# Basic Concept of Axiomatic Design by ewghwehws

VIEWS: 3 PAGES: 71

• pg 1
Multi-FR Design II

Prof. W. Hwang
Dept. of Mechanical Engineering
Postech

POSTECH ME PCM                               Chapter 06 Multi-FR Design 2   1
Example 1.1 Design and assembly of the injection-
molded vacuum cleaner wheel

The wheel must rotate freely and should withstand
a pulling force of 500N.
It must be easily assembled during the manufacturing
operation with an axial force of less than 50N.

Clean Vac Corp. has found that many parts of the
Cleaner was broken when it was assembled by a hammer.
Clean Vac is concerned about long-term durability.

Fig. 6.1 Cross-sectional View of the Wheel/Shank

For this purpose, you are asked to provide the following:
a. Define the FRs
b. Develop DPs
c. For your chosen DPs, determine the design matrix
d. Modeling the relationship between FRs and DPs
e. Optimize the design based on the design matrix and the model

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   2
Example 1.2 Design and assembly of the injection-
molded vacuum cleaner wheel

Determination of FRs and Cs

FR1 = Make the wheel rotate easily by maintaining low friction between the wheel and the vacuum
cleaner body and by making the torque exerted by the wheel/floor contact larger than the
friction of the contact of plastic components

FR2 = Retain the wheel in the vacuum cleaner body under 500N of pulling force

FR3 = Provide a means of easy assembly with an axial force of less than 50N

FR4 = Carry the weight of the vacuum cleaner and accidental load applied when people step on the
vacuum cleaner(200lbs)

POSTECH ME PCM                                                        Chapter 06 Multi-FR Design 2    3
Example 1.3 Design and assembly of the injection-
molded vacuum cleaner wheel

C1 = No fracture

C2 = No fatigue failure

C3 = No plastic yielding of the wheel

C4 = Torque due to the traction at the wheel and the floor > torque at the shaft
surface due to the friction between plastic components

C5 = Manufacturing considerations, e.g., injection-molded part should have
approximately a constant thickness to prevent secondary flow caused by
non-uniform cooling

C6 = Minimize the manufacturing cost

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   4
Example 1.4 Design and assembly of the injection-
molded vacuum cleaner wheel

Selection of DPs
DP1    ( D1  D2 ) / 2  (i.e., the clearance between the diameters of the wheel
DP2  t  ( D2  D3 ) / 2 and the vacuum cleaner body)
DP3  
DP4   D2t         (i.e., the area of the tubular stem without the axial cut)

Fig. 6.2 End View of the Shank for the Clean       Fig. 6.3 Free-body Diagram of one of the curved
Vac Design                                             Beams.

POSTECH ME PCM                                                       Chapter 06 Multi-FR Design 2     5
Example 1.5 Design and assembly of the injection-
molded vacuum cleaner wheel

Design equation
DP 1( )   DP 2(t )   DP 3()   DP 4( D 2 t)    FR1   A11   0   0   0   DP1 
 FR   0      A22 A23 A24   DP2 
FR 1     X          0          0            0
 2                             
FR 2     0          X          X            X                                          
FR 3     0          X          X            0           FR3   0     A32 A33 0   DP3 
FR 4     0          X          0            X           FR4   0
      

A42 0 A44   DP4 
     
To make the design a decoupled design we must make the off-diagonal elements
A23 and A24 zero.

A23 can be made to be zero if we make the total circular length of all the curved
beams remain constant by adding more sections of the curved beams, but this may
increase the manufacturing cost.

Another way of decoupling the design is to choose either the height h of the interlock
key or the length L of the beam as DP3.

A24 can be made zero by choosing the area of the unslitted section A as DP4.

POSTECH ME PCM                                                      Chapter 06 Multi-FR Design 2    6
Example 1.6 Design and assembly of the injection-
molded vacuum cleaner wheel

We will set the length L=2D2.
We will then select the height h of the interlock key as DP3.

 FR1   A11 0 0 0   DP1   
 FR   0 A 0 0   DP  t 
 2                  2        

22
                               
 FR3   0 A32 A33 0   DP3  h 
 FR4   0 0 0 A44   DP4  A
                             

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   7
Example 1.7 Design and assembly of the injection-
molded vacuum cleaner wheel

Modeling the relationship between the FRs and DPs
a. Evaluating A11

We will choose the clearance to be 0.010 inch(0.25cm) in each side.

The friction force F is

FR1  F  W
Then A11 is given by
W
A11      constant

If the friction coefficient is the same, C4 is satisfied as long as the diameter
of the wheel is larger than the shaft diameter

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   8
Example 1.8 Design and assembly of the injection-
molded vacuum cleaner wheel

b. Evaluating A22


We assume that there are three circular sections and that         is the included angle.

                   3
FR2  Fpull  3( )( D22  D32 )  ( )( D2  t )t
8                    2
If we want to avoid fatigue, then a good rule of thumb is that    
should not exceed  y / 2    .

FR2 3
A22          ( )( D2  2t ) y
DP2    4

POSTECH ME PCM                                                       Chapter 06 Multi-FR Design 2   9
Example 1.9 Design and assembly of the injection-
molded vacuum cleaner wheel

c. Evaluating A33

FR3  P tan 
DP3  h
The deflection at the end of the cantilever is given by

PL3
h
3EI

3EI
A33  (     3
) tan 
L

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   10
Example 1.10 Design and assembly of the injection-
molded vacuum cleaner wheel

d. Evaluating A44

y
FR4  W  A  A
2
DP4  A
y
A44  (       )
2

e. Evaluating A32

 3 tan  
3EIh
         
A32 
FR3
   L         3Eh tan  I
DP2         t          L3     t

POSTECH ME PCM                                            Chapter 06 Multi-FR Design 2   11
Example 1.11 Design and assembly of the injection-
molded vacuum cleaner wheel

We can numerically determine t, h, diameters and the length of the circular cantilever beam
after setting the value of      , which was set to be 100 degrees after trying several possibilities.

The material properties for nylon are:

Coefficient of friction = 0.4
E=362,590 psi
 y = 7,250 psi
The solution of the design equations for the dimensions are approximated as:

D2  0.375 inch
t  0.030 inch
D3  0.315 inch
D1  0.395 inch
h  0.0625 inch
D4  0.437 inch
POSTECH ME PCM                                                        Chapter 06 Multi-FR Design 2   12
1. The relationship between complexity
and information content

Next example proves the following statements.

(1)   Complexity is related to the probability of achieving the functional requirement.

(2)   Even the same design can have very different information content and
complexity, depending on the stiffness of the system.

(3)   Information content , Complexity , Probability of Success
Information content is a measure of design complexity.

(4)   A design that violates the Independence Axiom is more complex and requires

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   13
Example 2.1 Knob Design

FR1 = Grasp the end of the shaft tightly with axial force of 30N
FR2 = Turn the shaft by applying 15 N-m of torque
DP1 = Interference fit between the shaft and the inside diameter of the knob
DP2 = The flat surface
The design equation may be written as

FR1   X X  DP1 
                
 FR2   x X  DP2 

The lower-case x is used to signify the fact that the effect of DP1 on FR2 is much less
than the other effects indicated by upper-case X.

Eventually, when the grip force is less than the required force to keep the knob on the shaft
the knob will slide off the shaft. How do we solve this problem?

POSTECH ME PCM                                                      Chapter 06 Multi-FR Design 2   14
Example 2.2 Knob Design

Some will suggest that the solution to this coupled design problem is to make the outer
diameter of the knob shaft thicker, which will make the slot open up less and thus
minimize the reduction of the gripping force.

However, this solution has its cost; not only does it require more materials but also higher
information content, which ultimately means a higher manufacturing cost.
Depending on the stiffness, the same bell-shaped
distribution along the DP axis translates into very
different system distributions in the functional domain .

When the stiffness is lower, the system pdf fit in
the design range, but when the stiffness increases, the
system pdf is outside the design range.

When the thickness of the cylinder wall increases,
Fig. 6.4 Dependence of System pdf
on “Stiffness.”          the stiffness increase.

POSTECH ME PCM                                                    Chapter 06 Multi-FR Design 2   15
Example 2.3 Knob Design

New Design
The slot terminates where the flat part of the knob begins.
Since the flat surface is completely away from the slot,
the turning action does not force the slot to open and
therefore, the axia l grip is not affected.
This is a completely uncoupled design.
Fig. 6.5 A New Uncoupled Design.

How do we actually determine the wall thickness and the desired interference?

The thickness must be determined by considering two limiting factors:
Manufacturability and failure of the knob under stress.

Can it be manufactured by injection molding?

Does the maximum stress at the bottom corner of the slit, which is the stress
Concentration point, cause either fracture or plastic deformation?

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   16
Example 2.4 Knob Design

Fig. 6.6 Modified Shaft.                  Fig. 6.7 A Cantilever Beam Loaded at the
End
The maximum deflection is given by
FL3                      bh3
ymax                 where I 
3EI                      12
The maximum stress is given by                 The stiffness K   max / ymax          is given by
FLh 6 FL                                          3Eh
 max         2                                       K 2
2I  bh                                            2L
To minimize K for robustness, h should be made as small as possible.
The limit is reached when     max    reached  y .                      1
  yb  2
Then the smallest h is obtained as                             6 FL 
hmin        
      
POSTECH ME PCM                                                        Chapter 06 Multi-FR Design 2   17
The foregoing example illustrates the following aspects of
the Information Axiom and the Independence Axiom:

   Complexity is related to the probability of achieving the functional
requirement. The coupled design made it much more difficult to
make the knob.

   Even the same design can have a very different information
content and complexity, depending on the stiffness of the system.

   The greater the information content, the more complex is the task
of achieving the FR since the probability of success decreases.
Therefore, information content is a measure of design complexity.

   A design that violates the Independence Axiom, i.e., a coupled
than a design that satisfies the Independence Axiom.

POSTECH ME PCM                                         Chapter 06 Multi-FR Design 2   18
2.1 Determination of Information Content

   Uncoupled design

The probability that all m FRs are satisfied by uncoupled designs can be computed by the
product of the probabilities for each FR.

m
I     log 2 Pi      (m : the number of FR)
i 1

Theorem 13 (Information Content of the Total System)
If each DP is probabilistically independent of other DPs, the information content
of the total system is the sum of the information of all individual events associated
with the set of FRs that must be satisfied.

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   19
2.2 Determination of Information Content

   Decoupled design
The probability that all m FRs are satisfied by decoupled designs can be computed by
the product of the probabilities for each FR, provided that appropriate conditional
probabilities are used where necessary.

m
I     log 2 Pi| j      (m : the number of FR)
i 1

Theorem 12 (Sum of Information)

The sum of information for a set of events is also information, provided that proper
conditional probabilities are used when the events are not statistically independent.

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   20
2.3.1 Determination of Information Content

   For example, consider the following design matrix

FR1   A11 0  DP1 
                    
 FR2   A21 A22  DP2 


   The information content of FR1 can be determined by computing
the area of the system pdf in the common range just as for an
uncoupled design, since DP2 does not affect FR1.

   However, to compute the information content associated with FR2,
we have to include the change in the information content due to the
off-diagonal element.

POSTECH ME PCM                                            Chapter 06 Multi-FR Design 2   21
2.3.2 Determination of Information Content

Shift of Mean Value

The solid curve is the system pdf of FR2
when the off-diagonal element A21 is
equal to zero.

Fig. 6.7 Shift of the FR2 System pdf Due to change in DP1

However, the system pdf of FR2 for a decoupled design may be shifted
to the right or left by the off-diagonal element A21, which changes the
mean of the system pdf for FR2 when DP1 changes because FR2 is
affected by DP1.

POSTECH ME PCM                                                  Chapter 06 Multi-FR Design 2   22
2.3.3 Determination of Information Content

Change of Variance
In some case, the variance of the
system pdf can change as well as the
mean.

Fig. 6.9 Change in Variance of System pdf
of FR2 Due to change in DP1.

When the system pdf is symmetrical with its mean in the middle of
the design range, the effect of the off-diagonal element is to change
the spread of the system pdf, as shown in next example.

POSTECH ME PCM                                               Chapter 06 Multi-FR Design 2   23
2.3.4 Determination of Information Content

In the results, the information content of the decoupled design can
increase or decrease by the off-diagonal element.

However, in most cases, the information content of a decoupled design
is expected to be larger than that of an uncoupled design, since a
decoupled design can not be as robust as an uncoupled design.

POSTECH ME PCM                                        Chapter 06 Multi-FR Design 2   24
Example 3.1 Information Content of a Decoupled Design

FR1 = Turn the shaft               The design ranges for FR1 and FR2 are +/- 5%, i.e.,
FR2 = Grip the shaft                               FR1 = 1 +/- 0.05
DP1 = Flat surface                                 FR2 = 1 +/- 0.05
DP2 = Interference fit
 FR1   1 0   DP 

1
                     
 FR2   0.2 1   DP2 

Supplier A : +/- 5% for the DP1 tolerance
Supplier B: +/- 10% for the DP1 tolerance

Both suppliers : the same 6% for DP2 tolerance

Fig. 6.10 System pdf and Common Range             Fig. 6.11 System pdf and Common Range
of FR1 for Supplier A.                             of FR1 for Supplier B.

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   25
Example 3.2 Information Content of a Decoupled Design

From the measurements, it was determined that of the parts that were within the DP1
tolerance, only 90% were also within the DP2 tolerance.
Determine the information content of this design.
The DPs are determined as:

DP1  FR1 / A11  1.0
DP2  ( FR2  A21 DP1 ) / A22  (1  0.2(1)) / 1  0.8

DP1  FR1 / A11  0.05
DP2  (FR2  A21DP1 ) / 1  0.04

If the manufacturing process cannot hold the DP1 tolerance to within DP1,two things
will happen for those parts that are outside the DP1 tolerance.

(1)   FR1 will not be satisfied.
(2)   The established DP2 tolerance will be too large to satisfy FR2

POSTECH ME PCM                                                           Chapter 06 Multi-FR Design 2   26
Example 3.3 Information Content of a Decoupled Design

In order to calculate the probability that FR2 is satisfied, we must first determine the
probability that FR1 is satisfied.

Supplier A
I   log 2 (1)  log 2 (0.833)  0.263
Supplier B
I   log 2 [Pr(satisfy FR1 )]
 log 2 [Pr(satisfy FR2 | satisfy FR1 )]
Fig. 6.12 Specified tolerance for DP2      I   log 2 (0.5)  log 2 (0.833)  1.264
and actual pdf of DP2.

The probability that FR2 is satisfied must be computed conditional upon FR1 being
satisfied.

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   27
3.1 Accommodating “Noise” in the design process

   During manufacturing and use of a product, random variation from
various sources affects the performances of a machine or system.

   The variation so introduced is given the generic name “Noise”.

   There are five generic noise sources:
Manufacturing variation
Customer usage
Environmental variation
System-to-system iteration

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   28
3.2 Accommodating “Noise” in the design process

Recall example of one-FR design(Joining of aluminum Tube to
Steel shaft). Noise was introduced by the random variation of the
machining processes and by the temperature fluctuation in service.

Muti-FR designs must also accommodate noise by adjusting the
“stiffness”.

POSTECH ME PCM                                     Chapter 06 Multi-FR Design 2   29
3.3 Accommodating “Noise” in the design process

When the design is a decoupled design with a triangular design matrix [A], the variation
of FRi is caused by the random variations of many DPs, which may be expressed as
i
δ FRi      Aijδ DPj
j 1

To satisfy FRi, the elements Aij that correspond to large values of the DPs must be made
smaller.
δ FRi  M iδ DPi
Where Mi is defined as a module which is equal to
i
DPj
M i   Aij
j 1    DPi

To Minimize the effect of random noise, Mi must be decreased if the random variation
in FR is larger than the specified design range of FR.

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   30
4.1 Integration of DPs to Minimize the Information Content

In general, physical integration reduces the information content by
removing the uncertainly associated with assembling several physical
pieces.

Providing that the Independence Axiom is not violated, DPs may be
integrated in a single physical part under the following circumstances;

1)    DPs do not undergo relative motion
2)    DPs can be made of the same material
3)    Integration does not create a problem such as excess stress and fracture
4)    Integration does not violate a cost contraint
5)    The integrated parts can be manufactured

POSTECH ME PCM                                                Chapter 06 Multi-FR Design 2   31
4.2 Integration of DPs to Minimize the Information Content

The integration of the physical part must be consistent with the DP

hierarchy in the physical domain, where all leaf-level DPs are related

to other leaf-level DPs according to the specified relationship.

POSTECH ME PCM                                       Chapter 06 Multi-FR Design 2   32
5.1 Nonlinear Multi-FR Design

When the elements of the design matrix are not constants, but instead
are the function of DPs, the design is a nonlinear design.

There are three kinds of situation in nonlinear design.

(1)   The design matrix is always either diagonal or triangular regardless of how DP
change.

(2)   The elements of the matrix may vary, depending on the specific values of DPs,
so that the design behaves as a coupled, uncoupled, or decoupled design in
different parts of the design window.

(3)   The design is always coupled regardless of the specific values of DPs.

POSTECH ME PCM                                                Chapter 06 Multi-FR Design 2   33
5.2 Nonlinear Multi-FR Design

The difference between the linear and the nonlinear design of the second kind is that in
the case of nonlinear design, we may strive to find a better design window, because of
the elements of the design matrix changes as functions of DPs.

This can be illustrated graphically. FR1 and FR2 are independent each other by definition.
So they are mutual orthogonal.

Uncoupled linear case:
DP1 affects only on FR1 and
DP2 affects only on FR2.
So, DP1 parallels to FR1 and
DP2 parallels to FR2.

Fig. 6.13(a) A Completely Uncoupled Two-FR Design.

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   34
5.3 Nonlinear Multi-FR Design

Decoupled linear case:

DP1 affects only on FR1 but
DP2 affects on FR1 and FR2.
So, DP1 parallels to FR1 but
DP2 doesn’t parallel to FR2.

Fig. 6.13(b) A Decoupled Two-FR Design.
Coupled linear case:

DP1 affects on FR1 and FR2 and
DP2 affects on FR1 and FR2.
So, DP1 doesn’t parallels to FR1
and DP2 doesn’t parallels to FR2.

Fig. 6.13(c) A Coupled Design.

POSTECH ME PCM                                                Chapter 06 Multi-FR Design 2   35
5.4 Nonlinear Multi-FR Design

In nonlinear design, the lines of
constant DPs are curved since
the elements of the design matrix
are function of DPs as shown in
Fig. 6.13(d).

Fig. 6.13(d) A Case of Nonlinear Dsign.

Nonlinear Case
Region A – The magnitude of FR1 is small and that of FR2 is large.
Nearly uncoupled region. Diagonal elements are zero or very small.
Region B – Nearly decoupled region
Region C – Coupled region

POSTECH ME PCM                                                Chapter 06 Multi-FR Design 2   36
5.5 Nonlinear Multi-FR Design

When there are more than two FRs, it is difficult to use a graphical
means. As an alternate means of measuring the independence of FRs,
We define two scalar metrices – reangularity R and semangularity S.

2
                       2                                      
       n                                                    

                                                            
            Aki Akj 
           n                     
              
                                                    
1     k 1                                      A jj       
R                 n                                 S
 n                                      1/ 2 
                                          n               
            
i 1, n 1                                           j 1 
j  i  i, n  
  k 1
 
Aki2 

 k 1
Akj2  





        2
Akj 




                                             k 1           

These measures are useful when there are many FRs and DPs.

POSTECH ME PCM                                                                Chapter 06 Multi-FR Design 2   37
5.6 Nonlinear Multi-FR Design

Reangularity R measures the angular relationship between the DP axes.
Semangularity S measures the magnitude of the diagonal elements of a
normalized design matrix.

R=S=1              The design is an uncoupled design
R=S                The design approaches a decoupled design.
[When there are only two FRs and two DPs, R=S represents a decoupled design.]

In all other cases, the design is a coupled design.

In fig. 3.3 (d) , Region A :     1
R=S
Region B :     RS
Region C :     R< 1, S< 1

POSTECH ME PCM                                              Chapter 06 Multi-FR Design 2   38
6.1 Axiomatic Design Basis for Robust Design

Why robust?

1. The original design goals can be achieved easily and faithfully.

2. The product must be reliable and durable.

Necessary condition for robust design

The fulfillment of FRs within the bounds established by constraints under all

operating conditions

POSTECH ME PCM                                                  Chapter 06 Multi-FR Design 2   39
6.2 Axiomatic Design Basis for Robust Design

One-FR Design (Review)
Consider redundant design with one-FR

FR1  f ( DP a , DP b , DP c ,, DP n )
Task: Make this design robust under all conditions, if possible.

The desired change of FR,

f          f                f
FR        DP 
a
DP  ... 
b
DP n
DP a       DP b             DP n

In an ideal design for one-FR, only one DP is needed. All other DPs
are possible sources of rand variations

POSTECH ME PCM                                         Chapter 06 Multi-FR Design 2   40
6.3 Axiomatic Design Basis for Robust Design

Robust Design

1.    Make the coefficient (f/DP) associated with extra DPs to be zero.
– Immune to random variation
[one of the basic concept of robust design practiced in industry today]

e.g. Windshield Wiper – Robust Mounting Design

2.    Fix the values of all DPs except one DP chosen

e.g. Van Seat Assembly

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   41
6.4 Axiomatic Design Basis for Robust Design

The desired change of FR may be expressed as
f         i n
f
FR        DP  
c
DP i
DP c       i  a DP
i

i c

f
       DP c  [Extra terms]
DP c
 [Module]DP c  [Extra terms]
 [stiffness ]DP c  [Extra terms]

Where DPc is the DP chosen to satisfy FR.

POSTECH ME PCM                                         Chapter 06 Multi-FR Design 2   42
6.5 Axiomatic Design Basis for Robust Design

How to select DP?

   The magnitude of the term (f/DP)DP of the chosen DP should be larger than

the sum of the constant terms so that the accumulated errors can be

compensated.

   If the magnitudes of two or more terms are approximately the same, the one with

smaller f/DP should be chosen to minimize the sensitivity of FR to the

variation of DP.

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   43
6.6 Axiomatic Design Basis for Robust Design

Multi-FR Design
The robust design concept discussed with respect to one-FR design does apply, if the
design satisfies the Independence Axiom.

Consider redundant design with multi-FR
{FR} = [Square DM]{DP} + [Extra Matrix]{DP}extra

{DP} = the vector of DPs chosen to satisfy the vector {FR}
{DP}extra = the vector of the redundant DPs
[Square DM] = must be either diagonal or triangular matrix to satisfy
the Independence Axiom
[Extra Matrix] = can be any matrix, including a full matrix

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   44
6.7 Axiomatic Design Basis for Robust Design

Consider a special case of three FR design,
 DP4 
 DP 
 5
 FR1   X     0   0   DP1   X     X   X   . . . X   DP6 
                                                           
 FR2    0   X   0   DP2    X   X   X   . . . X  . 
                                           
 FR   0               DP   X
X  3                    . . . X  . 
 3           0                      X   X                 
 . 
 DP 
 7

The above design can be treated as uncoupled design, if the values of
DP4 through DPn are fixed.

POSTECH ME PCM                                            Chapter 06 Multi-FR Design 2   45
6.8 Axiomatic Design Basis for Robust Design

The desired change of each FRi may be expressed as

FRi          n
FRi
FRi       DPi            DPj
DPi        j  4 DPj

FRi
      DP  [Extra terms]
DPi

The above equation is similar to one-FR design case with similar implica-
-tions for compensation by fixing the values of the extra DPs.

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   46
6.9 Axiomatic Design Basis for Robust Design

If [Square DM] is a triangular matrix,
 DP4 
 DP 
 5
 FR1   X     0   0   DP1   X     X   X   . . . X   DP6 
                                                           
 FR2    X   X   0   DP2    X   X   X   . . . X  . 
                                           
 FR   X               DP   X
X  3                    . . . X  . 
 3           X                      X   X                 
 . 
 DP 
 7

The above design can be treated as decoupled design, if the values of
DP4 through DPn are fixed.

POSTECH ME PCM                                            Chapter 06 Multi-FR Design 2   47
6.10 Axiomatic Design Basis for Robust Design

The desired change of each FRi may be expressed as

FRi         3
FRi          n
FRi
FRi       DPi           DPj            DPk
DPi        j 1 DPj        k  4 DPk
j i

FRi        3
FRi
      DP           DPj  [Extra terms]
DPi       j 1 DPj
j i

In compensating for this design, DPj must be set first
According to the sequence defined by the triangular matrix.

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   48
6.11 Axiomatic Design Basis for Robust Design

How to select the primary DPs?

The selection of DP1, DP2, and DP3 in a multi-FR design
Must satisfying the same set of conditions as those discussed
For one-FR design; robustness and sensitivity.

POSTECH ME PCM                                     Chapter 06 Multi-FR Design 2   49
Example 4.1 Robust design of a micro-gyroscope

It is made of silicon by means of photo-lithograph and etching.
It measures the motion by resonant vibration responses of MEMS
in response to external motion.

The mechanism

1.   Measurement of bending resonance by means of electric potential
2.   Measurement of angular velocity
3.   Measurement of Coriolis acceleration
4.   Sensing of torsional resonance of the sensing plate
5.   Sensing by capacitance

It measures motions in one translational direction(the x-direction) and
the rotational motion about the x-axis.

POSTECH ME PCM                                              Chapter 06 Multi-FR Design 2   50
   Electrostatic Comb Drive

POSTECH ME PCM                  Chapter 06 Multi-FR Design 2   51
Example 4.2 Robust design of a micro-gyroscope

It measures motions in one translational direction(the x-direction)
and the rotational motion about the x-axis.

When three of these gyroscopes are mounted along the three
orthogonal directions, they can measure motion in six directions.

The driving force generates the translational motion, deforming the
four bending springs, which in turn induces the rotational motion of
the central plate(gimbal) that is attached to the translational plate by
the sensing spring.

The actual measurement of the relative motion is done by means of
the capacitance change between series of capacitor plates between
the stationary part and the moving part.

POSTECH ME PCM                                        Chapter 06 Multi-FR Design 2   52
Example 4.3 Robust design of a micro-gyroscope

The gyroscope is designed to have two specific natural frequencies,
f1   and f 2  . f1is the driving force mode and f2 is the sensing
mode.
These two frequencies must be exactly the same to have the best
response and provide the most accurate measurement.
Because the tolerance of the manufacturing processes is only 10%,
there is a frequency mismatch between the two modes.
The current design has large random variation in dimensional
tolerances due to the manufacturing accuracy, so tuning is extremely
difficult.
How can you manufacture more easily and reliably, increasing the
yield of good gyroscopes?

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   53
Example 4.4 Robust design of a micro-gyroscope

Fig. 6.14 Design of the original resonant   Fig. 6.15 Finite element model of the gyroscope
vibratory gyroscope

POSTECH ME PCM                                                Chapter 06 Multi-FR Design 2      54
Example 4.5 Robust design of a micro-gyroscope

FR1= Set the frequency of the driving mode-
the translational motion of the moving
plate- at f1
FR2=Set the frequency of the sensing mode-
the torsional motion of the central plate
- of the central plate at f2
FR3=Let the distribution of the frequency
difference f2-f1 be insensitive to geometric
tolerances of the gyroscope(i.e., set the mean
of (f2-f1) to be within the lower bound aL
and the upper bound aU)
DP1=Stiffness of the four bending springs
DP2=Stiffness of the two torsional springs

In the original design, DP3 was absent, and thus
Fig. 6.16 The driving mode(upper) and
the design was a coupled design.
the sensing mode(lower)

POSTECH ME PCM                                                  Chapter 06 Multi-FR Design 2   55
Example 4.6 Robust design of a micro-gyroscope

If we select another DP3, the design equation may be written as

 FR1   X       0     ?   DP 
1
                               
 FR2    0     X     ?   DP2 
               
 FR   ?              X   DP3 
 3              ?            

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   56
Example 4.7 Robust design of a micro-gyroscope

Finding DP3
To satisfy FR3, the variance of (f2-f1) must be made very small, the best being zero.
This can be done if we can make the design immune to the variations introduced by
manufacturing operations so that the variance of FR is equal to zero.
2            0
f   2  f1
The natural frequencies of both f1 and f2 are affected by the distribution of the
moments of inertia of the gimbal plate. the gimbal plate has three dimensions (a,b,c),
all of which affect the FR1, FR2 and FR3.

Fig. 6.17 Gimbal plate with three dimensions

POSTECH ME PCM                                                      Chapter 06 Multi-FR Design 2   57
Example 4.8 Robust design of a micro-gyroscope

 f1                              f1               f1              f1         DP1 
                  0                                                                   
 FR1   DP1                                a                b               c         DP2 
                         f 2             f 2              f 2             f 2        DP a 
 FR2     0
 3 
(a)
 FR                    DP2                a                b               c               
 3   ( f 2  f1 )    ( f 2  f1 )    ( f 2  f1 )     ( f 2  f1 )    ( f 2  f1 )   DP3b 
                                                                               
 DP1
               DP2                a                b               c         DP3c 
       

If the design equation elements are under following conditions,

f1       f       f
da  1 db  1 dc  0
a        b        c
The design is a decoupled design.
f 2      f 2      f 2
da       db       dc  0
a         b        c

POSTECH ME PCM                                                                      Chapter 06 Multi-FR Design 2   58
Example 4.9 Robust design of a micro-gyroscope

If the design equation elements are under following conditions,

( f 2  f1)       ( f 2  f1)        ( f 2  f1 )      ( f 2  f1)      ( f 2  f1)
dDP 
1               dDP2                da               db               dc  0
DP  1            DP2                   a                 b                c

Can be reduced

( f 2  f1)      ( f 2  f1)      ( f 2  f1)
da               db               dc  0
a                b                c

FR1 and FR2 can be satisfied by varying DP1 and DP2 to reach target frequencies
 and  .

POSTECH ME PCM                                                     Chapter 06 Multi-FR Design 2   59
Example 4.10 Robust design of a micro-gyroscope

Set up an orthogonal array “experiment” and FR3 was evaluated for 27 combinations of
(a,b,c) at three level of DPs as shown in Table ex4.1, using the finite element method.
By various trials, they recommended values as given in table ex5.2.

Table. Ex 4.1

Level            DP3a:a                   DP3b:b               DP3c:c
1             Lower bound            Lower bound           Lower bound
2               Current                  Current              Current
3             Upper bound            Upper bound           Upper bound

Table. Ex 4.2

Design        f1   f2   (f2-f1)/f1      (f2-f1)          (f2-f1)        Improvement

Original       bO   gO    0.287        334.33                 84.84
Recom.         bR   gR    0.119        127.15                 51.81           38.9%

POSTECH ME PCM                                                        Chapter 06 Multi-FR Design 2   60
Example 4.11 Robust design of a micro-gyroscope

They determined the probability of success of their proposed design.
The results show a significant increase.

Table. Ex 4.3

Design        Probability of success          Information content
Original               5.7%                           4.14
Rrecom.                86.0%                          2.18

POSTECH ME PCM                                                 Chapter 06 Multi-FR Design 2   61
Example 4.12 Robust design of a micro-gyroscope

Fig. 5 The fabricated de-coupled vertical gyroscope
(a) The perspective view (b) the wafer level vaccum packaged gyroscope
(c) The closed view of comb electrode (d) pad and interconnection feedthrough

POSTECH ME PCM                                                  Chapter 06 Multi-FR Design 2   62
7.1 Design of Dispatching Rules and Schedules

   Dispatching and scheduling are important tasks in many situations
such as production of mechanical parts in job shops, scheduling of
robot tasks in automated manufacturing system, and scheduling of
airline flights.

   So far, the mathematical tools of operations researcher simulations
of the actual situation have not always been successful.

   Because the design that violates the Independence Axiom can not
be improved through optimization, these techniques do not always
yield sufficiently improved results.

POSTECH ME PCM                                       Chapter 06 Multi-FR Design 2   63
7.2 Design of Dispatching Rules and Schedules

They must be designed right ( to satisfy the Independence Axiom )
before the parameters can be adjusted to obtain the correct FRs, and the
Information Axiom should be applied to minimize the information
content.

All dispatching and scheduling algorithms must satisfy the Independence
Axiom and the Information Axiom to be able to come up with a rational
strategy.

POSTECH ME PCM                                     Chapter 06 Multi-FR Design 2   64
7.3-1 Design of Dispatching Rules and Schedules

When an identical set of parts is processed through a variety of different
machines but the same set of processes, rational scheduling and
dispatching algorithms can be developed based on the Independence
Axiom so that the scheduling and transport of the part will be uncoupled
from the manufacturing processes.

In this case, we can come up with a “Push type” process that can
maximize the productivity.

POSTECH ME PCM                                       Chapter 06 Multi-FR Design 2   65
7.3-2 Design of Dispatching Rules and Schedules

When a random set of parts is process through a variety of different
processes, a “push” system can no longer maximize the throughput rate.
In this case, independence of FRs can be satisfied by designing a
cellular manufacturing system – a “Pull” system. This “pull” system
will control the production rate based on the demand rate, an approach
which satisfies the Independence Axiom.

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   66
7.4 Design of Dispatching Rules and Schedules

    Decoupler

The system must be designed correctly by decoupling the tasks from each other using

“ decoupler”.

The role of the decoupler is to eliminate coupling when more than one part requires

the attention of the same robot(or person) at the same time, or when a machine is
not ready to accept the next part which has just been complete by a preceding
machine.

POSTECH ME PCM                                                    Chapter 06 Multi-FR Design 2   67
8.1 Dispatching Rules and the Independence Axiom

There are several special cases of dispatching situations;

(a)   Frequency of dispatches for identical parts
Consider a manufacturing system where a certain part must be processed by N
machines in a sequential arrangement.

τ i  the processing time at each machine
τ t  the transport time between the machines
τ m  the longest processing time ( Machine m)

τ d  τ m τ t                (1)

Then, the part should be dispatched for processing at an interval   τ d given by,

POSTECH ME PCM                                                  Chapter 06 Multi-FR Design 2   68
8.2 Dispatching Rules and the Independence Axiom

If the dispatching rate must be increased to a higher rate than that given
equation(1), the number n of the slowest machines must be increased to

τ 
n  int  m 
τ                  (2)
 d
where int(x) is an integer rounded to the next whole number for any x.

(b) Dispatch rate when random parts are processed

Dispatch rate can not exceed the “dispatching” rate given by equation (1). When
the demand rate is greater than that given by equation (1), more machines must
be added according to equation (2).

POSTECH ME PCM                                                   Chapter 06 Multi-FR Design 2   69
9. Scheduling

Scheduling depends on whether an identical set of parts or different
random parts are being processed by the system.

POSTECH ME PCM                                      Chapter 06 Multi-FR Design 2   70
Summary

   The implications of the Independence Axiom and the Information

Axiom are presented with relevant theories that govern multi-FR

designs.

   The robust-design concept is given for the multi-FR case as well
as discussing the relationship between the complexity and the
information content of a design.

POSTECH ME PCM                                     Chapter 06 Multi-FR Design 2   71

To top