Control and noise factors
Don Clausing
Red border: emphasized slides
? Don Clausing 1998
16.881
Fig. 1
The engineered system
Noise
Signal System
Response
Control
factors
? Don Clausing 1998
16.881
Fig. 2
Ideal response
? Want Ideal Response to Signal – usually
straight-line function
? Actual response is determined by values of
control factors and noise factors
? If noise factors are suppressed early, then
difficult problems only appear late
? Introduce noises early!
? Don Clausing 1998
16.881
Fig. 3
Actual response
Ideal
response
Effect of
noises
RESPONSE
M
1
SIGNAL
M
2
? Don Clausing 1998
16.881
Fig. 4
Response depends on:
? Value of signal factor
? Values of control factors
– Engineers can select values
– Examples: dimensions, electrical characteristics
? Values of noise factors
– Engineers cannot select values
– Examples: temperature, part variations
? Don Clausing 1998
16.881
Fig. 5
Critical control parameters
? Strongly affect performance of the system
? IPDT can control (select) the value
? Complex systems have hundreds of critical
control parameters
? Fault trees help IPDT to identify
Note: IPDT is Integrated Product Development Team
? Don Clausing 1998
16.881
Fig. 6
Fault tree for paper feeder
FAIL TO FEED
SINGLE SHEET
MULTIFEEDMISFEED
JAM DAMAGE
BIG
SLUG
WEAK
SEPARATION
OVERLAP
SMALL
WRAP
ANGLE
INITIAL WEAR
WEAK
BELT
TENSION
LOW
RETARD
FRICTION
LARGE
FRICTION
DIFFERENCE
SMALL
RETARD
RADIUS
?μ
pp
α
0
α
t
μ
rp
T R
Identification of critical parameters – both control and noise
? Don Clausing 1998
16.881
Fig. 7
Critical parameters at base of tree
? Both control and noise parameters
? Example of control factors:
– Roll radius
– Belt tension
? Example of noise factor: paper-to-paper
friction, μ
p-p
? Don Clausing 1998
16.881
Fig. 8
Noises
? Affect performance – adversely
? IPDT cannot control – examples:
– Ambient temperature
– Power-company voltage
– Customer-supplied consumables
? IPDT must apply large magnitudes of noise
early in the development schedule
? Don Clausing 1998
16.881
Fig. 9
Role of control factors
? Values of control factors determine response
? Many combinations of control-factors values
will give same value for response
? One of these combinations will give the least
sensitivity to undesirable variations (noises)
? Improvement is achieved by searching through
the combinations of control-factors values to
find the one that gives best performance
? Don Clausing 1998
16.881
Fig. 10
Important steps in parameter
design
? Define ideal performance
? Select best SN definition
? Identify critical parameters
? Develop sets of noises that will cause
performance to deviate from ideal
? Use designed experiments to systematically
optimize control parameters
? Don Clausing 1998
16.881
Fig. 11
Three kinds of product noise
? Environment; e.g., ambient temperature
? Manufacturing – no two units of production
are exactly alike
? Deterioration – causes further variations in
the components of the system
? Don Clausing 1998
16.881
Fig. 12
Searching for robustness
1.Select one combination of control-factors
values
2.Test performance when a set of noise values
are applied: y
1
, y
2
… y
n
are performance
values for n combinations of noise
3. Change values of control factors; repeat 2.
4. Search through control-factor space, testing
each point with same set of noises
? Don Clausing 1998
16.881
Fig. 13
Introduction of noises
? Natural noises are often small in laboratory
environment; therefore slow to learn effect
? Therefore we consciously introduce large
magnitufes of noises to obtain quick
evaluation of effect of noises
? Example: (1) high temperature, high humidity,
(2) low temperature, low humidity
? Measure y
1
and y
2
? Don Clausing 1998
16.881
Fig. 14
Effect of noises, 1 & 2
FAILURE
MODE 2
FAILURE
MODE 1
A. GOOD
PERFORMANCE
FAILURE
MODE 2
FAILURE
MODE 1
Y
1
Y
2
B. BAD
PERFORMANCE
Y
Y
Y
1
Y
2
? Don Clausing 1998
16.881
Fig. 15
Understanding performance
? Case A: Is good performance due to:
– Good system?
– Small magnitude of noises?
? Winning approach:
– Apply large magnitudes of noises 1 & 2 to
create Case B
– Then improve values of control factors; B nullA
– Increase noises; repeat improvement, B nullA
? Don Clausing 1998
16.881
Fig. 16
Path to success
CON TROL
FAC TOR
SE T
NOISE 1 Y
1
NOISE 2 Y
2
Y
2
-Y
1
1C F SET 1 MODERATE 5 MODERATE 25 20
2C F SET 2 MODERATE 14 MODERATE 16 2
3C F SET 2 ST RONG 8 ST RONG 22 14
4C F SET 3 ST RONG 14 ST RONG 16 2
IMPROVEMENT PATH: CF1 nullCF2 nullCF3
? Don Clausing 1998
16.881
Fig. 17
Noises and failure modes
? Apply one noise, N
i
, for each failure mode
? Example: combination of resistances,
capacitances, transistor characteristics, and
temperature have strong effect on performance
– Adjust values, N
1
, to cause low voltage out, Y
1
– Adjust values, N
2
, to cause high voltage out, Y
2
? Then optimize control factors; minimize Y
2
-Y
1
? Don Clausing 1998
16.881
Fig. 18
Fundamental noise situation
EFFECT OF ALL
FAILURE
MODE 2
FAILURE
MODE 1
NOISES IN WORLD
TYPICAL LAB
VARIATION
Y
FORCE FAILURE MODES
PUT IN NOISES
16.881
? Don Clausing 1998
Fig. 19
Robustness improvement
FAILURE
MODE 2
FAILURE
MODE 1
PERFORMANCE
VARIATION
AFTER IMPROVEMENT
Y
PERFORMANCE VARIATION
BEFORE IMPROVEMENT
? Don Clausing 1998
16.881
Fig. 20
Important noise strategy
? Not all sources of noise need to be used
? Identify key noise functional parameter; e.g.
– Interface friction in paper stack, μp-p
– EM radiation in communications
? Specific source is not important
? Magnitude enables quick optimization
– Specs on noise are not important
– Worse noise in field is not important
? Don Clausing 1998
16.881
Fig. 21
Fig. 22
? Don Clausing 1998
16.881
INPUT
NOISE
N
IN
OUTPUT
NOISE
N
OUT
SYSTEM
NOISE
SOURCE
STRATEGY
? HOLD N
IN
CONSTANT
? MINIMIZE N
OUT
NOT IMPORTANT
? SPECIFIC SOURCE
? MAGNITUDE OF N
IN
Noise strategy
Example of noise strategy
? Paper feeder failiure modes:
? Sheet 1 arrives too soon: paper jam
? Sheet 1 arrives too late: misfeed
? Sheet 2 arrives too soon: multifeed
– Caused by large change in paper friction, ?μ
p-p
– Strategy for implementation during
improvement?
? Don Clausing 1998
16.881
Fig. 23
Introduce large ?μ
p-p
F
N
?μ
1-2
?μ
p-p
= ?μ
1-2
- ?μ
2-3
Sheet 2
?μ
p-p > 0
causes
F
N
?μ
2-3
sheet 2 to move
? Don Clausing 1998
16.881
Fig. 24
Noises for multifeeds
? Make ?μ
p-p
large, ≈0.1
? Will cause many multifeeds; enable quick
improvement
? Ignore
– Paper brands
– Customer usage, etc.
? Concentrate on creating paper stack with
large ?μ
p-p
? Don Clausing 1998
16.881
Fig. 25
Introduce product noises early
? Drive the performance away from ideal
? Do it early. Don't wait for the factory or
customers to introduce noises
? IPDT needs to develop the skill of
introducing these noises
? Management needs to design this into the
PD process and check that it is done to an
appropriate degree
? Don Clausing 1998
16.881
Fig. 26
Successful noise strategy
? Enables quick optimization
? Provides best performance inherent in
concept
– Even when future noise sources change
– Even when future noises are larger
– Even when spec changes
? Performance is as robust as possible
? Future improvements will require new
concept
? Don Clausing 1998
16.881
Fig. 27
Critical control-factor range
? IPDT judges best nominal value
? Then select larger value and smaller value –
feasible but significant changes
? Gives 3 values for each of 13 parameters
? 3
13
=1,594,323 candidate sets of values
? Must then choose trials to systematically
probe 1,594,323 candidates
? Don Clausing 1998
16.881
Fig. 28
Optimization
? Use designed experiments to select 27 out
of 1,594,323 control-factor options
? Subject each option to same set of noises
? Select option that gives best SN ratio
? Enter selected values for critical control
factors into critical parameter drawing;
become requirements for detailed design
? Don Clausing 1998
16.881
Fig. 29
Critical parameter drawing for paper feeder
WRAP ANGLE 45
o
BELT:
CONTACT:
ANGLE: 0
TENSION: 15 NEWTON
WIDTH: 50 MM
VELOCITY: 250 MM/SEC
DISTANCE: 12 MM
ANGLE: 45
RETARD:
PAPER
STACK
STACK FORCE:
VELOCITY: 300 MM/SEC
0.7 LB
GUIDE:
MOUTH OPENING: 7 MM
FRICTION: 1.0
RADIUS: 25 MM
FRICTION: 1.5
Fig. 30
16.881
Optimized values of critical parameters guide the detailed design
? Don Clausing 1998
Robustness
? Keeps the performance (response) of the
system acceptably close to the ideal
function
? Optimizes values of control factors to
minimize effect of noise factors
? Key to proactive improvement
? Don Clausing 1998
16.881
Fig. 31
Signal/noise ratio
? Measure of deviation from ideal
performance (noise here is N
out
)
? Based on ratio of deviation from straight
line divided by slope of straight line
? Many different types – depends on type of
performance characteristic
? Larger values of SN ratio represent more
robust performance
? Don Clausing 1998
16.881
Fig. 32
Manufacturing
? Machine-to-machine variation is one of
three types of noise that affects product
? Machine-to-machine variations can be
reduced by making production more robust
? N
out
for production is one type of N
in
for
product
? Don Clausing 1998
16.881
Fig. 33
Examples of manufacturing noise
? Temperature variations
? Humidity variations
? Cleanliness variations
? Material variations
? Machine-tool variations
? Cutting-tool variations
? Don Clausing 1998
16.881
Fig. 34
Noise strategy for production
? Simply operate in normal manner during
optimization of production robustness
? Don’t take special care
Would reduce magnitude of normal noises
? Assure that every trial is done in normal
manner – realistic noises are present and the
part variation is typical of actual production
? Don Clausing 1998
16.881
Fig. 35
Tolerance design
? Select economical precision
? Determines typical variation relative to
optimized nominal value
? Primary task is selection of production
process (or quality of purchased
component) -determines variation of
production
? Then put tolerance on drawing
? Don Clausing 1998
16.881
Fig. 36
Summary; control & noise basics
? Control factors are systematically varied
through large ranges seeking best combination
? Noise values are set at large values to enable
quick improvement
? Tolerance design selects variation range to be
used during production
Clarify these three variations in your mind;
you will be well on your way to Master
RD
? Don Clausing 1998
16.881
Fig. 37