Accelerating

Validation
Testing
Thomas N. Riley

Bringing new or improved products to conditions, one can use an accelerated test—
market sooner has long been proven profit- higher torques or forces at faster speeds—to
able for companies. One way to help shorten produce the same accumulated damage, there-
the time-to-market is to accelerate valida- by completing the test in a much shorter time.
tion testing. That is, shorten the test time
required to validate a new or improved prod- s = b•Nm (1)
uct.
Fatigue testing is a common validation Where:
test that often involves long periods of time. s = Stress
However, the time can be shortened by an N = Cycles
often forgotten—or unknown—concept of b = Coefficient
accumulated damage, which works espe- m = Exponent
cially well for mechanical components such
as gears and shafts. For many mechanical components, fatigue
Accumulated damage is based on a curves define allowable forces, torques or
fatigue curve for the material and Miner’s stresses as a function of cycles. Many inves-
rule: damage is the life already expended tigators have shown that power-function equa-
divided by the total life available according tions define fatigue properties for mechanical
to the fatigue curve for the material. If one components, as shown in Equation 1, where
knows the accumulated damage for any test stress is a function of cycles (Ref. 1).
A typical fatigue curve is often called an
S-N curve in which S = Stress and N = Cycles
(Fig. 1). Notice that it declines very sharply at
Stress versus Cycles (S-N) Curve the beginning and trails off at higher cycles.
One could allow the curve to continue forever,
which means there is always a fatigue limit
to any material. Some, however, declare an
endurance limit—either at one million cycles
s = 368,709 N–0.11448 or at ten million cycles. This means there is
Stress (ksi)

infinite life at or below a specific stress level.

Others change the curve at a specific number
of cycles to reflect the realities found from
testing. Whatever your experience, base your
decision on actual test results.
The lower end of the curve is asymptotic
to the vertical axis (approaching the vertical
axis, but never getting there, even going on
Cycles forever), which means, theoretically, there
could be an infinitely high stress if the com-
Figure 1—Typical S-N curve. ponent does not experience any cycles. This is
not physically possible in that to create stress,

52 GEAR January/February 2012 www.geartechnology.com

a load must be applied and at least one cycle.
Material has an ultimate stress level at one
cycle, but it is not infinitely high. The fatigue Stress versus Cycles (S-N) Curve
curve, then, is limited on the lower end to
the ultimate stress. To account for low-cycle
fatigue, the fatigue curve is stopped at 1,000
s = 368,709 N–0.11448
cycles.
It may be visually easier to interpret the
curve if it is plotted on log-log paper to pro-

Stress (ksi)
duce a straight line, along with the limit at
1,000 cycles (Fig. 2).
The S-N curve relates life (cycles) and
stress for a specific material, heat treatment,
special conditioning or any other process
that influences the fatigue property. Be sure
to define the fatigue curve from actual test
1.00E+02 1.00E+03 1.00E+04 1.00E+05 1.00E+06 1.00E+07
results.
Figure 3 is an exaggerated S-N curve that Cycles
shows Miner’s rule (Ref. 2)—linear damage Figure 2—Typical S-N curve on log-log graph.
accumulation—for calculating and accumu-
lating damage. The principle may be sim-
ple—damage is cycles used divided by cycles
allowable—but it is very powerful. Of course,
there are other damage-accumulation tech- Cycles Used N
Damage = = 1used
niques that readers could use if they wish. But Allowable Cycles N1allow
Miner’s rule will be used in this article.
If one tested a component to the stress
level and number of cycles of point P1 in
Stress (s)

Figure 3, it might take longer than desired; but

one can shorten the time by testing at a higher
stress.
Figure 4 illustrates the concept of acceler-
ated testing. A second point (P2) accumulates
the same amount of damage as P1, meaning
that P2 must be located on a curve that runs
through P1 and be parallel to the original S-N
N1used N1allow
curve.
The proportion of life used (damage) is the Cycles (N)
same for the two points shown in Figure 4; it
just may not visually look that way because Figure 3—Exaggerated S-N curve to illustrate damage.
the S-N curve (straight line) is drawn on log-
log coordinates.
Perhaps an example with numbers will point. But what if the duty cycle has many
improve clarity. The S-N curve’s equation in points? The same procedure is followed as
Figures 1 and 2 will be used for the example. shown in Figure 5.
Let us suppose the duty cycle requires There are five stress levels in Figure 5, so
200,000 cycles at a stress level of 91 ksi. To there are five different numbers of cycles to
complete the test in half the time (100,000 accumulate for a total of 200,000 cycles. The
cycles), the stress must be at 99 ksi to accumu- accelerated test in Figure 5 shows that testing
late the same amount of damage: for 100,000 cycles at 95 ksi accumulates the
same amount of damage, which means the
s(ksi) = 368,709 (200,000)–0.11448 = 91 ksi test time is reduced by half. Using previous
s(ksi) = 368,709 (100,000)–0.11448 = 99 ksi (2) equations for damage, stress (S) and cycles (N)
and then re-arranging them as needed, one can
The previous example is simple—i.e., one continued

www.geartechnology.com January/February 2012 GEAR 53

 make the calculations shown in Figure 5. more common duty cycle measurements—i.e.,
One can use two more relationships that torque (or force), speed and time.
are shown in Figure 6. Stress is proportional Equation 3 shows the use of torque, geom-
to a geometry factor, and the number of cycles etry, speed and time to replace stress and
equals speed X time. Although torque is used cycles (also shown in Fig. 6) within the fatigue
in Figure 6, force could just as easily have curve equation, where:
been used. To move from one point to another
on any curve one must use the curve’s equa- s = b•Nm
tion; it shows how the derivation creates a Torque • Geometry Factor = b(Speed • Time)m
simple method to go from one point to anoth- (3)
er on the same curve, and how one may use
If one wants to accelerate a gear test, for
example, then one knows that the gear tooth
stress equals the tooth force from the torque
Cycles Used N
Damage = = 1used = N2used X a geometry factor. That relationship is well
Allowable Cycles N1allow N2allow documented in technical literature such as
AGMA documents (Ref. 3).
The next step is to change each duty cycle
point (torque 1, speed 1 and time 1) to the
test condition (torque 2, speed 2 and time 2).
Stress (s)

The derivation to achieve that change is in

Equation 4:

T1 •J = b(S1 • t1)m
T2 •J = b(S2 • t2)m

N2used N1used N1allow N1allow T1 S1 • t1  m

=
T2  S2 • t2 
1 1
–  – 
Cycles (N) S1 • t1 • T1  m = S2 • t2 • T2 m
(4)

Figure 4—Exaggerated S-N curve with two points.

Cycles
Damage =
Allowable Cycles

s = 368,709 N–0.11448

1
 –
 s 0.11448
N= 
368,709

Cycles
Allowable Cycles =
Total Damage

s = 368,709 (Allowable Cycles)–0.11448

Figure 5—One accelerated test point replaces complex duty cycle.

54 GEAR January/February 2012 www.geartechnology.com

 Where:
S = Speed (rpm)
t = time (hours)
T = Torque (lb-ft)
J = Geometry Factor
m = Exponent

Stress (s) = Torque * Geometry

When each duty cycle point is converted to
the test condition, the test condition duplicates
the same damage as entire duty cycle.
Figure 7 shows an example of a gear duty
cycle converted to a gear test cycle. If the
gears were tested according to the duty cycle,
it would take 36,000 hours; but an accelerated
test would only take 22.6 hours—a fantastic
savings in test time. N2used N1used N1allow N1allow
Accelerated testing must be based on well-
known fatigue (S-N) curves and test experi- Cycles (N) = Speed * Time
ence to assure that new failure modes would
not be introduced. Without such knowledge
and experience, one can introduce failure Figure 6—Moving from one point to another.
modes that would not appear in real applica-
tions or pass tests that are not related to the
material’s real fatigue life. However, properly
used, accelerated testing can save a company s = 368,709 N–0.11448
time and money in getting products to market
much sooner. And that means bringing in new 1
revenue much sooner.
 –
 s 0.11448
N= 
References: 368,709
1. Shigley, J. E. and L.F. Mitchell. Mechanical
Engineering Design, Fourth Edition, McGraw- 1 1
–  – 
Hill, New York, 1983, Section 7–4, “High  m  m
S1 · t1 · T1 = S2 · t2 · T2
Cycle Fatigue.”
– 1  – 1 
T1

2. Juvinall, R.C. and K.M. Marshek. S1 · t1 · T1
 m
S1 · t1
 m
=
S2 · t2  T2
Fundamentals of Machine Component Design, t2 =

– 1 
 m
Fourth Edition, John Wiley and Sons, Inc., S2 · t2
Section 8.12 “Fatigue Life Prediction with
Randomly Varying Loads.”

3. ANSI/AGMA 2001–D04. Fundamental

Rating Factors and Calculations Methods for
Involute Spur and Helical Gear Teeth.
Figure 7—Example of duty cycle conversion to test condition.

Thomas N. Riley is chief engineer/CoE component analysis for Eaton Corporation Vehicle
Group’s truck components operation in Galesburg, Michigan. He has spent many years
designing and testing gears, shafts, bearings, enclosures and other components for geared
systems. In his career Riley has worked in several industries, including open-pit mining equip-
ment, aerospace and transportation. He also holds several patents as well as an Engineer of
the Year award from Eaton.

www.geartechnology.com January/February 2012 GEAR 55