Download as pdf or txt
Download as pdf or txt
You are on page 1of 36

DR ANNA BAKER ANNA.B.BAKER@BRISTOL.AC.

UK

D R M AT T H E W P E E L M AT T H E W . P E E L @ B R I S T O L . A C . U K

ENGINEERING SCIENCE
MENG10004

M AT E R I A L S - S E L E C T I O N
Contents
1 Translation and Design-limiting properties 3
1.1 Bar charts . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Bubble charts . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Design process . . . . . . . . . . . . . . . . . . . . . . . 3
1.4 Selection Process . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Selecting on Physical Properties 11


2.1 Electrical Properties . . . . . . . . . . . . . . . . . . . . . 11
2.2 Thermal Properties . . . . . . . . . . . . . . . . . . . . . 12

3 Stiffness-Limited Design 19
3.1 Constraints for Bending . . . . . . . . . . . . . . . . . . 20
3.2 Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

4 Strength-Limited Design 25
4.1 Strong, Light Tie . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Strong, Light Beams . . . . . . . . . . . . . . . . . . . . 26

5 Selecting Manufacturing Processes 31


5.1 Selection Strategy . . . . . . . . . . . . . . . . . . . . . . 31
5.2 Cost of Manufacture . . . . . . . . . . . . . . . . . . . . 34
1 Translation and Design-limiting properties

This lecture covers how we can turn a design idea into some critical
materials properties that should be used to make a rational choice of
material.

Intended Learning Outcomes


Skills Identification of design-limiting attributes.
Generate translation tables for designs.
Selecting on the basis of min/max attribute values on design charts.

1.1 Bar charts

Individual records are not useful for general selection. We prefer


to visualise many hundreds of materials simultaneously. The easi-
est plots are bar-charts of a single property such as elastic modulus
(Figure 1.1), tensile strength or density. Generally it is necessary to
use a log scale as properties often vary by orders of magnitude. E.g.
in figure 1.1 polymers are 2-5 orders of magnitude lower modulus
than metals or ceramics.

Figure 1.1: Single properties can be vi-


sualised in 1D plots. Log scales are usu-
ally used due to the many decades dif-
ference between materials.

1.2 Bubble charts

Bubble charts plot two properties forming a 2D property space. The


commonalities between families and classes of materials becomes
very obvious in such plots as families fall within boundaries on the
chart. Each material is shown as a bubble since the data encapsu-
lates all the varieties available. Simplified charts show fewer, larger
bubbles as they encompass a greater range of materials1 . 1
So the bubble for ’steel’ would be
much bigger than the bubbles for ’high
carbon steel’ and ’low carbon steel’.
1.3 Design process
Design starts with a market need, which is expressed as design re-
quirements. In some case design is entirely original, perhaps stimu-
lated by a new material. In most cases we are actually redesigning a
Figure 1.2: Two properties can be visu-
alised in 2D bubble plots. Materials are
grouped into family bubbles that oc-
cupy characteristic portions of the plot.

product to overcome some shortcoming. Reasons include:


1. Product-recall: overcome urgent failure.
2. Poor value for money: it works but performance needs to be increased.
3. Inadequate profit margin: cost of manufacture exceeds the price the market will bear.
4. Sustainable technology: response to consumer pressure to reduce harm to the environment.
5. The Mac effect: Style, character and feel dominate otherwise identical products.

Figure 1.3: The design flow chart show-


ing how material selection enters the
process. Information is needed at each
stage but only at appropriate levels of
detail.

Concepts that might fulfil the requirements are developed (embod-


ied) and refined (detailed) to produce a product specification. The
choice of material (and processing) must evolve in parallel and be-
come increasingly detailed as more is known. As an example, we
engineering science 5

might want to redesign a wine bottle opener. The need might be


expressed as ’a device to access wine in a bottle, conveniently, at
modest cost, and without contaminating the wine’. Initial concepts
(Figure 1.4) are high-level and material choice hardly figures but by
the end we need specific materials or we can’t do the calculations to
ensure it won’t break (Figure 1.5) .

Figure 1.4: Illustration of the process of


moving from a need to several general
concepts

Figure 1.5: Choosing a single con-


cept, trialling several embodiments be-
fore producing a detailed design.

1.4 Selection Process


Material selection follows a process:

1. Translation: The conversion of design requirements into attributes


that we can use to select our materials. This includes functions,
constraints, objectives and free variables.
Figure 1.6: Narrowing of search space
2. Screening: The constraints act as a filter by removing any materi-
as constraints are tightened
als that fail outright. The aim to to remove materials that can’t do
the job at all - they are subject to attribute limits.

3. Ranking: After clear failures have been removed we can use the
objectives to rank the remaining materials by how well they per-
form.
1.4.1 Translation

Translation requires us to identify design-limiting attributes. A col-


umn supporting a roof must support a high force in compression
so compressive strength is design-limiting. We don’t want a camera
lens to be easily scratched, but the optical properties are so impor-
tant that hardness may not be design-limiting. Then we need to pro-
duce a translation table. Note an important detail here though - the
limiting attribute for the column is the force it can support (load car-
rying capacity) and not the strength (property). Often, the limiting
attribute is related to a property but we can’t just assume that (e.g.
concrete has quite low strength but we can compensate by making
really big columns).

Aspect Description Examples


Function What does the component do?
Constraints Attributes that can’t be negotiated; the expected load, mass, cost, mode of
requirement is satisfied or not and the loading, length, shape, thickness,
material accepted or rejected. transparency, corrosion resistance.
Objectives Attributes that enhance performance. Mass, cost, heat loss, environmental
Optimising these is always desirable. impact.
Free Variables Attributes that can be altered. Material, thickness, shape

1.4.2 Screening and ranking

Typically, we try to make plots of only the most critical properties


for a given application based on the design-limiting attributes. In
Figure 1.7 the E vs density plot is suitable for an application where
a component must not deflect too much (stiffness-limited) but still
needs to be light-weight. The E vs price plot would be more help-
ful if we are more concerned with how much the material will cost.
Screening normally manifests as fixed lines separating suitable from
unsuitable materials. Ranking manifests as a list of materials or-
dered by the value of some property. Consider the difference in the
statements: "the material must have a modulus in excess of 60 GPa"
(screening) vs "we want the lowest price material possible that can
still do the job." (ranking - after screening).

Example 1.1 (Heat sinks) CPU chips generate heat and often run
close to 200◦C. Excess heat is removed by an attached heat sink.
The heat sink must be an electrical insulator or we short circuit the
chip.
engineering science 7

Figure 1.7: Changing one axes can alter


our perception of the best material for a
given case depending on our priorities.
Different materials are suitable for mass
optimised design (E vs density) or cost
optimised design (E vs cost per volume)

The operating temperature and electrical insulation are constraints,


while thermal conductivity is the objective. The operating temper-
ature is pass/fail, and increasing it does not help. The material is
either an electrical insulator and prevents shorts, or it is not, and
doesn’t. Conversely, a higher thermal conductivity is always better
since it will remove more heat. The dimensions are fixed so there are
no clever engineering tricks to exploit - just material properties.
We can explore the best material via a thermal conductivity vs
electrical resistivity bubble plot (Figure 1.9). Since the constraint and Figure 1.8: Heat sink on a CPU
objective are single properties they appear as vertical and horizon-
tal lines. The resistivity constraint is placed wherever we decide an
’insulator’ begins (presumably a function of design voltage). All ma-
Function Heat sink
Constraints Good electrical insulator Functional constraints
Max operating temperature >200◦C
Dimensions fixed Geometric constraint
Objectives Maximise thermal conductivity
Free Variables Choice of material

Figure 1.9: A bubble chart for selecting


a heat sink material for a CPU. Electri-
cal insulation is applied as a constraint
(threshold=1019 µΩ cm−1 ) - anything to
the right passes. Thermal conductivity
is a movable line - any material closer to
the top of the chart is better.

terials less resistive are rejected. The best material in the remaining
subset is the one closest to the top (highest thermal conductivity). If
aluminium nitride won’t do we can use alumina or silicon nitride as
the next best thing - although we might not be able to push the CPU
as hard.

1.5 Objectives

We often need to quantify the objective that we are seeking to opti-


mise. Sometimes, this is a material property of some kind e.g. max-
imise thermal conductivity for a heat exchanger. Often, we are inter-
ested in attributes that are not material properties e.g. volume, mass
or cost. We need equations to describe these that will depend on the
context. Some common objectives are:

1. Volume: We often need the smallest component that will do the


job. This will be determined from a combination of geometrical
constraints (say the length of a cylindrical shaft) and free variables
(say the radius of the same shaft). The objective equation is then
simply:

V = πr2 L

2. Mass: Very commonly we want the lightest component that will


engineering science 9

do the job. If we know the volume then we get the mass by

m = ρV = ρπr2 L

where ρ is the density of the material we choose.

3. Cost: If we want the cheapest component possible then we need to


include some measure of the material price. This can be found as a
function of price per unit volume (Cv ) or price per unit mass (Cm ).
In this course we mostly use the latter, so the objective equation
becomes, for the same shaft:

c = Cm ρV = Cm ρπr2 L

This only includes the raw material cost, of course, so issues of


manufacturing, transport, etc are missed out.

4. CO2 production: Increasingly, it is a concern how much pollution,


toxins or gasses contributing to global warming are released by
the materials we use. If we want to minimise these we can make
a basic attempt by using, say, the kg of CO2 released per kg of
material.

mCO2 = MCO2 ρV = MCO2 ρπr2 L

The property MCO2 is only an estimate from primary material pro-


duction (say digging up iron ore, transporting it then putting it in
a blast furnace to get the metal) and doesn’t capture regional or
technological variations.

5. Embodied Energy: In a similar manner, we might care how much


energy is consumed in turning a material from its raw form to a
form suitable for making something. If we want to minimise this
we can use the embodied energy of primary production (J/kg).

h = Hm ρV = Hm ρπr2 L

The property Hm is only the energy needed to make a basic ma-


terial (e.g. oil in the ground to polymer pellets to put in a press).
Further manufacturing will also contribute.

There are limitations on how confident we can be with such equa-


tions but they are often sufficient for obtaining a short list of suitable
materials. Importantly, these objectives are function of both a mate-
rial property (e.g. density) but also how much material you need.
This makes things more complex since we might choose a low den-
sity material but end up needing a lot of it since the material isn’t
very strong and it needs to support a load (constraint). Dealing with
coupling like this will be examined in later chapters.
2 Selecting on Physical Properties

Often we need to choose materials based on their physical proper-


ties. This includes thermal, electrical and optical properties. While
we can’t cover all the science we can investigate how selection can
proceed even on a basic treatment of the related phenomena.

Intended Learning Outcomes


Skills Select an appropriate physical property to describe an attribute.
Apply the properties as constraints or objectives
Rank materials on the basis of fulfilling objectives based around physical properties.

2.1 Electrical Properties

The resistivity of a material (ρe = A/LR) is essentially the resistance


(R = i/V) normalised by sample dimensions. The units are Ω.m but
µΩ.cm are reasonably common.

Figure 2.1: The electrical resistivity


plotted against the thermal conductiv-
ity.

The resistivity is related to the ease of movement of charged par-


ticles in a material. Metals have mobile electrons so they are typi-
cally excellent conductors (ρe ≈ 1µΩ.cm). Since free electrons carry
both charge and heat energy (see below) we find a linear relation-
ship between thermal and electrical conductivity for metals (Figure
2.1). Polymers, glasses and many ceramics have few mobile charged
particles and can’t carry meaningful currents under normal circum-
stances (ρe ≥ 1020 µΩ.cm).
We often want to exploit the electrical characteristics of materi-
als while also needing it to support loads without failing in some
way. An example is overhead power lines, which need low resis-
tivity while also supporting the weight of the wire. Pure copper
and aluminium offer excellent conductivity but have very low yield
strength (10s MPa), which limits the span between pylons. Increas-
ing the strength can be done by using alloys (Cu-Zn, Al-Mg) but the
alloying elements act like scattering centres in the lattice and get in
the way of the electrons. Precipitation strengthening is better, and
usually offers the best improvement in strength, but the effect on
conductivity is still notable. The best option is usually work hard-
ening as the extra dislocations have a very low impact on electron
movement and the resistivity is only slightly higher, even at high
strains. For this reason, most wires are very heavily cold worked -
which fits well with the ideal way of making a wire i.e. pull a thick
wire through a series of thinner dies until it’s thin and also very
strong.

Figure 2.2: Altering the strength of cop-


per and aluminium also alters the resis-
tivity.

2.2 Thermal Properties

Temperature is a measure of the average kinetic energy of atoms


and molecules. In a solid, they oscillate more around their mean
positions as the temperature increases. Heat is a flow of this energy
from regions of higher T to those of lower T.

2.2.1 Min-Max Service Temperature

Figure 2.3: The maximum sustained


temperature at which a material can be
used.

The minimum and maximum service temperature are the limit


for sustained, safe use of a material. Typically, this is constrained by
issues like oxidation, drops in strength or changes in chemical struc-
ture and as such the concept is surprisingly difficult to deal with.
We can easily read off a maximum service temperature but we won’t
know the reason unless we investigate in depth. Importantly, max
service temperature is usually much lower than the melting point of
engineering science 13

a solid.

Figure 2.4: The temperature depen-


dence of several properties. In this case
all are linear, or nearly so, with temper-
ature. Some increase, others decrease.

Many properties vary linearly with temperature, or nearly so, and


correcting the property is straightforward. Some properties, like
strength, can vary much less reliably and the change can be non-
linear and sometimes sudden. Particular care is needed for load
bearing materials at high temperature.

2.2.2 Thermal Conductivity

Materials that can move heat quickly from areas of high T to those
with low T have a high thermal conductivity. We measure this by
monitoring the heat flux through the a surface (W/m2 ) required to
maintain a temperature gradient between a hot side (T1 ) and cold
side (T2 ) over a plate of material of thickness x (Figure 2.5)

∆T T − T2
q = −λ =λ 1 (W/m.K ) (2.1)
x x Figure 2.5: Measuring the thermal con-
ductivity of a material.
All materials can move heat by the thermal vibrations being passed
to neighbouring atoms (phonons). These phonons essentially move
like elastic waves and move faster in stiff (high E) materials with
high speed of sound. However, they are also very easily scattered
by even slight defects in a crystal lattice. Some materials, notably
metals, have the advantage of also having free electrons that can
more easily carry thermal energy around. Even in metals, scatter-
ing by impurities has a major effect so pure metals (Fe = 80W/m.K)
are much more conductive than alloys (Stainless steel [Fe + 10-30%
Cr/Ni]=18W/m.K).

2.2.3 Specific Heat Capacity and Thermal Diffusivity

The specific heat capacity of a material is a measure of how much


heat needs to be supplied to increase the temperature (Figure 2.6).
This matters since heat costs money and resources. For solids we
tend to use the specific heat capacity at constant pressure C p (J/kg.K).

Figure 2.6: Measuring the heat capacity


of a material at constant pressure (C p )
Thermal conductivity governs heat flow under steady state condi-
tions (time is irrelevant) but often heat is transient. Such heat flow is
complex to model but generally it is governed by the thermal diffu-
sivity

λ
a= (m2 /s) (2.2)
ρC p
where ρ is the density1 . A material with low thermal diffusivity 1
Multiplying C p (J/kg.K) by the den-
will change its temperature more slowly and a heat wave will take sity (kg/m3 ) changes to energy ab-
sorbed per unit volume, rather than
longer to pass through. A plot of conductivity against diffusivity mass.
shows that most materials fall on a contour of constant ρC p (= λ/a).
I.e. the heat capacity per unit volume ρC p is about constant. The
main (important) exception is foams. These contain large air bubbles
and so have low conductivity but also a lower thermal capacity - they
transmit less heat but also change temperature quickly.

2.2.4 Thermal Expansion Coefficient

Thermal expansion is a strain arising from the increased separation


of atoms as the temperature increases2 . It is defined as 2
This occurs because the atoms oscil-
late more around their mean position
as they gain thermal energy.
1 dL −1
α= (K ) (2.3)
L dT
where L is some linear dimension of a body. As a strain per degree
we get units of K −1 . The expansion is resisted by the stiffness of the
atomic bonds and to a good approximation we find

1.6 × 10−3 0.02


α= = (2.4) Figure 2.7: Measuring the thermal ex-
E Tm pansion of a material.
I.e. stiffer materials with higher melting points expand less per
degree and all solids expand by about 2% between absolute zero and
melting.
engineering science 15

If a material heats up then stresses can accumulate if expansion is


constrained in some way, leading to distortions or failure. There are
two common ways for heat to cause problems:

1. Difference in expansion between two materials: Most structure are


made up of multiple materials that expand differently as the tem-
perature changes. For the simplest scenario of a small component
(Modulus E1 , expansion coefficient α1 ) connected to a much big-
ger one (α2 ), the stress in the smaller one will be about

σ = E1 (α1 − α2 )∆T (2.5)

Minimising differential stresses relies on choosing materials with


similar, not necessarily low, expansion coefficients.

2. Thermal gradients: If heat is applied suddenly then a heat wave will


move through the material and there will be a thermal strain/stress
mismatch over that gradient. The stress is more extreme if the
thermal gradient is greater. Better materials for low thermal dis-
tortion are those with high conductivity (spread the heat out, re-
ducing gradients) and low expansion coefficient (reduce the dis-
tortion for a given thermal gradient). Picking higher values of the
ratio λ/α is predictive of better performance.

Figure 2.8: The thermal expansion coef-


ficient plotted against the thermal con-
ductivity.

Example 2.1 (Energy Efficient Kilns) A lot of ceramics require a fir-


ing stage during production and this requires very high tempera-
tures for a long time leading to high energy costs. Part of this cost is Figure 2.9: A kiln with wall thickness
w, internal temperature T and external
the heat that conducts through the wall and is lost - this is reduced temperatre T0 .
by using a material with low thermal conductivity and making the
wall thick. The remainder is the heat absorbed by the kiln itself - this
is reduced by a low heat capacity and making the wall thin. What is
the best choice given the conflicting requirements on wall thickness?
Assume upon firing the internal temperature quickly rises from
the ambient T0 to operating temperature T where it is held for time
t. The approximate heat lost via conduction per second is given by
Eq 2.1 so the total heat lost is roughly

T − T0
Q1 = λ t (2.6)
w
The energy absorbed by the wall per unit volume will be C p ρ per
degree where C p specific heat capacity and ρ is the density. So the
total energy per unit area of wall will be

T − T0
Q2 = C p ρ w (2.7)
2
The total energy consumed per unit area of wall will be

T − T0 T − T0
Q = Q1 + Q2 = λ t + Cp ρ w (2.8)
w 2
A thin wall absorbs little but loses too much via conduction, a thick
wall does the opposite. The optimum thickness is obtained by differ-
entiating Eq 2.8 and equating to zero:

dQ T − T0 T − T0
= −λ 2
t + Cp ρ =0 (2.9)
dw w 2
Solving for w gives
 1/2
2λt
w= = (2at)1/2 (2.10)
Cp ρ

where a = λ/ρC p is the thermal diffusivity (Eq 2.2). The quantity


(2at)1/2 has dimensions of length and is a measure of the distance
heat can travel in time t. This suggests the most efficient kiln is one
where the outside of the wall just starts to get hotter at the end of
the cycle. Substituting Eq 2.10 into Eq 2.8 and separating material
properties gives:

Q = ( T − T0 )(2t)1/2 (λC p ρ)1/2 (2.11)

and hence the material index to be maximised is

a1/2
M = (λC p ρ)−1/2 = (2.12)
λ
Ideally we want low conductivity and low heat capacity. However,
by eliminating w we might pick a material requiring an unfeasibly
thick wall so we need to go back and calculate the thickness for any
material we pick. The best material options are found using Figure
2.10 with a selection line as shown. Any material lying on this line
has the same ratio of a1/2 /λ. Materials to the lower right of the line
are better than those to the upper left. Foams, polymers and compos-
ites look good but have a low maximum service temperature. A kiln
operating at 1000◦ C would need brick based on the figure. In real-
ity, specialist refractory bricks would be used but there is insufficient
space on the chart to show these.
engineering science 17

Figure 2.10: A selection chart suitable


for a kiln wall
3 Stiffness-Limited Design

Intended Learning Outcomes


Skills Identify when constraints and objectives are coupled.
Derive merit index for common stiffness-based scenarios.
Rank materials using selection lines for indices of varying type.
Stiffness-limited design occurs when the usefulness of a part is re-
stricted by how far it can elastically deflect. A limit on the maximum
deflection you can tolerate or the stiffness you need is an example of
a constraint. The component is either stiff enough or it is not.

Figure 3.1: a) a tensile load on a tie, b)


a panel in simple bending, c) a square
section beam in bending.

At the same time we usually want to minimise the mass or the


cost of the component - the objective. Both of these need equa-
tions to describe them. Three common scenarios for loading are
shown in Figure 3.1. Each has a different equation (constraint) for the
elastic deflection δ arising from a force F. Since they have different
shapes, they also have different equations to describe the objective
(volume/mass/cost). There is a complication in that the constraint
and objective equations will probably interact with each other. You
can reduce the deflection of a tie by increasing the area A but this
also makes it heavier.
Dealing with these coupled variables follows a regular pattern:

1. Define an objective and constraint equations.

2. Identify the coupled free variable (i.e. can change and appears in
both equations).

3. Solve the constraint equation for the free variable.

4. Substitute into the objective to eliminate the free variable.

5. Separate variables to identify material properties (the index).

Example 3.1 Light, stiff tie


Consider that we want a tie that has some acceptable stiffness
(S = F/δ). It has a fixed length L, since it needs to fit into some
structure, but let us assume the cross-sectional area A has scope to
vary. The mass should be as low as possible. We can complete the
translation table.

Function Support a tensile force F


Constraints Stiffness specified S Functional constraint
Length L specified Geometric constraint
Objectives Minimise mass m
Free Variables Choice of material
Cross-sectional area A Coupled variable

The objective equation is the mass m of the tie

m = ρAL (3.1)

where ρ is the density of the material. The constraint equation is the


tensile (spring) stiffness
F EA
S= = (3.2)
δ L
where E is Young’s modulus. The equations are coupled by the area,
A. We remove coupling by eliminating the shared free-variable by
solving the constraint equation and substituting it in the objective
equation. Only the material properties in the resulting equation are
variables for a given design.

m = ρAL
EA
S= (3.3)
L  
ρ
m = SL2 (3.4)
E
ρ
M= (3.5)
E
S and L aren’t in the index as they are fixed by the design and not
material properties. Minimising Equation 3.5 predicts the lightest
material for the tie. It is perhaps more intuitive to maximise
E
M= (3.6)
ρ
instead. This is the specific stiffness of a material i.e the modulus per
unit mass. Materials with high modulus and low density are best for
low mass design, as we would expect.

3.1 Constraints for Bending

The constraint equations for bending (and torsion, buckling, etc) are
a bit more complex than tension as shape plays an important role.
For selection purposes it is convenient to define a general constraint
equation for the stiffness of a beam
Figure 3.2: Values of C1 for 3 com-
F C EI
S= = 13 (3.7) mon beam scenarios. Note that in many
δ L cases the actual value is not important
to us.
where L is the length, F is the force, I is the 2nd moment of area and
C1 is some constant reflecting the type of beam (Figure 3.2). Using
engineering science 21

Equation 3.7 is possible for selection because all the details captured
by C1 and I are constant for any given design and so don’t alter
the ranking of materials. It is too simple to describe complex beams
with multiple loads but often we find that a good material for simple
scenarios is a good material for complex ones.

Figure 3.3: Summary of equations for


area, 2nd moments of section for 4 sec-
tion shapes. I (second moment of area)
governs bending, K (polar second mo-
ment of area) governs torsion.

Example 3.2 (Square-section beam) Consider the square beam shown


in Figure 3.1. The length is fixed while the height and breadth are
free but constrained to be equal (b). The objective equation is the
mass.

m = ρLb2 (3.8)

The constraint equation is Eq 3.7 and we use I for a rectangular sec-


tion beam and setting h = b (Figure 3.3)

F C EI C Ebh3 C Eb4
S= = 13 = 1 3 = 1 3 (3.9)
δ L 12L 12L
These are coupled via b2 (i.e. the area A) so we need to eliminate it.
1/2
12SL3

2
b =
C1 E
1/2
12SL3

m = ρL
C1 E

Extracting the material properties gives the merit index and the re-
ciprocal merit index. A high modulus is much less important than
having a low density.

ρ E1/2
M= or M= (3.10)
E1/2 ρ

Example 3.3 (Thin panels) A panel is a thin sheet that is usually


used to cover an area e.g. to cover a train carriage or clad a building.
The length and breadth are constrained but the thickness is variable.
Here we look at a panel loaded in bending. We want a light panel so
the objective equation is still the mass.

m = ρLbh (3.11)

Rather than consider area, we separate the thickness h and breadth b


since b is fixed. The constraint equation is Eq 3.7 and we use I for a
rectangular section beam (Figure 3.3)

F C EI C Ebh3
S= = 13 = 1 3 (3.12)
δ L 12L
These are coupled via h so we need to eliminate it.
1/3 1/3
12SL3 12SL3
 
h= so m = ρLb (3.13)
C1 Eb C1 Eb

Extracting the material properties gives the merit index and the re-
ciprocal merit index. A high modulus is much less important than
having a low density.

ρ E1/3
M= or M= (3.14)
E1/3 ρ

3.1.1 Ranking for ties, beams and panels

The equations 3.6,3.10 and3.14 are the merit (or material) indices for
light, stiff ties, beams and panels. They indicate how good (relatively
speaking) the material is in fulfilling the objective of minimising the
mass. It is much easier to rank materials using this number rather
than explicit calculation. Note that we don’t need to know how stiff
the component should be, nor its dimensions. We don’t really need
to know how the beam would be loaded (cantilever or centre load,
for example). So long as those factors are fixed, a high value of an
index indicates a better (lighter) material.
Rather than actually calculate the index for all the material options
we would often prefer to use a graphical method. On a selection
chart plotting property P1 on the y-axis and P2 on the x-axis, and
using log axes, an index of the form

M = P1α /P2 (3.15)

results in a line of gradient of 1/α. Hence, Eq 3.6 manifests as a


line of gradient 1, Eq 3.10 results in a gradient 2 and Eq 3.14 results
in gradient 3. These lines are plotted on Figure 3.4. Such a line
represents a constant value of M i.e. any material on the line has
equal performance (in the examples, this means equal mass). Since
the modulus is raised to different powers in the three indices we
find a change in emphasis on properties. So for a panel we get more
benefit from using a less dense material than from shifting to a higher
modulus material.
We can find better materials by shifting the selection line to higher
values of the index (i.e. increase E, reduce ρ). In Figure 3.4 materials
engineering science 23

to the upper-left are better and those to the lower-right are worse.
The more to the upper-right a material is, the higher ranked it is. We
would normally pick the top ranked materials for examining in more
depth.

Figure 3.4: Using selection lines to find


the best materials on selection charts.

3.2 Shape
The deflection of a beam is a function of the shape as reflected in the
2nd moment of area, which gets bigger if more material is far from
the neutral axis. Tubes and I-beams can be lighter than solid sections
for the same stiffness constraint.
We can describe this via a shape factor
I 12I
ΦeB = = (3.16) Figure 3.5: Converting a solid square
Isquare A2 section to tubes. One has the same area
and greater I, the other has smaller area
where Isquare is the second moment of area of a solid square beam and the same I.
of the same area as the shaped beam. The superscript denotes elas-
tic, the subscript denotes bending. The thinner the walls, the more
effective the tube - but make them too thin and it breaks for other
reasons. This has a big effect as not all materials can be fashioned
into efficient shapes.

Material Maximum shape factor Mass ratio by shaping


(stiffness relative to solid (relative to solid
square beam) square beam)
Steels 64 1/8
Al alloys 49 1/7
Composites 36 1/6
Wood 9 1/3

We can modify the merit index to account for the maximum shape
factor by noting

12I (ΦE)1/2
A2 = so M= (3.17)
ΦeB ρ
The maximum shape factor for a material is very complex to deal
with as it comes down to varying combinations of alternative modes
of failure, constraints on practical manufacturing methods or just
availability on the market place.
4 Strength-Limited Design

In many cases our design is limited by the load carrying capacity of


a component without it failing. In this case the material is limited
by it’s strength although this might be the yield strength, tensile
strength, compressive strength or any other suitable limitation on
maximum stress depending on the application.

Intended Learning Outcomes


Skills Deriving material indices for strength-limited scenarios.
Screening materials based on minimum strength.
Ranking materials on strength per unit mass or cost.

4.1 Strong, Light Tie


The index for a uniaxial tie under tension or compression is the sim-
plest derivation.

Table 4.1: Translation table for a tie


Function Support a tensile force
of minimised mass constrained to not
Constraints Support force F without yielding yield
Length specified
Objectives Minimise mass
Free Variables Choice of material
Cross-sectional area A
The objective (mass) equation is the same as for the light, stiff tie
and is independent of cross-sectional shape so we use area rather
than explicit dimensions (i.e. who cares if it is a cylindrical or a
square section? Solid or hollow?).

m = ρAL0

Given some (currently unknown) minimum force the component


must take without failing (fracture or yielding), the area must be
big enough to keep the stress below the failure stress, which changes
as we alter material. In this case no attempt is made to put an upper
limit on size but we need to keep in mind that very low strength
materials may need to be infeasibly large to prevent failure.
F
σf ≥
A
The objective and constraint equations are coupled by the area so we
eliminate this free-variable.
F
A=
σf
!
ρ
m = FL0
σf
ρ σf
M= or M= reciprocal merit index
σf ρ
The reciprocal merit index is the specific strength of the material
which is quite an intuitive index to use for a light, strong material.

4.2 Strong, Light Beams


Yielding in beams is more complex as the stress is not uniform. The
highest stress occurs at the location of maximum moment (Figure
4.1). The bending moment M generates a linear variation in longi-
tudinal stress σ across the section following the fundamental beam
equation
σ M
= (4.1)
y I Figure 4.1: Plastic bending of different
beams. The red spot indicates the loca-
where y is the distance from the neutral axis. The second moment of tion of yield. The value of the constant
area, I, captures the influence of the section shape (Figure 3.2). The C2 varies for each a) cantilever = 1, b)
simply supported centre point load =
maximum stress occurs at the upper and lower surface at the furthest 1/4, c) simply support uniform load =
distance from the neutral axis ym . 1/8.

Mym M C FL
σmax = = = 2 (4.2)
I Ze Ze
where C2 is a constant reflecting the type of loading and constraint.
The quantity Ze = ym /I is the elastic section modulus. If the stress
exceeds σy then the beam is bent but hasn’t failed entirely. Applying
a greater moment progrsses the plasticity through the section until
it is entirely plastic and we get sudden plastic collapse. This more
catastrophic critical moment (force) is predicted by the plastic section
modulus Z p . The two

Figure 4.2: A beam loaded in bend-


ing showing the stress state as plastic-
ity begins at a critical applied moment.
As the moment is increased, plasticity
moves through beam until the whole
cross section is plastic and we reach the
collapse moment.

As more load is applied the plastic zones move towards the neu-
tral axis and eventually meet. At this point they form a plastic hinge
engineering science 27

and the beam will undergo plastic collapse. If this is our failure cri-
teria we need to replace Ze with the plastic section modulus Z p . A
summary of these are shown in Figure 4.3.

Figure 4.3: A summary of elastic and


plastic section modulus for three com-
mon section shapes.

Table 4.2: Translation table for a square-


Function Beam in bending
section beam of minimised mass con-
Constraints Support moment M without any yield strained to not yield
Length specified
Square x-section (b=h)
Objectives Minimise mass
Free Variables Choice of material
Cross-sectional area A
Consider a generic square section beam. The moment must be
low enough, or the area big enough, to keep the stress below the
yield point at any point. The translation is shown in table 4.2. The
objective equation is the mass (density times volume)

m = ρAL0

The constraint equation is that the yield strength must exceed the
maximum stress due to the moment M
M 6M 6M
σy ≥ = 2 = 3
Ze bh h

The objective and constraint equations equations are coupled by the


area (since the height and breadth of the beam are the same) so we
eliminate this free-variable.

6M 1/3
 
h=
σy
6M 2/3
 
A = h2 =
σy
ρ
m = L0 (6M )2/3 2/3
σy

Extracting the material properties gives the index and reciprocal in-
dex
σy2/3 ρ
M= M = (4.3)
ρ σy2/3
A selection line for this index, and for a tie and panel, are shown
in Figure 4.4. For a tie, we get similar performance from steels, light
alloys (Al, Mg), some composites and high-end wood. In contrast, for
beams and panels we get a lot more benefit from lower density/lower
strength materials.

Figure 4.4: A bubble chart for choos-


ing strong, light components. Selec-
tion lines for ties, panels and beams are
shown.

Example 4.1 (Failure of a tubular shaft) Shafts are components un-


der torsion, which acts to twist them. They are commonly, but not
always, cylindrical cross-section.

Figure 4.5: Shafts develop shear stresses


that lead to yield. Helical springs are a
special form of torsion loading.

As with beams, the stress in a shaft is not uniform. A torque


applied to the ends of the shaft creates a shear stress on the plane
perpendicular to the axis of the bar. This increases linearly from zero
at the axis to a maximum on the outer radius r
Tr
τmax = (4.4)
K
engineering science 29

where K is the polar 2nd moment of area (Figure 3.3). Yield starts
at the outer surface when τmax exceeds the shear yield strength (typ-
ically σy /2). The shaft doesn’t fail outright at this point but it will
have a permanent twist.
Consider the selection of a hollow, cylindrical tube under torsion
that must not yield at all and be as light as possible. From Figure 3.3
we can get the moment K for a tube, which is simpler if we make the
thin-walled approximation (t«r) and so the constraint equation is:

T
K = 2πr3 t so τy =
2πr2 t
while the objective, given the cross-sectional area is ≈ 2πrt, will be

m = ρL2πrt

But we hit an issue here. Both the radius and wall thickness are free
and coupled. We need to constrain one to proceed. If we impose a
constant shape on the tube i.e. t = αr we get a performance equation
 2/3
T
m = ρL2πα
2πατy

Extracting the material properties and inverting so we can maximise,


gives the index

τy2/3 σy2/3
M= ≈
ρ ρ

The substitution of σy , the uniaxial tensile yield strength, for τy , the


shear yield strength, can be helpful since it is more commonly avail-
able and the difference is mostly just a factor, which doesn’t change
ranking. Note that the index is the same as for a square beam.
From Fig 4.4 we expect light alloys (aluminium, titanium) to work
well and would be easier to make as a tube via extrusion meth-
ods. Composites, especially CFRP, look good although this is a case
where the fibre direction would need to be examined carefully and
we might be misled. In theory, materials like rigid foams look ok but
it is likely that the radius/wall thickness needed would be impracti-
cal given the low strength.
5 Selecting Manufacturing Processes

In addition to choosing the best material for a job, we often need to


choose the correct manufacturing method. This is not easy since the
choice is a matter of many parameters and generally intersects with
economics and other factors. It is possible to make some general
observations though and introduce some considerations.

Intended Learning Outcomes


Skills Read processing selection charts and identify appropriate values.
Apply processing constraints to remove unsuitable processing methods.
Make reasonable evaluations of suitable processing routes.

5.1 Selection Strategy


The strategy for selecting a process matches that for properties: trans-
late design requirements into constraints and objectives; screen based
on the constraints, rank on the objectives and finally do a research
stage on the final options.

5.1.1 Process-Material compatibility

A principle constraint on process is whether it works with your cho-


sen material. Figure 5.2 is a summary of which processes (rows) are
suited to which materials (columns). Compatible combinations are
indicated by the dots. Some processes are specific to certain mate-
rial classes. Sand casting is limited to metals that can be melted and
poured as a liquid into the mould, layup methods are used for com-
posite materials involving fibres and a polymer matrix. Others, like
fixing with adhesives, are flexible and appropriate for many materi-
Figure 5.1: The general strategy for se-
als. lecting the best processing method to
Processing and material selection are a codependent. If you choose produce a component.

aluminium as your material and then want to cast the component


you will be limited to aluminium alloys that are castable. If you then
want to age-harden (precipitate strengthen) to improve the strength
then you are limited to the subset of aluminium alloys that are age-
hardenable and castable. You many need to jump around between
material and process selection to obtain a solution.

5.1.2 Size and Shape

A key constraint is whether a process can produce the size and shape
of component you need. Figure 5.3 illustrates typical mass1 ranges 1
We cold also use volume but mass is a
for different processes and includes colour coding for different ma- reasonable approximation for this - big
things tend to be heavy.
terials. Generally, processes span around 3 orders of magnitude in
mass with most being suited to making things you could comfort-
ably hold (0.1-10 kg). Producing very big or very small things is
more restrictive.
Similar charts can be generated for specimen thickness (Figure
Figure 5.2: Material-Process compati-
bility matrix summarising how a subset
of processes are suited to given material
classes.

Figure 5.3: The range of component


masses for which a process is suited.
These ranges represent typical capabil-
ities or the most common reasonable
mass. You might be able to go smaller
or larger but then you should expect to
incur additional costs.

5.4). The upper limit on mass and thickness tends to be the size of the
machine you can feasibly make or buy (although physical issues like
shrinkage can be limitations). Lower limits tend to be determined
by the physics of the process e.g. the ability of liquid to flow down
narrow channels or the precision with which a cutting piece can be
placed. Very small dimensions (10s of micrometres) end up being
limited by the stiffness of the machine itself. Very tiny components
need chemical or electrical methods to produce them, which limits
engineering science 33

the choice of material and often restricts production capacity.

Figure 5.4: The range of component


thickness for which a process is suited.
These ranges represent typical capabili-
ties.

5.1.3 Shape

Shape is challenging to classify but a useful scheme is illustrated


in Figure 5.5. There is a useful distinction between continuous and
discrete processes. The former usually have a consistent shape on
at least one axis (tubes, I-beams, sheet) and so long lengths can be
made, with gains in production volume. Discrete shapes can only be
made one at a time in a batch process. This can still be fast but com-
plex shapes often need many steps and production gets expensive.

Figure 5.5: One scheme for splitting


component shapes into 6 classes suited
Some process and materials are suited to certain shapes as indi- to different production methods.
cated in Figure 5.6.

5.1.4 Tolerances and Roughness

There are many issues that need to be considered in choosing the


correct process. A final example here is that of the achievable toler-
ance and surface roughness that is acceptable. Tolerances are how
close can the nominal dimensions of a component match the desired
value. High accuracy demands higher quality machines and often
more skilled users. Roughness is a measure of how much the lo-
cal surface dimension varies from the nominal surface (typically root
Figure 5.6: The shape-process compat-
ibility matrix. Combinations with dots
are suited to that combination of shape
and process for the indicated material
class.

mean square amplitude).

Figure 5.7: Tolerance and roughness ca-


pacities for different processing meth-
Very low roughness is demanded for contacting surfaces, in par- ods. High quality finishing processes
ticular, and required values are not normally achievable be shaping like grinding are included.
processes alone, which are rarely better than 0.1mm and often worse.
Very high tolerances and low roughness require special surface ma-
chining stages like grinding or lapping in a production chain of man-
ufacturing steps.

5.2 Cost of Manufacture


Typically, the objective to be minimised for the selection of a process
is the total cost. A true cost estimate, done by an expert, is highly
detailed and requires domain specific knowledge. Here we focus on
a general approach suitable for an initial cost estimate. Precision is
limited but this is sufficient for comparing processing routes.
Costs in manufacture arise from the use of resources: material,
equipment, time or knowledge. These are summarised in Table 5.1.
engineering science 35

Making an object requires a mass m of material and/or consumable


feedstock at a cost of some £/kg. Processing generally wastes some
fraction f so the material cost per unit is

mCm
C1 = (5.1)
(1 − f )
Capital investment has two main contributions: tooling, which
is specific to a given component, and equipment, which is used on
many components. Tooling (jigs, moulds, dies, etc) is a dedicated
cost that is used for this component. Its cost is written off against the
number of components made n. Equipment is used generally, with
different tooling making it a non-dedicated cost. The total outlay
can be made into an hourly rate by Cc /two where two is the time to
write off the cost (e.g. 5 years). The cost per item is obtained by
dividing this by the production rate per hour ṅ. Neglecting wear
(requiring multiple tool sets) and applying a load factor L to the
machine (fraction of time in use) gets the equipment cost per unit is

Ct 1 Cc
C2 = + (5.2)
n ṅ Ltwo

Table 5.1: Summary of symbols used


Resource Sym- Unit
when estimating processing costs
bol
Materials Including consumables Cm £/kg
Capital Cost of tooling Ct £
Cost of equipment Cc £
Time Overhead (inc labour, rent, Coh £/h
admin)
Energy Cost of energy (inc power, heat) Ce £/h
Information R&D, patents, royalties Ci £/year

Finally, we have the general background costs, which include ad-


min, labour, energy, rent, etc. It might also include the need to pay
for the right to use technology or knowledge. Lets assume this all
aggregates into one cost per hour Coh . The cost per item is

Coh
C3 = (5.3)

and the total cost for shaping a component is

mCm Ct 1 Cc C
Cs = + + + oh (5.4)
(1 − f ) n ṅ Ltwo ṅ

or, more simply,

Figure 5.8: The cost of shaping a com-


Cdedicated (Ċcapital + Ċoverhead ) ponent using sand casting, low pres-
Cs = Cmaterial + + (5.5)
n ṅ sure casting and die casting as a func-
tion of production volume.
Thus, there are three main contributions: a material cost per unit
that doesn’t change, a dedicated cost that varies with production vol-
ume and a gross overhead that changes with production rate. Figure
5.8 plots the variation in cost with the number of items made for
three casting methods using representative values. The cost is rel-
ative to the raw material cost of the product. At low volumes the
cost is dominated by setup - making the tooling and investing in the
needed equipment. In this regime, methods like sand casting - which
are easy to set up but limited in speed - can be the cheapest options.
As production volumes increase the initial cost is spread out over
more items and the cost of making an item approaches that of the
material required. At this point the higher production rates of low
pressure and die casting are a benefit. As a result the curves cross at
some production volume leading to the idea of an economic batch size
for each process - a range of production volumes that are likely to be
economically viable (Figure 5.9).

Figure 5.9: Economic batch sizes for dif-


ferent processes.

You might also like