Department of Mechanical and Aeronautical Engineering

University of Pretoria
Open Book Exam
(Allowed: Textbook, Handout notes, Class notes, downloads from the website)
Not allowed: Working in groups, discussing problems with classmates.
MSV 780 Fatigue – Examination
Time: 3 Hours Full Marks: 180
Examiner: Dr. Michiel Heyns Date: 6 July 2020
External Examiner: Prof N. E. Dowling, Virginia Tech, Backsberg, VA, USA 24061
Number of pages: 24
Instructions:
• You received this paper at the start time of the examination.
• You must submit your answers after 3 hours.
o If you e-mail it, make sure I receive it before 3 hours have expired.
o TurnitIn will automatically measure submission time.
• Late submissions:
o Will be penalised at 2% per minute.
• Clarifications or other information may be written on the blackboard -- check it before handing in
your paper.
• Logic and calculations leading to all answers must be given, and any points or lines used from
graphs must be marked and labelled where applicable.
• For discussion or explanation type questions, answers need to be in complete sentences and in
reasonably good English, and the logic must be clear.
• Numerical answers should have neither insufficient nor excessive numbers of significant figures.
• Correct units are required for full credit.

Good luck!

Page 1 of 24
1 STRESS-LIFE PROBLEM [60 Marks]
A RQC-100 steel quenched and tempered part (see Figure 1) is subjected to the non-rotating bending
nominal stress spectrum shown in Table 1. The nominal stress was determined at the notch in the
𝑀𝑟
smaller diameter of the component using the formula 𝑆 = . The stress spectrum was determined
𝐼𝑥𝑥
from calculated stress and Rainflow-counting.
The part dimensions are shown below.

Figure 1: Component subject to rotating bending stress

Assume that a fatigue curve for 50% probability of survival of a mirror polished unnotched specimen
has the following endurances and completely reversed bending stress amplitudes (𝑁; 𝜎𝑎𝑟 ):
• (1 000; 0.9𝜎𝑢 )
• (1 000 000; 𝜎𝑒𝑟𝑏 ), where the completely reversed endurance limit in bending is: 𝜎𝑒𝑟𝑏 = 0.5𝜎𝑢
The operating parameters are:
• Temperature 200 °C
• Surface finish: Machined
• Non-corrosive environment and the part is surface protected
1.1 Problem statement
Please answer the following:
1. Use the Budynas parameters to estimate fatigue limits and construct the S-N curve [Marks =3].
2. Use the modified Goodman mean stress compensation method and the Peterson equations for
notch sensitivity and calculate the following for a 99.9% probability of survival [Marks =52]:
a. Repetitions of the stress spectrum that can be applied before crack initiation [95% x
mark].
b. The fatigue life in years [5% x mark].
3. Comment on the use of the modified Goodman mean stress compensation in this case.
Comment on SWT and Walker mean stress compensation approaches and how they compare
with the modified Goodman approach. [5 Marks].
Table 1: Non-rotating bending stress spectrum over a period of 2 years

𝜎𝑎 𝜎
[MPa] [MPa] [cycles]
200 80 10 000
300 0 2 000
200 -80 5 000

Page 2 of 24
1.2 Solution
Question 1
From the table below, the material has a yield strength 𝜎𝑜 = 683 𝑀𝑃𝑎 and ultimate tensile strength
𝜎𝑢 = 758 𝑀𝑃𝑎.

Determine the values on the S-N curve

′
Because there is a notch involved, the equation for the fatigue strength at 1 000 cycles, 𝜎𝑎𝑟𝑏,10 3 , and
′
the endurance limit at 1 000 000 cycles, 𝜎𝑒𝑟𝑏 , is:
′
0.5𝜎𝑢
𝜎𝑒𝑟𝑏 =
𝐾𝑓
′ 0.9𝜎𝑢
𝜎𝑎𝑟𝑏,103 =
𝐾𝑓′
From the figure below, for the dimensions provided, the notch theoretical stress concentration factor is
𝐾𝑡 = 1.8.

Page 3 of 24
 Figure 2: Theoretical stress concentration factor

The fatigue notch factors are given by:

300 1.8
𝑎=[ ] × 10−3 × 25.4
𝜎𝑢 [𝑘𝑠𝑖]
1.8
300
=[ ] × 10−3 × 25.4
700
(6.89)
= 0.178 𝑚𝑚
The question request calculation with the Peterson equation as follows:
log 𝛼 = 2.654 × 10−7 𝜎𝑢2 − 1.309 × 10−3 𝜎𝑢 + 0.01103
𝛼 [𝑖 𝑚𝑚] = 10log 𝛼 (345 ≤ 𝜎𝑢 ≤ 2 070 𝑀𝑃𝑎)
−7 2 −3
= 102.654×10 𝜎𝑢 −1.309×10 𝜎𝑢 +0.01103
= 0.148 𝑚𝑚
The fatigue notch factor at the endurance limit is:
𝐾𝑡 − 1
𝐾𝑓 = 1 + 𝑎
(1 + 𝑟 )
1.8 − 1
=1+
0.148
1+
10
= 1.79

Page 4 of 24
 The fatigue notch sensitivity factor, 𝑞, is approximately 0.30 from the figure below:

Equation for the S-N curve

The fatigue notch factor at 1 000 cycles is then:
𝐾𝑓′ − 1
= 𝑓(𝜎𝑢 ) = 𝑞
𝐾𝑓 − 1
𝐾𝑓′ = 𝑞(𝐾𝑓 − 1) + 1 Allocate 10% of the mark for the
fatigue notch factor calculations
= 0.30(1.79 − 1) + 1
= 1.24
The modified S-N curve has the following points:
𝜎𝑒𝑟 = σ′erb 𝐶𝑠𝑖𝑧𝑒 𝐶𝑙𝑜𝑎𝑑 𝐶𝑠𝑢𝑟𝑓 𝐶𝑇 𝐶𝑟𝑒𝑙
′
0.5𝜎𝑢
𝜎𝑒𝑟𝑏 =
𝐾𝑓
′
𝜎𝑎𝑟,103 = 𝜎𝑎𝑟𝑏,103 𝐶𝑙𝑜𝑎𝑑 𝐶𝑇 𝐶𝑟𝑒𝑙
′ 0.9𝜎𝑢
𝜎𝑎𝑟𝑏,10 3 =
𝐾𝑓′

Page 5 of 24
 Modifying factor: size, 𝐶𝑠𝑖𝑧𝑒
For the shaft with diameter 100 mm, the size factor is as follows from the detail presented below
(note, this is non-rotating bending problem and 𝑑𝑒 = 0.37𝑑 = 37 𝑚𝑚:
𝐶𝑠𝑖𝑧𝑒 = 1.51𝑑 −0.157
= 1.51(37)−0.157
= 0.86

Allocate 5% of the mark

Source: (Dowling, 2013, p. 505)

Modifying factor: Load, 𝐶𝑙𝑜𝑎𝑑
In this case the problem asks for an analysis of a roller subject to bending. The S-N curve used, is
also for a specimen in bending. Therefore, 𝐶𝑙𝑜𝑎𝑑 = 1.0.
Allocate 5% of the mark who motivated this

Page 6 of 24
 Modifying factor: surface, 𝐶𝑠𝑢𝑟𝑓
The S-N curve was for a mirror polished specimen. The fatigue analysis is to be done on a hot rolled
part. Therefore, the reduction factor for surface finish is 𝐶𝑠𝑢𝑟𝑓 = 0.75. According to the Budynas the
factor for the machined surface is 𝐶𝑠𝑢𝑟𝑓 = 4.51𝜎𝑢−0.265 = 4.51 × 758−0.265 = 0.78. The latter will be used
in calculations.

Allocate 5% of the mark

Source: (Dowling, 2013, p. 505)

Page 7 of 24
 Modifying factor: reliability, 𝐶𝑟𝑒𝑙
The fatigue curve to be constructed uses as reference the strengths at 1 000 and 1 000 000 cycles for
50% probability of survival in log(N). To have a reliability of 99.9%, the modification factor is: 𝐶𝑟𝑒𝑙 =
0.753.

Allocate 5% of the mark

Modification factor: temperature, 𝐶𝑇

A modification factor for temperature 200 °C is given below, from which:
1.0 𝑓𝑜𝑟 𝑇 ≤ 450 ℃
𝐶𝑇 = {
1 − 5.8−3 (𝑇 − 450), 𝑓𝑜𝑟 450 < 𝑇 ≤ 550 ℃ Allocate 5% of the mark
= 1.0
Two points on the S-N curve
The two points to use on the modified S-N curve are:
• At 1 000 cycles:
′ 0.9𝜎𝑢
𝜎𝑎𝑟𝑏,10 3 =
𝐾𝑓′
′ Allocate 5% of the mark for the
𝜎𝑎𝑟,103 = 𝜎𝑎𝑟𝑏,10 3 𝐶𝑙𝑜𝑎𝑑 𝐶𝑇 𝐶𝑟𝑒𝑙
student who understands this
0.9𝜎𝑢 process and who obtained a
= 𝐶 𝐶 𝐶
𝐾𝑓′ 𝑙𝑜𝑎𝑑 𝑇 𝑟𝑒𝑙 fatigue strength within 10% of
0.9 × 758 the value
= × 1 × 1 × 0.753
1.24
= 414 MPa

• At 1 000 000 cycles, the endurance limit for completely reversed loading is:

′
0.5𝜎𝑢 Allocate 5% of the mark for the
𝜎𝑒𝑟𝑏 =
𝐾𝑓 student who understands this
𝜎𝑒𝑟 = σ′erb 𝐶𝑠𝑖𝑧𝑒 𝐶𝑙𝑜𝑎𝑑 𝐶𝑠𝑢𝑟𝑓 𝐶𝑇 𝐶𝑟𝑒𝑙 process and who obtained a
0.5𝜎𝑢 fatigue strength within 10% of
= 𝐶 𝐶 𝐶 𝐶 𝐶
𝐾𝑓 𝑠𝑖𝑧𝑒 𝑙𝑜𝑎𝑑 𝑠𝑢𝑟𝑓 𝑇 𝑟𝑒𝑙 the value
0.5 × 758
= × 0.86 × 1 × 0.78 × 1 × 0.753
1.79
= 107 MPa

With the fatigue strengths known at 1 000 and 1 000 000 cycles, determine S-N curve parameters
The equation for the S-N curve is assumed to be:
𝜎𝑎𝑟,1 𝑁1𝑏 = 𝜎𝑎𝑟,2 𝑁2𝑏
Therefore:

Page 8 of 24
 𝑁1 𝜎
𝑏 log = log 𝑎𝑟,2
𝑁2 𝜎𝑎𝑟,1
𝜎𝑎𝑟,2
log
𝜎𝑎𝑟,1
𝑏=
𝑁
log 1 Allocate 5% of the mark for the
𝑁2
𝜎𝑒𝑟 student who understands this
log
𝜎𝑎𝑟,103 process and who obtained a
=
𝑁103 fatigue strength within 10% of
log
𝑁𝑒 the value
107
log
= 414
1 000
log
1 000 000
= 0.196
The endurance at any completely reversed stress amplitude is then:
1
𝜎𝑒𝑟 𝑏
𝑁𝑅 = {(𝜎 ) 𝑁𝑒 𝜎𝑎𝑟 ≥ 𝜎𝑒𝑟
𝑎𝑟
∞ 𝜎𝑎𝑟 < 𝜎𝑒𝑟
Mean stress compensation
The problem statement mentioned that the modified Goodman method should be used for mean stress
compensation. From the modified Goodman mean stress compensation, the equivalent completely
reversed stress for positive maximum stress is:
𝜎𝑎
𝜎𝑎𝑟 = 𝜎
1− 𝜎
𝑢

Apply to the stress spectrum

𝜎𝑎 𝜎
[MPa] [MPa] [cycles]
200 80 10 000
300 0 2 000
200 -80 5 000
For amplitude 200 MPa, mean 80 MPa and 10 000 cycles, the damage per repetition is:
1. The mean stress corrected completely reversed amplitude:
𝜎𝑎
𝜎𝑎𝑟,1 = 𝜎
1− 𝜎
𝑢
= 224 𝑀𝑃𝑎
2. This completely reversed stress exceeds the endurance limit, 𝜎𝑒𝑟 , and the endurance is:
1
𝜎𝑒𝑟 𝑏
𝑁𝑅 = {(𝜎 ) 𝑁𝑒 𝜎𝑎𝑟 ≥ 𝜎𝑒𝑟
𝑎𝑟
∞ 𝜎𝑎𝑟 < 𝜎𝑒𝑟 5% for correct mean
1 stress compensation
107 0.196
=( ) × 1 × 106 5% for endurance within
224
= 23 064 𝑐𝑦𝑐𝑙𝑒𝑠 10%
3. The damage is then: 5% for correct damage
10 000
𝐷1 = = 0.43
23 064
For amplitude 300 MPa, mean 0 MPa, and, 2 000 cycles the damage is:
1. The mean stress corrected completely reversed amplitude:
𝜎𝑎𝑟,2 = 300 𝑀𝑃𝑎

2. This completely reversed stress exceeds the endurance limit, 𝜎𝑒𝑟 , and the endurance is:

5% for correct mean stress

compensation
Page 9 of 24
5% for endurance within 10%
5% for correct damage
 1
𝜎𝑒𝑟 𝑏
𝑁𝑅 = {(𝜎 ) 𝑁𝑒 𝜎𝑎𝑟 ≥ 𝜎𝑒𝑟
𝑎𝑟
∞ 𝜎𝑎𝑟 < 𝜎𝑒𝑟
1
107 0.196
=( ) × 1 × 106
300
= 5 195 𝑐𝑦𝑐𝑙𝑒𝑠
3. The damage is then:
2 000
𝐷2 =
= 0.38
5 195
For amplitude 100 MPa, mean -80 MPa, and 5 000 cycles, the damage is:
1. The mean stress corrected completely reversed amplitude:
200
𝜎𝑎𝑟,3 =
−80
1−
758
= 181 𝑀𝑃𝑎
2. The completely reversed stress range exceeds the endurance limit and the endurance is:
1
𝜎𝑒𝑟 𝑏
𝑁𝑅 = {(𝜎 ) 𝑁𝑒 𝜎𝑎𝑟 ≥ 𝜎𝑒𝑟
𝑎𝑟
∞ 𝜎𝑎𝑟 < 𝜎𝑒𝑟
1
107 0.196
=( ) × 1 × 106
181
= 68 427 𝑐𝑦𝑐𝑙𝑒𝑠

3. The damage is then:

5 000
𝐷3 = = 0.07 5% for correct damage
68 427
The total damage per repetition is then:
𝐷 = ∑ 𝐷𝑖
= 0.43 + 0.38 + 0.07
= 0.88
The number of repetitions to failure is then:
1
𝐵𝑓 =
𝐷
1
= 10% for answer within
0.88 10%
= 1.1
The period of one repetition was 2 years, therefore, the fatigue life for a 95% probability of survival is:
𝐿𝑖𝑓𝑒 = 𝐵𝑓 × 2 𝑦𝑒𝑎𝑟
= 2.2 𝑦𝑒𝑎𝑟𝑠 5% for answer within 10%

For a candidate who followed the process above, or used another equation to fit the S-N
curve:
1. 100% if all aspects have been considered and the answer is within 10%
2. Subtract for aspects not included in the assessment e.g.:
a. mean stress compensation;
b. temperature;
c. load;
d. surface finish, and,
e. size.
Page 10 of 24
Check answer with Excel spreadsheet
′
𝜎𝑢 758 MPa 𝜎𝑒𝑟𝑏 211.93 MPa
𝑘𝑡 ′
1.8 𝜎𝑎𝑟𝑏,10 3 551.72 MPa
𝑑 100 mm 𝐶𝑙𝑜𝑎𝑑 1.00
𝑇 550 𝐶 𝐶𝑠𝑖𝑧𝑒 0.86
𝑎 𝐶𝑠𝑢𝑟𝑓
0.148 mm 0.78
𝑟 𝐶𝑇
10 mm 1.00
𝐾𝑓 1.79 𝐶𝑟𝑒𝑙 0.75
𝑞 0.3 𝜎𝑒𝑟 106.77 𝑁𝑒 1.00E+06
𝐾𝑓′ 1.24 𝜎𝑎𝑟,103 415.44 𝑁103 1.00E+03
𝑏 1.97E-01

𝜎𝑎 𝜎 𝜎 𝑎 𝜎𝑎𝑟 𝑁 𝐷
[MPa] [MPa] [cycles] [ MPa] [MPa]
200 80 10 000 280 223.6 23329.4 0.43
300 0 2 000 300 300.0 5234.6 0.38
200 -80 5 000 120 180.9 68506.8 0.07
Total damage per repitition 0.88
Number of repetitions to crack initiation 1.13
Period per repetition 2 year
Life to crack initiation 2.26 year

Page 11 of 24
Question 2
The material used in this problem is steel.
• Modified Goodman was demonstrated in class to give a good correlation for steel, and, makes
provision for compressive stresses. It is a conservative method and can be used where there
are marginal uncertainties present in material fatigue properties and/or loading.
• Morrow relationship with 𝜎𝑓′ , the ordinate at one reversal (half-cycle), fits data well for steels.
However, in this case the modified fracture strength is not available.
• SWT was found to be a good choice for Aluminium.
o Because the SWT mean stress compensation does not rely on any material constant,
it is an attractive method to use that give reasonable results for Aluminium alloys.
o SWT gives reasonable results for steel.
Allocate marks according to the candidate’s reasoning of the above.

Page 12 of 24
2 STRAIN LIFE [60 Marks – one mark/minute]
Figure 3 shows strain vs life curves for RQC-100 steel. For each of the several tests, elastic, plastic,
and total strain data points are plotted versus life, and fitted lines are also shown. The total strain is
given as follows by the Ramberg-Osgood equation:
𝜀𝑎 = 𝜀𝑒𝑎 + 𝜀𝑝𝑎
1
𝜎𝑎 𝜎𝑎 𝑛′
= + ( ′)
𝐸 𝐻
The total curve is given by:
𝜎𝑓′ 𝑏 𝑐
𝜀𝑎 = (2𝑁𝑓 ) + 𝜀𝑓′ (2𝑁𝑓 )
𝐸
As shown, the plastic part of the relationship above dominates fatigue life (endurance) at high strain
amplitudes and low cycles to failure.
The elastic part of the relationship governs fatigue life (endurance) at low strain amplitude and high
cycles to failure.
Material properties of the material is given in Table 14.1 of the prescribed textbook.
Estimate:
1. The endurance limits for the strain amplitude spectrum shown in Table 2 below.[6/14 x 60 marks]
2. The total damage. [2/14 x 60 marks]
3. The number of repetitions to failure. [1/14 x 60 marks]
4. The transition point from strain-life to stress-life, 𝑁𝑡 and 𝜀𝑎𝑡 . [3/14 x 60 marks]
5. Say you used the Ramberg-Osgood stress-strain relationship to calculate the stress amplitude
and mean from the supplied strain, and you perform fatigue calculation with the stress-life part
of the total strain-life curve only, would the number of repetitions be lower or higher than that
calculated in the estimation in Question 3 above? Explain.
[2/14 x 60 marks]
Please submit the next page with your answering sheet if you used the graphical method. Clearly
indicate the levels with numbers on the graph.

Page 13 of 24
 PLEASE SUBMIT THIS PAGE WITH YOUR ANSWERING SHEET

STUDENT NUMBER: .

Table 2: Strain spectrum at a notch

𝜀𝑎 𝑖 𝑁𝑖 𝐷𝑖

9.E-03 200
4.E-03 1 000
3.E-03 1 000
Total damage
Number of reptititions

𝜎𝑓′ 𝑏
𝜀𝑎 = (2𝑁𝑓 )
𝐸

𝑐
𝜀𝑎 = 𝜀𝑓′ (2𝑁𝑓 )

Figure 3: Strain vs life curves for RQC-100 steel

Page 14 of 24
2.1 Solution [scale marks to 60]
For information, to interpolate on a logarithmic scale using a ruler:
log10 𝑁 − log10 𝑥1 𝑎
=
log10 𝑥2 − log10 𝑥1 𝑏
𝑎
log10 𝑁 = (log10 𝑥2 − log10 𝑥1 ) + log10 𝑥1
𝑏
𝑎
(log10 2 −log10 1 )+log10 1 )
𝑁 = 10(𝑏

𝑏
𝑎

𝑥1 𝑁 𝑥1

2.1.1 Endurance [6 marks]

In this case the endurances were determined from the S-N curve shown below:
For 𝜺𝒂 = 𝟎. 𝟎𝟎𝟗:
𝑎
(log 1 )+log10 1 )
𝑁 = 10(𝑏 10 2−log10 2 marks
= 1 000 𝑐𝑦𝑐𝑙𝑒𝑠
For 𝜺𝒂 = 𝟎. 𝟎𝟎𝟒:
𝑎
(log 1 )+log10 1 )
𝑁 = 10(𝑏 10 2−log10
= 7 000 𝑐𝑦𝑐𝑙𝑒𝑠
2 marks
For 𝜺𝒂 = 𝟎. 𝟎𝟎𝟑:
𝑎
(log10 2 −log10 1 )+log10 1 )
𝑁 = 10(𝑏
12.1 5 4 4
= 10(31.1(log 10 −log 10 )+log 10 ) 2 marks
= 24 494 𝑐𝑦𝑐𝑙𝑒𝑠

Page 15 of 24
 𝑎
(log10 2 −log10 1 )+log10 1 )
𝑁 = 10(𝑏
12.1 5 4 4
= 10(31.1(log 10 −log 10 )+log 10 )
= 24 494 𝑐𝑦𝑐𝑙𝑒𝑠

(0.004;24 494)
(0.009;1 000) (0.004;7 000)

2.1.2 Total damage [2 marks]

The total damage is:
200 1000 1000
𝐷= + + 2 marks
1000 7000 24494
= 0.384
2.1.3 Number of repetitions to failure [1 mark]
The number of repetitions to failure is:
1
𝐵𝑓 =
𝐷 1 mark
= 2.60
2.1.4 Transition point [3 marks]
The transition point on the fatigue curve is where the parts of the elastic and plastic components are
equal to each other. Therefore:
𝜎𝑓′
(2𝑁𝑡 )𝑏 = 𝜀𝑓 (2𝑁𝑡 )𝑐
𝐸
𝜀𝑓′ 𝐸
(2𝑁𝑡 )𝑏−𝑐 = ′
𝜎𝑓
1
𝜀𝑓′ 𝐸 𝑏−𝑐
( ′) 2 marks
𝜎𝑓
𝑁𝑡 =
2

Page 16 of 24
 1
1.38 × 200 000 −0.0648+0.704
( 938 )
=
2
= 3 640 𝑐𝑦𝑐𝑙𝑒𝑠
At this endurance, the transition strain is:
𝜎𝑓′
𝜀𝑎𝑡 = 2 × (2𝑁𝑡 )𝑏
𝐸 1 mark
= 0.0053

2.1.5 Stress-life only [2 marks]

The Ramberg-Osgood stress-strain relationship is as follows:
𝜀𝑎 = 𝜀𝑒𝑎 + 𝜀𝑝𝑎
1
𝜎𝑎 𝜎𝑎 𝑛′
= + ( ′)
𝐸 𝐻
If the stress amplitude is calculated from this relationship, the value would be 𝜎𝑎 .
The total fatigue curve is given by:
𝜎𝑓′ 𝑏 𝑐
𝜀𝑎 =(2𝑁𝑓 ) + 𝜀𝑓′ (2𝑁𝑓 )
𝐸
If the elastic part of the Ramberg-Osgood equation is equated to the first part of the total fatigue curve,
the answer is for, say 10 000 cycles:
𝑏
𝜎𝑎 = 𝜎𝑓′ (2𝑁𝑓 ) Mark as follows:
−0.0648
= 938(2𝑁𝑓 ) 2 marks if the student
= 494 𝑐𝑦𝑐𝑙𝑒𝑠 demonstrates understanding of the
For the plastic part of the curve we have: answer.
1 They will provide different answers
𝜎𝑎 𝑛′ 𝑐
( ) = 𝜀𝑓′ (2𝑁𝑓 ) – mark accordingly.
𝐻′
′ 1 mark for a student who answered
𝑐 𝑛
𝜎𝑎 = 𝐻 ′ (𝜀𝑓′ (2𝑁𝑓 ) ) “yes” with no logic explanation.
−0.704 0.0905
= 903 (1.38(2𝑁𝑓 ) )
−0.0637
= 929(2𝑁𝑓 )
= 495 𝑐𝑦𝑐𝑙𝑒𝑠
Therefore, the number of cycles to failure will be approximately the same as when the total stress-strain
relationship was used with the total fatigue curve.

Page 17 of 24
3 WELD FATIGUE [60 Marks – one mark/minute]
Note, to obtain maximum points for each problem clearly motivate solutions and equations used.
The steel H-beam shown in Figure 5 is joined by a complete joint penetration butt weld (made from both
sides) with no cope holes. The steel has yield strength 350 MPa and ultimate tensile strength 490 MPa
and welding was done with a perfectly matched electrode.
The web and flange thickness is 35 mm.
The height of the weld convexity is less than 10% of the weld width with smooth transition to the plate
surface.
Weld run-on and run-off pieces were used during welding and removed. The edges were then ground
flush in the direction of the stress. An NDT inspection revealed no defects in the weld.
The operating temperature is 250 °C and the surface is corrosion protected.
Your class notes have extractions of some detail categories from BS EN 1993-1-9.
Please answer the following:
1. What is the fatigue life for 95% probability of survival of the component, in years, for the nominal
axial stress signal given in Figure 4 below? Assume a high consequence of failure and damage
tolerant design and apply Rainflow counting to calculate the stress spectrum. [55 Marks]
2. Will this detail category benefit from weld toe dressing? [5 Marks]

Stress signal over 30

minutes

10 cycles

10 000 cycles

Figure 4: Stress signal over a period of 30 minutes

Page 18 of 24
 Complete joint penetration butt weld

Figure 5: Weld-spliced H-beam under axial loading

Page 19 of 24
3.1 Answers
3.1.1 Question 1 (16 minutes – allow 35 minutes) [55 Marks – one mark/minute]
3.1.1.1 Rainflow counting and the stress spectrum
From the figure below, the stress spectrum over 30 minutes is as follows:
𝚫𝝈 Number of cycles, 𝒏𝒊
[MPa]
125 1
75 10
25 10 000

Stress signal over 30

minutes

1x cycles of 125 MPa range

10x cycles of 75 MPa range

10 000 cycles of 25 MPa range

3.1.1.2 Partial factor for fatigue

From the formula sheet the partial factor for fatigue for high consequence of failure and damage
tolerant assessment method is 1.15.
Consequence of failure
Assessment method
Low consequence High consequence
Damage tolerant 1.00 1.15
Safe life 1.15 1.35
Source: BS EN 1993-1-9, 2005:11

Page 20 of 24
3.1.1.3 Detail category
The detail category is 90 because the welds were not all ground flush to the plate. This is the
strength at 2 000 000 cycles at room temperature for 95% probability of survival (the design Sr-N
curve)

3.1.1.4 Thickness compensation

The factor to compensate for thickness more than 25 mm is:
25 0.2
𝑘𝑠 = ( )
𝑡
25 0.2
=( )
35
= 0.93

Page 21 of 24
3.1.1.5 Temperature compensation
At 250 ℃ the compensation for temperature is 𝐶𝑇 = 0.90.

3.1.1.6 Constant amplitude fatigue limit

The characteristic strength is given at 2 million cycles, and the constant amplitude fatigue limit at 5
million cycles. Therefore, the thickness modified constant amplitude fatigue limit is:
1
2 3 Δ𝜎𝐶
Δ𝜎𝐷 = ( ) ( 𝑘𝐶 𝐶 )
5 𝛾𝐹 𝑠 𝑇 𝑃𝑊𝑇
1
2 3 90
=( ) × × 0.93 × 0.90 × 1.0
5 1.15
= 48.3 𝑀𝑃𝑎
3.1.1.7 Cut-off limit
The cut-off limit is:
1
5 5
Δ𝜎𝐿 = ( ) Δ𝜎𝐷
100
1
5 5
=( ) 53.6
100
= 26.5 𝑀𝑃𝑎
3.1.1.8 Sr -N curve
The fatigue curve for the weld detail is then shown below from which the endurance at any stress
range can be calculated:
Δ𝜎𝐶 1
( ) 𝑁𝐶 𝑚1 = 3 𝑓𝑜𝑟 1.5𝑓𝑦 > Δ𝜎𝑅 ≥ Δ𝜎𝐷
Δ𝜎𝑅
𝑁𝑅 = Δ𝜎𝐷 2
( ) 𝑁𝐷 𝑚2 = 5 𝑓𝑜𝑟 Δ𝜎𝐷 > Δ𝜎𝑅 ≥ Δ𝜎𝐿
Δ𝜎𝑅
{ ∞ 𝑓𝑜𝑟 Δ𝜎𝑅 < Δ𝜎𝐿

3.1.1.9 Calculations
For 𝚫𝝈𝑹 = 𝟏𝟐𝟓, 𝒏𝟏 = 𝟏
The endurance at this stress range is then:

Page 22 of 24
 Δ𝜎𝐷 1
𝑁𝑅 = ( ) 𝑁𝐷
Δ𝜎𝑅
48.3 3
=( ) × 5 × 106
125
= 288 457 𝑐𝑦𝑐𝑙𝑒𝑠
1
The damage is then: 𝐷1 = = 3.47 × 10−6
288 457
For 𝚫𝝈𝑹 = 𝟕𝟓, 𝒏𝟐 = 𝟏𝟎 𝚫𝝈 Number
[MPa] of
The endurance is: cycles,
Δ𝜎𝐷 3 𝒏𝒊
𝑁𝑅 = ( ) 𝑁𝐷
Δ𝜎𝑅 125 1
48.3 3
=( ) × 5 × 106 75 10
75
= 1 335 450 25 10 000
The damage is then:
10
𝐷2 =
1 335 450
= 7.49 × 10−6
For 𝚫𝝈𝑹 = 𝟐𝟓, 𝒏𝟐 = 𝟏𝟎 𝟎𝟎𝟎
This stress range is below the cut-off limit and will not do any damage. Therefore:
𝐷3 = 0
3.1.1.10 Total damage and fatigue life
The total damage over 30 minutes is:
𝐷𝑡𝑜𝑡𝑎𝑙 = 3.47 × 10−6 + 7.49 × 10−6
= 1.10 × 10−5
The number of repetitions to 5% probability of crack initiation in logN is then:
1
𝐵𝑓 =
𝐷𝑡𝑜𝑡𝑎𝑙
= 90 909
If one repetition was 30 minutes, then the fatigue life in days and years is:
𝐿 = 0.5/24 × 𝐵𝑓
= 1 893 𝑑𝑎𝑦𝑠
= 5.2 𝑦𝑒𝑎𝑟𝑠

Page 23 of 24
3.1.1.11 Verification in Excel
Detail category 90 Δ𝜎𝐶, 𝑜𝑑 65.9 MPa
𝛾𝑀𝑓 1.15 𝑁𝐶 2.00E+06
Thickness 35 mm 𝑁𝐷 5.00E+06
𝐶𝑡 0.9349199 Δ𝜎𝐷 48.5 MPa
𝐶𝑇 0.9 𝑁𝐿 1.00E+08
m1 3 Δ𝜎𝐿 26.7 MPa
m2 5
Δ𝜎𝑅 𝑖 𝑁𝑅 Damage Marking:
[MPa]
1 mark for gamma MF
125 1 292 405 3.41992E-06
1 mark for Ct for thickness
75 10 1 353 726 7.38702E-06
25 10 000 inf 0 1 mark for Delta Sigma D
Total damage 1.08069E-05 1 mark for Delta Sigma L
Period 30 minutes 3 marks for correct damage
Fatigue life 1 928 days 1 mark for total damage – mark with error
5 years
1 mark for fatigue life – mark with error

3.1.2 Question 2 [5 Marks – one mark/minute]

NO, in this case the surface was ground flush at the edges, and all stress concentrations were removed.
Then weld toe dressing will not necessarily improve the situation at the edges.
The weld detail will benefit from post-weld toe dressing if it was not ground because the weld toe is the
point of expected crack initiation. The nominal stress is also transverse to the weld.

Marking:
100% for motivation of answer. Consider both as acceptable answers if properly motivated –
taking into consideration that the weld is ground.
2 marks for yes as answer with motivation.

4 References
Dowling, N. (2013). Mechanical Behavior of Materials: Engineering Methods for Deformation,
Fracture, and Fatigue (4th ed.). Boston: Pearson.

Page 24 of 24