Statistics & Probability in Mechanical Design
Jason Wojack Monday, March 12th, 2012
MOTOROLA and the Stylized M Logo are trademarks or registered trademarks of Motorola Trademark Holdings, LLC. All other trademarks are the property of their respective owners. © 2010 Motorola Mobility, Inc. All rights reserved.
Why Statistical Methods
• Disciplined Approach •Repeatable Results • Quantifiable Decision Criteria • Optimization © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Statistics Review
Population Sampling Parameters:
µ, σ
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Probability Distributions
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Normal Distribution Normal (Gaussian) Distribution
0.45
Probability Density
0.4
0.35
ƒ (x) =
0.3
e-(x-µ)
2
/ (2σ2)
σ(2π)1/2
0.25
68.27%
0.2
µ = mean σ = standard deviation
0.15
0.1
0.05 0 -6 -6σ
-4.8
-3.6 -3σ-2.4
-1.2 1σ
-0
1.2 1σ
95.45% 99.7% 99.9999998%
2.43σ 3.6
4.8
6 6σ
Process Capability
Normal Distribution
Allowable Variation Cp = Actual Variation Allowable Normal (Gaussian) Distribution 0.45 0.4
Probability Density
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -6
LSL
-4.8
-3.6
-2.4
-1.2
-0
1.2
2.4
3.6
4.8
6
USL
Actual © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Process Capability
Normal (Gaussian) Distribution 0.45 Normal (Gaussian) Distribution 0.4
0.4
0.35
Probability Density
0.45
Probability Density
0.35 0.3 0.25 0.2
0.3 0.25 0.2 0.15
0.15
0.1
0.1
0.05 0
0.05
-6
0 -6
LSL
-4.8
-3.6
-2.4
-1.2
-0
1.2
2.4
3.6
4.8
6
LSL
-4.8
-3.6
-2.4
-1.2
-0
1.2
2.4
3.6
4.8
6
USL
USL
Cp =2.0
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
2.0 distributions can fit within the specification
13-Mar-12
Process Capability Poor Capability
Bad
< 1.0
1.0 – 1.5 Marginal
Capability
Good Capability
> 2.0
Motorola 6σ Capability
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Best
> 1.5
13-Mar-12
Process Capability Normal (Gaussian) Distribution 0.45 0.4
Probability Density
0.35 0.3 0.25 0.2 0.15 0.1
Cp
0.05 0 -6
-4.8
-3.6
-2.4
LSL
-1.2
-0
1.2
2.4
3.6
4.8
6
USL
Normal (Gaussian) Distribution 0.45
Independent of the
0.4
Probability Density
Cp = 2.0
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -6
LSL
-4.8
-3.6
-2.4
-1.2
-0
1.2
2.4
3.6
4.8
6
USL
Normal (Gaussian) Distribution 0.45 0.4
target.
Probability Density
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
LSL © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
-6
-4.8
-3.6
-2.4
-1.2
-0
1.2
2.4
3.6
USL
4.8
6
13-Mar-12
Process Capability; Cp & Cpk
USL - LSL Cp = 6σ
LSL
µ
Cp = 2.0 Cpk = 1.0
Cpk = min
USL
USL - µ , µ - LSL 3σ
3σ LSL
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
µ
Cp = 1.0 Cpk = 1.0
USL
13-Mar-12
Cp, Cpk Example
Chip set Stack Up
Chipset Stack Up
W3G (current)
OMAP4430
Package Height Prediction (post-reflow) Memory Mold Cap (2-die) Memory Substrate Thk. Memory Ball Ht. (after reflow)
Min. 0.330 0.110 0.140
Nom. 0.340 0.140 0.190
Max. 0.350 0.170 0.240
Tol. 0.010 0.030 0.050
Min. 0.340 0.110 0.240
Nom. 0.370 0.140 0.290
M 0 0 0
OMAP Mold Cap (Ref.) OMAP / W3G Die Thickness (um) Die-to-Substrate Gap (um) OMAP Top Ball Ht. OMAP/W3G Substrate Thk. (w/o SM) OMAP/W3G Ball Ht. (after reflow) OMAP/W3G Ball Ht. (before reflow)
0.000 0.095 0.020 0.000 0.408 0.168 0.180
0.000 0.100 0.025 0.000 0.448 0.206 0.230
0.000 0.105 0.030 0.000 0.488 0.244 0.280
0.000 0.005 0.005 0.000 0.040 0.038 0.050
0.220 0.095 0.020 0.160 0.244 0.110 0.140
0.250 0.100 0.025 0.190 0.284 0.160 0.190
0 0 0 0 0 0 0
Total 2-Package POP Height (post-reflow): Max. 2-Package POP Height (post-reflow): Delta to OMAP4
1.24
1.32 1.40 0.06 delta
1.40 RSS:
1.15
1.24 1.34
1
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
0.081
reference 13-Mar-12
Chipset Height Cp, Cpk +0.16mm
-6σ Cp = 2.8 Cpk = 2.7
Etna
Targa
+6σ
0.1154mm
Mechanical Development Process
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Tolerance Analysis
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Development Flow (simplified)
Concept
Tolerance Analysis
part / process design
no
TA Acceptable?
yes © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Tolerance Analysis
Components
A B
4.54
0.50 +/- 0.05
1.00 +/- 0.07
A
B
C
0.55 + 1.07 +
C
1.60
D +
1.35
1.50 +/- 0.10
Envelope Size
D
1.25 +/- 0.10
?? +/- 0.13 © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
4.67 A
B
C
D 13-Mar-12
Root Sum Squared (RSS)
Variances can be added…… 2 σ2 = σA2 + σB2 + σC2 + σD2 + σEnvelope
( Σ
Te + 3Cp
(
(
m
Tpi 3Cpi
2
(
σgap =
2
i=1
σgap = 0.035 © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Root Sum Squared (RSS)
Components
A
0.50 +/- 0.05
B
1.00 +/- 0.07
C
1.50 +/- 0.10
D
1.25 +/- 0.10
Gap Size
σgap = 0.035 3σgap = 0.105 6σgap = 0.210 Envelope = A+B+C+D+6σgap = 4.46
?? +/- 0.13 © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
A
B
C
D 13-Mar-12
RSS vs. Worst Case
Worst Case = 4.67
A
B
C
D
∆ = 4.5%
Statistical = 4.46
A
B
C
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
D 13-Mar-12
TA Example: uUSB and HDMI Connectors
Measurement Systems Analysis (MSA)
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Development Flow (simplified)
Good TA
spec / print
trial run
(requirement)
….. Measurement Systems Analysis
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Measurement Systems Analysis
WHY? Measurement Error
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Bad Decisions
13-Mar-12
Measurement Systems Analysis
Characteristics: Stability
Discrimination
Accuracy (Bias)
Linearity
Precision
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Measurement Systems Analysis
Total Variation:
Precision:
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
MSA; %GR & R
% GR&R 30%
Unacceptable
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Measurement Systems Analysis Example: Glue Weight
Measurement Size
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Sample Size • Dependent on the type of analysis to be performed • Apply applicable formula • Example: Sample Mean to a known population
n=
(Zα + Zβ) • σ2 δ2
α = level of acceptability of a false positive (0.05 is typical) β = level of acceptability of a false negative (0.10 is typical) σ = known standard deviation m = amount of difference that matters (practical difference)
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Development Flow (simplified)
MSA
MSA Acceptable?
Determine Quantity to Measure
yes
no Improve the Measurement System
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
5 or 20 or 100 ???
Perform Measurements
13-Mar-12
Data Analysis
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Data Analysis
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Development Flow (simplified) Concept
Tolerance Analysis
part / process design
yes Good?
spec / print
no
Process Control
trial run
Improve Measurement
no yes
Reliability Modeling
Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
Good?
trial run
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Comparative Analysis & Design of Experiments (DOE)
Comparative Methods
• Analytical method to evaluate changes & differences. • Examples: • Different vendors • Change to a process
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Comparative Methods
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Comparative Analysis Example: PCB Bow
Design of Experiments
• Efficient Experimental Method • Optimizes Processes and Designs • Allows for the Analysis of Interactions
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Process Control
Development Flow (simplified)
Concept
Tolerance Analysis
part / process design
Good?
yes
no
Process Control
trial run
no
yes Reliability Analysis
Good?
trial run Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
spec / print
Improve Measurement
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Process Control
Statistical Process Control
SPC Goals:
• Predicable process • Consistent σ (Cp) • Centered distribution © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
(Cpk)
13-Mar-12
Control Charts
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Control Charts
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Control Charts (X-bar / R)
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Development Flow (simplified)
Concept
Tolerance Analysis
part / process design
Good?
yes
no
Process Control
trial run
no
yes Reliability Analysis
Good?
trial run Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
spec / print
Improve Measurement
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Reliability Analysis
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Failure Rate
Reliability Analysis
Decreasing Failure Rate
Constant Failure Rate
Increasing Failure Rate
bad region
desired region
bad region
Normal Operation b=1 Infant Mortality b1 This region is not always perfectly flat.
13-Mar-12
Failure Analysis Example: Display Breakage
Development Flow (simplified) Analyze Data
Measurement Data
Normal (Gaussian) Distribution 0.45 0.4
Probability Density
0.35 0.3 0.25 0.2
Process Control
0.15 0.1 0.05 0 -6
-4.8
-3.6
-2.4
-1.2
-0
1.2
LSL
2.4
3.6
4.8
6
Cp = 2.0 Cpk =0.0
USL
Modify Specification 3.65
no
Improve Manf. Process
Data Acceptable?
yes Reliability Analysis
Cp= 2.0 Cpk = 2.0 © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Tools Used
DFSS Tools
Tolerance Analysis
Measurement System Analysis
Process Capability
Comparitive Methods
Design of Experiments
Reliability Modeling
Monte Carlo Simulation
DACE
Etc….
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Development Flow (simplified)
Concept
Tolerance Analysis
part / process design
Good?
yes
no
Process Control
trial run
no
yes Reliability Modeling
Good?
trial run Measurement Systems Analysis
Improve Manf. Process
Modify Spec.
spec / print
Improve Measurement
no
Good?
yes Analyze Data
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Measurements
Measure Quantity ?
13-Mar-12
Why Statistical Methods
• Disciplined Approach •Repeatable Results • Quantifiable Decision Criteria • Optimization © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Q&A
Back Up Slides
6 Sigma Methodology
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
DFSS – CDOV Process
DFSS – Design for Six Sigma
CDOV Concept •Prioritize Customer Needs •Select Superior Concept
Design •Baseline Design •Customer needs captured in Design Requirements
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
Optimize
Verify
•Optimize “Critical to Quality” Parameters
•Ensure long term performance
13-Mar-12
CDOV Process
Concept Major Steps
VOC
KJ Analysis
Initiate CPM
Pugh Analysis
Design MSA
Process Capability
Confidence Comparative Intervals Methods
Monte Carlo
CPM
FMEA
Control Charts
Optimize Regression
DOE
RSM
Robust Design
Tolerance Analysis
DACE
Verify Reliability Modeling
System Reliability
System Availability
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Measurement Systems Analysis
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Design of Experiments
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Design of Experiments
LevelFactors -> 22 = 4 Runs © 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Design of Experiments
2k -> LevelFactors
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12
Design of Experiments
© 2010 Motorola Mobility, Inc. - Motorola Confidential Proprietary
13-Mar-12