You are on page 1of 10

Zn alloys/ Mg alloys

Al2O3 ceramic

Compare these in a table:


First Selection Stage Second Selection Stage Third Selection Stage
Ceramics Silicon/ Invar Cast iron/ Carbon steels
CFRP Ceramics (SiC, WC, AlN) Al alloys
Glasses/WC/Ti, W, Mo, Cu alloys/ W alloys Zn alloys/ Mg alloys
steel, Mg alloys and Al
alloys
Woods/Rock, Stone, Al alloys, Mg alloys/ Ni Al2O3 ceramic
Cement/ GFRP. alloys, steel
Ti alloys/ CFRP

The candidate materials that make it through all three stages are STEELS,
Al ALLOYS, and Mg ALLOYS.
 The main advantage of this multiple stage selection process is that
the assumptions are simple and clearly stated regarding the rank
ordering of the performance indices. The disadvantage is that it is
still subjective in determining the overall rank ordering and the
position of the selection lines on each of the charts.
 The quantitative approach to multiple constraints combines the
decision matrices and selection stages with coupling equations
and/or penalty functions. These are topics we’ll look at next.

II. COUPLING EQUATIONS


Consider a design situation in which we have one design objective, two
constraints, and one free parameter (example1). We are over-constrained
in this situation.
By calculating a performance index analysis using the first constraint, we
end up with:
1 1 1 𝜎𝑓 2⁄3
𝑃= = 2⁄ = 2⁄3 5⁄3 2⁄3 ( )
𝜌𝐿𝑡 2 6𝐿𝐹 3 6 𝐿 𝐹 𝜌
𝜌𝐿 (
𝜎𝑓 )
𝜎𝑓 2⁄3
∴ 𝑀1 =
𝜌
8
With the second constraint, we have:

1 1 𝛿 1⁄2 𝐸 1⁄2
𝑃= = = 1⁄2 5⁄2 ( )
𝜌𝐿𝑡 2 4𝐹𝐿3
1⁄2 2𝐹 𝐿 𝜌
𝜌𝐿 ( )
𝐸𝛿

𝐸 1⁄2
∴ 𝑀2 =
𝜌

Now, the objective is the same, so we can equate these two

1 1 1 𝛿 1⁄2
= ⟹ 2⁄3 5⁄3 2⁄3 𝑀1 = 1⁄2 5⁄2 𝑀2 ⟹
𝑚1 𝑚2 6 𝐿 𝐹 2𝐹 𝐿

1.65 𝐹 1⁄6 𝛿 1⁄2


𝑀1 = 𝑀2
𝐿5⁄6
On logarithmic scales
1.65 𝐹 1⁄6 𝛿 1⁄2
log 𝑀1 = log 𝑀2 + log ( ) 𝐶𝑜𝑢𝑝𝑙𝑖𝑛𝑔 𝐸𝑞.
𝐿5⁄6
This describes a line of slope 1, in a position that depends on the value of
1.65 𝐹 1⁄6 𝛿 1⁄2
( ). We refer to this line as the coupling line and to
𝐿5⁄6
1.65 𝐹 1⁄6 𝛿 1⁄2
( ) as the coupling constant, symbol 𝐶𝑐 .
𝐿5⁄6

The relative weighting of the two performance indices is DETERMINED


BY THE DESIGN and not by our subjective judgments!

EXAMPLE 3: A Light Tie Rod


DESIGN ASSIGNMENT:
 cylindrical tie rod of length L, A is unknown
 minimum weight
 support a load F
 extension less than 𝛿𝑜

9
MODEL:
𝐹 𝐹𝐿𝑜
𝜎= , 𝛿𝑜 =
𝐴 𝐴𝐸

OBJECTIVE: minimum mass


𝑚 = 𝐴𝐿𝑜 𝜌

PERFORMANCE EQUATION ONE


1 1
𝑃= =
𝑚 𝐴𝐿𝑜 𝜌
𝐹 𝐹
𝜎= ⟹𝐴=
𝐴 𝜎
1 1 1 𝜎
∴𝑃= = = ( )
𝐴𝐿𝑜 𝜌 (𝐹 ) 𝐿 𝜌 𝐹𝐿𝑜 𝜌
𝜎 𝑜
𝜎
∴ 𝑀1 =
𝜌

PERFORMANCE EQUATION TWO

𝐹𝐿𝑜 𝐹𝐿𝑜
𝛿𝑜 = ⟹𝐴=
𝐴𝐸 𝛿𝑜 𝐸
1 1 1 𝛿𝑜 𝐸
𝑃= = = = ( )
𝑚 𝐴𝐿𝑜 𝜌 (𝐹𝐿𝑜 ) 𝐿 𝜌 𝐹𝐿𝑜 2 𝜌
𝛿𝑜 𝐸 𝑜

𝐸
∴ 𝑀2 =
𝜌

DEVELOP THE COUPLING EQUATION:

1 1 1 𝛿𝑜
= ⟹ 𝑀1 = 𝑀2 ⟹
𝑚1 𝑚2 𝐹𝐿𝑜 𝐹𝐿𝑜 2

10
𝛿𝑜
∴ 𝑀1 = 𝑀 𝐶𝑜𝑢𝑝𝑙𝑖𝑛𝑔 𝐸𝑞𝑢𝑎𝑡𝑖𝑜𝑛
𝐿𝑜 2
On logarithmic scales
𝛿𝑜
log 𝑀1 = log 𝑀2 + log ( )
𝐿𝑜

Result

The best performing material will be one in which 𝐸 ⁄𝜌 is maximized, 𝜎⁄𝜌


is maximized, and their ratio is held at 𝛿𝑜 ⁄𝐿𝑜

How do we apply this to a selection chart? We want to use a chart


𝐸 𝜎
for this example like chart − .
𝜌 𝜌

We want both of our performance indices to be maximized, so we'll be


looking at materials in the upper right hand corner of the chart. For our
particular design, we'll have a given value of 𝛿𝑜 ⁄𝐿𝑜 that is determined by
the constraints of the design. Let's say it is 100-1.
We will look at a straight line of slope 1 on the plot, and we want the line
for which the ratio of the performance indices is 100-1
By moving along this line of constant 𝛿𝑜 ⁄𝐿𝑜 , we improve our performance
by increasing the values of the performance index, and we simultaneously
maintain the weighting factor determined by the design.
Our best choice of material is CFRP. We probably want to open up the
search region a bit, to allow some materials other than CFRP in the mix, so
for this case we can use a rectangular search region centered on the line of
𝛿𝑜 ⁄𝐿𝑜 =100-1.
By moving away from the line, we shift toward STIFFNESS
DOMINATED designs (to the upper left) or toward STRENGTH
DOMINATED design (to the lower right).

NOTE: If you use coupling equations, you don't need to use multiple stage
selection processes, but you may have to generate your own Ashby
Selection Charts!

11
III. ACTIVE CONSTRAINT METHOD
Let's look at an example with a single design objective (MOP), but several
constraints, an OVERCONSTRAINED problem. One way to approach it
is to use a multiple selection stage process, as we did for the precision
micrometer example in the last lecture. A difficulty with the approach is
the ordering of the constraints and selection stages, and the subjective
placement of the selection line in the multiple stages. A more systematic
approach uses the "active constraint" approach. An example:

EXAMPLE 4: The support rod for an infrared-electronics cooling


cryogenic fluid container in a spacecraft is to be designed. The most
important characteristic of this tie rod is that it should carry a minimum
amount of conductive heat into the cryogenic container. The conductive

12
heat flow equation tells us that the conductive heat flow along this support
rod is:
𝑞 = 𝐶𝜆𝐴

where C is a constant (the temperature gradient), λ is the thermal


conductivity of the rod, and A is the cross sectional area of the rod.
There are three constraints on the rod:
First, that the loading due to the mass of the cryogenic fluid and container
should not exceed the failure strength of the tie rod (ignore the mass of the
rod).
Second, the deflection, δ, should be less than a critical value, δ max.
Third, the vertical frequency of vibration must be high enough to not affect
the measurements being made. In other words, f should be larger than a
critical frequency, fmin.

13
MODEL:
Assume a solid cylinder for the rod

2
𝐴2
𝐴 = 𝜋𝑟 , 𝐼=
4𝜋

𝐹 2 𝐸𝐼 𝐹𝐿𝑜
𝜎𝑓 > , 𝑓= √ > 𝑓𝑜 , 𝛿= < 𝛿𝑚𝑎𝑥.
𝐴 𝜋 𝑚𝐿4 𝐴𝐸

Objective: minimum heat flow into the cryogen:


𝑞 = 𝐶𝜆𝐴

PERFORMANCE EQUATION ONE: Start with the load constraint:


1 1
𝑝𝑒𝑟𝑓𝑜𝑚𝑎𝑛𝑐𝑒 𝑚𝑒𝑡𝑟𝑖𝑐 𝑃 = =
𝑞 𝐶𝜆𝐴

𝐹 𝐹
𝜎𝑓 > ⟹𝐴= ,
𝐴 𝜎𝑓
1 1 1 1 𝜎𝑓
∴𝑃= = = = ( )
𝑞 𝐶𝜆𝐴 𝐶𝜆 ( 𝐹 ) 𝐶𝐹 𝜆
𝜎𝑓
𝜎𝑓
∴ 𝑀1 =
𝜆

PERFORMANCE EQUATION TWO: Now use the deflection constraint:

𝐹𝐿𝑜 𝐹𝐿𝑜
𝛿= ⟹𝐴= ,
𝐴𝐸 𝛿𝐸
1 1 1 𝛿 𝐸
∴𝑃= = = = ( )
𝑞 𝐶𝜆𝐴 𝐶𝜆 (𝐹𝐿𝑜 ) 𝐶𝐹𝐿𝑜 𝜆
𝛿𝐸

𝐸
∴ 𝑀2 =
𝜆

14
PERFORMANCE EQUATION THREE: And finally the vibration
constraint: For a vibrating rod with a mass at the end, the fundamental
(lowest) frequency is

2 𝐸𝐼 𝐴2 2 𝐸𝐴2
𝑓= √ 4 , 𝐼= ⟹𝑓= √
𝜋 𝑚𝐿 4𝜋 𝜋 4𝜋𝑚𝐿4

𝐴 𝐸 2
𝜋𝑚
⟹𝑓= √ ⟹ 𝐴 = 𝜋𝐿 𝑓√
𝜋𝐿2 𝜋𝑚 𝐸

1 1 1 1 𝐸 1⁄2
∴𝑃= = = = ( )
𝑞 𝐶𝜆𝐴 𝜋𝑚 𝐶𝜋𝐿2 𝑓√𝜋𝑚 𝜆
𝐶𝜆 (𝜋𝐿2 𝑓√
𝐸 )

𝐸 1⁄2
∴ 𝑀3 =
𝜆

To perform the multiple selection stage process, we would set up two


stages, one for 𝑀1 and one for 𝑀2 , 𝑀3 . For the active constraint approach,
we have to know more about the design, especially the details of the values
of the fixed and constraint parameters. First, write out the equations for the
MOP using each of the constraints:

1 1 𝜎𝑓
𝑃𝑀𝐴𝑋 (𝑀1 ) = = =
𝑞 𝐶𝜆 𝐹 𝐶𝜆𝐹
𝜎𝑓
1 1 𝛿𝐸
𝑃𝑀𝐴𝑋 (𝑀2 ) = = =
𝑞 𝐶𝜆 𝐹𝐿𝑜 𝐶𝜆𝐹𝐿𝑜
𝛿𝐸
1 1 𝐸 1⁄2
𝑃𝑀𝐴𝑋 (𝑀3 ) = = =
𝑞 𝐶𝜋𝐿2 𝑓 𝜋𝑚 𝜆 𝐶𝜋𝐿2 𝑓 √𝜋𝑚𝜆
√ 1⁄ 2
𝐸
If we know, or can estimate, the values of the fixed and constraint
parameters, we can calculate numerical values of the measures of
performance for each material. Let's put some numbers down for this
design:

15
𝐹 = 196 𝑁
𝑚 = 20𝑘𝑔
𝛿𝑚𝑎𝑥 = 0.1𝑚
𝑓𝑚𝑖𝑛 = 100𝐻𝑧 = 100(1⁄𝑠)
𝐿 = 0.1𝑚
𝐶 = 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 = 300𝐾⁄0.1𝑚 = 3000 𝐾⁄𝑚

Now we can set up a spreadsheet table of values for the material properties
of the materials we're interested in and calculate the measures of
performance. I pulled the rough values from the Ashby selection charts:

For each individual material, we look at the SMALLEST value of P.


WHY? In order to satisfy all the constraints, we must satisfy the one that
most limits our performance. If we can satisfy that one (by choosing a
particular value of r, the free parameter), we will satisfy all of them,
In this example, the minimum performance for all of the materials is the
vibration constraint-- it is the ACTIVE CONSTRAINT for all of the
materials we have examined. If we don't satisfy it, the design will fail.
Now, we can pick the material with the LARGEST value of the active
constraint performance (P3 in this example) to be the optimal performer
for the design, in this case CFRP.
The Last Step: REALITY CHECK: Let's plug back into the constraint
equations to find the value of the cylinder radius in each case:
16
17

You might also like