aoe 3054 - uncertaintyaborgolt/aoe3054/classes/foundationsclass3handouts.pdfdetermining uncertainty...
TRANSCRIPT
![Page 1: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/1.jpg)
AOE 3054 - Uncertainty
• Relevant to everything from now on.• See lab manual, “Basic Concepts in
Experiments” and “Experiment 4”
(or, How To Estimate Error)
![Page 2: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/2.jpg)
Accuracy...• Accuracy (same as error):
• Precision
• Repeatability
• Uncertainty
Thoughts
![Page 3: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/3.jpg)
Boeing AH-64A Apache
![Page 4: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/4.jpg)
Uncertainty
• Concept of an uncertainty interval• Determining uncertainty
– in a primary measurement– in a result
• Examples• Using a computer to estimate
uncertainty
![Page 5: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/5.jpg)
Uncertainty Interval• Definition:
The 95% condition is sometimes referred to as “20:1 odds”. Depending on how critical the information being handled is, higher odds are sometimes specified, e.g. 100:1 or 99%
• Representation:-
---- Error bars...
-5
0
5
10
15
20
25
0 1 2 3 4
Distance (cm)
Velo
city
(m/s
)
![Page 6: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/6.jpg)
Determining Uncertaintyin a Primary Measurement
• Primary measurement: E.g. Distance from a dial gage, height from a manometer (not velocity from a Pitot probe)
• Make an informed guess based on :---
-
- other factorsvalidity of measurement schemeoperating outside the design/calibration rangeenvironmental conditionsblunders
- experience
![Page 7: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/7.jpg)
Determining Uncertaintyin a Primary Measurement
Example:
Uncertainty estimate = ??
![Page 8: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/8.jpg)
Determining Uncertaintyin a Primary Measurement
• What you consider an “uncertainty” may not always be an error in the traditional sense. It depends on what you’re hopingto determine from the primary measurement.
E.g. Suppose you aim to determine the mean value of a fluctuating quantity, like the position of a wing spar vibratingin flight, by taking a series of samples…
![Page 9: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/9.jpg)
Determining Uncertaintyin a Derived Result
• Consider an experimental result R that is determined from one or more primary measurements a, b, c… by means of a function f( ), i.e. any derived result.
,...),,( cbafR =• The uncertainty in R due to each of the primary measurements may be determined by multiplying the uncertainty in that measurement by the sensitivity of R to that measurement. Since each of the primary measurements will be independent, the total uncertainty in R is thus,
![Page 10: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/10.jpg)
Example 1: Uncertainty in Velocity from a Pitot-Static
U
patmTatm
Determine the uncertainty in velocity measurements made with the Pitot-static probe and digital manometer used in experiment 3
EquationReadingspo-p = 0.22kPa = 220Papatm = 941.1mBar = 94110PaTatm= 17.0oC = 290.0K
Result
smU /72.19)0.290/287/94110(
2202=
×=
0.22
po-p
![Page 11: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/11.jpg)
Step 1: Primary Uncertainties
Thermometer ?Barometer ?
Manometer ?
p-po δ(p-po) = patm δ(patm) = Tatm δ(Tatm) =
![Page 12: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/12.jpg)
Step 2: Partial Derivatives
)//()(2
atmatm
o
TRpppU −
=
![Page 13: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/13.jpg)
Step 3: Combine Uncertainties
[ ] [ ] [ ]
sm
d
U
/80.0000027.000029.0803.0
50000105.05.00340.0200448.0
)(
222
=++=
×−+×+×=
=δ
000105.0
0340.0
0448.0)(
−=∂∂
=∂∂
=−∂
∂
atm
atm
o
pU
TUpp
U
δ(p-po) = 20Pa δ(patm) = 50Paδ(Tatm) = 0.5oC
![Page 14: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/14.jpg)
Step 4. Interpretation
•
• Since it is (hopefully) a small fraction of the measurement, the accuracy of the uncertainty estimate itself is not that important. (e.g. it wouldn’t change any conclusions here if the uncertainty were 0.6m/s. This is why we can tolerate informed guesswork in determining the uncertainty in the primary measurements.
sm
ppUT
TUpp
ppUU atm
atmatm
atmo
o
/80.0000027.000029.0803.0
)()()()(
)(222
=++=
⎥⎦
⎤⎢⎣
⎡∂∂
+⎥⎦
⎤⎢⎣
⎡∂∂
+⎥⎦
⎤⎢⎣
⎡−
−∂∂
= δδδδ
![Page 15: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/15.jpg)
Example 2: Uncertainty in a Statistical Result
Problem: The deflection ‘y’ of a helicopter rotor blade is measured, once per revolution, over N rotor revolutions. The data is used to estimate the mean and standard deviation σ of the deflection. Assuming the error in the individual deflection measurements is negligible, determine the uncertainty in the mean value.
∑ ==
N
i iyN
y1
1Mean
The HELINOISE Aeroacoustic ROTOR Test in the DNW
40 % geometrically and dynamically scaled highly instrumented main rotor model of the BO-105 helicopter
![Page 16: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/16.jpg)
Step 1: Primary Uncertainties
Step 2: Partial Derivatives
Nyyyy
Ny NN
i iK++
== ∑ =21
1
1
![Page 17: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/17.jpg)
Step 3: Combine Uncertainties
δ(yi) = 2σyNyy
i
1=
∂∂
![Page 18: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/18.jpg)
Step 4: Interpretation
• The uncertainty in the mean value goes as the square root of the number of samples. Most statistical results behave in this way.
![Page 19: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/19.jpg)
Determining Uncertaintyusing a computer
Suppose we write a computer program to determine our result R from the primary measurements a, b, c… We have:
or, as before:
The uncertainty in R may therefore be calculated usingthe same expression:
The only problem is determining the partial derivatives.
![Page 20: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/20.jpg)
Determining Uncertaintyusing a computer
The δ’s musn’t be too large a fraction of the measured quantity – up to 20% is probably OK. (If δ’s are bigger, then just use a smaller fraction and multiply the result accordingly)
This can be done by perturbing, in turn, each of the inputs to our program by its uncertainty, and recording the change in output that results. (The modified program that does this is usually called a jitter program).
![Page 21: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/21.jpg)
ExampleComputing uncertainty in measured Reynolds number
a, b, c
R
Program
,...),,( cbafR =
![Page 22: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/22.jpg)
ExampleCopy calculation into 3 spare columns
![Page 23: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/23.jpg)
Example
a+δa, b, c
R+(∂R/∂a)δa
Program
(∂R/∂a)δa
![Page 24: AOE 3054 - Uncertaintyaborgolt/aoe3054/classes/FoundationsClass3Handouts.pdfDetermining Uncertainty in a Derived Result • Consider an experimental result R that is determined from](https://reader036.vdocuments.net/reader036/viewer/2022071415/61114ab37e31ca17c211c1fd/html5/thumbnails/24.jpg)
Example
.)()()()(222
⎥⎦⎤
⎢⎣⎡∂∂
+⎥⎦⎤
⎢⎣⎡∂∂
+⎥⎦⎤
⎢⎣⎡∂∂
= ccRb
bRa
aRR δδδδ
(∂R/∂a)δa(∂R/∂b)δb
(∂R/∂c)δc