Measurement System Analysis (MSA)

Validating
g yyour measurement systems
y
Measurement Uncertainty

LSL USL

Measurement
point pass or
reject?

• Imagine you measure a component and find it to be at the point

shown within the specification limits
• Would you pass or reject the component?
Measurement Uncertainty

LSL USL

Measurement
point pass or
reject?

• You now discover that is has a measurement error associated with it

as shown by the error bar above
• Would you pass or reject the component?
Main sources of variation
• Materials
• M th d
Methods
• Machines
• People
• Environment
Measurement System Analysis - MSA
• A scientific and objective method of analysing the validity
of a measurement system
system.
• A tool which quantifies
– Equipment
q p variation
– Appraiser (operator) variation
– The total variation of a measurement system
• Examples
E amples of meas
measurement
rement ssystems
stems
– Micrometer
– Shadowgraph
g p
– Go/no-go gage
– Data collection form
– S r e
Survey
– On-time delivery report
 Variation
Part Measurement Total
Variability Variability Observed Variability

Process A

σ²Actual (Part) + σ²Meas. syst. = σ²Observed Total

Process B

Which process is best? Which is easier to improve?

Components of Measurement Error

1. Resolution/Discrimination
2. Accuracy (bias)
3. Linearity
4. Stability (consistency)
5. Precision – Repeatability & Reproducibility
1. Resolution
What is the length of this
component to the nearest 10th?
• Definition:
D fi iti Th
The capability
bilit tto
detect the smallest acceptable
change 1         2        3        4
• C
Cause: IInadequate
d t Poor resolution
measurement units
• Guideline “10 bucket rule” 1        2         3        4
– Increments in the measurement Better resolution
system should be one tenth of the
product specification or process
variation What percent of requests are issued
within 4 hours of receipt??
• Actions: Customer Name ________________
– Change measurement device Date Received ________________
Date Issued ________________
– Record sample averages
Poor resolution
– Live with it but understand its
Customer Name ________________
limitations
Date Received ________________
Time Received ________________
R
Resolution
l ti 1/10th Date Issued ________________
Time Issued ________________

Better resolution
2. Accuracy /Bias
Master value
Reference standard

• Accuracy/Bias:
A /Bi Difference
Diff
between the observed average
value and the master reference X X
X X
• Master value is an accepted, X
X
traceable reference standard X X
X X
• Actions: Less accurate
– Calibrate regularly
Master value
– Use operations instructions Reference standard
– Review
R i specifications
ifi ti ffor
resolution X X
– Validate data systems input X X
accuracy
X
X X X
– Create operational definitions X X
More accurate
3. Linearity

• Measurement is
M i ““true”” and/or
d/

Temperature deg C
consistent across the range of
the “gauge”
• Actions Linear
– Check gauge specification range
– Rebuild/replace gauge
Gauge reading deg C
– Use within restricted range
– Use correction factor Form Linearity
Super Outstanding 10
Outstanding 9
Incredible 8
Excellent 7
Great 6
Very Good 5
Good 4
OK 3
Fair 2
Poor 1
4. Stability
Time 1 Time 2

• Measurements remain
M i
constant and predictable over Shift

time i.e. accuracy remains

constant
• No drifting, sudden shifts or
cycles
• Actions
– maintain and service
equipment
– use control charts
– use SOP
– ensure adequate training
– regular audits
5. Precision – Repeatability & Reproducibility
Master value
Reference standard

• Repeatability
R t bilit - variation
i ti ththatt
occurs when repeated X X
measurements are made of the X X
same item under
nder identical X X X X
conditions X X
• Actions
Poor precision
– repair, replace, adjust
– SOP Master value
Reference standard
• Reproducibility – variation that
results when different conditions
are used to make the same
measurements XXXXXXX
XXX
• Actions
– training
– SOP
Good precision
Gauge R&R Studies
• Method of assessing Repeatability & Reproducibility of a
measurement system
• A number of appraisers (usually two or three) measure a
number of parts (or process output) (usually 5 to 20) a
number of times (usually two or three)
• The results are compared within each appraiser
(Repeatability) and between appraisers (Reproducibility)
• Randomisation is critical for repeat
p measurements to
avoid learning or copying.
Gage R&R Study – Continuous Data
Three appraisers, two 
measurements each
measurements each

% of Tolerance

% of Total Variation

EV=Equip’t Variation
AV=Appraiser Variation
PV=Part Variation
R&R = √(EV²+AV²)
Gage R&R Study – Attribute Data

Two appraisers, two 
measurements each h

Appraiser Variation

Appraiser vs Control

AV=Appraiser Variation
AC=Appraiser vs Control
R&R = √(AV*AC)
 Measurement Error Matching Exercise
Time 1
Time 2
A. B. X X
Shift
1. Resolution/Discrimination X X
X
X X X
2. Accuracyy (bias)
( ) X X

3. Linearity

4. Stability (consistency)

5. Precision – Repeatability &

Reproducibility

deg C
C
C. D
D. E
E.

mperature d
XXXXXXX
XXX

Tem
1        2         3        4

Gauge reading deg C