decarburization (c). After tempering, the coil spring is shot
Other sources of defects include improper peened. The shot peening process is benefi-
heating prior to coiling. The control of the cial for two reasons: it cleans the surface of
prior-austenite grain size is an important step defects and scale caused by quenching, and
in coil manufacturing. Figure 8 shows the introduces compressive residual stresses at the
difference between a small grain size and a surface. Figure 9 shows a typical residual stress
large grain size. This example was taken from distribution formed by shot peening. When a
identical materials processed with different load is applied to a coil spring, the net stress
parameters. Although not reflected by other is the superposition of the beneficial residual
mechanical properties, larger prior-austenite stress due to shot peening and the applied
grain size has proved to be less advantageous stress.
to fatigue life than that of a small size. Some Shot peening is followed by setting, which
argue that this is due to the fewer number usually does not have a detrimental effect on
of the grain boundaries passed during crack the coil. Coating is typically the last step of
propagation.4 coil-making. The process of coating consists
Once the raw material is heated properly, of two major steps: pre-treatment and coat-
the coil is then formed. Physical defects due ing application. The main ingredient in a
to coiling sometimes cause the coil to fail pre-treatment is usually zinc, which works as
early. Following coil formation, a heat treat- a sacrificial anode to protect the steel. After
ment process of quenching and tempering is pre-treatment, either a powder coat with spray

Figure 8

Identical raw materials heated with different heating conditions.

September 2008   ✦   39
Figure 9

Typical residual stress distribution caused by shot peening.

application or e-coat is applied, followed by thin or heavy, in addition to composition and


baking to cure the coating. The combination shape. Type A is sulfide-type with a boundary
of the pre-treatment and baking quality typi- of thin and heavy classification of 4µ, Type B is
cally determines the corrosion resistance of aluminate-type with a boundary of 9µ, Type C
the coating, which in turn determines the cor- is silicate-type with a boundary of 5µ, and Type
rosion resistance of the coil. D is globular oxide with a boundary of 8µ. It
is also worth noting that thin inclusions rarely
Failure Analysis Case Studies cause a coil spring to fail early.
Raw Materials Defect — A typical raw material Another defect that is not detectable once
defect is the existence of a foreign material the raw material becomes a product is an
inside the steel matrix, such as nonmetallic excessive constituent of bainite. The raw mate-
inclusions. Figures 10a and b show the frac- rial that is intended to have ferrite pearlite
ture surface and the SEM fractograph, as well microstructure sometimes has large amount
as the EDS spectrum of an inclusion located of bainite. Bainite can potentially cause crack-
~1 mm below the surface. This particular coil ing during cold processing.
failed early, initiating at the defect, despite all
other parameters being normal. Surface Imperfections — Surface imperfec-
In general, there are two types of foreign tions can occur as small hardening cracks, tool
materials that can become trapped inside marks, scale embedded to the base material
the steel solution: large imperfections such during cold drawing, or surface flaws inherited
as spinels, and smaller imperfections such by the raw material. Figure 11 shows a surface
as inclusions that are caused by alloying ele- flaw deep enough to cause a coil spring to fail
ments. ASTM differentiates inclusion types as early; this surface imperfection was inherited

Figure 10a

Fracture surface of a coil that failed early due to an inclusion (left), and its SEM appearance (right).

40   ✦  Iron & Steel Technology


Figure 10b

EDS spectrum mapping of the inclusion.

from the raw material. This type of defect can structure. This defect usually occurs when the
occur when the surface flaw detector does not heating system does not operate normally.
function normally. It is usually easy to deter- Again, referring to Figure 13, the coil on the
mine if such a flaw was inherited from the raw left-hand side has a much shorter lifetime
material and is not due to coil manufacturing. than that of the right side.
A pre-existing defect usually has surround- Bainitic formation is another form of
ing decarburization after the raw material is improper heat treatment. Unlike martensite,
heated during coil manufacturing, whereas a
surface defect caused by coil manufacturing is
often not accompanied by decarburization. Figure 11
Poorly shot peened surfaces can also be
classified as surface imperfections. Figure
12 shows a comparison between two differ-
ent coils that failed at similar locations, but
possessed completely different fatigue lives.
On the left side, the surface was poorly shot
peened and therefore exhibited a shorter life.
On the right side, the surface was shot peened
sufficiently and therefore had a longer life.

Improper Heat Treatment — Improper heat


treatment can often be overlooked, since a
temperature difference in heating does not
relate directly to the hardness of the mate-
rial. Extensive evaluations are usually needed
to identify this type of problem. Improper
heat treatment can also result in the micro-
structure becoming pearlite instead of the
required martensite. This latter type of defect
is easier to identify due to the clear difference
in hardness. Figure 13 shows two different Surface defect inherited from raw material.
coils of the same product with varying micro-

September 2008   ✦   41
Figure 12

Surface imperfections due to poor shot peening condition.

bainitic ferrite usually contains only a slight tures.5 Tempering induces the decomposi-
excess of carbon in ferrite solution. Most of tion of the retained austenite into a mixture
the carbon in a transformed sample of bainite of ferrite and carbides. Figure 14 shows the
is in the form of cementite particles, which microstructure of bainite steel.
in turn tend to be coarser than those associ-
ated with tempered martensite. The effects of Corrosion — Corrosion is a more common
tempering are therefore always milder than cause of spring breakage than is usually
is the case when the microstructure is mar- understood by users. However, recent coat-
tensite. Furthermore, bainitic structures are ing technology has reached a point where
usually accompanied by a greater percentage there is now a greater ability to withstand
of retained austenite than martensitic struc- even the harshest stone impacts without

Figure 13

Improperly heat treated sample (left) vs. properly heat treated sample (right).

42   ✦  Iron & Steel Technology


Figure 14

chipping. Figure 15 shows the appearance of


a coil failed due to corrosion. The right side
of the figure depicts the cross-section of the
coil, with the red line illustrating its approxi-
mate original dimension.

Decarburization — Decarburization may be


considered as the least severe offender in the
entire list of defects. Partial decarburization is
usually permissible in spring wire. However,
when a complete, full-ferrite ring around the
circumference is found, the wire is always
subject to rejection. Figure 16a shows a case
where a coil spring failed due to an excessive
decarburized layer. The fractography of the
sample also shows that the first outer layer
consisted of softer layer where almost no stria- Bainitic structure.
tions were visible.

Figure 15

Coil spring that fractured due to corrosion.

Figure 16a

Fracture surface and SEM fractograph of a coil that failed due to excessive decarburization.

September 2008   ✦   43
Figure 16b

Cross-sectional metallograph of the broken coil (left) and micro-hardness profile near the surface (right).

Further evaluation of the longitudinal and result of a global model with coarse meshing.
cross-sections revealed that the sample had Boundary conditions for the sub-model will
significant decarburization. Subsequently, be automatically interpolated from the global
when the micro-hardness near the surface was model solution. As shown in Figure 17, the
profiled, apparent decarburization was also sub-modeling technique was used twice for
found (Figure 16b). this study.
Sub-model 2 was modified to apply to vari-
Analysis ous defects. For meshing, either the quadratic
Procedure — A finite element analysis was brick element (C3D20) or quadratic tetrahe-
performed to check the local stress distribu- dron element (C3D10) was used. For material
tion around a given defect using a typical coil specifications, typical spring steel properties,
spring (Figure 17a). First, the overall stress E = 210 GPa and ν = 0.3, were used, except
distribution was checked without any defect for the decarburized layer (see Table 1).
in the material. Then, at the location where The commercially available FEA software,
the highest stress was found, each defect ABAQUS, was used here to study each stress
was added. Since the size of the defect is distribution.
significantly smaller than the whole model,
a sub-modeling technique6 was used. This FEA Result of Model Without Defect —
technique is used to study a local part of a For comparison with the defect model, sub-
model with refined meshing based on the FEA model 2 was analyzed first without any defects.

Figure 17

(a) Global model

(b) Sub-model 1 (c) Sub-model 2

FEM model of a coil spring and its sub-models.

44   ✦  Iron & Steel Technology


Figure 18

Boundary conditions were interpolated from


the result of sub-model 1 at the inner and
side surfaces of sub-model 2. Figure 18 shows
the Von Mises stress distribution. The highest
stress was found at the outer surface and the
lowest at the inner surface of the wire. The
maximum principal stress was about 1,200
MPa (the highest Von Mises stress was about
1,715 MPa), which matches the stress level of
the global model. The gray area around the
outer edge shows a stress concentration; how-
ever, this stress is ignored, since it is where the
boundary condition was applied.

Defect FEA Models and Results Von Mises stress result of no-defect model.

• Inclusion — A cubic hole was placed about


1 mm below the outer surface; its size was
50 µm (Figure 19a, red dot is the inclusion). mesh was used at the center and a coarser
Instead of using a foreign material for the mesh was used at the outer area (Figure
cubic area, it was left as a hole for simplifica- 19b).
tion. Since a higher stress concentration was The stress distribution is shown in Figure
expected around the inclusion area, a finer 20. As expected, a local stress concentration is

Figure 19

(a) (b)

Part model with inclusion (a) and FEM model with inclusion (b).

Figure 20

Von Mises stress result of inclusion model.

September 2008   ✦   45
Figure 21

(a) (b)

Part model with imperfection (a) and FEM model with imperfection (b).

observed at the inclusion area, and the high-


est Von Mises stress reached was 2,000 MPa,
Figure 22 which was higher than the outer surface stress
level. Stress on other areas, such as the outer
surface, was at the same level as the no-defect
model.
• Imperfection — A model was created based
on a surface imperfection inherited from the
raw material (Figure 21a). A crack (50 µm
wide, 500 µm deep) along the centerline of
the wire was applied to sub-model 2, as shown
in Figure 21b.
The stress distribution is shown in Figure
22. A high stress concentration is observed at
the crack location, and the Von Mises stress
exceeded 4,000 MPa, which is much higher
than the outer surface stress level. Therefore,
the product would likely fail originating at
this location. A stress concentration is also
observed at the vertical edge; however, this
concentration occurred due to the boundary
Von Mises stress result of imperfection model. condition and should be ignored.
• Corrosion — Instead of modeling the
actual corroded part, a simple oval shape

Figure 23

(a) (b)

Part model with corrosion (a) and FEM model with corrosion (b).

46   ✦  Iron & Steel Technology


Table 1 Figure 24

Material Property of Decarburization


Model7
Decarburized
Original (α-Fe values)

Young’s modulus, E [GPa] 210 124.7

Yield stress, σYS [MPa] 1449 359

Poisson’s ratio, ν 0.3 0.3

was removed from the outer surface to sim-


plify the FEA model. Its size is approximately
300 µm deep, 500 µm high and 1 mm wide.
Finer meshing was used around the corrosion
area, since a higher stress concentration was
expected there. Figure 23 shows the model of
the corrosion at the surface of the wire. Von Mises stress result of corrosion model.
The stress distribution is shown in Figure
24. As expected, a local stress concentration
is observed at the bottom edge of the cor- stress; therefore, no plastic deformation was
rosion area. The Von Mises stress was about observed below the decarburized layer.
3,450 MPa, which is, again, much higher than
the outer surface stress level. This high stress Analysis Result Summary — Table 2 shows a
concentration will cause early failure from this summary of analysis results. As expected, a
location. local stress concentration was observed in the
• Decarburization — The decarburization inclusion, imperfection and corrosion defect
model is shown in Figure 25. A softer material models at each defect area, and those stress
property from Table 1 was used for a 0.15-mm values were much higher than the model
layer on the outer surface. The inner-side without any defects. These high stress concen-
material is the same as the original, except the trations will be expected to cause premature
yield stress was specified this time for elastic- failure; hence, the material needs to be free
plastic analysis. of these defects as much as possible.
Analysis results are shown in Figure 26.
The stress level on the decarburized layer Concluding Remarks
reached the yield stress and remained at that Failure analyses of suspension coil springs
value because the material was assumed to were performed and summarized in this
be perfectly plastic, and the plastic defor- paper. Finite element analyses of represen-
mation occurred in the decarburized layer. tative cases were also modeled. Integrating
The rest of the part never reached the yield finite element modeling in metallurgical

Figure 25

(a) (b)

Part model with decarburization (a) and FEM model with decarburization (b).

September 2008   ✦   47
Figure 26

(a) (b)

Von Mises stress result (a) and plastic deformation result (b).

Table 2 Acknowledgments
All the samples used in this paper were provid-
FEA Summary ed by NASCO (NHK of America Suspension
Defect Summary
Components Inc.), Bowling Green, Ky. The
fatigue tests were all performed by the experi-
None No stress concentration. The highest stress was found on the mental team at NHK’s Wixom, Mich., lab.
outer surface. Max. principal stress ≈ 1,200 MPa. Von Mises
stress ≈ 1,715 MPa. No plastic deformation occurred.
References
Inclusion Stress concentration is observed at the inclusion area. Von Mises
stress = 2,069 MPa. Max. principal stress = 1,922 MPa. 1. Taihei Sugano, ed., Design, Manufacture, and
Testing Methods of Springs, Japan Society for Spring
Imperfection Stress concentration is observed at the crack location. Von Mises Research, Nikkan Kogyo Shimbunsha, 2001 (in
stress = 4,195 MPa. Max. principal stress = 2,670 MPa. Japanese).
Corrosion Stress concentration is observed at the bottom edge of corrosion
2. A.M. Wahl, Mechanical Springs, McGraw-Hill,
surface. Von Mises stress = 3,453 MPa. Max. principal stress 1984.
= 3,286 MPa. 3. Masayoshi Simozeki, ed., FEM for Springs,
Japan Society of Spring Engineers, Nikkan Kogyo
Decarburization On decarburized layer, the stress reached the yield point, and a Shimbunsha, 1997 (in Japanese).
plastic deformation occurred.
4. Y. Prawoto, “The Effect of Residual Stress on
Fatigue Crack Propagation,” Journal of Practical
Failure Analysis, ASM International, 2(5), 2002.
5. H. Badhesia, Bainite in Steels, The Institute of
Materials, 1992.
failure analysis synergizes the power of failure 6. ABAQUS User’s Manual, Volume II: Analysis,
analysis into convincing quantitative analysis. ABAQUS Inc, 2006.
This presumably will be the trend in failure 7. W.D. Callister Jr., Materials Science and
analysis. Engineering, John Wiley and Sons Inc. 2003. F

This paper was presented at the 2007 AIST Steel Properties & Applications Conference, Detroit, Mich.,
and published in the Conference Proceedings.

Did you find this article to be of significant relevance to the advancement of steel technology? If so,
please consider nominating it for the AIST Hunt-Kelly Outstanding Paper Award at www.aist.org/huntkelly.

48   ✦  Iron & Steel Technology

Das könnte Ihnen auch gefallen