Uncertainty Analysis

2141-375
Measurement and Instrumentation
Measurement Error
M
e
a
s
u
r
e
d

v
a
l
u
e
,

x
x
' x
True data
Bias error
Precision
error in x
i
Measurement number
Uncertainty defines an interval about the measured value within
which we suspect the true value must fall
We call the process of identifying and quantifying errors as
uncertainty analysis.
Design-Stage Uncertainty Analysis
Design-stage uncertainty analysis refers to an initial analysis
performed prior to the measurement
Useful for selecting instruments, measurement techniques and
to estimate the minimum uncertainty that would result from the
measurement .
Design-Stage Uncertainty Analysis
%) (
2 2
0
P u u u
c d
+ =
RSS method for combining error
Design-state uncertainty
2 2
0 c d
u u u + =
Design-state uncertainty
Interpolation error
0
u
Instrument error
c
u
Zero-Order Uncertainty (Interpolation Error)
Even when all error are zero, the value of the measurand must be
affected by the ability to resolve the information provided by the
instrument. This is called zero-order uncertainty. At zero-order, we
assume that the variation expected in the measurand will be less than
that caused by the instrument resolution. And that all other aspects of
the measurement are perfectly controlled (ideal conditions)
(95%) resolution 2 / 1
0
= u
Design-Stage Uncertainty Analysis
Instrument Uncertainty, u
c
This information is available from the manufacturers
catalog
x
y
resolution
y
o
uncertainty
1/2 resolution
Design-Stage Uncertainty Analysis
0-1000 cm H
2
O
15 V dc
0-5 V
0-50
o
C nominal at 25
o
C
0.5%FSO
Less than 0.15%FSO
0.25%of reading
0.02%/
o
C of reading from 25
o
C
0.02%/
o
C FSO from 25
o
C
Operation
Input range
Excitation
Output range
Temperature range
Performance
Linearity error e
L
Hysteresis error e
h
Sensitivity error e
S
Thermal sensitivity error e
ST
Thermal zero drift e
ZT
Specifications: Typical Pressure Transducer
The root of sum square approach:
2 2
3
2
2
2
1 n rss
e e e e e L + + + = (95%)
Example: Consider the force measuring instrument described by the catalog data that follows.
Provide an estimate of the uncertainty attributable to this instrument and the instrument design
state uncertainty.
Known: Instrument specifications
Solution:
Assume: Values representation of instrument 95% probability
Design-Stage Uncertainty Analysis
Force measuring instrument
Resolution: 0.25 N
Range: 0 - 100 N
Linearity: within 0.20 N over range
Repeatability: within 0.30 N over range
Design-state uncertainty
2 2
0 c d
u u u + =
Design-state uncertainty
0
u
c
u
Resolution = 0.125 N
N 36 . 0 3 . 0 2 . 0
2 2 2 2
= + = +
r l
e e
N 38 . 0 36 . 0 125 . 0
2 2
= + =
d
u
Example: A voltmeter is to be used to measure the output from a pressure transducer that outputs
an electrical signal. The nominal pressure expected will be ~3 psi (3 lb/in2). Estimate the design-
state uncertainty in this combination. The following information is available:
Known: Instrument specifications
Solution:
Assume: Values representation of instrument 95% probability
Design-Stage Uncertainty Analysis
Voltmeter
Resolution: 10 V
Accuracy: within 0.001% of reading
Transducer
Range: 5 psi
Sensitivity: 1 V/psi
Input power: 10 Vdc 1%
Output: 5 V
Linearity: within 2.5 mV/psi over range
Repeatability: within 2 mV/psi over range
Resolution: negligible
Design-Stage Uncertainty Analysis
Design-state uncertainty
( ) ( )
2 2
P
d
E
d d
u u u + =
Design-state uncertainty
Design-state uncertainty
( ) ( ) ( )
2 2
0
E
c
E E
d
u u u + =
Design-state uncertainty
Design-state uncertainty
( ) ( ) ( )
2 2
0
P
c
P P
d
u u u + =
Design-state uncertainty
Error Propagation
Computation of the overall uncertainty for a measurement system consisting
of a chain of components or several instruments
Let R is a known function of the n independent variables x
i1
, x
i2
, x
i3
, , x
iL
) , , , (
2 1 L
x x x f R K =
L is the number of independent variables. Each variable contains some
uncertainty (u
x1
, u
x2
, u
x3
,, u
xL
) that will affect the result R.
%) ( ' P u R R
R
=
Application of Taylors expansion gives, (neglect the higher order term)
xL
L
x x
L xL L x x
u
x
f
u
x
f
u
x
f

x x x f u x u x u x f R R

+ +

+ =
...
) ,..., , ( ) ,..., , (
2
2
1
1
2 1 2 2
1
1
The best estimate value, R
Where ) ,..., , (
2 1 L
x x x f R =
Error Propagation
( ) %) (
1
2
2 2
2
2
2
1
1
P u
u
x
f
u
x
f
u
x
f
u
L
i
xi i
xL
L
x x R

=
=
|
|

\
|

+ +
|
|

\
|

+
|
|

\
|

K
The combination of uncertainty of all variables (probable estimate of u
R
)
Where
i
is the sensitivity index relate to the uncertainty of x
i
i
i
x
f

=
Example: For a displacement transducer having a calibration curve y = KE, estimate the
uncertainty in displacement y for E = 5.00 V, if K = 10.10 mm/V with u
k
= 0.10 mm/V and u
E
=
0.01 V at 95% confidence
Solution: Find u
y
Error Propagation
Known: y = KE
E = 5.00 V u
E
= 0.01 V
K = 10.10 mm/V u
k
= 0.10 mm/V
( ) ( )
2 2
K K E E y
u u u + =
K
E
y
E
=

= E
K
y
K
=

=
u
E
= 0.01 V
u
K
= 0.10 mm/V
y y
u KE u y y = = '
( ) ( )
( ) ( ) mm 51 . 0 mm/V 10 . 0 V 5 V 01 . 0 mm/V 10 . 10
2 2
2 2
= + =
+ =
K E y
Eu Ku u
Sequential Perturbation
A numerical approach can also be used to estimated the propagation of
uncertainty. This refers to as sequential perturbation. This method is
straightforward and uses the finite difference to approximate the
derivatives (sensitivity index)
1) Calculate the average result from the independent variables
) ,..., , (
2 1 L
x x x f R =
2) Increase the independent variables by their respect uncertainties
and recalculate the result based on each of these new values. Call
these values
+
i
R
) ,..., , (
) ,..., , (
), ,..., , (
2 1
2 2 1 2
2 1 1 1
L L L
L
L
u x x x f R
x u x x f R
x x u x f R
+ =
+ =
+ =
+
+
+
3) Decrease the independent variables by their respect uncertainties
and recalculate the result based on each of these new values. Call
these values

i
R
Sequential Perturbation
4) Calculate the difference for each element
R R R
R R R
i i
i i
=
=

+ +

5) Finally, evaluate the approximation of the uncertainty contribution from

each variables
i i
i i
i
u
R R
R


+
=
+
2
The uncertainty in the result
( )
2 / 1
1
2
(

=

=
L
i
i R
R u
) ,..., , (
) ,..., , (
), ,..., , (
2 1
2 2 1 2
2 1 1 1
L L L
L
L
u x x x f R
x u x x f R
x x u x f R
=
=
=

Example: For a displacement transducer having a calibration curve y = KE, estimate the
uncertainty in displacement y for E = 5.00 V, if K = 10.10 mm/V with u
k
= 0.10 mm/V and u
E
=
0.01 V at 95% confidence
Solution: Find u
y
Error Propagation
Known: y = KE
E = 5.00 V u
E
= 0.01 V
K = 10.10 mm/V u
k
= 0.10 mm/V
i
u
i
x
i
+u
i
x
i
-u
i
R
i
+
R
i
-
R
i
+
R
i
-
R
i
1 E 5 0.01 5.01 4.99 50.60 50.40 0.10 -0.10 0.10
2 K 10.1 0.1 10.20 10.00 51.00 50.00 0.50 -0.50 0.50
x
i
( ) ( )
2 2
K E y
R R u + =
y y
u KE u y y = = '
( )( ) mm 50 . 50 5 10 . 10 = = = KE y
Steps in measurement process
1) Calibration
2) Data-acquisition
3) Data-reduction (Analysis)
Error Sources
Calibration
error
K
, 12 , 11
e e
Data-acquisition
error
K
, 22 , 21
e e
Data-reduction
error
K
, 32 , 31
e e
e
ij
i = Error source group
i = 1 for Calibration Error
i = 2 for Data-acquisition Error
i = 3 for Data-reduction Error
j = Elemental error
Calibration Error Source Group
Element (j) Error Source
1 Calibration curve fit
2 Truncation error
Etc.

Data-Acquisition Error Source Group
Data-Reduction Error Source Group
Element (j) Error Source
1 Measurement system operating conditions
2 Sensor-transducer stage (instrument error)
3 Signal conditioning stage (instrument error)
4 Output stage (instrument error)
5 Process operating conditions
6 Process installation effects
7 Environmental effects
8 Spatial variation error
9 Temporal variation error
Etc.

Element (j) Error Source
1 Primary to interlab standard
2 Interlab to transfer standard
3 Transfer to lab standard
4 Lab standard to measurement system
5 Calibration technique
Etc.

Multiple-Measurement Uncertainty Analysis
The procedure for a multiple-measurement uncertainty analysis
Identify the elemental errors in each of the three source groups
(calibration, data acquisition, and data reduction)
Estimate the magnitude of bias and precision error in each of the
elemental errors
Estimate any propagation of uncertainty through to the result
Calibrate
e
11
, e
12
,...
Data acquisition
e
21
, e
22
,...
Data reduction
e
31
, e
32
,...
e
1j
=P
1j
+B
1j
e
2j
=P
2j
+B
2j
e
3j
=P
3j
+B
3j
This section develops a method for the estimate of the uncertainty in the
value assigned to a measured variable based on repeated measurements
Multiple-Measurement Uncertainty Analysis
Consider the measurement of variable, x which is subject to elemental
precision errors, P
ij
and bias, B
ij
in each of three source groups. Let i = 1, 2,
3 refer to the error source groups ( calibration error i = 1, data acquisition
error i = 2, data-reduction i = 3) and j = 1,2,,K refer to each of up to any K
error elements of error e
ij
Source Precision index P
i
| |
2 / 1
2 2
2
2
1
...
ik i i i
P P P P + + + =
Source Bias limit B
i
| |
2 / 1
2 2
2
2
1
...
ik i i i
B B B B + + + =
| |
2 / 1
2
3
2
2
2
1
B B B B + + =
Measurement Precision index P
| |
2 / 1
2
3
2
2
2
1
P P P P + + =
Measurement Bias limit B
3 , 2 , 1 = i
3 , 2 , 1 = i
Multiple-Measurement Uncertainty Analysis
The measurement uncertainty in x, u
x
( ) (95%) ,
2
95
2
P t B u
v x
+ =
The degrees of freedom, v (Welch-Satterthwaite formula)
( )

= =
= =
|
|

\
|
=
3
1 1
4
2
3
1 1
2
/
i
K
j
ij ij
i
K
j
ij
v P
P
v
Multiple-Measurement Uncertainty Analysis
Measurand, x
Identify elemtal errors
in measurement, e
ij
e
ij
=P
ij
+B
ij
Source precision index, P
i
Source bias limit, B
i
Measurement bias limit, B Measurement precision index, P
Measurement uncertainty, u
x
| |
2 / 1
2 2
2
2
1
...
ik i i i
B B B B + + + = | |
2 / 1
2 2
2
2
1
...
ik i i i
P P P P + + + =
| |
2 / 1
2
3
2
2
2
1
P P P P + + = | |
2 / 1
2
3
2
2
2
1
B B B B + + =
( ) | | (95%)
2 / 1
2
95 ,
2
P t B u
v x
+ =
Measurement bias limit, B Measurement precision index, P
Measurement uncertainty, u
x
Example: After an experiment to measure stress in a load beam, an uncertainty analysis reveals
the following source errors in stress measurement whose magnitude were computed from
elemental errors
B
1
= 1.0 N/cm
2
B
2
= 2.1 N/cm
2
B
3
= 0 N/cm
2
P
1
= 4.6 N/cm
2
P
2
= 10.3 N/cm
2
P
3
= 1.2 N/cm
2
v
1
= 14 v
2
= 37 v
3
= 8
If the mean value of the stress in the measurement is 223.4 N/cm
2
, determine the best estimate of
the stress
Solution: Find u

Known: Experimental error source indices
Assume: All elemental error have been included
Multiple-Measurement Uncertainty Analysis
| |
2 / 1
2
3
2
2
2
1
P P P P + + =
| |
2 / 1
2
3
2
2
2
1
B B B B + + =
( ) | | (95%)
2 / 1
2
95 ,
2
P t B u
v x
+ =
Propagation Uncertainty Analysis to a result
The measurement uncertainty, u
R
The degrees of freedom, v
| |
| | { }

=
=
|

\
|
=
L
i
xi xi i
L
i
xi i
v P
P
v
1
4
2
1
2
/

( ) (95%) ,
2
95
2
R v R R
P t B u + =

=
=
L
i
xi i R
P P
1
2
] [

=
=
L
i
xi i R
B B
1
2
] [
where
Consider the result, R which is determined from the function of the n independent
variables x
i1
, x
i2
, x
i3
, , x
iL
%) ( ' P u R R
R
=
Example: The density of a gas, , which is believed to follow the ideal gas equation of state, =
p/RT, is to be estimated through separate measurements of pressure, p, and temperature, T. the
gas is housed with in a rigid impermeable vessel. The literature accompanying the pressure
measurement system states an accuracy to within 1% of the reading an that accompanying the
temperature measuring system suggest 0.6
o
R. Twenty measurements of pressure, N
p
= 20, and
ten measurements of temperature, N
T
= 10, are made with the following statistical outcome:
Where psfa refers to lb/ft
2
absolute. Determine a best estimate of the density. The gas constant is
R = 54.7 ft lb/lb
m
o
R
Solution: Find
Known:
Assume: Gas behaves as an ideal gas
Propagation Uncertainty Analysis to a result
psfa 91 . 2253 = p psfa 21 . 167 =
p
S
R 4 . 560
o
= T
R 0 . 3
o
=
T
S
T p
S T S p , , ,
R lb/lb ft 7 . 54 /
o
m
= = R RT P

u + = '
Propagation Uncertainty Analysis to a result
( ) | | (95%)
2 / 1
2
95 ,
2
P t B u
v
+ =

( ) ( )
2 2
T T p p
B B B + =
( ) ( )
2 2
T T p p
P P P + =
( ) ( ) | |
( ) ( )
T T T p p p
T T p p
v P v P
P P
v
/ /
4 4
2
2 2


+
+
=
where
RT p
p
1
=

R lb/lb ft 7 . 54 /
o
m
= = R RT P
2
RT
p
T
T
=

where