EG-262 Stress Analysis 1

Failure Theories and Stress Concentration Effects

4.1 Failure Mechanisms

In this module, we will consider ways in which design can predict and aviod failure. It is
therefore logical to consider the extensive number of different ways in which components can
fail.

 Mechanical overload (or underdesign)

 Ductile failure
 Brittle fracture
 Elastic yielding (due to applied force and/or temperature)
 Fatigue (high cycle, low cycles, thermal, corrosion, fretting….)
 Creep
 Corrosion (chemical, galvanic, cavitation, pitting…)
 Impact
 Instability (buckling)
 Wear (adhesive, abrasive, corrosive…)
 Vibration
 Environmental (thermal shock, radiation damage, lubrication failure)
 Contact (spalling, pitting, seizure)

4.2 Ductile Failure

When elastically strained, a ductile material will plastically deform and necking will be
observed. Small cavities will form in the internal cross sectional area and will grow and join
together upon further application of stress. These cavities will continue to link together and the
crack will propagate parallel to its major axis until eventual fracture by shear deformation at an
angle of about 45o to the applied stress. In many cases it is possible to detect ductile crack
growth prior to, and thus preventing, catastrophic failure.

1
4.3 Brittle Fracture
Only very limited plastic flow occurs during brittle fracture. A crack, once initiated, can
propagate very rapidly without warning. The crack will spread perpendicular to the applied
stress and is classed as unstable. Usually, the crack develops by cleavage whereby atomic bonds
are broken.

2
The crack can propagate through the grains (transgranular fracture) or along the grain
boundaries (intergranular facture). Intergranular propagation is usually only seen in a few
specific circumstances such as in the presence of creep, stress or corrosion – aggressive
environments.

Transgranular Fracture Intergranular Fracture

The main differences between ductile and brittle fracture are shown in the table below:

Ductile Brittle
Fracture Stress Greater than yield strength Lower than yield strength
Energy High Low
Absorption
Nature of Necking, rough fracture surface, No necking, shiny granular
Fracture linking up of cavities surface, cleavage or
intergranular
Type of Material Metals Ceramics, glasses
Crack Slow Fast
Propagation
Nature of Failure Plastic deformation warning, Little deformation, more
less catastrophic catastrophic

It is important to note that the working temperature of a material will have an effect on its
ability to absorb energy prior to fracture. For example, steel has a high toughness and behaves
in a ductile manner at room temperature, but will exhibit brittle behaviour at lower
temperatures. Some materials have what is termed as a ductile-brittle transition temperature.
It is imperative to take into account the likely working temperature during the design stage.

3
4.4 Fatigue
So far we have only considered constant loads (forces, moments, torques etc). The behaviour of
engineering materials under static loading conditions is well understood. However, in practice
most engineering components are subjected to loading that varies with time (wind, waves,
bending of a crankshaft, pulsations in pipes…)

It has been discovered that many components subjected to a cyclic load, fail at stresses well
below the anticipated tensile strength, and sometimes even below the elastic limit of the
material. These failures generally occur after a large number of stress cycles (𝑁) and are hence
known as fatigue failures. We look at fatigue is more detail later in the module.

Low cycle fatigue (LCF) is when 𝑁 is < 10,000 (104) cycles

High cycle fatigue (HCF) is when 𝑁 is > 10,000 (104) cycles

4
4.5 Failure Theories
How do we test the strength of a material? Firstly it is useful to qualify to what type of strength
we are referring:

 Tensile
 Compressive
 Shear

In addition to this, as many materials are ductile, we also need to consider the following
strengths:

 Yield
 Ultimate tensile strength

In a simple tensile test, a dog bone specimen is often used with the weakest part in the middle.
A force is applied and the material will begin to yield at a given section. We measure the force
and extension and transform these to stress and strain (force to stress by dividing by cross-
sectional area, extension to strain by dividing by original length). The resultant stress (𝑦-axis)
strain (𝑥-axis) plot allows us to determine the UTS and yield stress.

In the simple case of a member subjected to uniaxial stress, it is common to base failure
prediction for ductile materials upon the yield stress.

In a one-dimensional stress state, there is one principal stress, 𝜎1 and yielding will start when:

𝜎1 = 𝜎𝑦𝑖𝑒𝑙𝑑

Or as we saw in our first lecture:

1
𝜏𝑚𝑎𝑥 = (𝜎 − 𝜎3 )
2 1
1 1 1
𝜏𝑚𝑎𝑥 = 𝜏𝑦𝑖𝑒𝑙𝑑 = (𝜎1 − 𝜎3 ) = 𝜎1 = 𝜎𝑦𝑖𝑒𝑙𝑑
2 2 2

We have seen that stress has multiple components. In the above, we are only measuring the
stress in one direction. In practice materials aren’t often loaded in such a simplistic manner, the
strengths of materials under complex stress systems are not generally known but such stress
systems are the most frequently encountered in reality. So we need a mechanism to allow us
to predict the failure (yielding or rupture) for loading cases when there is more than one stress
component.

5
Let us consider a simple example, the comparison between a simple tensile bar test and a
pressure vessel with an internal pressure.

We need to develop ways to use a yield stress calculated based on one stress component to
compare to a more complex real-life situation. This is where failure theories and failure
criterion come into play. They allow us to determine allowable working stresses and thus avoid
failure. Such theories aim to predict, from the behaviors of materials in a simple tensile test,
when elastic failure will occur under any condition of applied stress.

Predicting the yielding of ductile materials or the fracture of brittle materials depends on a
Theory of Failure.

A good failure theory requires:

 Observation of a large number of test results in different load conditions (empirical

knowledge)
 A theory explaining the microscopic mechanisms involved which must be consistent with
the observations

In this way, we can have a high level of confidence that not only does a given theory of failure
work for a loading condition that we have observed, but that it will also apply to other
conditions that have not been exhaustively tested.

Various theoretical criteria have been proposed to obtain adequate correlation between the
estimated component life and the actual life achieved in service. We will consider two of the
most commonly used theories for ductile materials.

Our chosen failure theory needs to be appropriate to the type of failure that will occur – for
example we couldn’t apply the same failure theory to glass and aluminium. Ductile failure
characterized by yielding tends to occur by shearing of material so we need a failure theory
which is based around shear stress.

6
4.5.1 Tresca Criterion (1868) – Maximum Shear Stress Theory
This theory considers that failure (yielding) will occur when the maximum shear stress in the
complex stress state becomes equal to the material’s limiting shear strength in a simple tensile
test.

We are essentially extrapolating the results from a simple tensile test to give us information
about failure under a more complex stress situation.

Since the maximum shear stress is equal to half the greatest difference between two principal
stresses:

1
𝜏𝑚𝑎𝑥 = (𝜎 − 𝜎3 )
2 1

And since the maximum shear in simple tension is equal to half the tensile stress at yield:

1
𝜏𝑚𝑎𝑥 = 𝜏𝑦𝑖𝑒𝑙𝑑 = 𝜎𝑦𝑖𝑒𝑙𝑑
2
Equating these gives:

1 1
𝜎𝑦𝑖𝑒𝑙𝑑 = (𝜎1 − 𝜎3 )
2 2

𝜎𝑦 = 𝜎1 − 𝜎3

Where the value of 𝜎3 is algebraically the smallest value (taking into account the sign and the
fact that one stress may be zero).

E.g. if 𝜎1 = 50𝑀𝑝𝑎 and 𝜎3 = −20𝑀𝑃𝑎 then 𝜎1 − 𝜎3 = 70𝑀𝑃𝑎.

Remember that in 2D analysis where the two principal actresses are 50MPa and 15MPa, the
other principal stress is zero thus:

𝜎1 − 𝜎3 = 50 − 0 = 50𝑀𝑃𝑎 NOT 50 − 15 = 35𝑀𝑃𝑎

This criterion has been shown to give reasonably accurate prediction of failure, especially for
ductile materials.

7
4.5.2 Von Mises Criterion (1913) - Maximum Shear Strain Energy Theory
The strain energy of a stressed component can be divided into two components, volumetric
strain energy (associated with volume change but no distortion) and shear strain energy
(associated distortion of the stressed elements). This theory states that failure will occur when
the maximum shear energy component in the complex stress system is equal to that at the
yield point in a simple tensile test.

Defining a von Mises equivalent stress 𝜎𝑒𝑞 where:

For Principal Stresses:

1
𝜎𝑒𝑞 = √(𝜎1 − 𝜎2 )2 + (𝜎2 − 𝜎3 )2 + (𝜎1 − 𝜎3 )2
√2

For components of stress:

1
𝜎𝑒𝑞 = √(𝜎𝑥𝑥 − 𝜎𝑦𝑦 )2 + (𝜎𝑦𝑦 − 𝜎𝑧𝑧 )2 + (𝜎𝑥𝑥 − 𝜎𝑧𝑧 )2 + 6(𝜏𝑥𝑦 2 + 𝜏𝑦𝑧 2 + 𝜏𝑥𝑧 2 )
√2

It can be shown that yielding occurs when:

𝜎𝑒𝑞 = 𝜎𝑦𝑖𝑒𝑙𝑑

When 𝜎𝑦𝑦 = 𝜎𝑧𝑧 = 𝜏𝑦𝑧 = 𝜏𝑥𝑧 = 0 i.e. there is one direct stress, 𝜎𝑥𝑥 and one shear stress, 𝜏𝑥𝑦
only (such as in bending and torsion)

𝜎𝑒𝑞 = √(𝜎𝑥𝑥 )2 + 3(𝜏𝑥𝑦 2 )

This theory is widely regarded as the most reliable basis for design and has received significant
practical verification, particularly for ductile materials.

8
Example 1:
If 𝜎1 = 200𝑀𝑃𝑎 and 𝜎2 = 100𝑀𝑃𝑎, calculate the limiting value of 𝜎3 to avoid yielding in
accordance with Tresca and Von Mises. The yield stress 𝜎𝑦 = 300𝑀𝑃𝑎.

Example 2:
A 2m length of 20mm diameter steel bar is subjected to a torque of 150Nm. What additional
bending moment can be applied to the bar before the material started to yield according to the
Tresca and Von Mises criteria? The yield stress 𝜎𝑦 = 300𝑀𝑃𝑎.

9
4.6 Safety Factor:
It is common practice in design to apply a safety factor (SF). Introducing a safety factor into the
above criteria gives;

Tresca:
𝜎𝑦
𝜎1 − 𝜎3 ≤
𝑆𝐹
Von Mises:
𝜎𝑦𝑖𝑒𝑙𝑑
𝜎𝑒𝑞 ≤
𝑆𝐹

The value used for the SF depends on the confidence in the accuracy of the stress values.

10
Example 3:
A thin mild steel tube has a mean diameter of 100mm and a thickness of 3mm. Determine the
maximum torque that the tube can transmit according to the Tresca and von Mises criteria and
using a Safety Factor of 2.25.

Assume a constant shear stress (thin tube) and a yield stress of 230 MPa.

11
4.7 Graphical representation of failure theories for 3-dimensional stress systems

4.7.1 Von Mises:

The most reliable theory seen above gives:

1
𝜎𝑒𝑞 = √(𝜎1 − 𝜎2 )2 + (𝜎2 − 𝜎3 )2 + (𝜎1 − 𝜎3 )2
√2

Squaring both sides:

2𝜎𝑦 2 = 2𝜎𝑒𝑞 2 = (𝜎1 − 𝜎2 )2 + (𝜎2 − 𝜎3 )2 + (𝜎1 − 𝜎3 )2

This equation defines the surface of regular prism having a circular cross-section. A cylinder
with its central axis along the line 𝜎1 = 𝜎2 = 𝜎3 . The axis therefore passes through the origin of
the principal stress coordinate system and is inclined at the same angle to each axis.

The yield locus is defined by the surface of the inclined cylinder. Any point that lies within the
cylinder represents safe conditions and any point that lies outside the cylinder indicates failure
conditions.

12
Example 4:
Determine whether the below are in the elastic of plastic zone.

i) 𝜎𝑦 = 200𝑀𝑃𝑎, 𝜎1 = 150𝑀𝑃𝑎, 𝜎2 = 200𝑀𝑃𝑎, 𝜎3 = 250𝑀𝑃𝑎

ii) 𝜎𝑦 = 200𝑀𝑃𝑎, 𝜎1 = 150𝑀𝑃𝑎, 𝜎2 = 200𝑀𝑃𝑎, 𝜎3 = 400𝑀𝑃𝑎

iii) 𝜎𝑦 = 200𝑀𝑃𝑎, 𝜎1 = 150𝑀𝑃𝑎, 𝜎2 = −200𝑀𝑃𝑎, 𝜎3 = 400𝑀𝑃𝑎

13
4.7.2 Tresca:
Although not considered as reliable as Von Mises, in the majority of cases the simpler Tresca
theory can be reasonably applied to assess the yielding of most ductile materials.

The Tresca theory can be shown to produce a prismatic three-dimensional yield locus with
hexagonal cross-section. The central axis of this diagram will again lie on the line 𝜎1 = 𝜎2 = 𝜎3 .
Any point that lies within the hexagon represents safe conditions and any point that lies outside
the indicates failure conditions

4.8 Comparison of Von Mises and Tresca Surfaces:

14
Experimental testing suggests that:

 Behaviour in tension is the same as in compression

 There is no volume change associated with yielding
 Yielding will never occur if all three principal stresses are the same (prove it)!

The Tresca criterion is always more conservative than the von Mises criterion, except at six
positions on the surfaces (where the hexagon and circle coincide).

15
4.9 Stress Concentrations
Mechanically and/or thermally loaded components have stress distributions set up in them
which maintain equilibrium with externally applied loads.

To date, the problems analysed have resulted in stress distributions which are either uniform or
vary smoothly and gradually over the section. A uniform cross-section bar subjected to an axial
tensile or compressive load is assumed to experience uniform stress across the section.

However, in practice the presence of any sudden change of section can cause the local stress to
rise significantly and rapidly over a short distance.

The stress gradient (𝑑/𝑑𝑥) in one area of a loaded component can be orders of magnitude
greater than in other areas of the same component. For example, in the vicinity of the point of
application of a concentrated load where the maximum stress can be much higher than the
average or nominal value.

This situation is called a stress concentration. Stress concentrations are also produced at
geometric discontinuities in a component such as holes, notches, keyways, fillets and material
flaws.

The effect of a stress concentration is only local and may not generally be a problem under
static loading since local yielding will cause a redistribution of stress which will relieve the high
stresses.

Also St. Venant’s Principle states that the actual distribution of the load over the surface of its
application will not affect the distribution of stress or strain on sections of the body which are
at an appreciable distance (relative to the dimensions) away from the load.

e.g. for a rod in simple tension

The two loading conditions are statically equivalent and at the Section AA the stress is
reasonably uniform and equal to 𝑙𝑜𝑎𝑑/𝑎𝑟𝑒𝑎 so long as ‘𝑥’ is greater than 3 times the diameter.

16
4.9.1 Geometric Discontinuities
Abrupt changes in geometry give rise to stress concentrations and, as already stated, these high
stresses may not be a problem under static loading conditions.

However, failure can very often be attributed to a stress concentration due to a hole, fillet,
keyway etc. This is because stress concentrations play a significant part in the failure of
components due to fatigue (and to an extent brittle fracture). The fatigue life of a component,
in terms of the number of cycles to failure, is a function of the stress range being experienced
by the component.

Fatigue calculations which ignore the presence of a stress concentration will therefore over-
predict the life of that component since the stress range in the region of the stress
concentration will be significantly greater than the nominal value.

The stress concentration due to a geometric discontinuity is a function of the shape and
dimensions of the discontinuity and is expressed in terms of an elastic stress concentration
factor, 𝐾𝑡 .

This is the ratio of the maximum stress occurring near the discontinuity to the nominal stress at
the section in the absence of a stress concentration.
𝜎𝑚𝑎𝑥
𝐾𝑡 =
𝜎𝑛𝑜𝑚

This factor is constant within the elastic range of the material.

A typical example of a stress concentration is the classical hole-in-a-plate problem. A hole

machined in a plate produces both an increase in the mean stress and also a peak in stress at
the surface of the hole. This peak stress can be up to three times the mean value.

17
The presence of a stress concentration can be explained clearly using a stepped bar as an
example. At the step there is a discontinuity in stress, due to the change in section.

There is a significant increase in the stretching of the flow lines in the region of the
discontinuity; hence the strain and stress are higher in this region than along the rest of the
lines.

The larger the fillet radius at the step, the smoother the transition and hence the less
noticeable is the stress concentration. Therefore, the smaller the fillet radius the greater is the
stress concentration. The maximum stresses occur at the surface of the fillet where maximum
lateral displacements and hence maximum stretching occurs.

The theoretical analysis of stress concentrations can be very complex. Many classical problems
have been solved and results are available.

18
Where theoretical solutions are impossible, photoelastic and finite element techniques have
been used to obtain results and again these are available.

“Peterson’s Stress Concentration Factors” by Pilkey contains many results for a large number of
component configurations and loading conditions.

Some typical examples are shown below for a flat bar with holes, and a flat bar with fillets:

19
4.9.2 Static Load Design
This simplifies the work of a designer. All they need to do in order to determine the maximum
stress occurring near a discontinuity in a given member subjected to an axial load, P is calculate
the average or nominal stress 𝜎𝑛𝑜𝑚 = 𝑃⁄𝐴 in the critical section. This result is then multiplied
by the appropriate stress concentration factor, 𝐾𝑡 .

𝜎𝑚𝑎𝑥 = 𝐾𝑡 𝜎𝑛𝑜𝑚

One of two possible design models can then be applied.

Elastic only design model when all the stresses must be elastic and the nominal stress must be
such that the maximum stress does not exceed yield. A factor of safety is often employed.

Limited plasticity where local yielding is allowed in the region of the stress concentration and
hence the nominal stress can be significantly closer to yield within some factor of safety.

4.9.3 Fatigue Design

The effect of a notch (due to a hole, fillet etc.) on the fatigue strength of a component depends
on the material and the notch geometry and is generally less than the effect that would be
predicted by the straight application of an elastic stress concentration factor. The notch
sensitivity relates fatigue notch factor (used to determine fatigue life) with the elastic stress
concentration factor:
𝐾𝑓 = 𝑞(𝐾𝑡 − 1) + 1

Where:

𝑞 is the notch sensitivity

𝐾𝑓 is the fatigue notch factor
𝐾𝑡 is the elastic stress concentration factor

Notch sensitivity data is available (see Pilkey for references to 𝑞 data).

𝐾𝑓 is then used to determine the elastic stress range used for fatigue life calculations.

20
Example 5:
Determine the stresses in the fillets of the component shown below due to an axial load of
100N. The thickness of the part is 1mm.

21