Beruflich Dokumente
Kultur Dokumente
VARIATION: THE
MAIN ENEMY
Breaking Through to the Next Level of Customer
Expectation
•Conclusions
2 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Following
July Scrap Projects Customer Focus
Focusing on the customer is essential to the success of
any business. We all recognize this idea. But do we really
internalize this idea? Do we really believe that such a focus has
the potential to drive business growth and improve the quality of
our day-to-day work as well as improve the chance that our
business, and hence our jobs, will flourish?
Business growth depends on how well we meet customer
expectations in terms of quality, price, and delivery. In turn, our
ability to satisfy these needs with certainty is controlled by process
capability and the amount of variation in our processes. Then
variation has a direct impact on our business results in terms of
cost, cycle time, and the amount of defects that affect customer
expectations.
L S L U S L L S L U S L
Centered Reduced
Process Spread
Target
Defects
LSL USL
Reduce
Reduce Variation
Variation && Center
Center the
the Process
Process —
—
Customers
Customersfeel
feelthe
thevariation
variationmore
morethan
thanthe
themean.
mean.
h(t) constant
Manufacturing Wearout
Defects Failures
Time
Random
10 Johnson Controls, Inc. © Sep 2006 R. Failures
Arévalo C. SS/BB
Following
July Scrap Projects
¾Outside-In: Focus
¾Outside-In: Focus on
on meeting
meeting customer’s
customer’s CTQ’s
CTQ’s
¾Insight through
¾Insight through variance:
variance: Statistical
Statistical design
design to
to reduce
reduce
performance
performance variability
variability
12 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Following
July Scrap Projects An Opportunity to Break Through “the Wall”
DFSS is
Process Improvements
6
Plus DFSS essential
5 DFSS Benefit to exceed
4
Z (σ) customer
3
Process Improvements
expectations
2
Only
1
0
Time
DFSS
DFSS Gets
Gets Us
Us Beyond
Beyond Improving
Improving Marginal
Marginal Designs
Designs // Processes
Processes
13 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Design Opportunity Assessment
Following
July Scrap Projects
$
Maximize LEVERAGE in the Quality
Efforts here!
Biggest
Biggest Opportunity
Opportunity for
for Impact
Impact is
is Design
Design Stage!
Stage!
14 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Following
July Scrap Projects
Moving From Reactive to Proactive
Reactive Predictive
Design Transition Design
Quality to ... Quality
FROM TO
Evolving Requirements CTQ Flowdown from Customer
Expectations
Design rework / tweaking Control Critical Design Parameters
Design Changes
Proactive Company Reactive Company
(6 σ Company Experience) (Typical Experience)
Start Production
Project Time Line
• Engineering Knowledge
• Market Analysis
• Physics
• Models & Simulations
• Empirical DoE to Generate Regression
Equation
• DoE on Computer Model to Generate
Equation
20 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Transfer Functions
Following
July Scrap Projects
X1 Y1
Factors CTQ’s
X2 Transfer Functions Y2 (Dependent
(the
Independent Yi = fi (X1, X2, … Xm) Variables)
Variables)
XM YM
FanModulePareto
Robust Design
Operability
FODTolerance
Legend
Sf press_to_fill time_cycle cost_total Design for
Reliability
0.534 80 17.5 2.22
X1 0.532 75
2.2
press_to_fill
time_cycle
17
cost_total
0.53 70
Clearance, %Diameter 2.18
Sf
246.1 0.528 65
16.5
0.526 60 2.16
Durability 260
0.524 55 16 2.14
273.9 245 250 255 260 265 270 275 245 250 255 260 265 270 275 245 250 255 260 265 270 275 245 250 255 260 265 270 275
GasPathLeakage
0.54 74 18.5 2.2
Unbalance X2
0.535
0.53
72 18
2.19
press_to_fill
time_cycle
70 17.5
cost_total
0.525
68 17 2.18
Sf
Airflow 60
0.52
0.515
66 16.5
2.17
0.51 64 16
71.1
Maintainability 82.2
0.505
60 65 70 75 80 85
62
60 65 70 75 80 85
15.5
60 65 70 75 80 85
2.16
60 65 70 75 80 85
dTemp_Mold dTemp_Mold dTemp_Mold dTemp_Mold
DovetailWear&Cracking
Efficiency 0.534 90 22 2.4
X3 0.532 20
Speed 80 2.3
press_to_fill
time_cycle
cost_total
0.53 18
70 2.2
Sf
1 0.528 16
RotorThrust 5.5
0.526
60
14
2.1
0.524 50 12 2
LifeLimitedComponentPart Life 10 0 2 4
Time_Inj
6 8 10 0 2 4 6
Time_Inj
8 10 0 2 4 6
Time_Inj
8 10 0 2 4 6
Time_Inj
8 10
LRURemovalTime
ModuleIntegration 0.6 68.5 17 2.178
2.176
X4 68 16.9
press_to_fill
time_cycle
0.55 2.174
cost_total
35 0.5 2.17
67 16.7
2.168
42.5
A1
X2 Transfer Function
A1
Y
Y = f (X1, X2, … Xm)
5.00 6.84 8.67 10.51 12.34
A1
XM
0.70 0.85 1.00 1.15 1.30
Predicted
Factor CTQ
Probability Probability
Distributions Distribution
Linear Non-linear
Sensitivity coefficient is a
Part or process step Targeted Mean value Std. Dev. - initially based on tolerance relative weighting of the
factor description field initial setting Data collection divided by 6. Input this directly influence of this factor on DPU is the probability that parts will be
enter the X's here period after measuring the factor. changing the response produced outside the tolerance range.
Analysis Table
Variable Information Tolerance Dist. Type % Cont.
Factor Short or Lower Upper % or Normal or to σ Sensitivity
Description Mean Long Term Std Dev Limit Limit Actual Uniform Response Coef. DPU ZST
Block 1 4 s 0.6667 2 2 A N 35.00% -1 2.1E-02 3.53
Block 2 2 s 0.5774 1 1 A U 25.00% -1 0.0E+00 N/A
Block 3 4 s 0.0533 4 4 % N 40.00% -1 2.1E-02 3.53
Envelope 13 s
Left buton will execute macro to calculate sensitivity
coeficients. Right button will clear current
Tolerance - the Actual (A) number of units above and below mean
sensitivities. Always clear previous sensitivity values Two options; N (normal) or U (uniform).
that the 'X' is allowed to vary. Enter all positive numbers or an
before calculating new ones On parts with a uniform distribution this
allowed percentage (%) of the mean value
probability is shown to be zero, with
Mean Response→ 3 truncation outside the tolerance range.
Response Upper Spec Limit 6
Response Lower Spec Limit 0 Push the appropriate buttton to hide or unhide all rows. Percent contribution that the
There are 20 total available for X's. individual component (X) has on
Enter the equation that Summary Table the Standard Deviation for the
defines how the X's are Response Components entire system.
related to the response Worst Case
Mean Std Dev DPU ZST Lower Limit Upper Limit DPU ZST
3 1.149 0.009 3.866 -0.160 6.160 4.20E-02 3.81 Rev 2.0a
Crystal Ball
OptQuest
Perform optimization over multiple Monte Carlo
simulations
Genetic Algorithm based optimizer
Crystal Ball Developer’s Kit
Write advanced macros
Build custom reports
Automate post-simulation analysis
Crystal Ball Extenders
Tornado charts, Correlation matrices, Decision
tables, ...
Before performance
Baseline Process Sigma Level (Short-term) = 4
Baseline PPM (Long-term) = 7,298
45
40
35
High
Response 30
25 Low
20 Upper +DL
15 Upper -DL
10 Low er +DL
5 Low er -DL
0 Reassembly
Reassembly
UpperCover
Extrusion
Cushion
Results
Elbow
LowerCover
Initial
#1
#2
Step
Before performance
Baseline Process Sigma Level (Short-term) = 2.47
Baseline PPM (Long-term) = 166,314
31 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Tools Used
Following
July Scrap Projects
Normal Probability Plot of the Standardized Effects
(response is BKT THK, Alpha = .05)
99
Effect Ty pe
E
Not Significant
95 AE Significant
90 F actor N ame
A M elt Time
80 B Bladder P ressure
C Bladder D elay
70 D S eal Time
Percent
60 E C ore S ide
50
40
30
20
10
5 BDE
A
1
-5.0 -2.5 0.0 2.5 5.0 7.5
Standardized Effect
The Differences
Results and performance (After supplier controlling)
Baseline Process Sigma Level (Short-term) =
On going
Baseline PPM (Long-term) = On going
32 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Ford S197 D-Ring misalignment
Following
July Scrap Projects
Problem Statement.
During the end of April´2006 the AAI-Ford customer reported D-ring
misaligned during the visor vs. car marriage, resulting as a campaign by a 3rd. Party
during 3 months after.
Before performance
Baseline Process Sigma Level (Short-term) = 4.0
Baseline PPM (Long-term) = 4,173
Total Envelopment
aprox.= 17.10 mm
“Z”
Poin
t
Ga
“A p
First Core
Attachme
3.0 mm
” nt
Y1
D-Ring
posts Second
Limiting Core
Ga
up Attachme
p 0.60 mm
position nt4
Y2
Before performance
Baseline Process Sigma Level (Short-term) = 1.58
Baseline PPM (Long-term) = 467,062
R eas s em bly #2
Initial R es ults
Ex trus ion/Spring
Inner Spring
0
Elbow as s y
As s y
BoB=0.6°
The
The significant
significant component
component isis the
the Elbow
Elbow TUBE,
TUBE, itit makes
makes
the
the difference
difference between
between WoW
WoW and and BoB
BoB (( Noisy/
Noisy/ Not
Not noisy)
noisy)
.013 132
.009 88
.004 44
Max: 12.65
Min Inner Max:
Gap=13.21
0.20 0.35 0.50 0.65 0.80
13.21-13.15=0.06 mm Gap mm
Before performance
Baseline Process Sigma Level (Short-term) = 3.12
Baseline PPM (Long-term) = 898
37 Johnson Controls, Inc. © Sep 2006 R.
Arévalo C. SS/BB
Tools Used
Following
July Scrap Projects
Pareto Chart of the Standardized Effects
(response is Average Thk, Alpha = .05)
S c a tt e rp lo t o f W o b b le v s A v era g e T h k
1 .4
2.002
Factor Name
A Melt Time
1 .2 BC B General Pressure
C Blader Delay
D Blader Pressure
1 .0 CD
0 .8
Wobble
Term
0 .6
A
0 .4
C
0 .2
0 .0 B
1 5 .0 0 1 5 .2 5 1 5 .5 0 1 5 .7 5 1 6 .0 0 1 6 .2 5 1 6 .5 0 1 6 .7 5
A v e r a g e Th k 0.0 0.5 1.0 1.5 2.0 2.5
Standardized Effect
NO OK
Before performance
Baseline Process Sigma Level (Short-term) = 2.48
Baseline PPM (Long-term) = 6411
Add pockets or
Harness Routing windows and
Change routing
trapped below harness change cut
method
clip and clip troubles fabric
interference
Results and performance (After supplier controlling)
Baseline Process Sigma Level (Short-term) =
Proposals under validation in Holland
Baseline PPM (Long-term) = T.B. Validated
Failure Mode:
Bezel Unseated
Effect: Cached
Before performance Extender
Baseline Process Sigma Level (Short-term) = 2.50
Baseline PPM (Long-term) = 6331
A n a lys is T a b le
V a ria b le In fo rm a tio n T o le ra n c e D is t. T y p e % C o n t.
F a c to r S h o rt o r % or N o rm a l o r to σ S e n s itiv ity
D e s c rip tio n M ean L o n g T e rm S td D e v Lower U pper A c tu a l U n ifo rm R esp on se C o e f. DPU Z LT
B e ze l 3 .6 2 S 0 .0 3 8 0 0 .1 0 .1 A N 4 2 .4 2 % 1 4 .2 9 4 E -0 2 1 .7 2
C o re 3 .1 3 S 0 .0 4 2 0 0 .1 1 0 .1 1 A N 5 1 .8 2 % -1 4 .3 9 4 E -0 2 1 .7 1
F a b ric 0 .4 S 0 .0 1 4 0 0 .0 1 0 .0 1 A N 5 .7 6 % -1 5 .8 2 7 E -0 1 -0 .2 1
M e an R e sp on se→ 0 .0 9
R e s p o n s e U p p e r S p e c L im it
Failure Mode:
R e s p o n s e L o w e r S p e c L im it 0
S u m m a ry T a b le
Response C o m p o n e n ts
W o rs t C a s e L im its
Bezel
M ean S td D e v DPU Z LT Low er Upper DPU Z LT
0 .0 9 0 .0 7 5 8 4 7 1 .1 7 7 E -0 1 1 .1 9 -0 .1 3 0 .3 1 6 .6 9 6 E -0 1 0 .7 6 R e v: G E A 9 /1/9 8 B .M allory
D r. M a u ric e L . B e rrym a n , 1 9 9 6 . A ll rig h ts re s e rv ed .
M ean R esponse→ 0 .4 7
.009 89 R e s p o n s e U p p e r S p e c L im it
R e s p o n s e L o w e r S p e c L im it 0
S u m m a r y T a b le
R esponse C o m p o n e n ts
W o r s t C a s e L im its
.004 44.5 M ean S td D e v D PU Z LT Low er U pper D PU Z LT
0 .4 7 0 .0 7 5 8 4 7 3 .9 1 1 E -1 0 6 .1 5 0 .2 5 0 .6 9 6 .6 9 6 E -0 1 0 .7 6 R e v : G E A 9 / 1 / 9 8 B . M a llo r y
D r . M a u r ic e L . B e r r y m a n , 1 9 9 6 . A ll r ig h t s r e s e r v e d .
.000 0
Forecast: Gap(Modified)
10,000 Trials Frequency Chart 0 Outliers
.023 233
.017 174.7
.012 116.5
.006 58.25
.000 0
0.25 0.36 0.48 0.59 0.70
mm
Here in, we had seen an over view about the effort we can do and we are making focus
in reach and exceed the customer expectations, not so far from other greatest world
wide companies and in the commodity of our facilities to make quickly response.
As well as we are using Six Sigma to improve our current processes, we can use
another systematic methodology that enables us to Design products and processes that
meet customer expectations (CTQ´s) to a 6s level before SOP by means of using
statistical methods to predict and improve quality and performance before building
prototypes whose can be validated using pilot runs, pre-production and production units
Although each of the DFSS tools can be applied to the design process in and of itself the
true '"global"' benefit can only be realized through an integrated application across the
board..
Simulation & DOE can be effectively used to identify leverage variables, establish
sensitivities of the response variable, and define tolerances through the deliberate
arrangement of competing settings among one or more independent variables. The goal
here is to provide the most efficient and economical means for arriving at valid and
relevant conclusions – resulting in reduction of variation at minimal cost.