Practical Finite Element Analysis

Fatigue Analysis 17
17.1 Why Fatigue Analysis?
 Fatigue accounts for 90 % of service failures.

 Manufacturers give warranty on the components (in terms of kms or years). Fatigue analysis
helps in predicting life of the component in design phase itself. Static or dynamic analysis
can tell us about stress, displacement, acceleration etc but not how long the component will
survive.

 Many a times static or dynamic analysis predicts location of failure not matching with lab
test or field failure and then analyst keep on thinking whether some thing is wrong with
boundary conditions or material properties or geometry of the component. But when
fatigue analysis is carried out using same static or dynamic results, it reveals correct location
of failure.

 Failures or crack usually initiates at surface. Life of the component depends on surface
condition (like grinding, induction hardening, shot preening etc). Static or dynamic analysis
can not take in to account these details while fatigue can.

17.2 Static, Dynamic and Fatigue Analysis Comparison

Static Dynamic Fatigue
Force is static (dead wt.) Force vary with respect to time / Force vary with respect to time /
frequency frequency
Displacement is static Displacement is function of time / Calculations for life of the structure when
frequency subjected to repetitive load
No velocity & acceleration Velocity & acceleration develop due Static & dynamic analysis can not tell how
due to constant or fixed to variation of displacement. long the component will survive for given
displacement load. Also there is no consideration for
Damping force, inertia force due to factors like surface finish, heat treatment,
kx=F velocity & acceleration decarburizing, alloying elements, realistic
representation of spot & arc welds.
m ẍ + c ẋ + k x = F(t)
If force is doubled static & dynamic
Solution Time Solution Time stress will also be doubled but fatigue life
would be reduced by a factor of 210.
Output – Stress, Output – Stress, Displacement, Output – Life of the component or safety
Displacement Velocity, Acceleration with respect factors
to time / frequency

17.3 What is Fatigue?

“Life of structure when it is subjected to repetitive load”

Failures were observed even after designing the components with maximum stress value well

297
Fatigue Analysis

below yield / ultimate stress. Later it was concluded that the failure is because variation of load
with respect to “time” not being taken in to account. Tests were then carried out for time varying
loads. Results proved that the component fails at values below yield stress when subjected to
time varying load. It was also observed that below a specific stress value components were not
failing at all. This stress value was termed as endurance limit. For example yield stress for general
steel is around 250 N/mm2 and endurance limit is 160 N/mm2.

Objective of fatigue analysis is to calculate Life of structure when it is subjected to repetitive load.
Does it mean that for random load (i.e. not repetitive) fatigue calculations are not possible. Any
random load history can be converted to series of constant amplitude sine waves using various
techniques like rain flow counting. Life is calculated for each individual constant amplitude load
and then superposition of results using minor’s rule would lead to fatigue life for random load.

Fatigue accounts for 90% of failures in mechanical engineering application. Typically these
failures are observed at stress concentration points or welded, bolted joints etc.

FEA based fatigue analysis of metals (like iron, steel, aluminum etc.) has been well established.
But same is not true with non metals (like polymers etc.). The nonlinear fatigue behaviors is not
yet completely understood and is still in research phase.

17.4 History of Fatigue

• The term “Fatigue” was suggested by French engineer Monsieur Poncelet.

• German language word “Betriebsfestigkeit” (operational strength) is a better descriptor of

Fatigue phenomenon.

• 1820’s - It was recognized that the metal subjected to repetitive or time varying load will
fail at a stress much lower than that required to cause fracture on a single application of
load.

• 1870’s - Wohler conducted experiments and proved a safe alternating stress below which
failure would not occur.

• 1880’s - Bauschinger developed mirror extensometer capable of measuring strain upto

one micro strain. He suggested a natural elastic limit (measured in cyclic tests) below which
fatigue would not occur. Elastic limit measured from uniaxial test was not found equal to
the one predicted by cyclic test. This set basis for monotonic & cyclic yield strength of the
material.

• 1900’s – Ewing & Humphrey used optical microscopy to study the same region of the
specimen at various stages of the fatigue life. They suggested ‘To and Fro slip theory’ for
fatigue failures.

• 1920’s - Jekin, suggested “spring slider model” for simulating the stress strain behavior of
metals & to study cyclic deformation. Griffith published paper on fracture & proved that
fatigue failure is because of brittle fracture caused by cyclic growth of a fatigue crack to an
unstable length. Moore and Kommers published book “The fatigue of Metals” based on
their practical applications for real life problems. This book provided guidelines for design
298
 Practical Finite Element Analysis

applications.

• 1950’s - Coffin and Manson proved effect of plastic strain on fatigue life.

• 1960’s - Irwin and others work helped in the development of fracture mechanics.

• 1970’s - Fatigue analysis theory was quite established and commercial softwares either
based on FEA or practically measured strains were launched in the market.

17.5 Definitions
What is Durability , Reliability and Fatigue ?

Durability, Reliability and Fatigue are often loosely used for describing Fatigue related analysis.
There is slight difference in these three terms.

Durability describes overall life requirement, like to last for 2 years (or warranty period).

Reliability includes a probability of failure, such as to have a 95 % chance of survival (i.e. if we test
100 specimen, 95 will pass and 5 might fail).

Fatigue is the failure caused by application of repetitive load by the process of initiation of cracks
and growth.

What is Life ?

Total Crack Crack

Life = Initiation Life + Growth Life

= +

Crack growth life – Ductile material

Brittle material

What is the criteria for transition from crack initiation to crack growth life?

Life till crack of the size 2 mm detected is crack initiation life and remaining life after detection of
crack is crack propagation or crack growth life.

S – N Curve

299
Fatigue Analysis

High cycle fatigue

Alternating
stress

Endurance
strength

Low cycle fatigue

105 2 * 106 No. of cycles

- Life (abscissa) is always plotted on log scale while alternating stress on either linear or log.

Low cycle fatigue : Life of component is less than 100000 cycles, applicable for heavy duty
application loading,.

High Cycle fatigue : Component subject to less sever loads and life > 105 cycles.

Infinite Life : Stress level below which material never fails is known as endurance limit or fatigue
limit. Never fails or infinite life is a relative term. For steel, test is stopped after 2 * 106 cycles (in
case if till then failure is not detected) and said to have infinite life. This is the point where S-N
curve slop changes and it becomes parallel to x-axis.

Unlike steel, non ferrous alloys have no specific endurance limit (S-N curve never become parallel
to x-axis). Pseudo-endurance limit for these materials is stress value corresponding to life = 5x108
cycles (some thing similar to proportionality limit for brittle materials).

S-N curve shown above is based on constant amplitude rotating bending test (Shaft subjected to
pure alternating bending stress). Similar test could be conducted for Tension, compression, shear
and torsional stress. Bending fatigue strength is higher than tension / compression and torsional
fatigue strength is the lowest.

Damage and Endurance Factor of Safety

 Damage is calculated for Low cycle fatigue applications.
Damage = n/N = no. of cycles applied/ Total life
Damage < 1 ⇒ safe Damage > 1 ⇒ fail

 Endurance Factor of safety is calculated for High cycle fatigue applications.

Endurance factor of safety = Endurance strength/ FE stress
Endurance factor of safety < 1 ⇒ fail
300
 Practical Finite Element Analysis

Endurance factor of safety > 1 ⇒ safe

17.6 Various Approaches in Fatigue Analysis

Fatigue Analysis

Experimental data based Finite Element analysis based

n-code, FEMFAT Strain FEMFAT, MSC FATIGUE, LMS,
Physical prototype is necessary, FE SAFE etc.
Expensive test set up No physical prototype required
Many iterations could be carried out
at less cost as well less time

Stress Life Strain Life Crack Propagation Vibration Approach

Approach Approach Approach Frequency Fatigue
High cycle fatigue Low cycle fatigue Fracture Mechanics Resonance effect
Total Life Crack initiation life LEFM, EPFM Input data for
First fatigue analysis Developed in 1960’s Rate of crack growth fatigue calculations
is dynamic analysis
method to be Elastic and plastic Life left results based on
developed strains Could be used in Transient,
Stress and strain elastic Strain Vs. No. of cycles combination with Frequency domain
Uses stress Vs No. of Data (ε-N curve) strain life approach to or Power Spectral
cycles plot (S-N curve) predict total life Density input

Fatigue Analysis

Static Fatigue Dynamic Fatigue

• When the excitation frequency is very • Excitation frequency > 1/3 rd

less than fundamental frequency. fundamental frequency

• No resonance effect. • Vibration fatigue.

• Input data for fatigue analysis is static • Input data is dynamic stress results
stress results
• Accurate Load data (variation
• Time is not important for input load w.r.t. time or frequency) should be
& excitations in the form of full or half specified
cycle is sufficient

My company does not have any Fatigue analysis software but I wan’t to have approximate
301
Fatigue Analysis

idea about the life of the component. Can I achieve it using linear static results?

Yes, for basic calculations one can use S-N curve. But the results would be approximate as it can
not take in to account welding, bolted joints, localized effects etc.

For example, say static stress results for a steel component = 260 N/ mm2 and application of load
is alternating in nature. Amplitude stress for this case = {260-(-260)} / 2 = 260. Plot the stress on
S-N curve for steel as shown below.

Alternating
Stress

Life for stress 260 N/mm2 = 3*105 cycles

260

3 * 105 No. of cycles

17.7 Stress Life Approach

Stress life or S-N method was the first method used for fatigue calculation. It was the standard
fatigue design method for 100 years before developments of other methods like Strain life and
LEFM.

Advantages of Stress Life approach:

- Easy to use and simple approach based on S-N Curve (also known as Wohler diagram) i.e.
alternating stress ‘ S ’ versus No. of cycles ‘N’. The curve is generated by conducting rotating
bending test (constant amplitude, uniaxial loading).

- Work very well for high cycle fatigue (Stresses within elastic limit)

Limitations :

- It works with engineering stress and ignores true stress – strain behaviour (treats all strains
as elastic). It has been proved now that failure is caused due to localized plastic strain and
failure mechanism is to-fro slip.

302
 Practical Finite Element Analysis

σ max

σ A

σ min

stress range : σA = σmax - σmin stress amplitude : σa = (σmax - σmin) /2,

mean stress : σm = (σmax + σmin)/2 stress ratio : R = σmin /σmax
amplitude ratio : A = σa /σm

Mean stress consideration

The standard S-N curve used for design is based on pure alternating load (constant amplitude).
Mean stress for this test is zero. Presence of mean stress affects the results and should be
considered in the design. Mean stress would be present for all the loading conditions other
than pure alternating. It is also generated due to processes like rolling or heat treatment, bolt
pre-stresses or constant (dead weight) loading applications. Tensile mean stress decreases life
while the compressive mean stress increase the life. Various curves (theories) are in used for
consideration of tensile mean effect as shown below. Goodman curve is most commonly used
due to mathematical simplicity and slightly conservative results.

Goodman Diagram

Alternating Stress σa
σen
Gerber paratbola

Goodman line

Soderberg

0 Mean stress σm σy σu

Compression Tension

Soderberg : σa / σen + σm / σy = 1
Goodman : σa / σen + σm / σu = 1
Gerber : σa / σen + (σm / σu)2 = 1

303
Fatigue Analysis

Haigh Diagram : (σa and σm) Vs. Life

103

Lines of
Alternating Stress

104
Constant Life

105

106

Mean Stress

By plotting amplitude and mean stress on above graph one can easily know the range of fatigue
life. Same graph could be displayed in slightly different format as below (this diagram is also
known as master diagram). In the past material fatigue properties used to be represented by
master diagram.

4.0 2.33 1.5 A-I 0.67 0.43 0.25 0.11 0

-0.6 -0.4 -0.2 R-C 0.2 0.4 0.6 0.8 10
0
18

100 C
A-∞
0

100 C
16

R - -1.0 0 C
& 10
0

0 C
14

10
Maximum stress (ksi)

120
12

10 C
0
12
0

100 10 C
0
10

0
10

0 C
0

Al 10 AISI 4340
80 te
r na 0 C& s0 - 158 ksi. S1 - 147 ksi
10
80

80

tin RT, 2000 cpm

60 g
st
re
si)
60

s Unnotched
st 60
(k

s(
ss

ks Notched
re

40 i)
40

40

K1 - 3.3. p - 0010
ea
M

20
20

20

0
-120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 140 160 180
Minimum stress (ksi)

304
 Practical Finite Element Analysis

Miner’s Rule:

Formulas discussed so far were based on the assumption of constant amplitude loading. Miner’s
rule is used for calculating the damage for variable amplitude loading.

A3 M3 n3
A2 M2 n2
A1 M1 n1

N1 N2 N3

Amplitude
Stress Amplitude (log)

A1
A2 - +
SIN1 Mean
A3
SIN1
SIN1
ni
di = —
N (log) Ni

Total Damage :
D = ∑ di = d1 + d2 + d3 + ..... + dn

n1 n2 n3

Total damage = n1 / N1 + n2/N2 + n3/N3

n = applied cycles
N = cycles to failure

Different types of loads:

1. Constant: Bolt torque, Pressure, Shrink fit etc.

305
Fatigue Analysis

2. Pulsating : Name pulsating is derived from pulse (the one which doctors check when we are
ill).

3. Alternating : Imagine a rod being pulled as well as pushed at equal time interval. Load varying
as: Zero – Maximum-Zero – Minimum - Zero.

4. Random

Factors affecting Fatigue Analysis:

 Following factors will reduce value of endurance limit:

- Tensile mean stress
- Large section size – Large test specimen size shows less Endurance limit
- Rough surface finish – Machined surfaces have better life than the casting parts.
- Chrome and nickel plating
- Decarburization (due to forging and hot rolling) : Decarburization is loss of carbon
atoms from the surface, causes low strength & tensile residual stresses
- Stress concentration, geometrical discontinuity
- Corrosive Environment

306
 Practical Finite Element Analysis

 Following factors will increase endurance limit:

- Nitriding
- Flame and induction hardening
- Carburisation
- Shot peening - Compressive residual stress
- Cold rolling - Rolled threads have higher strength than the cut or machined
threads

Fatigue Results are very sensitive to FE-Mesh and static / dynamic stress values:

Following figure shows effect of mesh density and element type on fatigue analysis. Converged
static or dynamic stresses should be used as input for fatigue analysis. Sub- modeling could also
help in achieving better results.

260 TETRA10 4093 TETRA10 576 HEXA8

552 Nodes 6655 Nodes 819 Nodes
σv.Mises = 164 σv.Mises = 208 σv.Mises = 192

X = 0.077 X = 0.166 X = 0.205

SFEndu = 1.05 SFEndu = 0.90 SFEndu = 1.00
1. Coarse tetra model 2. Fine tetra model 3. Coarse hex sub model
based on results of 1

1280 HEXA8 5300 HEXA8

1683 Nodes 6424 Nodes
σv.Mises = 200 σv.Mises = 205

X = 0.249 X = 0.254
SFEndu = 0.99 SFEndu = 0.97
4. Fine hex sub model 5. Fine hex model based
based on results of 1 (accurate results)

Material type: nodular cast iron, UTS 400 MPa

Loading: pulsating bending
Courtesy: FEMFAT Software (Engineering Center Steyr GmbH & Co KG, Austria)

What is the effect on fatigue life when load is doubled

Consider plate with hole fixed at one end and subjected to load on other. Plate dimensions: 100
x 100 x 1 mm, Hole diameter: 12 mm. Material – Steel, σut = 450 N/mm2, σy = 240 N/mm2.

307
Fatigue Analysis

Total Force:
Load Case 1 - 3333.33 N
Load Case 2 - 6666.66 N

Static results for load = 6666.66 Damage plot for load = 6666.66 N

Life calculations: Applied cycles = 1E5 cycles,

damage = n / N => Life N= 1E05 / damage

Life = 1e05 / 0.0102

= 9.80E06 cycles

Load, N 3333.33 6666.66

Static stress, N/ mm2 100 200
Fatigue life, cycles 2.99 E 11 9.80 E 06

Static stress doubled when load is doubled but fatigue life reduced by a factor of 30,510 !

Alternating load is more sever than pulsating load:

Above problem was solved for load = 6666.667 N for alternating and pulsating load cycles.
Results show alternating loading is more sever than pulsating

Alternating Cycle Life Pulsating Cycle Life

9.80 E 06 2.35 E 13
308
 Practical Finite Element Analysis

Better surface finish means better fatigue life:

Crack initiates at outer surface, and better surface finish results in better fatigue life. Please note
just machining a cast surface shows around 1.37 times better fatigue life. Similarly shot peening
induces compressive residual stresses and is recommended.

Casting 200 μ machining 60 μ Polished 2 μ Shot peen

1.15 E 07 1.58 E07 7.12E07 2.30 E 08

Which type of stress (Max. principal or vonMises or max. shear stress) is used for fatigue
calculations.

Commercial softwares provide option for max. principal, vonMises as well as max. shear stresses.
Fatigue calculations are based on absolute max. principal stress or signed vonMises or signed
max. shear stresses .

Signed principal stress or absolute principal stress : This term is commonly used in fatigue
analysis. Fatigue calculations are based on amplitude and mean stress. It has been observed that
if the calculations are just based on only maximum principal stress or only minimum principal
stress then stress range (σmax. – σmin.) is less and leads to higher fatigue life. Remedy is to find
max. value out of the two at a point over given period of time and then find the stress range or
amplitude and mean stress based on this data(say max. principal stress at a node at time 1 sec.
is + 250 and min. principal is – 400 then absolute principal stress = -400, collect the data over a
period of say 10 sec and then find stress range based on max. and min absolute stress values out
of the 10)

Signed vonMises stress: vonMises and max. shear stress values are always positive. If these
values are used for fatigue calculations then stress range would be reduced to half resulting in
higher fatigue life. Remedy is to find out sign of absolute principal stress at the point at a given
time instance and assign it to corresponding value of vonMises or max. shear stress. Say absolute
max. principal stress is -300 at time 2 sec and corresponding vonMises stress at time = 2 sec is 315
then signed vonMises stress would be -315.

In general max. principal stresses are recommended over vonMises and max. shear. VonMises and
max shear stress are not directional i.e. direction of crack propagation could be better answered
by using max. principal stress.

For fatigue calculations whether Average stress or unaverage, Element (centroidal) stresses
or nodal stresses?

Nodal stresses are recommended over elemental similarly un-average over average. In some
situations this might lead to lesser fatigue life than actual but from design point of view it is
always safer and recommended.

309
Fatigue Analysis

17.8 Strain Life Approach

Strain life approach is also known as Crack initiation approach or Local Stress Strain approach or
Critical Location approach

 Stress life approach is based on stress, Strain life approach is based on strain. It is now well
accepted that fatigue is strain & not stress controlled.

 This method calculates crack initiation life.

 It is based on strain history at a point where failure is likely to occur.

 Consideration for plastic strain

 Recommended for Low Cycle Fatigue

In the High Cycle Fatigue region, stress and strain levels are low and they are linearly related.
However, when stress is higher than yield (low cycle fatigue), strain based approach, which takes
in to account plasticity and non linear relation between stress - strain, gives better result.

Research has shown that crack always initiates in the plastic region & damage is dependent on
plastic deformation or strain. In strain life approach the plastic strain or deformation is directly
measured and quantified.

F
Stress

Stress

Strain
Away from crack
Strain
location

F
At crack tip

Away from the crack tip structure remains elastic and stress value below yield. At crack tip due
to geometrical discontinuity, stress may go in elastic – plastic region giving rise to plastic strains
and reversals as shown in the figure.

S-N (alternating stress Vs. life) curve is generated via rotating bending test is the base for stress
life approach, while ε - N (Alternating strain Vs. Life) is the base for strain life calculation. Smooth
specimen is loaded in fix grip and subjected to tension-compression loading. Test is conducted
in strain control environment (component’s deformation and original length kept constant, so
that the strain is controlled)

310
 Practical Finite Element Analysis

ε - N Curve :
εt
Strain amplitued (log scale)

c σa = σf (2 Nf )b
1 σa = True cyclic stress amplitude
σ1/E Total σf = σa intercept at 2Nf = 1, known as Fatigue strength coefficient
2Nf = no. of reversals for failure
b = slope known as fatigue strength coeff.
Elastic b
1
Plastic
N1

2N1
Reversals to failure (log scale)

Total strain ( εt) = Elastic strain ( εe) + Plastic Strain ( εp)

εf = {(σf (Nf )b / E} + {εf (2 Nf )c}

2 Nf = Number of reversals to failure

c= regression slop called fatigue ductility exponent

Differnce in reversal and cycle?

1 cycle = 2 reversals, reversal is more commonly used with strain history (hysteresis loop, counting)
while cycles are more common with stress histories and stress life approach.

Steps for Strain – Life calculations

1) Material properties obtained from smooth specimen strain controlled test (cyclic stress
strain data & strain life data)
2) Stress Strain history at the critical location
3) Cycle counting
4) Mean stress correction
5) Damage calculations

Elastic and plastic strains

Total strain (εt) = Elastic strain (εe) + Plastic strain ( εp)

εe = σ / E
Plastic strain as per Ramberg –Osgood empirical formula
εp = (σ / K)1/n
K = Strength coefficient, n=strain hardening exponent

311
Fatigue Analysis

Elastic
Unlading
σ

εp εe ε
εt

Stress-strain (Hysteresis) Loop

∆εe ∆εp
σ
B
∆σ - Stress range
A
∆σ ∆ε - Total strain range
— ε∆e – elastic strain range
2 ε∆p – plastic strain range

ε Area inside the loop = dissipated

0 energy per unit volume (plastic
(ε )
∆ —
2 (ε )
∆ —
2 ∆σ
work done)
—
C 2

How to determine plastic stress / strain at critical area ?

Strain life method requires local stresses and strains (plasticity consideration) for fatigue
calculations. This could be achieved by

• Conduct Test : Strain gauge measurement at the critical location.

• Non Linear Analysis : Elastic - Plastic finite element analysis (non linear solver)

• Linear Analysis and Neuber’s rule : Assumption to linear statc analysis is stress-strain curve is
straight line even after crossing the yield point, usually this leads to very high and unrealistic
stresses, Neuber’s rule helps in determining equivalent non linear stress – strain from linear
FEA results.

312
 Practical Finite Element Analysis

Elastic – Plastic Correction : Nueber’s equation

linear σ - ε curve
σlinerar Neuber Hyperbola

nonlinear σ - ε
curve
σnonlinerar

Neuber Equation
∆σ∆ε=E∆εe2

Stress

Strain

Kt 2 = Ke * Kσ
Kt = Theoretical (based on geometry shape, linear elastic behavior) stress concentration.
Ke = True strain concentration factor = Notch (plastic) strain / Nominal strain
Kσ = True stress concentration factor = Notch (plastic) stress / Nominal stress

17. 9 Fracture Mechanics Approach

Fracture mechanics is used for calculating “remaining life” after crack initiation i.e. crack
propagation life. There are two methods namely LEFM (Linear Elastic Fracture Mechanics)
and EPFM (Elastic Plastic Fracture Mechanics). For Automobile industries usually detection of
crack it self is considered as failure and the subject is more important for Aeroplanes designs
and maintenance. We will just discuss the basics now and reader is requested to refer books
mentioned in the references for further details.

In strength of materials we assume material is free from all the defects (and cracks), in fracture
mechanics the starting point it self is to assume presence of a finite length crack.

Strength of materials deals with stresses (normal & shear) developed due to various forces and
moments while fracture mechanics is all about calculating stress intensity factor and crack
growth rate for basic three crack opening modes as shown below.
Modes of Crack Opening

Mode 1 Mode 2 Mode 3

Opening Mode Sliding Mode Tearing Mode

313
Fatigue Analysis

Stress intensity factor = {K = Y S (π a)1/2}

Y – compliance (geometry) function, a- crack length) then ∆K from max & min. stress

• Rate of crack propagation: da/dN = C (∆K)m da/dN= current, C & m material const.

17.10 Cycle Counting

In real life components are very rarely subjected to purely constant amplitude loading. But
problems is most of the available data & empirical formulas are based on constant amplitude
loading. So the question is how to use this data for solving real life problem subjected random
or variable amplitude loading.

This is achieved by cycle counting. Converting random load history to a number of constant
amplitude events is termed as cycle counting.

Different methods for cycle counting

1) Level crossing counting

2) Peak counting
3) Simple- range counting
4) Rain flow counting

Level Crossing Counting

Real Life variable amplitude data
6
2
5
4 8
3 Level cross Count
4
2 6 0
1
0 5 1
-1
1 6
-2 4 2
-3 3
-4 3 2
-5 7
5 8
-6 2 3
Constant Amplitured conversion via cycle counting
1 3
0 3
-1 3
6
5 -2 3
5 4
4
3 -3 3
2
2
1 -4 3
0
-1
-5 1
-2 -2 -6 0
-3
-4
-5 -4
-6 -5
Level Crossing Method

• Divide Strain /Stress axis in to number of level (increments)

314
 Practical Finite Element Analysis

• A count for each time a positively sloped portion of strain history crosses a level
• A count for each time a negatively sloped portion of strain history crosses a level
• Counts are combined to form completed cycles as follows (from most damaging combination
of counts) : First form a largest possible cycle followed by next largest possible cycle & so
on

Peak Counting
Real Life variable amplitude data
6
2
5
8
4 Peak Count
3 4
2 6 0
1
0 5 1
-1
1 6 4 1
-2
-3 3
-4 3 0
-5 7
5 8 2 0
-6
Constant Amplitured conversion via cycle counting
1 0
0 0
-1 0
6 -2 1
5
5 4
4 -3 0
3
2
2 -4 2
1
0 -5 1
-1
-2 -2 -6 0
-3
-4
-5 -4
-6 -5
Level Crossing Method

• Based on peak counts i.e. max. & min. stress / strain values

• Axis divided in to number of levels (increments)

• Positions of all max. (peak) strain & all min. (valley) strain values are tabulated.

• Counts are combined to form completed cycles as follows (from most damaging combination
of counts) : First form a largest possible cycle by combining largest peak and smallest valley
followed by next largest possible cycle & so on.

Limitation of early methods of counting

- No consideration of order in which cycles are applied. Due to nonlinear relation between
stress & strain (plastic behavior) e.g.

315
Fatigue Analysis

Cycle A Cycle B
Starting Point Tensile Loading Starting Point Compressive Loading

Cycle C Cycle D
Small amplitude at the beginning Large amplitude at the beginning

Counting as per above methods will show exactly same cycles and same damage (fatigue life) for
A, B, C & D. But in practice different lifes are observed for above cases.

Rain Flow Counting

Japanese engineers Matsuishi & Endo were carrying out research on cycle counting and fatigue.
They got idea of this method while watching flow of rain flowing through Pagoda roof house and
hence its named as rain flow counting.
Cycles
7 4
ε1 Strain ε2 3 2
5 6
1 ε1 Counting of Hystersis Loop ε2
2
2
4
4

6 3
6
7 Time
1
5
Rain Flow Counting

Draw Stress / strain time history with time axis oriented vertically and greatest magnitude of
strain value at the start and end of strain history curve (this step eliminates counting of half
cycles)

316
 Practical Finite Element Analysis

1) A flow of rain begins at each valley and peak (strain reversal) and allowed to flow unless

a) The rain began at local peak falls opposite a local max. point greater than that
from which it originated.
b) The rain began at valley and falls opposite a local min. point greater than that
from which it originated.
c) It come across previous rain flow.
i. Rain flows from pt. 1 over points 2 & 4 and continues to the end of history
since none of the conditions for stopping rain flow are satisfied
ii. Rain flows from pt. 2 over 3 and stops opposite pt. 4 since both 2 and 4 are
local max. and the magnitude of 4 > 2
iii. Rain flows from pt. 3 and must stop upon meeting the rain flow from pt. 1
(and continuing through 2) .

There are 3 cycles of constant amplitude 1-4, 2-3, 5-6 (closed hysteresis loops) each having its
own mean stress & range value. Cumulative damage could be determined by using Miner’s rule
D = 1/ N1 + 1/ N2 + 1 /N3

At present many rain flow techniques are in use like-

- Original rain flow method

- Range-pair counting
- Three point cycle counting
- Four point cycle counting
- Hysteresis loop counting
- Race track method
- Ordered overall range counting
- Range pair range counting
- Hayes method etc.

17.11 Multi-Axial Fatigue

X-axis input Y- axis input Z-axis input

317
Fatigue Analysis

Stress life approach, strain life approach and other topics covered so far were based on assumption
of uniaxial loading with constant amplitude cycles. In real life seldom components are subjected
to this kind of loading. Most of the times it’s a multiaxial loading with variable amplitude. Popular
fatigue analysis softwares provide a separate module to take care of multiaxial applications.

Stress life, strain life or LEFM concentrates only on a single stress usually normal stress. Stress is a
tensor & has 9 components. For real life applications more than one of 9 stresses are non zero.

Fatigue Analysis, Nature of input stress

Uni-Axial Multi Axial Proportional Multi Axial Non Proportional

- Principal plane fix - Stresses vary in simple - Direction of Principal plane
proportion so that directions vary during the cycle &
• existing theories like of the principal plane remains function of time.
stress or strain life constant with time.
approach • Critical plane approach
• Equivalent stress
approach like theories
of failures in static
analysis

Limitations of proportional loading theory :

Proportional loading was the only method used for multi axial fatigue calculations before 70’s. It
has following limitations

- There is no consideration for “fatigue being a directional process”. Damage (crack) takes
place on particular plane.
- No consideration for phase angle.
- Whether or not the loading is proportional and how much it deviates from proportional
can be determined by observing the variation with time, stress ratios & principal stress
direction.

Non-proportional loading:

Though there are several theories for non proportional loading but most commonly used in
fatigue analysis softwares is “Critical plane approach”. It recognize fatigue as directional process
and calculates damage for all possible planes (at say 10º intervals) & the worst or critical plane is
reported.

17.12 Welding analysis

Many a times static or dynamic analysis predicts a location of failure not matching with actual
field or test failure. In particular this problem is quite common when failure is at weld locations.
For example consider wheel disc and rim analysis, Static analysis will always show failure at either
bolt locations or vent holes (elliptical holes) but never at the weld at the junction of disc and rim.

318
 Practical Finite Element Analysis

But in real life failures are usually observed at welding location and we keep on thinking what is
wrong with analysis. Whether constraints are wrong or loading. But when fatigue analysis using
weld module is carried out based on same static / dynamic analysis as input data, it clearly shows
failure at welding.

SPOT Weld Fatigue Analysis

Commercial softwares provides special provision for spot and arc weld fatigue analysis.

We will be carrying out a simple exercise to show “Spot weld is stronger in shear and weak in
normal loading”.

76.00 76.00
16.50 Total Force Total Force = 1000 N
= 1000 N
44.50

Spots Spots Extrude length = 200 mm

44.50

R5.00 No. of spots = 5

Spot pitch = 40 mm
R2.85 16.50 Starting spot = 20 mm from edge
Total Force = 1000 N
44.50

61.00
Clamped Clamped
Spot: Normal stress Shear stress

319
Fatigue Analysis

Shear load on spot Normal load on spot

(Courtesy: FEMFAT Software, Engineering Center Steyr GmbH & Co KG, Austria)

In the original mesh spots are represented by beam elements. Prior to linear static analysis spot
weld elements were specially pre-processed in FEFAT. The pre-processor creates special pattern
of elements with appropriate material properties around the spots.

Pre-processed mesh Boundary Condition Plot

Static Stress Fatigue Life

Normal loading 346 1.55 E 07
Shear loading 48.1 5.52 E 12

Spot welds are stronger in shear and weak in normal (tension, compression, bending)
loading. Many a times just rearranging spots (say orientation of spot changed by 90 degree)
works well and solve the problem.

Arc Welding :
Static or dynamic analysis cannot take into account various aspects of welding like type of
welding, heat zone, surface finish etc. For example a butt joint between two small thickness parts
(via shell mesh at the mid plane) will be represented as below

320
 Practical Finite Element Analysis

It does not differentiate between various joint types like square or single V, double V, Single U,
double U etc. and hence analysis result will be the same for all. But this is not the case in real life.
Fatigue analysis softwares provides provision for most of the variables and hence gives realistic
answers.

17.13 CAE (Fatigue) and Test data correlation

Courtesy: FEMFAT Software (Engineering Center Steyr GmbH & Co KG, Austria)

What do you feel, is this correlation acceptable ?

In static analysis, correlation is said to established if difference is < 15 %. For Fatigue analysis
acceptance criteria is factor of 3 (i.e. if test result is 3*105 cycles and fatigue analysis software
is predicting life in the range of 105 to 9*105 cycles). Acceptable difference may go as high as
factor of 10 when loading, material properties and localized effects such as welding not taken
in to consideration properly. Fatigue is quite a complicated phenomenon and even two test
results (if we conduct test on two identical specimen) won’t match. If the test is conducted on
several identical components then a factor of 3 is observed (standard deviation, scattered data).
Accuracy of fatigue analysis depends on

321
Fatigue Analysis

1. Meshing ( coarse mesh would lead to lower life, convergance of stress / strain result is
recommended before exporting it for fatigue analysis)
2. Load data and boundary conditions
3. Material properties
4. Process effects and residual stresses, localized effects like weld, bolt, riveted joints etc.

Importing load history data in Fatigue analysis sotwwares

Following are the common formats supported by most of the commercial softwares

• RPC ASCII File

• ADAMS Request file
• Tec Math ASCII file
• DIAdem data file
• ADAMS spread sheet FIC file
• RPC binary file
• n Code DAC file
• Rainflow load spectrum (TechMath-ASCII-RFM)

References
1) Fundamentals of Metal Fatigue analysis - J.A. Bannantine, J.J. Comer, J. L. Handrock, Prentice Hall
2) Finite Element Based Fatigue Calculations - NAFEMS - The international Association for the Engineering Analysis
community
3) Fatigue Testing and analysis, Theory and Practice: Yung-Li Lee, Jwo Pan, Richard Hathaway, Mark Barkey, Elsevier
Publication
4) Failure Fracture Fatigue, An Introduction: Tore Dahlberg, Anders Ekberg, Overseas Press (India) Pvt. Ltd.
5) Mechanical Metallurgy – G. E. Dieter, Mc Graw Hill
6) Elementary Engineering Fracture Mechanics, D. Broek, Martinus Nijhoff, The Hague
7) FEMFAT 4.6 Help.
8) Atlas of Stress -Strain Curves - ASM International,The Materials Information Society
9) Atlas of Fatigue Curves - ed. by H. E. Boyer, 1986, ISBN 081702142

322