1. Six Sigma Green Belt Project
Syed Salman Abbas
Perpendicularity Cylinder Bore vs. Base
Team:
Syed Salman Abbas, Bay Sirivong, Orele Laurenard, Peter
Zarkar, Pedro Fernandes.
3. Base flange
Bore
Bore centerline
0.12/100 A
Datum A
Maximum perpendicularity deviation allowed between Base
flange and Centerline of Bore is 0.12mm in a bore of length
100mm, with respect to Datum A.
Definition of problemDefinition of problem
DEFINE
Model 43ZD1
4. Problem StatementProblem Statement
Specific
(create statement from IS portion of Is/Is Not
from the bottom up)
Reduce scrap cost due to perpendicularity issue for the 4180âs 4-Mix cylinders from
Flange Machining at Lathe to Final Inspection at the Trumbull Plant of the small
engine components division of Mahle Engine Components USA, Inc.
Measurable (Document the historical
performance and goals)
As-Is
7725 ($89,000) in 2009 (for all cylinder models)
Sigma level: 2.439
Desired State
3860 ($45,000) in 2010 (for all cylinder
models)
Sigma level: 2.68
Achievable (Validate estimated project
duration and dates with a Gantt chart)
End by October 2010
Relevant (Does your project support any
initiatives? What are the potential savings?)
Supports company goal of reducing scrap to less than 1% of production.
Time Bound (Sponsor date; compare to
Gantt chart for any variances)
Project Problem Statement
Improve process from sigma of 2.439 to 2.68, resulting in reduced scrap cost due to perpendicularity issue for the
4180âs 4-Mix cylinders from Flange Machining at Lathe to Final Inspection at the Trumbull Plant of the small engine
components division of Mahle Engine Components USA, Inc. Project will be completed by 10/01/10. This project
supports the goal of reducing scrap to half with an estimated annual savings of $45,000.
DEFINE
5. Problem StatementProblem Statement
Project Problem Statement
Reduce scrap cost due to perpendicularity issue for the 4180âs 4-Mix cylinders from Flange Machining at Lathe to Final
Inspection at the Trumbull Plant of the small engine components division of Mahle Engine Components USA, Inc. Project
will be completed by 10/01/10. This project supports the goal of reducing scrap to half with an estimated annual savings of
$45,000.
Project Team
Core % Time Extended % Time
Salman Abbas 50% Pedro Fernandes 5%
Orele Laurenard 10% Ed Jones 5%
Bay Sirivong 10%
Andrei Aderca 10%
Project Timeline
Phase Est. End Date
Define 05/01/10
Measure 07/01/10
Analyze 08/01/10
Improve 09/01/10
Control 10/01/10
DEFINE
6. Stakeholders (Identify names for each functional area identified)
Suppliers Process Owners Customers
Raw materials
Steven Jock/ Logistics
Plant Manager:
ď§ Pedro Fernandes
Shift Supervisors:
ď§ Pedro Fernandes
ď§ Robert Augustine
ď§ Ron Traver
Support:
ď§ Orele Laurenard
Operators and Packers:
ď§
Next step
Stihl
Of data & information
Dipesh Jadav/ Maintenance
Marlene Martinez/ Quality
Downstream
Stihl dealer
Of human resources
Debbie Matsis/ HR
Consumer
Of financial resources
TJ Hicks/ Controller
Regulatory
Emission regulation
Key Metrics
Measurable Inputs, xs Process Metrics, xs and Ys, ys Measurable Outputs, Ys, ys
Casting dimensions Cp, No. of scrap parts, Cost of scrap parts, No rejections, No returns
Stakeholder AnalysisStakeholder Analysis
DEFINE
7. Force Field AnalysisForce Field Analysis
Positive forces (+) for
Project Success
Negative forces (-)
against Project Success
Action Plan to Overcome
Negatives (Roadblocks)
Management commitment.
High awareness.
Problem is not consistent.
This means that permanent
improvements can be made
with the current process.
More reliable measuring
system (Zeiss) than used
previously.
Hard to establish baseline for
measurements. Different
gaging methods give
different results.
Use only the measuring
method used by customer.
(Use only Zeiss CMM.)
Data in Daily sheet filled by
operators is not reliable/
accurate.
Don't use historical data
coming from op 10.
Problem is not consistent. It
comes and goes. Difficult to
gather data.
More emphasis to be given
to Control phase.
Fixturing at op 10, a critical
operation, is not adequate.
New fixturing is capital
intensive.
Make better use of available
resources.
DEFINE
8. Process Map â Model 43ZD1Process Map â Model 43ZD1
Flange machining
Lathe
Left
. A
. B
. C
. D
. E
Machine 4 Machine 5
A = Main bore
B = Mounting holes on flange, 4x
C = Top center hole
D = Rocker arm hole
E = Pulsation hole
Right
. A
. B
. C
. D
. E
Operation 10
Left
. A
. B
. C
. D
. E
Right
. A
. B
. C
. D
. E
Pallet 2
Left
. A
. B
. C
. D
. E
Right
. A
. B
. C
. D
. E
Left
. A
. B
. C
. D
. E
Right
. A
. B
. C
. D
. E
A
Pallet 1
Operation 10
Pallet 2Pallet 1
. Turn flange
. Deburr sharp
edges
Deburr Station
Deburring
Wash Machine
. Wash chips
Washing
Brushing Machine
. Brushing
Brushing
Plating Area
. Plating Nikasil
in bore
Plating
DEFINE
9. Process Map â Model 43ZD1Process Map â Model 43ZD1
. Install nipple
. Install valve seats
. Install valve
guides
Press Station
Component
Assembly
Honing
. Hone bore dia to
final specification
Sunnen 3&4
Final Inspection
. Perpendicularity
. Bore diameter
and roundness
. Camhole size
. Thread size
. Visual
. others
Inspection Station
Machine 2 Machine 6
A = Face cam
B = Face spark plug
C = Face left boss
D = Drill and tap spark plug
E = Drill rough cam holes
F = Drill finish cam holes
G = Finish oil seal boring bar
H = Finish bearing journal
boring bar
I = Rough valve seats and valve
guides
Left
. A
. B
. C
. D
. E
. F
. G
. H
. I
Machine 18
. Valve guide dia
. Finish valve seat
Operation 30
Washing
. Wash chips
Wash Machine
Right
. A
. B
. C
. D
. E
. F
. G
. H
. I
Left
. A
. B
. C
. D
. E
. F
. G
. H
. I
Right
. A
. B
. C
. D
. E
. F
. G
. H
. I
Left
. A
. B
. C
. D
. E
. F
. G
. H
. I
Right
. A
. B
. C
. D
. E
. F
. G
. H
. I
Left
. A
. B
. C
. D
. E
. F
. G
. H
. I
Right
. A
. B
. C
. D
. E
. F
. G
. H
. I
A
Operation 20
Pallet 2Pallet 1
Operation 20
Pallet 2Pallet 1
DEFINE
10. Quick WinsQuick Wins
⢠Increase operator efficiency
Operators are required to check perpendicularity every 1/2 hour. But the process is stable
enough that such frequent inspection is not required.
Reduce inspection frequency for perpendicularity from 1/2 hour to 1 hour.
⢠Implement lessons learnt from 50zd5 clamp study
It was proven by experiments on model 50zd5 that changing clamp from toggle-type to
screw-type reduces the process variation.
Replace toggle clamps with screw clamps. See next slide for data.
⢠Modify gaging method
Current gaging system used at op10 and final-inspection does not correlate with
CMM measurements.
Use different gaging method. This system is also easier to read for operators.
DEFINE
11. Quick WinsQuick Wins
⢠Verify parts rejected at final inspection
Gage at final inspection does not correlate perfectly with customer's measurement system
(using Zeiss CMM).
Recheck parts rejected by final inspection using Zeiss CMM.
DEFINE
12. ⢠Null hypothesis: no difference in variation between
toggle- and screw- clamps.
⢠Result: p=0.001. p<0.05 for f-test. Thus null hypothesis
is rejected.
⢠Conclusion: Less variation in data when screw clamp is
used.
Test for Equal Variance
Quick Wins (Lesson learnt)Quick Wins (Lesson learnt)
Model 50ZD5
Toggle vs Screw clamp
DEFINE
2-Sample t-test
⢠Null hypothesis: no difference in mean between
toggle- and screw- clamps.
⢠Result: p=0.146. p>0.05 for t-test. Thus null hypothesis
is not rejected.
⢠Conclusion: No change in sample mean.
13. MSAMSA
GR&R for Zeiss CMMGR&R for Zeiss CMM
MEASURE
CONCLUSION:
âş Number of distinct categories is
greater than 6. %StudyVar of Total
GR&R is less than 10%.
GR&R is acceptable.
Study Var %Study Var
Source StdDev (SD) (6 * SD) (%SV)
Total Gage R&R 0.0014947 0.008968 2.53
Repeatability 0.0010410 0.006246 1.76
Reproducibility 0.0010726 0.006435 1.81
Oper 0.0000000 0.000000 0.00
Oper*Part 0.0010726 0.006435 1.81
Part-To-Part 0.0591092 0.354655 99.97
Total Variation 0.0591281 0.354769 100.00
Number of Distinct Categories = 55
14. Data Management PlanData Management Plan
MEASURE
1. Data Description
Metric
Operational Definition (Verbal) or
Formula (Symbols)
Family of
Measure
Data Type
Perpendicularity
dimension
Process capability Cp/Cpk per print Q Continuous
Scrap (units) Non-conforming parts P Ordinal
15. Data Management PlanData Management Plan
MEASURE
2. Data Collection and Validation
Data Source or
Location
Collector Sampling Plan Stratification Plan MSA Plan
Operation 10 Quality Engineer 4 parts sample
by part location, shift,
and casting lot
GR&R
SAP data Quality Engineer Weekly data
16. Data Management PlanData Management Plan
MEASURE
3. Graphical Display and Project Validation
of Central Tendency
and Variation
over Time Main Desirability Validate Project
Histogram X and R Variability under control
t-test
F-test
Defects chart p chart Decrease mean
20. CorrelationsCorrelations
ANALYZE
⢠Selected 28 parts rejected by final inspection. Also
selected some parts awaiting inspection. So there
were mixed good/bad parts.
⢠Measured for perpendicularity, flatness, bore angle,
roundness, and concentricity.
21. Project Definition TreeProject Definition Tree
ANALYZE
Perpendicularity of
main bore
PartMeasurement
Rationale
FlangeBore
Roundness
Bore angle
(X, Y)
OthersConcentricity
⢠GR&R is acceptable. (see slide
12).
⢠No contrast between flange
flatness of good and bad parts.
(see next slide).
⢠Very good correlation
between Bore Angle and
Perpendicularity readings. No
correlation for others. (see
next slide)
bore
flange
Continue on slide 22âŚ
Perpendicularity problem results when main bore of cylinder is machined
at some angle.
23. Project Definition TreeProject Definition Tree
ANALYZE
Bore Angle
Rationale
At Honing
(After Plating)
At Operation 10
(Before plating)
Flatness of fixture
plate
Spring strength
(used in half
moon)
Part locators
in fixture
Part to part
variation in
castings
Clamp torque
Angle of pallet
⢠Correlation between Bore
Angle and perp remains the
same before and after plating.
Play in clamp
screw
Continued from Slide 20
Perpendicularity problem
occurs at Operation 10.
24. CorrelationsCorrelations
ANALYZE
⢠Selected 56 parts from Operation 10.
⢠Measured for perpendicularity, flatness, bore angle,
roundness, and concentricity.
⢠Ran these 56 parts through all operations.
⢠Measured for same dimensions again.
25. P=0 rejects the null
hypothesis. This means that
there is difference between
op 10 and honing, and that
honing can contribute in
perpendicularity.
CorrelationsCorrelations
(Change in perp between Op 10 and Honing)(Change in perp between Op 10 and Honing)
ANALYZE
DiffBwOp10andHoning.mtb
Paired T for before - after
N Mean StDev SE Mean
before 56 0.07995 0.04336 0.00579
after 56 0.09734 0.04103 0.00548
Difference 56 -0.01739 0.01130 0.00151
95% CI for mean difference: (-0.02042, -0.01436)
T-Test of mean difference = 0 (vs not = 0): T-Value = -11.52 P-Value = 0.000
Average change in
perpendicularity reading
between op 10 and
honing is 0.017 mm.
26. CorrelationsCorrelations
(Change in perp between Op 10 and Honing)(Change in perp between Op 10 and Honing)
ANALYZE
âş Honing contributes to perp problem by making the part go worse 0.017 mm on average.
âş This is not a significant change keeping in view the tolerance range of 0 to 0.12 mm.
Perpendicularity problem starts from Op 10.
Honing does not contribute significantly to perpendicularity problem.
DiffBwOp10andHoning.mtb
27. Project Definition Tree âProject Definition Tree â
ResultResult
ANALYZE
⢠Root cause of perpendicularity problem is machining
the main bore at an angle at Op 10.
⢠Operation 10 needs to be analyzed more closely.
32. ⢠Process capability analysis was run on the 17 data
points of 2L5. Result shows capable process.
Register to register variationRegister to register variation
(Process capability of 2L5 on the same data)(Process capability of 2L5 on the same data)
ANALYZE
Cpk=2.20
33. Register to register variationRegister to register variation
(Process capability of 2L5 on new set of data)(Process capability of 2L5 on new set of data)
ANALYZE
â˘
⢠Full blown process capability also shows process to be capable.
Cp = 2.23
Cpk = 1.84
Pp = 2.38
Ppk = 1.88
35. ⢠Plotting same data on IMR chart also shows out-of-control points.
Defect Trend Analysis - 2009Defect Trend Analysis - 2009
ANALYZE
IMR - 40ZD38 IMR â 43ZD1
IMR â 50ZD5
36. Analyzing Operation 10 âAnalyzing Operation 10 â
ConclusionConclusion
ANALYZE
⢠The process variation is acceptable.
⢠Defect spikes can happen anytime. This shows
the lack of control at the time of setup.
⢠There are several factors during process setup
which can effect this issue.
⢠These factors need to be studied individually,
and proper controls need to be placed.
38. Project Definition TreeProject Definition Tree
ANALYZE
Bore Angle
Rationale
At Honing
(After Plating)
At Operation 10
(Before plating)
Cleanliness of
fixture plate
Spring strength
(used in half
moon)
Clamp torqueIndexer and BrakesIndexer and Brakes
⢠Correlation between
Bore Angle and perp
remains the same before
and after plating. (see
slide 13)
Play in clamp
screw
Slide AA
Continued from Slide 20
39. Photos of Op 10 fixturePhotos of Op 10 fixture
IMPROVE/ CONTROL
Clamp screw
Pallet
âHalf moonsâ
Locator
Spring under
âhalf-moonâ
40. Indexer and Brakes (Machine)Indexer and Brakes (Machine)
IMPROVE/ CONTROL
Solution:
Probing system to be installed at op 10. This will ensure that
the pallet angle is good.
Problem: The rotation of pallet is not repeatable. The pallet
sometimes does not stop at the precise location.
Control:
Frequency of measurement to be 1x a shift.
Control plan to be revised.
FMEA to be revised.
41. Play-in clamp screwPlay-in clamp screw
IMPROVE/ CONTROL
Solution:
Short term: Implement âLayered Auditâ form at Operation 10.
Long term: Use heat treated clamp plate, clamp threads,
and pins. This will result in less wear and long life of the
fixture.
Problem:
Clamp screw is made of soft material. It therefore wears out
quickly. A worn clamp screw will cause play in screw, and
thus part wonât be able to clamped securely.
Control:
Documents related to Layered Audit.
42. Cleanliness of fixture plateCleanliness of fixture plate
IMPROVE/ CONTROL
Problem:
Some chips left on the steel pallet get fused in aluminum. This
results in part sitting crooked on pallet, causing perpendicularity
issue.
Solution:
Implement âLayered Auditâ form at Operation 10.
Control:
Documents related to Layered Audit.
43. Effect of Clamping torqueEffect of Clamping torque
IMPROVE/ CONTROL
Perpendicularity value decreases as torque increases.
Further study needed for range of torque values as marked
above.
Clamp nut
R2
= 93.0%
REGRESSION TORQUE.mtb
44. Effect of Clamping torqueEffect of Clamping torque
IMPROVE/ CONTROL
Not much change in R2
value when quadratic and cubic
equations are used to fit the regression line.
R2
= 95.9%
Using quadratic
equation
Using cubic
equation
R2
= 97.4%
45. 0
0.02
0.04
0.06
0.08
17 19.1 21.2 23.3
Perpendicularity(mm)
Torque (N-m)
Part 1 Part 2
Effect of Clamping torqueEffect of Clamping torque
IMPROVE/ CONTROL
Regression shows that
there is no correlation.
This means that in the
given range of torque,
perpendicularity does not
change as torque is
increased.
R2
= 21.1%
46. Effect of Clamping torqueEffect of Clamping torque
IMPROVE/ CONTROL
Conclusion:
⢠Perpendicularity improves as torque increases.
⢠Perpendicularity does not improve after torque goes somewhere
above16 N-m.
Solution:
Use torque wrench pre-set at 18 N-m.
Control:
Torque wrench has been ordered.
Control plan to be revised.
47. Effect of spring inEffect of spring in âhalf-moonââhalf-moonâ
IMPROVE/ CONTROL
Spring in
âhalf-moonâ
Half-moon
⢠Effect of spring on perpendicularity is not
known.
⢠At the time of setup, spring is selected
randomly from a box containing different
types of spring.
48. Effect of spring inEffect of spring in âhalf-moonââhalf-moonâ
IMPROVE/ CONTROL
Spring in
âhalf-moonâ
Half-moon
⢠Selected three random springs from the box.
⢠Calculated spring rate âkâ for each spring.
k = (applied force) / (displacement)
spring 1: k = 38.38 lb/in.
spring 2: k = 28.81 lb/in.
spring 3: k = 7.62 lb/in.
⢠Ran parts with each spring in fixture.
49. Effect of spring inEffect of spring in âhalf-moonââhalf-moonâ
IMPROVE/ CONTROL
Pearson coefficient (r)
of Perp and k (lb/in.) = 0.823
⢠High value of ârâ suggests that correlation exists between
spring âkâ and perpendicularity. Perpendicularity improves
for low values of âkâ.
⢠Operator prefers lighter spring, because such a spring
does not require the part to be pressed down hard.
⢠Therefore spring with low âkâ value is desirable.
50. Effect of spring inEffect of spring in âhalf-moonââhalf-moonâ
IMPROVE/ CONTROL
Problem:
⢠Spring force can have effect on perpendicularity, but spring is
selected randomly at the time of setup.
Solution:
⢠Standardize the spring to be used in fixture. Use only the
standard spring.
⢠A spring with low âkâ value is selected (see last slide).
⢠25 parts are run with this spring, and variability analyzed.
Control:
Establish SAP number for spring for purchasing.
51. Problem:
Parts are checked once-per-shift on Global CMM. But
the results of perpendicularity readings are not reliable.
Solution:
Start measuring parts on Zeiss CMM (the same method as
used by the customer).
Daily part inspectionDaily part inspection
IMPROVE/ CONTROL
Control:
CMM program.
53. ComparisonComparison
IMPROVE/ CONTROL
2009 GOAL 2010
2011
(on
04/20/2011)
Sigma
Level 2.439 2.68 2.671 2.782
Number of
rejected
parts
7725 3860 4353 659
Cost of
scrap $89,000 $45,000 $53,000 $10,240
⢠Achieved the six sigma goal.
⢠The sigma level is expected to go up further after
implementation of suggested changes.
56. ⢠Perpendicularity problem originates from op 10.
⢠Lack of adequate measuring system compounded the loss. The
measuring system was modified during the course of the project.
⢠The process capability is acceptable.
⢠The root cause of problem is part not sitting flush against the
fixture at op 10.
⢠Several reasons were identified, and controls were suggested
for each of the reason.
⢠Similar controls will be put in place for other models as well.
Summary of projectSummary of project
IMPROVE/ CONTROL