STRESS - STRENGTH

INTERFERENCE

PROBABILITY OF FAILURE
John Toksoy
Cummins Inc.
November 7, 2002

1
 QUIZ 1!

DESIGN = Strength
MARGIN Design Stress

§ Which one of the 50

materials would 45
40
you choose for a

Fatigue Strength (ksi)

35

higher fatigue 30
25
strength? 20

§ Would you pay a

15
10

premium for 5
0
material “D” over A B C D

material “A”?
Material Types

Stress & Strength Interference 2

 VARIATION IS THE ENEMY…
§ Would the additional information like the variation in
strength change your answers?
§ Is material “C” 50
better then “B”? 45 $2.6 / kg

§ How would you

VARIATION
40 $2.2 / kg

Fatigue Strength (ksi)

quantify the 35

difference 30

between these 25 $1.5 / kg $1.7 / kg

materials? 20

§ Is design margin
15
10
alone a good 5 AVE

criteria to select a 0
material? A B C D
Material Types

Stress & Strength Interference 3

 Early Practice - Design Margin
§ Early practice in stress-strength relationship
dealt almost entirely along the lines of
design margin. Factor of safety!

§ Design margin approach use the mean value

of stress & strength ignoring the natural
scatter that each may possess.
§ Utilization of design margin is justified when
n It is based on considerable experience
n Component design changes are not too different
than the existing design.
Ø Geometry, processing, function

Stress & Strength Interference 4

 Recent Practice – Probability of Failure
§ The variation in stress and strength results
in a statistical distribution and a natural
scatter in these variables.
§ When these two distributions interfere,
that is when stress becomes higher than
strength, failure results.
§ Means of expressing these distributions in
a practical engineering sense and means of
calculating the resulting interference
(probability of failure) is the heart of this
seminar.
Stress & Strength Interference 5
 Outline
§ Definition of failure - Unreliability
§ Reliability in simple terms
§ Part Strength & Stress
§ Normal Distribution
§ Probability of failure
§ Reliability quantified
§ Example

Stress & Strength Interference 6

 Definition of Failure - Unreliability
§ Failure
n The inability to meet customer required function

§ Failure Mode
n The manner in which the item fails, not the
display of the failure
Ø It is very important to identify the root cause
and separate the failure modes

§ Mission Disabling Failure

n When mission is interrupted such that the item
cannot or should not be operated until repair
occurs

Stress & Strength Interference 7

 How Do Customers Talk About Reliability

§ “ A system that does what I want

(function), when and where I want to use
it (conditions), for as long as I want to use
it (time) ”
§ “ No surprises - no unscheduled downtime”
§ “ Get me up and running quickly when
failures occur ”
n This is as important as not having
a failure in the first place.

Stress & Strength Interference 8

 Definition of Reliability
§ “The ability of an item to perform a
required function under stated conditions
for a stated period of time”
§ “Quality over time”
§ “It is also defined and/or measured as the
probability that an item performs…”
n It is with this definition that
we can quantify Reliability.

Stress & Strength Interference 9

 Reliability Measures
§ Cumulative failure rate at a stated time
n Repairs Per Hundred (RPH) within warranty
period

§ Instantaneous failure rate or hazard rate

n Failure rate per hour, month, mile in service

§ Time it takes to fail

n Mean Time Between Failures (MTBF)
n B10 life – time at which 10% of the items have
failed

§ Probability of failure
Stress & Strength Interference 10
 Causes of Different Failure Types
§ Infant Mortality
n Manufacturing & assembly issues
n Quality control issues
n Supplier Issues
§ Random Failures Titanic

n Interference of inherent strength and

experienced stress during operation
n Misapplication and/or abuse
§ Wear Out Failures
n Fatigue, wear and part deterioration
n Preventive maintenance issues
n Service issues
Stress & Strength Interference 11
 Reliability Bathtub Curve

30
Unit-to-Unit Deterioration
External Noise Factors
Noise Factors Noise Factors
Failure Rate

20

10 Warranty
Period

0
0 5 Time 10 15 20

Stress & Strength Interference 12

 Outline
§ Definition of failure - Unreliability
§ Reliability in simple terms
§ Part Strength & Stress
§ Normal Distribution
§ Probability of failure
§ Reliability quantified
§ Example

Stress & Strength Interference 13

 Stress Acting on a Part
§ The operating stress imposed on the part
is random.
§ Stress acting on a part changes with
n Time
Ø Climbing a hill with full, part & no load
Ø City vs. high way driving
n Ambient conditions (temperature, pressure)
n Part to part – variation in geometry
n User to user
Ø 18 year old driving dad’s Porsche

Stress & Strength Interference 14

 Part Strength
§ A given part has certain physical properties
which, if exceeded, will cause failure.
§ Part strength is a random variable that can
be represented by a statistical distribution.
§ A parts strength varies from
n Lot to lot – Difference in chemical composition
n Manufacturer to manufacturer – Process
n Ambient conditions
Ø Change in material properties with
temperature and humidity
Ø At low temperatures parts may shrink and
reduce sealing pressure (Space shuttle failure)

Stress & Strength Interference 15

 Stress & Strength Distribution
§ Random variation of stress and strength can be
expressed with different distributions
n Normal, Log Normal, Exponential, Weibull

§ Both stress and strength can be represented with

any combination of the above distributions.
n Normal – Normal, Normal – Weibull,
Log Normal – Exponential, Weibull – Weibull

§ For the purpose of this seminar we will assume a

normal distribution for both stress and strength
n Math behind normal-normal distribution is simpler

Stress & Strength Interference 16

 Outline
§ Definition of failure - Unreliability
§ Reliability in simple terms
§ Part Strength & Stress
§ Normal Distribution
n Probability Density Function

§ Probability of failure
§ Reliability quantified
§ Example
Stress & Strength Interference 17
 Normal Distribution
IQ IQ
(Intervals of size 20) (Intervals of size 20)
40 0.02
0.39 0.0195

30

Area=

Density
Percent

0.0195*20
20 0.01
=0.39

10

0 0.00

55 75 95 115 135 55 75 95 115 135

Percent Histogram
IQ
Probability = Area
IQ
of Rectangle

IQ IQ
(Intervals of size 10) (Intervals of size 5)
0.03
0.02

0.02
Density

0.01 Density
0.01

0.00 0.00

55 65 75 85 95 105 115 125 135 50 60 70 80 90 100 110 120 130 140

IQ IQ
Decrease interval size Decrease interval size more
Stress & Strength Interference 18
 Normal Distribution Characteristics
§ Symmetric, bell-shaped IQ

curve.
(Intervals of size 5)
0.03

§ Shape of curve depends 0.02

Density
on population mean µ and 0.01

standard deviation σ.
§ Center of distribution is µ.
0.00

50 60 70 80 90 100 110 120 130 140

IQ

§ Spread is determined by standard deviation σ.

§ Most values fall around the mean, but some values
are smaller and some are larger.

Stress & Strength Interference 19

 Normal Distribution:
Effect of Mean & Standard Deviation
§ The mean and standard deviation affect the
shape of the normal distribution

Smaller Larger
standard standard
Stress & Strength Interference
deviation deviation 20
 Probability Density Function (PDF)
§ The curve describes Bell-shaped curve

probability of getting 0.08

Mean = 70 SD = 5

any range of values

0.07

0.06

0.05

P(X > 120), P(X>75),

Density
n 0.04
Mean = 70 SD = 10

P(65 > X > 75)

0.03

0.02

§ Probability is the area

0.01

0.00

40 50 60 70 80 90 100

under the curve Grades

Probability student scores higher than 75?

§ Area under the whole 0.08

curve is 1
0.07

0.06
0.05

§ Probability of getting Density

0.04 P(X > 75)

0.03

specific number is 0
0.02

0.01

0.00

n P(X=120) = 0 55 60 65 70 75 80 85
Grades

Stress & Strength Interference 21

 Probability = Area under curve

§ Probability of all grades

0.08

0.07

0.06

falling between 65 & 70.

0.05

Density
0.04 P(65 < X < 70)

P (65 < X < 70)

0.03
n 0.02
0.01

0.00

55 60 65 70 75 80 85
Grades

§ Probability of all grades

falling below 65 0.08

0.07

n Is always a function of the 0.06

instructor!
0.05

Density 0.04
0.03

n Has nothing to do with 0.02 P(X < 65)

how much you study!

0.01

0.00

55 65 75 85
Grades

Stress & Strength Interference 22

 Probability = Area under curve
§ Integral? Calculus?! I am kidding, right?
§ But somebody did all the hard work for us!
§ We just need a table of probabilities for
every possible normal distribution.
§ But there are an infinite number of normal
distributions (one for each µ and σ)!!
§ Solution is to “standardize”.

Stress & Strength Interference 23

 Standard Normal Curve
§ Take a normally distributed Standard Normal Curve
value X
§
0.4
Subtract its mean µ from it
§
0.3
Divide by its standard

Density
deviation σ. 0.2

§
Tail probability
Call the resulting value Z. 0.1 P(Z > z)

(X - µ) 0.0

Z= σ -4 -3 -2 -1 0
Z
1 2 3 4

§ Z is called the standard normal. Its mean µ is 0 and

standard deviation σ is 1.
§ Probability of failure, Unreliability is calculated from
standardized normal distribution as Failure=P(Z).
§ Reliability = (1 – Unreliability)
Stress & Strength Interference 24
 Outline
§ Definition of failure - Unreliability
§ Reliability in simple terms
§ Part Strength & Stress
§ Normal Distribution
§ Probability of failure
§ Reliability quantified
§ Example

Stress & Strength Interference 25

 Stress & Strength Interference
§ X (Strength) and Y (Stress) are normally
distributed with mean values µx and µy and
variances s 2x and s 2y
§ Define I = X – Y à (Strength – Stress)
n Mean value µI = µx - µy
n Variance s 2I = s 2x + s 2y

§ Normalize function I=(Strength-Stress) so

that standard statistical tables can be used
(I - µI) I – (µX - µY )
Z = σI = s 2x + s 2y
Stress & Strength Interference 26
 Stress & Strength Interference
§ Part stress must be equal or exceed part
strength for failure to occur
n Stress Y >= Strength X
n I (X – Y) =< 0

§ Area under the normalize function where

I = (Strength-Stress) = 0 is consequently
the probability of failure

0 – (µX - µY )
Unreliability = Z =
s 2x + s 2y

Stress & Strength Interference 27

 Interference of
Two Normal Distributions
Part stress must exceed strength for failure to occur

I = Strength - Stress

Stress Strength

Stress & Strength Interference 28

 Outline
§ Definition of failure - Unreliability
§ Reliability in simple terms
§ Part Strength & Stress
§ Normal Distribution
§ Probability of failure
§ Reliability quantified
§ Example

Stress & Strength Interference 29

 Example - Probability of Failure
§ A component has a strength which is normally distributed
with a mean value of 5000 N and standard deviation of 400 N.
The load it has to withstand is also normally distributed with a
mean and standard deviation 3500 N and 600 N. What is the
reliability of this component under the given load application?

0 − (5000 − 3500 )
Z=
400 2 + 600 2
Z = 2.08
Unreliabil ity = 0.0188
Re liability = 1 − 0.0188
= 0.9812

§ Probability of failure is 1.88 out of 100

Stress & Strength Interference 30
 Example – Repairs Per Hundred
§ Unreliability = 0.0188 & Reliability = 0.9812
§ RPH (Repairs Per Hundred) = 100 * Unreliability
= 100 * 0.0188
RPH = 1.9
n 1.9 parts out of 100 will fail
n 19 parts out of 1000 will fail

§ RPH will be 0.4 if the load standard deviation is

reduced to 400N from 600N.

§ Design Margin for both cases is 5000

= 1.42
3500

Stress & Strength Interference 31

 Example – Cost of Robustness
§ If the repair cost of such a failure is $750 and
annual engine build rate is 40,000, How much
premium can the manufacturer pay for reduced
standard deviation - robustness?

Number of Failures =
(1.88 − 0.4 )
× 40,000 = 592 eng/year
100
592 engines * 750 $

Cost per Engine = year engine

= 11.10 engine
$

40,000 engines
year

§ Up to $11.1 per engine can be paid to reduce the load

variability (standard deviation) from 600N to 400N
n Larger crank damper to reduce torsional amplitudes

§ What is the price for 592 happy customers?

Stress & Strength Interference 32