f5763552 Microsoft Word - Uncertainty Analysis Guideline 2012d
f5763552 Microsoft Word - Uncertainty Analysis Guideline 2012d
f5763552 Microsoft Word - Uncertainty Analysis Guideline 2012d
Again, design-stage uncertainty estimate is intended only as a guide for selecting instrumentation before a test, and is
never used for reporting results. More detailed analysis is possible once actual measurements are taken.
Example 1: Consider a force-measuring instrument being considered that has the following catalog data.
Range: 0-100N
Linearity: within 0.2 N over the range
Hysterisis: within 0.3 N over the range
Resolution: 0.25 N
In this problem three instrument errors exist. Resolution, Linearity, and Hysteresis. Combine these errors using the RSS method.
There is no need to differentiate between systematic and random error during design-stage uncertainty analysis.
U
x,design
=(0.125
2
+0.2
2
+0.3
2
)
0.5
= 0.4 N
Note: Interpolation errors are estimated from the given resolution, 0.25N. The interpolation error is the resolution.
2.7 Combining Systematic Errors for a Measurement
Systematic error is the portion of the total error that remains constant in repeated measurements throughout the conduct of
a test. A systematic error may cause either a high or a low offset. Systematic error is usually derived from calibration
information (e.g. accuracy, linearity, hysteresis, offset, and known environmental errors). The total systematic uncertainty
is estimated using the root-sum-squares (RSS) method to combine the individual error sources.
The systematic uncertainty, B, is assumed to have 95% CI. These are the values of systematic errors that are often
reported in literature or in instrumentation calibrations. The value b is the estimate of the systematic standard
uncertainty which has a confidence interval of 1 standard deviation (68%).
MEAUREMENT
SYSTEMATIC STANDARD
UNCERTAINTY
/
where
K = Total number of systematic error sources
b
k
= Estimate of the systematic error of the k
th
elemental error.
7
It is worth noting that the uncertainty depends on the squares of the individual uncertainties. If the uncertainty from one
source is much larger than the others then the larger uncertainty values dominate the values. Many times, the other
smaller sources of uncertainty can be neglected.
2.8 Combining Random Errors for a Measurement
Estimates for random error are derived using repeated measurements and statistical methods. The random uncertainty is
determined from the standard deviation of the mean. The standard deviation of the mean is used with the assumption the
errors are normally distributed. The interval has a confidence of one standard deviation, equivalent to a probability of
68% for a population. The random uncertainty at a desired confidence level is defined by the interval
,
.
RANDOM UNCERTAINTY IN
A MEASUREMENT
S
X
=
s
x
N
where
s
x
= standard deviation
N = number of samples or readings
In the event that you have K random errors for a measurement then S
x
is estimated by the RSS method;
S
X
= (S
x,k
)
2
k=1
K
|
\
|
1/2
where each elemental source of random error may have its own sample size.
S
x,k
=
S
x,k
N
k
2.9 MultipleMeasurement Uncertainty Analysis
This section presents a method for estimating the uncertainty in the value assigned to a variable. The procedure assumes
that the errors follow a normal probability distribution and that there are sufficient repetitions to assess random error. The
total uncertainty in a measurement is then determined by combining the total systematic error and random error for a
measurement in the following manner.
MEASUREMENT UNCERTAINTY
The degrees of freedom can be calculated exactly using methods described in ASME PTC 19.1 and using the combined
degrees of freedom equation shown below. However, a value of 2 is assumed for the t-value in this course to simplify the
calculation method. The combined degrees of freedom,, can be calculated using the following equation.
8
Example 2 Ten repeated measurements of force are obtained from a force transducer that has two sources of systematic
error. B
1
=0.01 N and B
2
=0.15 N. Estimate the random, systematic, and total measurement uncertainty.
Find the systematic error, random error, and total error, U
X
, in the measurement
.
Express the answer in both absolute and
relative terms. Recall, B=2b.
b
X
=(0.05
2
+0.075
2
)
1/2
=0.09 N
S
X
=
s
x
N
=
3.04
10
= 0.96N
2 0.09
0.96
1.92
X=120 2 N at a 95% Confidence
or
X=120 1.7% Relative Uncertainty at a 95% Confidence
Reading Force, N Systematic Error Force, N
1 123.2 b1 0.05
2 115.6 b2 0.075
3 117.1
4 125.7 Total Systematic
5 121.1 Btotal 0.09
6 119.8
7 117.5
8 120.6
9 118.8
10 121.9
Average 120.13
Std Dev 3.04
Count 10
DOF 9
t value 2.2621572
sx_bar 0.96
Measurement Uncertainty 1.9
9
Example 3. A calorimeter is used to measure the heating values of samples. The manufacturer of the calorimeter states
that the device has an accuracy of 1.5% of the full-scale range where the full-scale range is 0 to 100,000 kJ/kg. The
measured heating values for 10 samples are:
Sample Heating Value, kJ/kg
1 48530
2 48980
3 50210
4 49860
5 48560
6 49540
7 49270
8 48850
9 49320
10 48680
Calculate the following:
(a) The systematic and random uncertainty of the mean of the measurements.
a. Systematic error is determined from the accuracy value given in the problem. 1.5% of FS. Meaning, B
X
=100,000*0.015=1500 kJ/kg.
b. Random error is determined using statistics. Calculate the standard deviation, standard deviation of the
means. The standard deviation of the means for the data above is: s
x
=179 kJ/kg
(b) The total uncertainty of the mean value using a 95% confidence interval.
a. The total uncertainty is determined using the equation:
2 750
179
1542
X=49200 1500 kJ/kg at a 95% Confidence
Reading Heating kJ/kg Systematic Error Heating kJ/kg
1 48530 b1 750.00
2 48980
3 50210
4 49860 Total Systematic
5 48560 Btotal 750.00
6 49540
7 49270
8 48850
9 49320
10 48680
Average 49180
Std Dev 566.31
Count 10
DOF 9
t value 2.262157163
sx_bar 179.08
Measurement Uncertainty 1542.2
10
3 Uncertainty of a Result
The uncertainty of a result is not obtained directly from measurements. Instead, the individual measurements and
associated individual measurement uncertainties are used to determine the uncertainty in the result using propagation of
error techniques. In general, the uncertainty of a result, R, is a function of k measured variables, x
1
, x
2
, x
3
,..x
k
;. or
R=f(x
1
, x
2
, x
3
,..x
k
)
The root sum of squares (RSS) method is used to propagate the individual measurement uncertainties into the result
uncertainty. The partial derivative of the result, with respect to each measured variable, is used along with the uncertainty
in the measured variable to determine the overall uncertainty in the result using the following equation. Another common
form of the RSS equation utilizes the sensitivity coefficient, , which is defined as the instantaneous rate of change in a
result to a change in a parameter.
U
R
=
R
x
1
U
x
1
|
\
|
2
+
R
x
2
U
x
2
|
\
|
2
+ ... +
R
x
k
U
x
k
|
\
|
2
(
(
where
U
R
= Uncertainty in the result, either systematic or random
U
x
k
= Uncertainty in the measured parameter, x
k
, either systematic or random
A special case formulation of the above RSS equation exists for cases where the result is dependent only on the product
or quotient of the measured variables. This form of the equation is typically easier to use but both methods, the partial
derivative method and special case formulation, yield the same result.
R=C x
1
a
, x
2
b
, x
3
c
,..x
k
n
U
R
R
=
aU
x
1
x
1
|
\
|
|
2
+
bU
x
2
x
2
|
\
|
|
2
+ ..... +
nU
x
k
x
k
|
\
|
|
2
All measurements and the associated error must be independent of each other. That is, an error in one variable must not
correlate with an error in another variable. If the variables are not independent, then an alternative approach is required
(this alternative approach is not discussed in this course but see the listed references for more information on these
alternate approaches).
In either method, the partial derivative or the special case formulation, the uncertainty estimates are typically computed
separately for the systematic error and random error components and then combined at the end to yield an estimate of the
total uncertainty in the result. However, the effect of combining the systematic and random errors earlier in the
calculation does not affect the final total uncertainty in the result.
11
The following examples demonstrate the RSS method and are not concerned with keeping track of the systematic and
random error components. Other examples will follow that demonstrate the handling of these errors separately.
Example 4. The measurement of power consumption in a simple resistive circuit is determined using the equation,
P=V*I. The voltage, V, and current, I, are the measured variables in this problem and the result is the power. The
measured quantities and associated uncertainty are:
V=100 3 V (relative uncertainty=3/100=3%)
I=10 0.2 A (relative uncertainty=0.2/10=2%)
Confidence levels of both measurements are assumed to be the same (i.e. @95% CI)
Both the partial derivative method and the special form method are shown below. Both methods yield the same results.
In both cases the power measurement is reported as: P= 1000 40 W (95% CI).
Example 5: A pressurized air tank is nominally at ambient temperature (25C). How accurately can the density be
determined if the temperature is measured with an uncertainty of 2C and the tank pressure is measured with an accuracy
of 1%? The data reduction equation solve for pressure is:
=
P
RT
U
P
/P=0.01
U
T
/T=2/298=0.00671 (Need to covert to the appropriate units firstIn this case degrees Kelvin)
Assume no uncertainty in the gas constant R.
U/=?
Use the relative uncertainty form of the RSS equation to solve this problem.
U
=
U
T
T
|
\
|
|
2
+
U
P
P
|
\
|
|
2
= 0.00671 ( )
2
+ 0.01 ( )
2
= 0.012 =1.2%
12
Example 6: The overall heat transfer coefficient U of a system of two fluids separated by a wall of negligible thermal
resistance is determined using the relationship:
U =
h
1
h
2
h
1
+ h
2
where h
1
and h
2
are the individual heat transfer
coefficients for the two fluids.
If h
1
=15 W/m
2
with an error of 5% and h
2
=20 W/m
2
with an error of 3%, what will the error be in U? Meaning
U
h1
=0.05*15=0.75 W/m
2
and U
h2
=0.03*20=0.60 W/m
2
.
The nominal value of U=8.6 W/m
2
This problem requires the long version of the RSS equation. Meaning, the partial derivatives must be calculated.
If
f (x) =
h(x)
g(x)
Then
d
dx
f (x) = f (x) =
h (x)g(x) h(x) g (x)
g(x)
2
therefore
U
h
1
=
h
2
(h
1
+ h
2
) h
1
h
2
h
1
+ h
2
( )
2
=
h
2
2
h
1
+ h
2
( )
2
= 0.32653
U
h
2
=
h
1
(h
1
+ h
2
) h
1
h
2
h
1
+ h
2
( )
2
=
h
1
2
h
1
+ h
2
( )
2
= 0.183673
Now insert the partial derivatives and uncertainty of each variable into the RSS equation.
U
U
=
U
h
1
U
h
1
|
\
|
|
2
+
U
h
2
U
h
2
|
\
|
|
2
(
(
(
or
U
U
= 0.32653 0.75 ( )
2
+ 0.183673 0.60 ( )
2
[ ]
= 0.27W / m
2
Report value as (8.6 0.3) W/m
2
Converting the absolute uncertainty to relative uncertainty is then 0.27/8.6=3%.
13
Example 7: Consider a counter-flow heat exchanger where there is a solution of ethylene glycol on the cold side of the
heat exchanger and pure water is on the hot side of the heat exchanger. Thermal losses are assumed to be negligible.
Therefore it is reasonable to expect that the observed energy exchange on both sides of the heat exchanger will be equal.
The heat exchange to the cold stream, Qc, should equal that from the hot stream, Qh, within the uncertainty of each
quantity. Is this the case?
A listing of all the measured values and corresponding uncertainty is presented in the following table.
Side Parameter Nominal Value Uncertainty
Cold
c
p1
(J/kg-C) 4186 5%
& m
1
(kg/s)
60 10%
T
1
(C) 1.1 0.5 C
T
2
(C) 7.8 0.5 C
Hot
c
p3
(J/kg-C) 3768 5%
& m
3
(kg/s)
56 10%
T
3
(C) 12.8 0.5 C
T
4
(C) 4.4 0.5 C
Hot and cold side energy balances are not identical but they overlap when their respective uncertainties are included.
Cold Side Range=1424 to 1942 kW. Hot Side Range=1624 to 2020 kW.
Hot Side
Cold Side
3 4
2 1
Partial Derivative Method
Measurement, x Units Average x Relative Ux Absolute Ux Partial, dR/dx Numerical Partial, dR/dx dR/dx* Ux
Specific Heat j/kg-C 4186 5% 209.3 m*(T1-T2) -402 -84138.6
Mass Flow Rate kg/s 60 10% 6 cp*(T1-T2) -28046.2 -168277.2
T1 C 1.1 0.5 m*cp 251160 125580
T2 C 7.8 0.5 m*cp 251160 125580
Calculated Result
Qc kW 1683
Total U
R
258.7 Watts
Relative Uncertainty 15.4%
Partial Derivative Method
Measurement, x Units Average x Relative Ux Absolute Ux Partial, dR/dx Numerical Partial, dR/dx dR/dx* Ux
Specific Heat j/kg-C 3768 5% 188.4 m*(T1-T2) 470.4 88623.36
Mass Flow Rate kg/s 56 10% 5.6 cp*(T1-T2) 31651.2 177246.72
T3 C 12.8 0.5 m*cp 211008 105504
T4 C 4.4 0.5 m*cp 211008 105504
Calculated Result
Qh kW 1772
Total U
R
248.1 Watts
Relative Uncertainty 14.0%
14
3.1 Single Measurement Experiments
An experiment is considered to be a single measurement experiment when the result is run only once at each test
condition. In this case, there are not sufficient test results to calculate the standard deviation of the result. Estimates for
the random uncertainty need to be obtained through other sources, such as manufactures data or auxiliary tests.
Estimates for systematic uncertainty are obtained using the equation:
/
The random portion of the uncertainty in the result is also obtained using a similar equation.
s
R
=
R
x
i
s
X.i
|
\
|
2
i=1
n
1/2
The total uncertainty in the result is again estimated using the RSS equation.
The degrees of freedom can be calculated exactly using methods described in ASME PTC 19.1 and using the combined
degrees of freedom equation shown below. However, a value of 2 is assumed for the t-value in this course to simplify the
calculation method. The combined degrees of freedom,, can be calculated using the following equation.
15
3.2 Uncertainty in Result for Multiple Measurement Experiments
In experimental situations where a test is repeated multiple times to determine M values of the result, R
1
, R
2
, R
M
a
slightly different approach is required to estimate the experimental uncertainty.
The bias error is estimated using the RSS equation or the special function form of the equation.
/
In the multiple measurement case, the average result, R , and standard deviation of the result,
,
is computed using the
multiple result values. The random uncertainty component of the result is estimated using the equation for the standard
deviation of the means.
s
R
=
s
R
M
=
standard deviation of Result values
Number of Result values
The total estimated uncertainty in the result value is again determined by combining the systematic uncertainty and
uncertainty using the RSS method. Degrees of freedom, , is determined from the number of result values, M-1.
16
Example 8. The thermal efficiency of a natural gas engine is determined using the equation
=
P
m
f
HV
Where P is the power output in kW, m
f
is the natural gas flow rate in kg/s, and HV is the heating value of natural gas in
kJ/kg. Average values of P, m
f
, and HV are 50kW, 0.2 kg/minute, and 49180 kJ/kg, respectively. The systematic
uncertainties are given as 0.2 kW, 0.003 kg/minute, and 1500 kJ/kg, respectively.
To establish the mean value for the efficiency of engines from a production line, five engines are tested under similar
conditions. The results for the calculated resulting efficiency are 31.0, 30.5, 30.8, 30.6 and 30.2 percent. The mean value
of the result is 30.6 percent and the standard deviation is 0.303 percent. Meaning, the random portion of the uncertainty in
the result is then,
s
R
=
0.00303
5
= 0.00136
Bias error is estimated using the special function form of the RSS equation because the result is a product/quotient of the
three measured variables.
The value for the efficiency is 0.306 (30.6%).
0.1
50
0.0015
0.2
750
49180
0.0171
0.0171*.306=0.00523
Total Uncertainty in the Efficiency Measurement is then:
2 0.00523
0.00136
0.011 1%
Efficiency is reported as: 31 1% (Relative Uncertainty ~3.5%)
(Systematic uncertainty clearly the dominant source of error perhaps instrumentation improvements could be made)
17
3.3 Summary of Procedure for Uncertainty Analysis and Error Propagation
Uncertainty analysis provides the values on your test results. Performing uncertainty analysis on tests can be
complicated at times and requires a consistent approach and systematic bookkeeping. The following steps summarize the
activities in performing an uncertainty analysis.
(1) Define the measurement process. Identify all measured values and the relationship between the measured
values and the test results.
(2) List all elemental sources. Make a complete list of all possible error sources for each measured parameter.
It is sometimes helpful to group uncertainties into categories based on their source (e.g. calibration, data
acquisition, etc.)
(3) Estimate elemental errors. Estimate the systematic and random errors. If data is available to estimate the
standard deviation of a parameter, or the error is known to be random in nature, then it should be treated as
random uncertainty; otherwise, classify it as a systematic uncertainty.
(4) Calculate the systematic and random uncertainty of each measured variable. Total/sum the systematic
and random uncertainties for each of the measurement variables.
(5) Propagate the systematic uncertainties and random uncertainties all the way to the result(s). Utilize the
RSS equation to propagate the systematic and random uncertainties of the measured variables to the final test
results. Keeping the systematic errors and random errors separate allows you to determine which type of
error is dominating the measurement uncertainty.
(6) Calculate the total uncertainty of the result(s). Utilize the appropriate RSS equation (for Multiple
Measurements or Single Measurements) to combine the systematic and random errors to obtain the total
uncertainty in the result.
18
Example 9: The efficiency of a pump is determined using the formula
pump. the into input power = W
pump the of outlet and inlet e between th al differenti pressure = P
rate flow volumetric = Q
efficiency pump
=
=
where
W
P Q
The following equipment is used:
Differential Pressure Gage Range 0-1200 kPa
Accuracy 0.2% of span (includes linearity, hysteresis, and repeatability)
Stability 0.2% of span
Flowmeter Range 1200 LPM (liters per minute)
Accuracy 1.5% of reading
Power Accuracy 0.07 kW
The average values and standard deviations of the means of the measured values are as follows.
Measurement Mean Value Standard Deviation Mean N
Pressure 702 kPa 10 kPa 20
Flow 340 lpm 5.6 lpm 15
Power 5.1 kW 0.15 kW 10
PART 1: Given the repeated measurements for each measured value to calculate the mean and standard deviation determine the
following: (Note: This is a Single Measurement Experiment approach)
(a) Efficiency of the pump
(b) Random and systematic uncertainty of the efficiency of the pump
(c) Total uncertainty of the efficiency of the pump at a 95% confidence interval.
(a) Pump efficiency is calculated by using the given relationship and converting all measured values to the proper units.
=
Q P
W
= 0.780
(b) Systematic and random uncertainty is determined using
B
R
=
R
x
i
B
i
|
\
|
|
2
i=1
n
`
)
1/ 2
and s
R
=
R
x
i
s
X.i
|
\
|
2
i=1
n
1/2
(c) Total uncertainty determined using RSS method.
U
R
= t
,
B
R
2
+ s
R
2
2 B
R
2
+ s
R
2
19
Below are the templates of the solution method and results using both the relative uncertainty method and the partial derivative
method.
Relative Uncertainty Method
Partial Derivative Method
Combining the Systematic and Random Components for Each Measurement Prior to Analysis
In the end, the final answer is the same using any of the above methods. However, only with the first two methods, where the
systematic and random components are kept separate until the end, can you determine which source of uncertainty is dominating the
uncertainty. That is, are systematic errors dominating or random errors?
Efficiency is presented as: 0.78 0.06 at 95% CI
Measurement, x Units Range Average x Bx bx N DOF t-value Sx_bar
bx/X,
Relative
Uncertianty
sxbar/X,
Relative
Uncertianty
Differential Pressure kPa 1200 702 3.39 1.70 20 19 2.09 10 0.002417459 0.014245014
Volumtric Flow L/minute 1200 340 5.1 2.55 15 14 2.14 5.6
Volumtric Flow,Q m^3/s 0.02 0.0056667 9E-05 0.00 15 14 2.14 9E-05 0.0075 0.016470588
Power kW 5.1 0.07 0.04 10 9 2.26 0.15 0.006862745 0.029411765
Calculated Result
Efficiency 0.780 Rel Uncert 0.010449468 Rel Uncert 0.0365958
Total b
R
0.008 Total s
R
0.029
Total Uncertainty 0.059
Total Rel. Uncertainty 7.6%
Systematic Uncertainty Random Uncertainty
Measurement, x Units Range Average x Bx bx N DOF t-value Sx_bar
Partial,dR/d
x
Numeric
Partial dR/dx*bx
dR/dx*sx_b
ar
Differential Pressure kPa 1200 702 3.39 1.70 20 19 2.09 10 Q/W 0.001111 0.001886 0.01111111
Volumtric Flow L/minute 1200 340 5.1 2.55 15 14 2.14 5.6 0
Volumtric Flow,Q m^3/s 0.02 0.0056667 9E-05 0.00 15 14 2.14 9E-05 DP/W 137.6471 0.00585 0.01284706
Power kW 5.1 0.07 0.04 10 9 2.26 0.15 (-Q*DP)/W^2 -0.15294 -0.00535 -0.0229412
Calculated Result
Efficiency 0.780 Total B
R
Total s
R
0.008 0.029
Total Uncertainty 0.059
Total Rel. Uncertainty 7.6%
Measurement, x Units Range Average x Bx bx N DOF t-value Sx_bar Ux
Partial,dR/d
x
Numeric
Partial dR/dx*Ux
Differential Pressure kPa 1200 702 3.39 1.70 20 19 2.09 10 20.286 Q/W 0.001111 0.02253995
Volumtric Flow L/minute 1200 340 5.1 2.55 15 14 2.14 5.6
Volumtric Flow,Q m^3/s 0.02 0.0056667 9E-05 0.00 15 14 2.14 9E-05 0.0002 DP/W 137.6471 0.02823256
Power kW 5.1 0.07 0.04 10 9 2.26 0.15 0.3081 (-Q*DP)/W^2 -0.15294 -0.0471148
Calculated Result
Efficiency 0.780 Total U
R
0.059
Total Uncertainty 0.059
Total Rel. Uncertainty 7.6%
20
3.4 Uncertainty Analysis Using Sequential Perturbation
A numerical approach can also be used to estimate the uncertainty of a result. The method known as sequential
perturbation is often used or preferred when direct differentiation is too cumbersome/complex and/or the number of
variables involved in the calculation of uncertainty is large.
The method utilizes a finite difference method to approximate the derivatives in the RSS equation and can be easily
implemented in a spreadsheet software program or using a hand calculator. The steps in the method are as follows:
1. Calculate the result R
0
=f(x
1
, x
2
, x
3
.x
i
) using the nominal values for all the independent measured variables.
2. Next, increase each independent variable by their respective uncertainties and recalculate the result. Only increase
one variable at a time and leave the other independent variables set to their respective nominal values. These
values are called R
i
+
. That is,
R
1
+
=f(x
1
+U
1
, x
2
, x
3
, x
i
)
R
2
+
=f(x
1
, x
2
+U
2
, x
3
, x
i
)
R
3
+
=f(x
1
, x
2
, x
3
+U
3
, x
i
)
R
i
+
=f(x
1
, x
2
, x
3
, x
i
+U
i
)
3. Next, decrease each independent variable by their respective uncertainties and recalculate the result. Again,
decrease one variable at a time and leave the other independent variables set to their respective nominal values.
These values are called R
i
-
. That is,
R
1
-
=f(x
1
-U
1
, x
2
, x
3
, x
i
)
R
2
-
=f(x
1
, x
2
-U
2
, x
3
, x
i
)
R
3
-
=f(x
1
, x
2
, x
3
-U
3
, x
i
)
R
i
-
=f(x
1
, x
2
, x
3
, x
i
-U
i
)
4. Now calculate the difference R
i
+
and R
i
-
for each independent measurement.
R
i
+
= R
i
+
-R
0
and R
i
-
= R
i
-
-R
0
5. Evaluate the approximated uncertainty for each variable.
R
i
=
R
i
+
R
i
R
x
i
U
i
6. The uncertainty in the result is then
U
R
= R
i
( )
2
i =1
k
`
)
1/ 2
21
Example 10. The measurement of power consumption in a simple resistive circuit is determined using the equation,
P=V*I. The voltage, V, and current, I, are the measured variables in this problem and the result is the power. The
measured quantities and associated uncertainty are:
V=100 3 V (relative uncertainty=3/100=3%)
I=10 0.2 A (relative uncertainty=0.2/10=2%)
1. Calculate the result R
0
=f(x
1
, x
2
, x
3
.x
i
) using the nominal values for all the independent measured variables.
R
1
=V, R
2
=I in this case.
R
0
=V*1=1000
2. Next, increase each independent variable by their respective uncertainties and recalculate the result.
R
1
+
=f(100+3, 10) = f(103, 10) = 1030
R
2
+
=f(100, 10+0.2) = f(100, 10.2) = 1020
3. Next, decrease each independent variable by their respective uncertainties and recalculate the result.
R
1
-
=f(100-3, 10) = f(97, 10) = 970
R
2
-
=f(100, 10-0.2) = f(100, 9.8) = 980
4. Now calculate the difference R
i
+
and R
i
-
for each independent measurement.
R
1
+
= 1030-1000 = 30
R
2
+
= 1020-1000 = 20
R
1
-
= 970-1000 = -30
R
2
-
= 980-1000 = -20
5. Evaluate the approximated uncertainty for each variable.
R
1
=
R
1
+
R
1
2
=
30 (30)
2
= 30
R
V
U
V
R
2
=
R
2
+
R
2
2
=
20 (20)
2
= 20
R
I
U
I
6. The uncertainty in the result is then
U
R
= 30 ( )
2
+ 20 ( )
2
0.5
= 36.06
A copy of a spreadsheet layout for this problem is shown below.
In both cases, Example 4 and Example 10, the power measurement is reported as: P= (1000 40) W (95% CI).
P=V*I
Variable Nominal U_I U+ U- R+ R- dR+ dR- dR
V 100 3 103 97 1030 970 30 -30 30
I 10 0.2 10.2 9.8 1020 980 20 -20 20
P 1000 U
P
36.06