Design for Reliability Approach in Magnetic Storage Industry
A. Parkhomovsky, R. M. PelstringReliability Engineering, Motor Design Division,
Seagate Technology
Outline• Introduction
i. Early Reliability Failure Detectionii. Design for Reliability Approach
• Reliability Risk Assessmenti. FMEAii. Fault Tree Analysis
• Predictive Reliability Modelingi. Understanding of physical processes in the productii. Identification of critical to reliability parameters and possible failure modesiii. Design for Reliability Modeling using DOE and first principles approachiv. Reliability Risk Assessment using predictive models
• Customized Accelerated Stress Tests• Summary
Spindle Motor Cross SectionSpindle Motor Cross Section
Journal Bearing
SleeveShaft
Journal Gap
Hub
Design for Reliability Definition
• The tool set that supports product and process design (during the Product Development Cycle) to ensure customer expectations for reliability are fully met.
• Initial
• After Current Stressing
Tk
E
n
a
eJ
ALife sticCharacteri
DFSS vs. DFR
DFR focuses on achieving high quality over time and across stress levels.
DFSS DFR
VOC
MSA
DOE
Control Plans
ANOVA
QFD
FMEA
Regression Flowdown
Environmental & Usage Conditions
Life Data Analysis
Physics of Failure
Accelerated Life Testing
Reliability Growth
Warranty Predictions
FA recognition
General Linear Model
Tolerancing
Sensitivity Analysis Modeling
Hypothesis Testing
Identify and Design
Optimize
FMEA
S = ?O = ?E = ? Fault
Tree
Critical to Reliability Parameters (CTR) and Supplier Capability
ReliabilityModels
StatisticalReliabilityPrediction
Scard79 Analyst: # CTQ's
Program 3
3Scorecard 3
Reliability 3Last Updated 3
3Seagate Confidential Rev 6.0 3
Lower UpperSpec Limit Spec Limit
Motor SeizureDrive Performance FailureDrive Contamination
CTQ Name Units Mean PNCStandard Deviation
Gage %(P/P)
PNC ZSTMaturity
Level
Default Threshold ZST
Menu GuideInput Long Term Mean and Std Dev
(for Normal Data) OR Long Term PNC
(for Non-Norm al/Attribute data)
New CTQ NameDefault Maturity Level
Missing ZST
Missing Gage %(P/P)
Parameter
Total
User Input
Guide
Maturity Level < 2
Top DTop L MDS
Design for Reliability
Validate
• Motor Design Limits Testing
• Concept Verification
Control
SPC
Post-TransferControl MeasuresVerify
System Margin and Robustness
Product Development and Life Cycle Process
Design For
Reliability
Reliability Verificatio
n
Product and Process Analysis
• Physics of Failure understanding and modeling
• FMEA, design risk analysis, Fault Tree
• Design, process and product analyses
• Failure Analysis
• Early Reliability Tests
• Design Limit Tests
• Field Data analysis
Gap Closure though interrelated concurrent activities
Ensuring Reliability in the Product Development Process
Concept
Evaluation
Design Maturity
Transition
ProductionProduct
Development
Phases
Fault Tree, FMEA,
Design Rules
Early Reliability Tests
Reliability Limit Tests
Reliability Limit Tests
Ongoing Reliability
Tests
1. Design Out Failure Mechanisms
2. Reduce Variation in Product Strength
3. Reduce Effects of Usage/ Environment
4. Increase Design Margins
Utilization of the design, product and process knowledge
Design for Reliability Approach Strategies
Design for Reliability Implementation Benefits
• Seagate benefits:• Significant Reduction in Cost of development.• Increase in the number of orders for disc drives.• Reduction in the reserve and storage needs.• Customer integration failures reduced.• Field failures reduced.
• Supplier benefits:• Larger allocation of business for suppliers commodity.• Improved designs and specifications allowing more opportunity
for optimization of the supplier’s process.• Improved yields with more predictability.• Less negative surprises.
Best Practices Define Success• Reliability must be designed into products and
processes, using the best available science-based methods.
• Knowing how to calculate reliability is important, but knowing how to achieve reliability is equally if not more important.
• Design for Reliability practices must begin early in the design process and be well integrated into the overall product development cycle.
Comparative Resource Commitment
Actual Resource
Level
Post ReleaseProblem Teams
Time
Planned Resource Level
Reso
urce
Lev
el
Expected Resource Levelwith Design for Reliability
Many Changes Few Changes
Shorter Development CyclesEfficient Use of Resources
Reliability Model Feedback LoopEvent Reduced fly height EventDescription DescriptionCond PNC 1 Cond PNC 1Cum PNC 0.0001 Cum PNC 0.0001Function AND Function AND
Event Wear occurs in CP grooves EventDescription DescriptionCond PNC 0.0001 Cond PNC 0.0001Cum PNC 0.0001 Cum PNC 0.0001Function AND Function AND
Event Contact occurs in thrustDescriptionCond PNC 1Cum PNC 1Function AND
EventDescriptionCond PNC 1Cum PNC 1Function
Event Op ShockDescription 1000 g'sPNC 1Function
DLC contamination causes wear / seizure
Contact stress exceeds DLC strength
Restoring force does not prevent contact
Product
Op-shock
250 g’s 2 ms
MobileMarket Requirement
Fault Tree Analysis
Design opportunity and model gap identified to “break” failure chain.
24.524.023.523.0
300
200
100
0
Force Distribution
Fre
qu
ency
Histogram of Force Distribution
1.585 1.595 1.605 1.615 1.625 1.635
0
100
200
Stress
Fre
qu
ency
Histogram of Stress
Model Developmentand Results
FImpact scontact
DesignImprovement
Contact relief to reduce contact stress.
Fault Tree Model – Shock Failure
Fault Tree general skeletons are developed, then they are easily adapted to the particulars of each design.
FMEA – Test Linkage: Example
Motor Design FMEAItem Part Potential Failure Mode Effects of Failure S Potential Cause O Design Verification E RPN
11 Sleeve/ Thrust Cup assy
excessive wear on thrust surface
motor seizure 10 High runout, contamination (ECM)
2 runout measurement 2 40
15 Bearing assemblycomponents rubbing while spinning motor lock up, oil leakage 9
parts tolerance allow contact or not meeting print.
4
Min Gap model includes all surface and diameter parameters, bearing drag test will be correlated to journal gap.Performance testing.
1 36
18 Bearing assembly journal wearchange in performance, oil degradation, motor lock up & oil leak from gyro test
8wear from operating tests, gyro scopic wear, CSS
5design validated through testing and run more that 60k cycles 2 80
23 EM EM bias force too highreduced fly height, increased wear rate 5
Misalignment of stator, magnet or bias ring. Incorrect magnetization
2In-process height measurements, drawings/tolerance studies, magnetization
3 30
The Design FMEA is developed based on critical failure modes from the fault tree analysis.
Motor Reliability Design Limit Test (RDLT) Plan Test Test Groove Depth Shaft DLC
FMEA Duration Test Test Orientation Temp. CSS Thrust Journal Thickness
Savvio Motor Design Variables Item # RPN # (month) Qty +VSA -VSA HSA ( 0C ) (cycles) (mm) (mm) (mm)
Motor Wear Test 11,15,18,23 81 2 15
Norminal design (control ) 3 3 3 70 72K 7.5 3.0 1.0
Max thrust cup to shaft runout+ max thrust groove + max magnetic bias +
low oil fill + max disk load/imbalance 3 70 72K 9.5 3.0 1.0
Largest journal gap, thin DLC + max journal groove
depth + max disk load/imbalance + low oil fill 3 70 72K 7.5 4.3 0.75
Reliability tests used are developed to address high risk items in the FMEA.
Design limit variables (e.g. groove depth, coating thickness) are selected based upon failure mode sensitivity.
Acceleration and stress factors (e.g. temperature, load, orientation) are selected based on design knowledge and product environment.
Design Limits Test Development
Total Failures by Mode – Customer Integration
Data represents a < 5 % FA of all Customer Integration Failures
0
5
10
15
20
25
30
35
Failu
re M
ode1
Failu
re M
ode 2
Failu
re M
ode 3
Failu
re M
ode 4
Failu
re M
ode 5
Failu
re M
ode 6
Failu
re M
ode 7
Plan to attack these failure modes in the ORT
plan
AB
C D E
QTY
Selection of top 5 Field Failure Modes
Total Failures by Mode – Field Returns
Data represents a < 5 % FA of all Field ARR Failures
0
5
10
15
20
25
30
35
Failu
re M
ode 1
Failu
re M
ode 2
Failu
re M
ode 3
Failu
re M
ode 4
Failu
re M
ode 5
Plan to attack these failure modes in the ORT
plan
HG
I J
F
QTY
Failu
re M
ode 6
Failu
re M
ode 7
Selection of top 5 Field Failure Modes
Defining Acceleration Factors
)(
)(
stressdaccelerateL
stressusageLAF
Acceleration factor (AF)is the ratio of the characteristic life at the use and accelerated test conditions:
Multiple Stressor Acceleration Factor Calculation
21total AFAFAF *=
)(
)(
2
11 timeLife
timeLifeAF
test
spec
timeLife
timeLifeAF
spec
spec
2
22
Where:
AF1 is the acceleration factor for stressor 1
AF2 is the acceleration factor for stressor 2
Lifespec – the motor life per specification
Typical Stressors• Variable Speed profile• Time/Number of Cycles• Temperature• Humidity• Operating and non operating shock• Electrical bias• Load
A failure is defined as a significant change in the motor performance parameter over time/cycles.
Definition of FailureP
aram
eter
Capillary Seal Analysis Meniscus Surface Area Calculation
Shock direction
Shock direction
Capillary Seal Non-operating Shock Analysis
Capillary Seal Fill Process Trade off
Gravitational Sag and Shock limitedEvaporation limited
Radial GapRadial Gap
Model based
Capillary Seal Gap Design Trade off
Model based
Gravitational Sag limitedEvaporation limited
Radial GapRadial Gap
Oil Sag due to gravity, margin to fill hole
0.2 0.25 0.3 0.35 0.40
50
100
150
200
250
300
Sag Margin millimeters
Seal Volume(ul) : 3.32
0.2 0.25 0.3 0.350
50
100
150
200
250
300
Sag Margin millimeters
Seal Volume(ul) : 3.5
0.1 0.15 0.2 0.25 0.3 0.350
50
100
150
200
250
300
Sag Margin millimeters
Seal Volume(ul) : 3.68
27
Autocatalytic Reactions• An Autocatalytic reaction is the reaction where the product of the
reaction is also a reactant. • The approach to an autocatalytic rate equation:
o
o
oo
oo
oo
o
A
Pb
kPAa
where
and
xPxAkdt
dx
so
xBB
xAA
and
BAkv
BA
atbe
ate
oPx
1
1
The rate of Change in concentration of the component(s) in an autocatalytic reaction and is described through
the logistic equation
28
Sigmoid Logistic Curve• In Case of the oil (ester) hydrolysis which is auto catalyzed by acids:
RCO2R’ + H2O → RCO2H + R’OH (a)RCO2R’ + RCO2H + H2O → 2 RCO2H + R’OH (b)
• The general rate change equation of the autocatalytic reaction:
o
o
oo
A
Pb
kPAa
where
atbe
ate
oPx
1
1Autocatalysis Logistic Curve
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14
Adjusted Time Unit ([A]o+[P]o)kt
No
rmal
ized
Co
nce
ntr
atio
n c
han
ge
x/[P
]o
Logistic Curve
29
Run Current Analysis of the Lubricant Hydrolysis
• Assume linear dependence between the Irun and the concentration increase of the hydrolysis reaction.
• Fit the Logistic Curve into the existing Irun versus time equation:
)exp(1
1)exp(
)(
2 ktk
kt
I
I
or
tfI
I
orun
run
orun
run
Understanding Wear• Wear is the erosion of material from a solid surface by the action of another solid.
There are four principal wear processes:1. Adhesive wear 2. Abrasive wear 3. Corrosive wear4. Surface fatigue
• Also wear can be classified as dry wear, semi-lubricated wear and lubricated (wet) wear.
• Wear is a complex phenomenon that is a result of generation of thermal or/and chemical energy.
• Wear in the bearing is generated as a result of the contact forces acting between the wear couple components. The work of wear can be calculated from the relation below if the spin down profiles and the forces acting on the bearing components are known. We assume that the wear depth is proportional to the contact pressure in place of contact.
Low
Par
amet
er3
Hi
33
24
28
20
42
30
31
22 26
3840
44
Hi Parameter2 Low
Lo
w
Pa
ram
ete
r1
H
i
Orientation 1
14
12
8
183
34
35
19
15
43
29
1
21 27
3732
1016
13
11
75
Failures are marked in red
Induce motor failures by testing beyond customer specifications
Responses: 1. Wear 2. Time to failure
Factors: 1. Parameter 1 2. Parameter 2 3. Parameter 3
Categorical: 1. Orientation
Stress Tests to induce failuresOrientation 2
Lo
w
Pa
ram
ete
r1
H
i
Hi Parameter2 LowLow
P
aram
eter
3 H
i
Typical Wear RateWear rate vs. sliding distance
Contact (sliding) Distance
We
ar
Rat
e
L
Assume that the wear coefficient is a constant (average wear coefficient) for a given material pair to simplify wear
experiments.
Critical Parameter ScorecardScard
79 Ana lyst: # CTQ's
Program 21Dakota/Firebird - Nidec_DLC Design 21
Scorecard 2121 Z Ma rgin
Last Updated 21 < 011-Oct-04 21 0.0 to 0.5
Sea ga te Confidentia l Re v 6.0 21 > 0.5Low e r Uppe r
Spe c Lim it Spe c Lim it> Pe rform a nce> Ele ctrica lParameter 1Parameter2Parameter3Parameter3Parameter4Parameter5Parameter6
> Me cha nica lParameter 1Parameter2Parameter3Parameter3Parameter4Parameter5Parameter6
> Re lia bilityParameter 1Parameter2Parameter3Parameter3 - -Parameter4Parameter5Parameter6 - -
- -
Thre shold ZST
PNC a nd ZST Ca lcula tor
Pa ra m e te r
Tota l
Us e r Input Guide
Ma turity Le ve l < 2
Missing ZST Z Ma rgin Sum m a ryMissing Ga ge %(P/P)
M e nu GuideInput Long Te rm M e an and Std De v
(for Norm al Data) OR Long Te rm PNC (for Non-Norm al/Attr ibute data)
Ne w CTQ Na m eDe fa ult Ma turity Le ve l
De fa ult Thre shold ZST
CTQ Na m e Units Me a n PNCSta nda rd De via tion
Ga ge %(P /P)
PNC ZSTMa turity
Le ve l
Top DTop L MDSScard
79 Ana lyst: # CTQ's
Program 21Dakota/Firebird - Nidec_DLC Design 21
Scorecard 2121 Z Ma rgin
Last Updated 21 < 011-Oct-04 21 0.0 to 0.5
Sea ga te Confidentia l Re v 6.0 21 > 0.5Low e r Uppe r
Spe c Lim it Spe c Lim it> Pe rform a nce> Ele ctrica lParameter 1Parameter2Parameter3Parameter3Parameter4Parameter5Parameter6
> Me cha nica lParameter 1Parameter2Parameter3Parameter3Parameter4Parameter5Parameter6
> Re lia bilityParameter 1Parameter2Parameter3Parameter3 - -Parameter4Parameter5Parameter6 - -
- -
Thre shold ZST
PNC a nd ZST Ca lcula tor
Pa ra m e te r
Tota l
Us e r Input Guide
Ma turity Le ve l < 2
Missing ZST Z Ma rgin Sum m a ryMissing Ga ge %(P/P)
M e nu GuideInput Long Te rm M e an and Std De v
(for Norm al Data) OR Long Te rm PNC (for Non-Norm al/Attr ibute data)
Ne w CTQ Na m eDe fa ult Ma turity Le ve l
De fa ult Thre shold ZST
CTQ Na m e Units Me a n PNCSta nda rd De via tion
Ga ge %(P /P)
PNC ZSTMa turity
Le ve l
Top DTop L MDS
Summary
• A successful implementation of Design for Reliability (DFR) approach in high volume spindle motor development and manufacturing demonstrated a significant benefit in identifying and addressing critical failures and accelerating design stages.
• We have developed, validated and implemented a number of physics and DOE based predictive reliability models to address the design CTQ early in the concept phase.
• In addition to this, a suite of highly accelerated stress tests was successfully developed to identify critical failure modes in the prototype build stages.