Materials Selection

Chapter 3
Multiple Constraints and
Objectives
 Procedure for deriving material indices (from Ashby)
Step Action

1 Define the design requirements

(a) Function: what does the component do?
(b) Objective: what is to be maximized or minimized?
(c) Constraints: essential requirements which must be met: stiffness, strength, corrosion resistance, forming
characteristics.
2 Develop an equation for the objective in terms of the functional requirements, the geometry and the material
properties (the objective function)
3 Identify the free (unspecified) variables.

4 Develop equations for the constraints (no yield; no fracture; no buckling, etc.)

5 Substitute for the free variables from the constraint equations into the objective function.

6 Group the variables into the three groups: functional requirements (F), geometry (G), and material properties (M); thus
Performance characteristic ≤ 𝒇𝟏 (F) 𝒇𝟐 (G) 𝒇𝟑 (M)
Performance characteristic ≥ 𝒇𝟏 (F) 𝒇𝟐 (G) 𝒇𝟑 (M)

7 Read off the material index, expressed as a quantity M, which optimizes the performance characteristic.
 Materials Selection:
Review
Function
Objective
Tie
Constraint

Beam Minimum cost

Index
Stiffness

Minimum weight E 1/ 2
M1 =
Shaft 
Strength
Maximum stored
energy
Column
Fatigue
Index
Minimum
environmental impact  f2 / 3
Mechanical
Geometry M2 =
Thermal 
Electrical…..
Trade-Offs

• I want to be rich but I want to be

happy.
• Need to balance between the two
“objectives”
• Materials for commercial aircraft
needs to be light and cheap.
• Light materials are not cheap.
Strength Constraint

Function Tie rod

Constraints • 𝐿
𝐹𝑓
• = 𝜎𝑓
𝐴
Objective • Minimize 𝑚 Linked
Free variable • 𝐴
• Materials

𝜌
𝑚 = 𝐿 𝐹𝑓 • There is one link between constraint
𝜎𝑓
and objective.
• There is one free variable.
𝜎𝑓
𝑀=
𝜌
Stiffness Constraint

Function Tie rod

Constraints • 𝐿
𝐶 𝐸𝐼
• 𝑆 = 13
𝐿
Objective • Minimize 𝑚 Linked
Free variable • 𝐴
• Materials

𝜌
𝑚 = 𝐿2 𝑆
𝐸 • There is one link between constraint
and objective.
• There is one free variable.
𝐸
𝑀=
𝜌
Stiffness and Strength Constraints

Function Tie rod

Constraints • 𝐿
𝐹𝑓
• = 𝜎𝑓
𝐴
𝐶1 𝐸𝐼
• 𝑆=
𝐿3 Link 1
Objective • Minimize 𝑚
Free variable • 𝐴
• Materials Link 2

• There are two links between constraints

and objective.
• There is one free variable.
• “Overconstraint”
Selection with Multiple Constraints

• Minimum-mass
tie rod
• Two “linked”
constraints 𝑚1

• Stiffness
• Strength 𝑚2

• Two equations
• Mass
• Materials index

Notice that the indices are to be minimized now.

 Selection with Multiple
Constraints
• If stiffness is very important
• Use 𝑀1
• Mass = 𝑚1
• If strength is very important
• Use 𝑀2
• Mass =𝑚2
• If both are important
• Use 𝑀1 and 𝑀2
• Mass (𝑚)
ഥ = max(𝑚1 , 𝑚2 )
• Mass is the largest of the Analytical method
two.
Analytical method

max(𝑚1 , 𝑚2 )

𝐦𝐢𝐧(max(𝑚
Graphical Method
This line divides
the graph into 2
regions.

Objective:
Minimize 𝑚2

Objective:
Minimize 𝑚1
The nearer the
box is pulled to
the bottom left,
the smaller is 𝑚.
ഥ
Graphical Method

Problem
• The values of 𝑚1 and 𝑚2 are
specific to the choice of 𝑆, 𝐿, and
𝐹𝑓 .
• If you choice of , 𝐿, and 𝐹𝑓
changes, we need a new graph.
Deriving Indices

𝑃 = 𝑓1 𝐹 × 𝑓2 𝐺 × 𝑓3 𝑀

Structural Index Material Index

Graphical Method
𝜌 𝜌
𝑚1 = 𝐿2 𝑆 = 𝐿2 𝑆𝑀1 𝑚2 = 𝐿𝐹𝑓 = 𝐿𝐹𝑓 𝑀2
The axis is now dependent on materials 𝐸 𝜎𝑦
properties, not the values of 𝑆, 𝐿, and 𝐹𝑓 .

𝑚1 = 𝑚2

𝐿2 𝑆𝑀1 = 𝐿𝐹𝑓 𝑀2

𝐿𝑆
𝑀2 = 𝑀
𝐹𝑓 1

𝐿𝑆
log 𝑀2 = log 𝑀1 + log
𝐹𝑓

log 𝑀2 = log 𝑀1 + log 𝐶

(Line of slope = 1)

Coupling constant.
Graphical Method

• Changing the values of 𝑆,

𝐿, and 𝐹𝑓 only changes
the coupling constant.
• No need for new graph.
General Procedures
• Determine objective
• i.e. minimize 𝑃
• Use the free variable to derive a set of
performance equations
• 𝑃1 = 𝑓1 𝐹 ⋅ 𝑔1 𝐺 ⋅ 𝑀1
• 𝑃2 = 𝑓2 𝐹 ⋅ 𝑔2 𝐺 ⋅ 𝑀2
• …
• The final values of “𝑃” is
• 𝑃ത = max(𝑃1 , 𝑃2 , … )
General Procedures

• If 𝑃1 is more restrictive (active constraint)

• Choose materials with lowest 𝑀1 .
• If 𝑃2 is more restrictive (active constraint)
• Choose materials with lowest 𝑀2 .
• If any cases, you need to check with
• 𝑃ത = max(𝑃1 , 𝑃2 , … )
General Procedures

• You can use a chart with 𝑀1 and 𝑀2 as axis.

• Draw the line from
𝑓1 𝐹 ⋅𝑔1 𝐺
• 𝑀2 = ⋅ 𝑀1
𝑓2 𝐹 ⋅𝑔2 𝐺
• 𝑀2 = 𝐶𝑐 ⋅ 𝑀1
• log 𝑀2 = log 𝑀1 + log 𝐶𝑐
• 𝐶𝑐 is the coupling constant
• Specified by the design
Conflicting Objectives
• Real-life materials selection requires a compromise
between conflicting objectives.
• Typical conflicting objectives
• Minimizing mass
• Minimizing volume
• Minimizing cost
• Minimizing environmental impact
• Other objectives
• Maximizing power-to-weight
• Minimizing mass per power output
• Maximizing energy storage
• Minimizing volume per stored energy
Case Study

• Assume that the

objectives are
• Minimizing cost
• Minimizing mass
• Assume that we can
plot the chart of cost vs
mass.
Case Study
• Dominated solution
• Far from optimal
• Non-dominated
solution
• Possible solution
• Trade-off surface
• Collection of non-
dominated solutions
• “Pareto set”
• There are 3 strategies
to continue …
Strategy 1: Basic

• Identify small list of

solutions near trade-off
surface
• Use supporting
information to rank the
solutions
Strategy 2: Cheating
Make one objective (cost) to be a constraint.
Strategy 3: Advanced
• Define a penalty function
• 𝑍 = 𝛼1 𝑃1 + 𝛼2 𝑃2 + 𝛼3 𝑃3 + ⋯
• The best choice is the smallest value of 𝑍.
• 𝛼𝑖 is the “exchange constant”.
𝜕𝑍
• 𝛼𝑖 =
𝜕𝑃𝑖 𝑃 ,𝑗≠𝑖
𝑗
• Measures change of 𝑍 per change of 𝑃𝑖
• In our case study, we have
• 𝑍 = 𝐶 + 𝛼𝑚
• Unit of 𝑍: Price
• Unit of 𝛼: Price/mass ($/Kg)
Strategy 3: Advanced
• From
• 𝑍 = 𝐶 + 𝛼𝑚
• We have
1 1
• 𝑚=− 𝐶+ 𝑍
𝛼 𝛼
• This is a linear equation
1
• Slope: − 𝛼
𝑍
• Intercept: 𝛼
• We choose the materials
that intersect the line
with the smallest values
of 𝑍.
Strategy 3: Advanced

Linear Scale Log Scale

Relative Penalty Functions
• We want to choose a material that is better
than the existing one.
• We define a relative penalty function
𝐶 𝑚
• 𝑍∗ = 𝐶 + 𝛼∗ 𝑚
0 0
• 𝑍∗: No unit
• 𝛼 ∗ : relative exchange constant
• The fractional gain in value for a fractional gain in
performance
• If 𝛼 ∗ = 1
• We can offset the 50% increase in cost by 50%
decreased in mass.
Relative Penalty Functions
• Region A
• Best
• Region B
• Heavier but cheaper
• Region C
• Light but more
expensive
• Region D
• No interest
Values for the exchange
constant
• 𝛼 is similar to the weight coefficient.
• Control how important the particular factor is.
• In the case study, 𝛼 indicates the impact of
reducing mass.
Range of the exchange
constant
• How cyclic value
reduction in weight?
• Ordinary cyclist
• $20/kg
• Pro cyclist
• $2000/kg
• Large range of 𝜶
Choice of Materials

Solution for given range of 𝛼

Three Objectives

• 𝑍 = 𝐶 + 𝛼1 𝑃1 + 𝛼2 𝑃2

𝑃2