Lecture 4
Lecture 4
Reproducibility Analysis
Gage R&R
1
Collect
baseline Validate
data on Develop a your Analyze Determine
defects Sampling measurement patterns process
and Strategy system in data capability
possible using
causes Gage R&R
2
The Objective
4
Foundation of the Tools
Data is derived from objects, situations, or phenomenon in the
form of measurements.
6
Knowledge to be obtained
7
Measurement Analysis
An evaluation of the measurement system MUST be
undertaken to ensure effective analysis of any
subsequent data generated for a given process/product
characteristic
Observed value = True value + Measurement Error
Measurement error is a statistical term meaning the net effect of all sources of
measurement variability that cause an observed value to deviate from the
master
9
Possible Sources of Process
Variation
12
Sources of Measurement Variation
Why do Measurement Vary?
Vibration
Temperature Fluctuation
Environment Methods
Environment
13
Sources of Variation
Product Variability
(Actual variability)
Measurement
Variability
Total Variability
(Observed variability)
14
Effects of Measurement Error
Measurement
System Bias -
Averages Determined through
“Calibration Study”
Measurement System
total
2
product
2
measurement
2
15
Terminology
Accuracy related terms:
True value
Bias
Linearity
Precision-related terms
Repeatability
Reproducibility
Linearity
Stability (over time)
16
Accuracy
• Accuracy - Does the average of the
measurements deviate from the true value?
• True value:
– Theoretically correct value
– NIST standards
• Bias
– Distance between average value of all
measurements and true value
– Amount gage is consistently off target
– Systematic error or offset
17
BIAS Definition
Observed
Average
Accuracy is linked to Bias
Value
18
Bias Example:
What is the Bias of the following situation
One part measured 10
X1=0.75 mm X6=0.8 mm
times by one Appraiser X2=0.75 mm X7=0.75 mm
X3=0.8 mm X8=0.75 mm
X4=0.8 mm X9=0.75 mm
X5=0.65 mm X10=0.7 mm
Bias = 0.75-0.8= -0.05
X
X = 0.75
% Bias=100[0.05/0.70]=7.1%
10 This means 7.1% of the process variation is BIAS
19
Precision
Total variation in the measurement system
REPEATABILITY
21
Reproducibility Definition
Operator-B
Operator-A
Reproducibility 22
Stability (drift) Definition
Stability: Is the total variation in Time-1
the measurement obtained with a
measurement system (test / gage) on
the same master parts when
measuring a single characteristic over
an extended time period
Time-2
23
Resolution Definition
X Adequate Resolution
X There is enough resolution in
X X the measurement device so that
X X X the product can have many
X X X X different values.
X X X X X
5.1 5.2 5.3 5.4 5.5
24
The Nature of Process Variation
1 2 3 4 5
6 * MS
P /T
Usually expressed
in percent
Tolerance
27
Capability Index (2)
%R&R
MS
% R& R
Usually expressed
in percent
tot
28
Conducting the Gage R&R
29
Types of R&R studies
30
Procedure for performing an R&R study
1: Calibrate the gage, or assure that it has been calibrated
2: Have the first operator measure all the samples once in random
order
3: Have the second operator measure all the samples once in
random order
4: Continue until all operators have measured the samples once
(this is Trial 1)
5: Repeat steps 2-4 for the required number of trials
6: Use the form provided to determine the statistics of the R&R
study
Repeatability
Reproducibility
%R&R
Standard deviations of each of the above
% Tolerance analysis
7: Analyze results and determine follow-up action, if any
31
Other Issues
Number of operators
If process uses multiple operators, choose 2-4 at random
If process uses only one operator, or no operators, perform
study without operator effects (reproducibility effects ignored)
Number of samples
Select enough samples so that
(number of samples) X (number of operators) > 15
If not practical or possible, choose number of trials so that:
if S x O < 4, trials = 6
if S x O < 5, trials = 5
if S x O < 8, trials = 4
32
Gage R&R study report
Objective
Gage Description
Gage study procedures
Results:
Calibration
Accuracy
Precision
Conclusions
Recommendations for improvements
33
Variable Gage R&R
Guidelines
34
Real Life Example
• Remember the Bolt Slot Cap from the Assignment?
• A Gage R&R study is peformed by 3 operators on 10 parts and
resultes are recorded in Bolt_slot_cap_gagerr .mtw
36
Gage R&R study
37
Gage R&R study - improved
38
How a Gage R&R Study Works
– Select units or items for measuring that represent the full range of
variation typically seen in the process
• Measurement systems are often more accurate in some parts of the
range than in others, so you need to test them over the full range
– Have each operator measure those items repeatedly
• In order to use Minitab to analyze the results, each operator must
measure each unit the same number of times
• It is extremely desirable to randomize the order of the units and not let
the operator know which unit is being measured
– Minitab looks at the total variation in the items or units measured
39
How a Gage R&R Study Works, cont.
– Minitab then estimates the proportion of the total variation that is due to
1. Part-to-part variation: physical or actual differences in the units being
measured
2. Repeatability: Inconsistency in how a given person takes the measurement
(lots of inconsistency = high variation = low repeatability)
3. Reproducibility: Inconsistency in how different people take the measurement
(lots of inconsistency = high variation = low reproducibility)
4. Operator-part interaction: An interaction that causes people to measure
different items in different ways (e.g., people of a particular height may have
trouble measuring certain parts because of lighting, perspective, etc.)
– If there is excessive variation in repeatability or reproducibility (relative to part-to-
part variation), you must take action to fix or improve the measurement process
• The goal is to develop a measurement system that is adequate
for your needs
40
Adequate vs. Inadequate Measurement
Systems
Adequate Inadequate
• Most of the variation is accounted for • Variation in how the measurements are
by physical or actual differences taken is high
between the units
– You can’t tell if differences between units
– What Minitab calls part-to-part are due to the way they were measured,
variation will be relatively large or are true differences
– All other sources of variation will – You can’t trust your data and therefore
be small shouldn’t react to perceived patterns,
special causes, etc.—they may be false
– You can have higher confidence that signals
actions you take in response to data are
based on reality
• The measurement system has sufficient • The measurements fall into less than four
precision to distinguish at least four categories
groups or ―categories‖ of measurements
41
Note on Calibrating Measurement
Equipment
• Measurement instruments should only be recalibrated when
they show special cause evidence of drift. Otherwise,
variation could be increased by as much as 40%. This is
because adjusting for true common cause variation adds
more variation (Deming’s rule 2 of the funnel).
42
Checking a Measurement System
for Discrete Data
43
The Inspection Exercise
Task: Count the number of times the 6th letter of the alphabet appears in the
following text.
The Necessity of Training Farm Hands for First
– For use in situations where all that is available is defect data. This can be in
a manufacturing facility or in a business environment to evaluate defects
– Most often this data is collected by people that inspect for quality, and for
this reason it is unreliable.
– Minitab (version 13.31 and later) has the capability to analyze the
reliability of the human inspection process.
– As with the continuous gage study, it will evaluate the repeatability and
reproducibility of inspectors. It will also evaluate how effective the
inspector is as compared to a known standard or attribute.
45
Gage R & R Attribute, Cont
• As with the Continuous study, the Attribute study is a set of
trials where:
46
Planning and Executing the study
• Identify the process to be studied and which attributes to include in the study
• Determine the sample size. Because this is attribute date, a larger sample will
be necessary. A minimum would be 30 samples, but more is better. The larger
the sample size the smaller the confidence intervals in the study. A
determination will need to be made to balance sample size and the
confidence intervals
• Collect a set of samples with an equal number of good and bad products or
any attribute we want to verify, such as call type or error type. Collect samples
that are not obvious as to the correct answer. The study will determine how
well the inspector can make the “tough” decision
• Perform at least two trials (Minitab will accommodate more trials, but two is
standard) having the inspectors “blindly” inspect the samples
47
Known Population Y/N
Sample # Attribute Try 1 Try 2 Try 1 Try 2 Try 1 Try 2
1 A A A A A A A Y
2 D D D D D D D Y
Attribute 3
4
5
A
A
A
A
A
A
D
A
D
A
A
A
A
A
A
A
A
A
A
A
A
N
Y
N
GR&R 6
7
8
A
A
D
A
A
D
A
A
D
A
A
D
A
A
D
A
A
D
A
A
D
Y
Y
Y
9 D D D D D D D Y
10 A A A A A A A Y
11 A A A A A A A Y
12 D D D D D D D Y
13 A D D A A A A N
14 A A A A A A A Y
15 A A A A A A A Y
16 A A A A A A A Y
17 D D D D D D D Y
18 A A A A A A A Y
19 A A A A A A A Y
20 A A A A A A A Y
21 D D D D D D D Y
22 D D D D D D D Y
23 D D D D D D D Y
24 A A A A A A A Y
25 A D D D D D D N
26 D D D D D D D Y
27 A A A A A A A Y
28 A A A A A A A Y
Total Score = 80% 29 A D D A A A D N
30 A D A A A A A N
Appraiser Score: 24/30= 80% 29/30= 97% 28/30= 93% 80%
48
How an Attribute R & R works
• Minitab estimates the following:
– Between Appraisers: the ability of all appraisers to agree with each other
– All Appraisers to the Standard: This is the overall capability of the gage
49
Data for the Study
• Data must be balanced for Minitab – each operator must measure
each unit the same number of times
• Operators should randomly and “blindly” test the units; they should
not know which unit they are measuring when they record the data
50
Minitab Output
51
Action Guidelines
• There are no absolutes for an Attribute R & R. The following are
guidelines as to what actions to take:
– > 90% adequate gage for most applications – no action required unless
team determines it is necessary
• The above are only guidelines. In some situations, teams have been
able to move forward with projects when the gage has been less
than 70%. It is still recommended that efforts be made to improve
the gage for future uses.
52
M & M Exercise for
Attribute Gage R & R
– Conduct an attribute gage R & R study with the
materials ( M & M ‘s ) provided by the instructor.
– Determine who will be the subject matter experts –
Subject matter experts will pick 30 M & M’s for the
exercise – half good & half bad
– Determine the three inspectors
– Use the criteria below for a good M & M
• Has a “shiny” surface
• M is correct and complete
– Conduct the experiment having each inspector
perform the inspection two times
– Collect the data and evaluate the results.
53
Summary
• Attribute Gage R & R can be used to evaluate
a measurement system with Discrete data