Sei sulla pagina 1di 87

Measurement

Systems Analysis

Design for Six Sigma


Week 2

© 2003-2007 Cummins Inc. All Rights Reserved.


Certain content © 2000-2002 Sigma Breakthrough Technologies, Inc. Used with permission.
Objectives

 Introduce Measurement Systems Analysis


 Define basic measurement terms
 Outline procedure for performing a Gage Study (Measurement
Systems Analysis)
 Perform an exercise to practice methodology

2
DFSS Work Flow
 RACI Diagram  Discussion Guide
Determine Voice of the Customer  Market Perceived Value Profile  Image KJ
 Customer Selection Matrix  Weakness KJ
C
CONCEPT
Develop requirements and  Translation Worksheet  QFD HOQ
measurable top level CFR(s)  Requirements KJ  CPM Scorecard
 Importance Questionnaire

Generate and select  Concept Generation  DFA Analysis Worksheet


the solution concept  Pugh Screening Matrix  DFMEA

Flow down top level CFR(s) to  Critical Parameter Tree  Correlation


D
DESIGN
lower level critical parameters  Cause & Effect Matrix  Process Mapping

 Regression  Basic Stats / Graphs


Determine strategic transfer functions  Design of Experiments  Test For Means
 Response Surface Design  Test For Variances
 ANOVA
Determine CFR(s) capability using  Monte Carlo Simulation  Control Charting
transfer functions and X distributions  Empirical Tolerancing x

Measurement Systems Analysis
Dimensional Variation Analysis
 Response Optimization  Regression Applied to Life Data
O
OPTIMIZE
Optimize for robust performance
 Robust Design 

Accelerated Life Testing (ALT)
Confidence Intervals
Is CFR(s) capable?   P-Diagram
Y/N Capability Analysis
 Life Data Analysis
 Sample Size Planning
Determine actual CFR(s) capability.  2 Proportions Comparison
 Capability Analysis
C
CAPABILITY
Compare to predicted
capability where possible. Y/N  Control Plan
 CPM Scorecard

3
Measurement Process

 The ideal measurement system produces “true” measurements


every time
 Quality of the measurement system is characterized by
statistically evaluated sources of measurement variation
 The measurement process should include:
– Design and certification
– Capability assessment over Time
– Control
– Repair and re-certification
 Properties of a Good Measurement System
– Must be in Statistical Control over time
– Variability must be small compared to product specifications
– Variability must be small compared to process variation
– Resolution should be about one-tenth of product specification or
process variation
4
Possible Sources of Observed
Process Variation
Observed Process Variation

Actual Process Variation Measurement Variation

Long-term Short-term Variation Variation due Variation due


Process Variation Process Variation w/i sample to gage to operators

Repeatability Calibration Stability Linearity

To address actual process variability, the variation due to the


measurement system must first be identified and separated from that
of the process

We will look at “repeatability” and “reproducibility” as primary


contributors to measurement variation

5
Measurement Process Model

Process Measure Value Analyze Decision

σact σMS σobs


σ2obs = σ2act + σ2MS

σobs Observed process variance


σact Actual process variance
σMS Measurement system variance
6
Effect of Measurement System
Variation – Single Measurement

“True” value

σMS

Range of observed measured results

7
Effect of Measurement System
Variation – Poor Capability

Actual Product or
Process Variability

Measurement
Variability

Total Variability
(Observed
variability)

LSL USL

8
Information to be obtained

 How big is the measurement variation?


 What are the sources of measurement variation?
 Has the gage been calibrated?
 Is the gage stable over time?
 Is the gage capable for this study?
 How do we improve the measurement system?

9
Sources of Measurement
Variation
 Measurement variation can come from many sources.
 A gage R&R study, if conducted correctly, can capture many of
the sources, if the studies are designed to look at both short and
long term sources of variation.
 There are sources of measurement variation that can not be
captured through repeated measurements and must be
evaluated separately.

10
Categories of Measurement
Variation

Reference Environment
Standard

Measurement Measurement Measurement


Set-Up Equipment Subject

Total
Measurement
Variation
Definition of Procedures Software
Measurand Calculations

Physical Personnel
Constants

Note: Examples of these categories of variation can be found in the Appendix


11
Now a few Definitions

12
Measurement Accuracy vs.
Precision
 Accuracy
– Closeness of the agreement between the result of a measurement
(Observed) and the true value of the measurand (reference value).

 Precision
– The net effect of discrimination, sensitivity, repeatability over the
range of interest of the measurement system.
– In some organizations precision is used interchangeably with
repeatability.

13
A B

Accurate & Precise Precise but not


Accurate

C D

Accurate but not Not Accurate


Precise and not Precise
14
Resolution

 The smallest difference between indications that can be


meaningfully distinguished.
 Must be suitable for needs.

Poor Resolution Improved Resolution

X X
15
Repeatability

 Closeness of agreement between the results of successive


measurements of the same measurand carried out under the
same conditions of measurement.

16
Reproducibility

 Closeness of agreement between the results of measurements


of the same measurand carried out under changing conditions of
measurement.
 A valid statement of reproducibility requires specification of the
conditions changed.

17
Bias

 An observed difference between the expected value and the


measured value, estimated over an appropriate number of
repeated measurements of calibration standards or calibrated
parts.

Bias

Measurement Expected
Result Value
18
Stability

 The ability of a measuring instrument to maintain a constant bias


over time.

Bias t3

Bias t2
Time

Bias t1

Expected Measurement Results


Value
19
Linearity

 Characteristic of a measurement scale in which the spacing


between each scale increment is proportional to the measurand.
Measurand

Non-Linear

Measurement Results
20
Sources of Variation that can be
Evaluated by a Gage R&R

Repeatability

Reproducibility

Linearity

Stability

Bias

21
Sources of Variation that
“Typically” Evaluated by R&R

Repeatability

Reproducibility

Linearity

Stability

Bias

22
A Measurement Capability Index
- P/T

Precision to Tolerance Ratio

6.00 *σ MS Usually expressed


P /T = as percent
Tolerance
 Addresses what percent of the tolerance is taken up by
measurement variation
 Includes both repeatability and reproducibility
– Operator x Unit x Trial experiment
 Best case: 10% Acceptable: 30%
 The P/T ratio will determine whether the measurement system
can sort parts acceptably against the product specifications.

Note: 6.0 standard deviations accounts for 99.73% of MS variation.

23
Precision to Tolerance Ratio

LSL USL
Product Tolerance

P/T = 20% measurement


system variation

P/T = 100%

P/T = 200%

24
Uses of P/T Ratio (% Tolerance)

 The P/T ratio (% Tolerance in MINITAB) is the most common


estimate of Measurement System precision
– This estimate may be appropriate for evaluating how well the
Measurement System can perform with respect to specifications
– Specifications, however, may be too tight or too loose
– Generally, the P/T ratio is a good estimate when the measurement
system is only used to classify production samples
• Even then, if process capability (Cpk) is not adequate, the P/T ratio may
give you a false sense of security

25
Another Measurement Capability
Index -- %R&R or Gage R&R

σ MS
% R& R = × 100 Usually expressed
σ Total as percent

 Addresses what percent of the Total Variation is taken up by


measurement variation
 Includes both repeatability and reproducibility
– Operator x Unit x Trial experiment
 As a target, look for %R&R < 30%

26
% Repeatability &
Reproducibility

Observed Process Variation

%R&R = 20% measurement


system variation

%R&R = 75%

%R&R = 100%

27
Uses of %R&R (% Study
Variation)
 The %R&R is the best measure for the Design Specialist
– This estimates how well the Measurement System performs with
respect to the overall process variation
– %R&R is the best estimate when performing Design studies

Care must be taken to use samples representing full, but


typical, process variation

28
LSL USL
Product Tolerance

Observed Process Variation

P/T = 20% %R&R = 20%

P/T = 50% %R&R = 50%

P/T = 100% %R&R = 100%

measurement system variation


29
LSL USL
Product Tolerance

Observed Process Variation

P/T = 50% %R&R = 25%

P/T = 100% %R&R = 50%

P/T = 200% %R&R = 100%

measurement system variation


30
LSL USL
Product Tolerance

Observed
Process Variation

P/T = 10% %R&R = 20%

P/T = 20% %R&R = 40%

P/T = 50% %R&R = 100%

measurement system variation


31
Effect of Measurement Variation
on Capability Indices

The higher the Measurement Variation the more dramatic the


impact on your ability to understand the true design/process
capability!

32
Gage R&R Study Set-up

 Generally two to three operators


 Generally 10 units to measure
 Each unit is measured 2-3 times by each operator
– This will quantify variation due to repeatability and reproducibility

33
Designing Your MSA Study

 When evaluating the stability or linearity of the measurement


system, assure the calibration standard is capable e.g.
– calibration standard’s stability < 25% measurement system’s stability
 Where practical get at least 30 measurement values to assure a
statistically valid sample.
 Measurement should be spread across trials that represent the
major sources of variation.

34
Potential Sources of Variation in
Your MSA Study

Category Variation Measurements Conditions

Environment Temp changes 3 Am, noon, pm


dramatically during the
day

Equipment Suspect calibration issue 1 5 measurements of


calibrated part and set-
up

Set-Up Part mounting on fixture 10 30 set-ups with same


may not be repeatable, operator
but is not influenced by
operator
Parts Surface finish may be 10 10 parts
issue

35
Gage R&R Study Sample
Selection
 For an accurate Gage R&R calculation, samples should be pulled from
the process that span the normal variation of the process
– Example: If you produce a material with a mean thickness of 0.010” and a
sigma of 0.001”, get samples that have thicknesses from 0.007 - 0.013”
(99% of the range)
– If the part-to-part variation is too small the GR&R value would be artificially
high
 Do not compare different part numbers
– If you produce the same base material with different assays with the same
process, sub-group them and perform the R&R study
– Example:
• A process produces a material with thicknesses of 0.025, 0.050, and 0.080 all with
a +/- 0.005 tolerance, all measured with the same system
• Perform 3 studies - one for each thickness
– If you lumped the above samples together, the GR&R value would be
artificially low
 In either case, the σMS will still be accurate

36
What if the Test is Destructive?
Sample Selection Guidelines
 Sample selection is an issue with destructive and on-line
measurement systems
 True repeatability is not possible
 Sample selection can be done to minimize between part
variability in the following manner
– Collect the number of ‘Master’ samples necessary for your study
– Sub-divide them into enough smaller samples to cover the Operator
x Sample x Trial value
 Assumption
– The variability between the smaller samples is negligible and can be
ignored

37
Destructive Testing
Sample Selection Example

Assume we have a 2 Operator, 2 Trial, 10 Sample study we wish


to conduct
We would need 10 Master Samples and 2 x 2 x 10 = 40 smaller
samples 1 2 3 4 5

6 7 8 9 10

38
Destructive Testing Methods

 Procedure
– For each Operator / Trial / Sample combination, randomly pick a
smaller sample from the respective group
– Each operator performs the measurement independently

DO NOT MIX ALL SMALLER SAMPLES TOGETHER!

39
General Procedure for
Performing an R&R Study
 Calibrate the gage, or assure that it has been calibrated
 Have the operator 1 measure all samples once in random order
 Have the operator 2 measure all samples once in random order
 Continue until all operators have measured samples once (this is
Trial 1)
 Repeat steps 2-4 for the required number of trials
 Use the form provided to determine the statistics of the R&R
study
– Repeatability
– Reproducibility
– Standard deviations of each of the above
– %R&R
– P/T Ratio
 Analyze results and determine follow-up action, if any 40
Gage R&R Example

 Let’s use MINITAB to analyze some data


 Open worksheet Thickness GR&R in file MSA.MPJ
 We will be using MINITAB’s Gage R&R Study functions

41
Gage R&R Example: Laminate
Thickness Measurement
 This is a gage study for an over-arm micrometer to measure the
thickness of a material
 The values represented are multiplied by 1000 for ease of data
entry (i.e. 8.9 = 0.0089”)
 The specification value is needed to determine %P/T. For this
example the specification is +/-0.7, or a tolerance of 1.4
 Let’s start analyzing the data
– Stat>Quality Tools>Gage Study>Gage R&R Study (Crossed)

42
MINITAB GR&R Study

43
Options / Info Screens

Enter the specific gage


information here for records
Important when doing a
number of iterations

Enter the Process Tolerance


in the Options screen
REMEMBER-tolerance is the
entire range of the
specifications!
You can add a descriptive
Title as well

44
MINITAB Output

 MINITAB produces both analytical and graphical analysis


information
– Analytical Results
• ANOVA table(s)
• Components of Variation
• Percent contribution table
– Graphical Results
• X-Bar / R chart
• Components of Variation
• Operator*Part Interaction Plot
• By Operator and By Part plots
 Let’s look at the analytical first and then return to the graphical

45
Gage R&R Report

Study Var %Study Var %Tolerance


Source StdDev (SD) (6 * SD) (%SV) (SV/Toler)
Total Gage R&R 0.0496522 0.297913 54.67 21.28
Repeatability 0.0476514 0.285908 52.47 20.42
Reproducibility 0.0139528 0.083717 15.36 5.98
Operator 0.0139528 0.083717 15.36 5.98
Part-To-Part 0.0760485 0.456291 83.73 32.59
Total Variation 0.0908224 0.544934 100.00 38.92

Number of Distinct Categories = 2

• % R&R (for Design efforts)


• % P/T (for acceptance to spec efforts)
• BE CAREFUL - watch for low Distinct Categories! Must
be at least 5 for Design use!
46
Distinct Categories

LCL
LSL UCL
USL

100% Noise

LCL
LSL
UCL
USL

Attribute Data
LCL
LSL
UCL
USL

Low Grade Variable Data


47
Graphical Output

48
Gage R&R X / R Chart

• The Range chart


can help identify
inadequate
discrimination

D4=2.575 for 3 trials


Avg. Range=0.0567
D4*Avg.R=0.1459

Based on Noise of the Gage

• We want to see variability in the X chart outside the Control Limits


• This indicates Part-to-Part variability
• If there was none, you probably did not get samples that cover the
normal range in production
49
X-Chart Indicators

 If the average for each operator is different, the reproducibility is


suspect
 We want more averages to fall outside the control limits… but
consistently for all operators
– This indicates more part-to-part variability which is what we want
 We want to see the majority of the points on the chart outside the
control limits
– If this is the case and the R-Chart is in control, then we will be able
to determine the percent of the design or process variability that is
consumed by the measurement system

50
R-Chart Indicators

 Suspect inadequate Discrimination if:


– the range chart has less than 5 distinct levels within the Control
Limits
– 5 or more levels for the range but more than 1/4 of the values are
zero
 Repeatability is questionable if the range chart shows out-of-
control conditions
 If the range for an operator is out-of-control and the others are
not, the method is suspect
 If all operators have ranges out-of-control, the system is sensitive
to operator technique

51
Operator-Part Interaction Plot

• A line is plotted for each operator by part


• Significant interactions are indicated by diverging or crossing lines
between operators
• We desire parallel lines for all operators, all parts
• If there exists Operator-Part interactions, we need to understand and
resolve them
52
Components of Variation

• A graphical representation of the data discussed before


• We want the Gage R&R bars to be as small as possible, driving the
Part-to-Part bars to be larger

53
By Operator

• The By Operator graph shows the average value (crossed


circle) and the spread of the data for each operator
• We want the grouping to be similar across all operators and a
flat line across the means for the operators
54
By Part

• The graph shows the average (crossed circle) and spread of the
values for each part
• We want to see minimal variability for each part and large
variability between parts (large part-to-part variation)
55
P/T and GR&R Metrics -
Guidelines

Number of
Distinct
P/T R&R Categories
%Study %Contribution (Discrimination
Ratio Variation (Ratio of Variances) Index)
Red

30% 28% 7.7% 5


Yellow

8% 14% 2% 10
Green

R&R % Contribution is more statistically correct than %Study Variation

56
Exercise

 Take a look at the next worksheet New Gage


 Tolerance RANGE for this file is 1.4 (values represented are
multiplied by 1000 for ease of data entry as before)
 Run the Gage R&R study
 Take 5 minutes and draw your conclusions on:
– Gage acceptability for Design
– Gage acceptability for Product Acceptance
– Improvements made to gage system

57
Exercise

 How to do a Run Chart on


the data in the New Gauge
worksheet and select Part,
Operator and Thickness.

58
Gage Run Chart
Gage Run Chart of Thickness by Part, Operator
R eported by :
G age nam e: T olerance:
D ate of study : M isc:

1 2 3 4 5 O perator
D oug
7.7 Rich
Rick

Mean 7.6
Thickness
s

7.5

6 7 8 9 10

7.7

7.6 Mean

7.5

Ope rator
Panel variable: Part

The chart shows the average values for each Sample by Trial by Operator
Look for anomalies - we want flat lines at the same value for all operators!
59
Measurement System Analysis
Questions
 Written inspection/ measurement procedure?
 Detailed process map developed?
 Specific measuring system and set-up defined?
 Trained or Certified Operators?
 Instrument calibration performed in a timely manner?
 Tracking Accuracy?
 Tracking R&R?
 Tracking Bias?
 Tracking Linearity?
 Tracking Discrimination?
 Correlation with supplier or customer where appropriate?

60
Measurement System Analysis
Questions
 Have you picked the right measurement system? Is this
measurement system associated with either critical inputs or
outputs?
 What do the precision, accuracy, tolerance, P/T ratio, %R&R and
trend chart look like?
 What are the sources of variation and what is the measurement
error?
 What needs to be done to improve this system?
 Have we informed the right people of our results?
 Who owns this measurement system?
 Who owns trouble shooting?
 Does this system have a control plan in place?
 What’s the calibration frequency? Is that frequent enough?
 Do identical systems match?
61
Variables MSA

How will a “good” measurement system look?

How will a “bad” measurement system look?

What is the impact to the customer if the system is “bad”?

62
GR&R Exercise

 The Department of Health has decided to let you check your own
waste water cleanliness provided you can measure the water
volume with acceptable repeatability and reproducibility
 Due to the nature of the materials being measured, if you touch
the container while measuring you will be exposed to potentially
high levels of biological and chemical toxins.
 If you touch the container while measuring, you will not be
permitted to perform your own measurements, costing the
company thousands of $$ in labor and testing by outside
laboratories
63
GR&R Exercise

 Equipment needed
– 10 plastic containers
– 1 measurement device
 Specifications
– See instructor
 Procedure
– Have each of three operators independently measure the height of
the water three times (use proper MSA techniques), WITHOUT
touching the container and following the order of the worksheet
• REMEMBER - Maintain one operator per trial!
– Record the measurements in the “MSA Exercise” worksheet in the
MSA.MPJ MINITAB file.
– The measurement order of the containers is randomized (i.e.,
1,3,7,5,2,...) for each operator
 Analyze the results and present to class
64
Summary

 Measurement systems are important to analyze BEFORE


embarking on Design activities
 Be careful when picking samples - watch for correct sub-
grouping and sample size requirements
 Analyze the measurement system for Operator, Part, and Trial
effects
 Make sure that the gage system has enough discrimination to be
useful in determining different levels in the measured attribute
 Always generate a GR&R Report to document findings, methods,
and improvement opportunities
 Total Variation includes Measurement Error - try to minimize the
controllable error in the gage system

65
Appendix

 Variation Sources and GR&R Study Matrix Design Methods

66
Categories of Measurement
Variation

Reference Environment
Standard

Measurement Measurement Measurement


Set-Up Equipment Subject

Total
Measurement
Variation
Definition of Procedures Software
Measurand Calculations

Physical Personnel
Constants

67
Example Sources of
Measurement Variation
 Measurement Set-Up
– Fixturing/part location
– Contact angle/Cosign error
– Flexure
– Probe type/size
 Reference Standards
– Gage blocks/rings/pins
– Ball bar/hole plate
– Torque transducer

68
Example Sources of
Measurement Variation
 Measurement Equipment
– Range
– Sensitivity
– Resolution
– Linearity
– Stability
– Frequency response
– Digital interface

69
Example Sources of
Measurement Variation
 Environment
– Temperature
– Humidity
– Vibration
– Electrical power quality
– Note: Environmental sources of variation can many time be very
dependent on time of day or shift.
 Measurement Subject
– Cleanliness
– Surface finish

70
Example Sources of
Measurement Variation
 Definition of Measurand
– Referenced the correct datum
– GD&T call out
 Physical Constants
– Temperate coefficient of material
 Measurement Procedures
– Number of points measured
– Location of measurement points
– Settling time
– Checking calibration prior to measurement

71
Example Sources of
Measurement Variation
 Personnel
– Properly trained
– Fail safes
 Software/Calculations
– Algorithms verified
• Example calculating a cone
– Filtering
• Frequency filtering
• Sampling frequency

72
MINITAB Gage R&R Data Matrix

 Assume we want to set up a table for a 3 Operators, 10 Sample


per trial, 2 Trials gage study
 We want to create a MINITAB file with the following columns
– Part or Sample
– Operator
– Trial
– Measurement
 There are two ways to create the data table
 First, let’s set up the table to have the correct Trial / Operator /
Sample combinations using the Make Patterned Data procedure

73
Data Matrix Development

The general format for generating a design for:


# Trials = a
# Operators = b
# Samples = c

Trials The syntax to the left


(1:a)bc are the MINITAB
Session commands for
Operators
setting up the table
a(1:b)c
Samples We will generate the
table with pull down
ab(1:c) menus instead

Let’s look at this example. In this case a=2; b=3; c=10

74
Making Patterned Data

75
Pull Down Menus
Trials
(1:a)bc

1:a

bc

76
Operators
a(1:b)c

1:b

77
Samples
ab(1:c)

1:c

ab

78
MINITAB Gage R&R File

79
Coding Variables

 You can change the


Operator values to the
actual person’s name and
still perform the analysis
using a Coding procedure

80
Coding Variables

81
Last Checks...

 Before running your study, make sure that you have the right
number of cells for each operator and sample
 To do this we use Cross Tabulation

82
Cross Tabulation

83
Table Statistics

 The table should contain equal counts for all operators and
all parts before continuing

Tabulated statistics: Trial, Operator

Rows: Trial Columns: Operator

Fred Joe Mary All We look for equal


cell counts in all
1 10 10 10 30 combinations
2 10 10 10 30
3 10 10 10 30
All 30 30 30 90

Cell Contents: Count

84
Mathematical Terms

 GRR – Repeatability and Reproducibility


 EV – Equipment Variation
 AV – Appraiser Variation
 PV – Part Variation

85
Mathematical Relationships

 Gage R&R
– Total variation of measurement system, GRR = EV 2 + AV 2
equipment and appraiser
 Total Variation
– Total variation of study
TV = GRR 2 + PV 2

86
Mathematical Relationships

 % of Equipment Variation
– Ratio of gage and fixturing variation [
% EV = 100 * EV
TV
]
compared to total variation
 % of Appraiser Variation
– Ratio of appraiser variation compared to
total variation
% AV = 100 * AV[ TV
]
 % Gage Variation
– Ratio of measurement process variation
compared to total study variation
[
%GRR = 100 * GRR
TV
]
 % Part Variation
– Ratio of variation of the parts compared
to total study variation
[
% PV = 100 * PV
TV
]

87

Potrebbero piacerti anche