This document discusses how John Deere used Six Sigma and Design for Six Sigma (DFSS) principles to optimize the performance and reliability of ground speed sensors on its tractors. It describes modeling the sensor system to better understand sources of error and improve design. It also details how vibration testing and the physics of failure approach helped identify an intermittent failure mechanism in some radar sensors caused by loose connections, and how implementing a spring clip solved the problem. The document shows how these quality improvement methods enhanced system design and understanding.
Automating the Point-to-Point Testing of Hundreds of Substations
Using Six Sigma to Optimize Performance and Reliability
1. 1
Operational Excellence
Using Six Sigma to Optimize
Performance and Reliability
Tim Williams
Deere & Co. World Headquarters
Liang Jiang
John Deere Waterloo Works
Operational Excellence
Setting the stage…
• John Deere
– Tradition of Excellence
– Customer Expectations
• Design for Six Sigma
– Addresses the needs of the customer
– Improves the way we do business
– Delivers better (optimized) results
2. 2
Operational Excellence
Adopting a Six Sigma Approach Means…
Reactive
Design Quality
Predictive
Design Quality
DFSS
From
• Evolving design requirements
• Extensive design rework and
rework and rework…
• Performance assessed by
“build and test”
• Performance and assembly
problems fixed after product
is released for production
• Quality “tested in”
To
• Disciplined identification and
flow-down of customer CTQs
• Optimize and verify design
performance using transfer
functions, modeling and
simulation
• Designed up front for robust
performance and assembly
• Quality “designed in”
Operational Excellence
The role of simulation in DFSS
• Predict performance and reliability
• Speed up design cycle
• Reduce reliance on expensive field
testing
• Unearth new knowledge
4. 4
Operational Excellence
Situation…
• Ground Speed on John Deere tractors
(most) is measured with RADAR
• Need to understand why we don’t get
the performance we expect
• Important enough that we are willing to
pay more for a better sensor (i.e., GPS)
– RADAR 2% GPS 1.1%
– But at an increase of about 4 times per
piece cost!
Operational Excellence
CTQ Flow Down
5. 5
Operational Excellence
CTQ Consolidated CTQ
Consolidated
All Displays Indicate
the Same Tractor
Ground Speed
System Delay
Must Be
Acceptably Low
System
Accuracy
Fault Tolerance
Reliability for
Planter-specific
Applications
Calibration
Signal delay /
response time
Update rate
Different speed
sources
Different
information
processing
Resolution of
display
Minimal lag in
shutting down
planter when
tractor stops
Minimal lag at
tractor start up
Ground speed
signal lags
Motion sensor
(planter) approx.
1 sec delay
Quick Start will
control rate
when ground
speed signal is
less than 2 kph
Ground speed
signal lags
when ground
speed signal is
greater than 2
kph
±2% or less
(per planter -
seeds/acre)
System tolerant
of sensor drop-
out
Appropriate
cross-checking
logic for
different speed
sources
Software checks
for realistic
tractor speed
change
Missing ground
speed source
(radar or bus
message) data
Wheel speed
sensor drop-
outs properly
managed
through fault
logic
A single drop-
out shorter than
0.25 seconds is
OK
No drop-outs
longer than 0.5
seconds (total)
Two
consecutive
drop-outs are
unacceptable
(0.75 seconds)
Software must
provide correct
calibration when
the procedure is
carried out
correctly
No mis-leading
calibration error
indications
Operator
manual must
have accurate
descriptions
Operator
manuals should
be consistent
across
platforms
Operational Excellence
Understanding and
Modeling the System
6. 6
Operational Excellence
Profiling the Radar Performance
(based on recorded data)
0
2
4
6
8
10
12
7.11
7.16
7.21
7.25
7.3
7.35
7.39
7.44
7.49
7.54
7.58
7.63
7.68
7.72
7.77
7.82
7.86
7.91
7.96
8.01
8.05
8.1
8.15
8.19
8.24
8.29
8.33
8.38
8.43
8.48
8.52
8.57
8.62
8.66
8.71
8.76
8.8
8.85
8.9
8.95
8.99
9.04
9.09
9.13
Collection Time
Speed
At Steady State Speed of 10.0 kph
Output Values: 9.9 kph min. / 10.1 kph max
Measured accuracy of about 1.1%
Operational Excellence
Monte Carlo Simulation
System CharacteristicsSystem Characteristics
Part CharacteristicsPart Characteristics
7. 7
Operational Excellence
True Ground
Speed
Radar Error
(2%)
CCU Error
(0.1%)
TECU Error
(0%)
True Ground
Speed
Radar Error
(2%)
CCU Error
(0.1%)
TECU Error
(0%)
First Input Signal Second Input Signal
Change in Tractor Speed
Between Updates
Ground Speed per
GSD
Ground Speed per
GSD
SystemTimeDelay
Tractor Ground Speed System
Operational Excellence
Results of Statistical Modeling
(Constant Speed of approx. 8 kph)
Mean System Time Delay 372 ms (Range 117 - 633 ms)
System Error ± 4%
8. 8
Operational Excellence
Seed Spacing –
6 inch spacing at 8 kph
Seed Spacing –
6 inch spacing at
maximum specified
acceleration
(4.5kph/s)
Effects on Seeding Application
Operational Excellence
What if the ground speed sensor had
No Error?
The System Performance Would Not Be Any Better!
9. 9
Operational Excellence
So What Did We Learn?
Operational Excellence
The Decision…
Stay with RADAR and not pursue GPS
• Keep the cost of the system lower for the
customer
• Avoid risk of new product/technology
introduction
• Savings in test and verification
• Better understanding of system performance
influencers
How to Improve the System
10. 10
Operational Excellence
Another leg of the DFSS journey…
Operational Excellence
Define The Problem
• Ground Speed Sensor (Radar)
generates erratic ground speed values
• Field returned units were ‘HALT’ed with
combined vibration and thermal
cycling; not able to duplicate this
intermittent failure mode
11. 11
Operational Excellence
Develop Measurement System for
Suspected Failure Mechanism
Analog
Circuit
DSP Circuit
IF1, IF2 DiodesTransceiver
Failure detection
Capability !
Operational Excellence
Hypothesize Failure Mechanism
Diode Pin Force
-10
0
10
20
30
40
50
60
70
0.000 0.002 0.004 0.006 0.008 0.010
DisplacementDifferential(inches)
EccsorbForce(lbs)
101481
108749
101436
109644
unitB1
unitB2
unitB3
unitB4
unitB5
S
S
S
Suspected Failure Mechanism:
Bent cover plate reduces the normal force on the IF1 and IF2 diodes
causing degradation of the mechanical / electrical connections
12. 12
Operational Excellence
Analyze – Different Testing Techniques
• HALT
– Stimulate
– Most energy at high
frequency
– Can’t correlate to real
environment
• Failure Detection
– Monitor output after DSP
– Sample 10% of data
• Stresses Used
– Vibration
– Thermal Cycling
• Potted Units No
Precondition
• ED Shaker
– Simulate
– Representative profile
– Can correlate to real
environment
• Failure Detection
– Monitor output right after
two diodes and after DSP
– Examine 100% of data
• Stress Used
– Vibration only
– Step test
• Un-potted Units First with
Different Size of Shims
Operational Excellence
Physics of Failure
13. 13
Operational Excellence
Improvement
Operational Excellence
Control
• Intermittent connection shows up as
dropouts or unstable speed output
• Failure mechanism is not time-
dependent
• Vibration test (10X) conducted on
potted-unit with spring clip
• The improved part is in production
• Gained knowledge used on new radar
design