measurement theory intro to measurement discussion of standards & traceability prac example...
TRANSCRIPT
Measurement Theory
Intro to measurementDiscussion of standards & traceabilityPrac exampleUncertaintyExamples of measurementExercisesCautions
Measurement
The process of determining the value of
some quantity in terms of a
standard unit.
Standards
There is a hierarchy of standards – that is agreed unitsSome of these are artifacts ie the kgSome are “realised” eg temperatureAt the top of the hierarchy are Primary standardsRIC currently hold Primary standards for P, T and Radiation
True Temperature Scale
Agreed international scale of temperature – ITS-90Comprised of points on the scale that are realized – that is made up temporarily using physical systems
Interpolation between the points is via Pt resistance thermometers
Pt resistance thermometers are approximately linear between points on the true temperature scale
Substance Temperature K Temperature °C State
Mercury, Hg 234.3156 -38.8344 Triple Point
Water, H20 273.16 0.01 Triple Point
Gallium, Ga 302.9146 29.7646 Melting Point
Indium, In 429.7485 156.5985 Freezing Point
Tin, Sn 505.078 231.928 Freezing Point
Zinc, Zn 629.677 419.527 Freezing Point
Aluminium, Al 933.473 660.323 Freezing Point
Silver, Ag 1234.93 961.78 Freezing
Water Triple Point Cell
Ultra pure water is sealed under vacuum into a glass vesselThe apparent air gap above the liquid is entirely composed of water vapour whose pressure is determined by the temperatureIt forms a sealed system at equilibrium
Contd.WTP defined to be at 0.01oCThe ice must be as a moveable mush ie. It must freely rotate in the cellThe WTP maintenance bath can keep the cell at this temperature for daysThe kelvin, unit of thermodynamic temperature, is the fraction 1/273.16 of the thermodynamic temperature of the triple point of water.
Gallium Melting Point
Defined to be at 29.7646 degrees C
As can be seen in the graph it is a plateau
Energy is going into breaking bonds – hence no temperature rise until all of the Ga has melted
Can be drawn out for about 30 hours 5 6 7 8 9 10
29.765
29.766
29.767
29.768
29.769
Te
mp
era
ture
(o C
)
Time (hours)
0 5 10 15 20 2527.0
27.5
28.0
28.5
29.0
29.5
30.0
30.5
31.0
End of Melt
Start of Melt
Te
mp
era
ture
(o C
)
Time (hours)
Traceability
Traceability is the unbroken chain of calibration/verification from a primary standard to the device in questionThis chain may have one link or several depending on the deviceAt each stage of must be fully documented
Pressure TraceabilityInternational
HO Transfer
Regional Kew
Regional Transfer
Station
0.03 hPa
0.07 hPa
0.10 hPa
0.15 hPa
0.20 hPa
National
RA V (WMO)
Dimensional
Pressure
Total 0.27 hPa
• The degree of doubt about a measurement!
• Parameter, associated with the result of a measurement, that characterises the dispersion of the values that could reasonably be attributed to the measurand. (International Vocabulary of Basic and General Terms of Metrology)
Uncertainty
Uncertainty
Low AccuracyHigh Precision
xxx
x
xxx xxx
xxx
x
x
x
xxx
x
x
x
x x
Medium AccuracyLow Precision
AccuracyThe closeness of the experimental mean
value to the true value.
High accuracy = Small systematic error.
PrecisionThe degree of scatter in the results.
High precision = Small random error.
1
High Repeatability / Low Reproducibility
Golfer OneGolfer One
Drift in an instrument
2
Low Repeatability / Low Reproducibility
Golfer TwoGolfer Two
Low Precision
3
High Repeatability / High Reproducibility
Golfer ThreeGolfer Three
Low Uncertainty
What’s Normal?
The outcome of most natural processes is normally distributedThis results from the central limit theorem
Significance of Differences
Xa Xb
U95
Not Significant
Significant
Xa Xb
Confidence
How many samples do you have to take to be “confident” you have estimated the mean value correctly?The mean we determine will have an expected value – in this case the mean of the population and a varianceHow well we estimate the mean depends on how many samples we take.
Xa
Temperature Prac
Use the two IR thermometers to take the victims temp.Take 7 measurements with each deviceForm an averageMax and Min
Making a measurement
Any single measurement is a “selection” from a distribution of possible valuesMore measurements give you greater “confidence” in estimating population parametersCan’t make an infinite amount of measurements because the system being tested may not be stable
Test Uncertainty Ratio (TUR)
It is intuitive that in order to measure something you need to measure it with something more accurateThis is the TUR – the ratio of the uncertainty in your reference to the uncertainty of the device under testUsually a TUR value of 4 or better is used
Xa
Xa
Contd.
You can work with TURs less than 4The barometers are calibrated with a TUR of approximately 1!You need to take a lot of samples! Xa
Xb
Instrument Properties
Linearity – Accuracy of response over measurement rangeStability – short and long term (drift)Response time – how fast it respondsPrecision Hysteresis
Linearity
Opposite are plots of True versus probe temperature for AWS Temp probesNote they are all approximately linear in responseThey each have a slightly different line
-10 0 10 20 30 40 50 60-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
0.06
0.08 RTDs
Cor
rect
ion
(o C)
Temperature (oC)
Stability
Humidity probe short term drift (2 hrs)Humidity Probe medium term drift (20 days)
0 5 10 15 200.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Probe 1 Probe 2 Probe 3 Probe 4
80 % RH
Co
rrec
tion
(%
RH
)
Day
40 60 80 100 12049.5
49.6
49.7
49.8
49.9
50.0
50.1
50.2
% R
H
Time (Min)
Response Time
Opposite are plots of RH versus time for a humidity probe.RH was changed “instantaneously” “Response time” is defined as the time taken for the instrument to read 63% of the step change
410 420 430 440 450 460 470
30
40
50
60
70
80
% R
H
Time (sec)
1980 1990 2000 2010 2020 2030 2040
30
40
50
60
70
80
% R
H
Time (sec)
Precision
Opposite are the plots Temp versus time for two probesThe two probes have differing systematic errors (y axis shift)The two probes have different precisions (y axis spread) 200 300 400
21.90
21.95
22.00
22.05
22.10
22.15
22.20
Tem
per
atur
e (o C
)
Time (sec)
Probe 1 Probe 2
Contd.
The probes exhibit a systematic error – offset or biasBoth probes have approximately the same precision
0 500 1000 1500 2000 2500 3000 3500
55.13
55.14
55.15
55.16
55.17
55.18
55.19
55.20
55.21
Tem
pera
ture
(o C
)
Time (sec)
Probe 1 Probe 2
2000 2500 300055.1700
55.1725
55.1750
55.1775
55.1800
55.1825
55.1850
55.1875
55.1900
55.1925
55.1950
Oceanus 6
Tem
pera
ture
(o C
)
Time (sec)
Quantization
0 20 40 60 80 1002.6
2.7
2.8
2.9
3.0
3.1
3.2
3.3
Pe
rio
d B
twn
Tip
s (s
ec)
Sample No.
Pump System Gravity System
Quantised measurements take discrete levelsImportant to know how they were quantizedWere they rounded or truncated?No necessarily less accurate than analogue data
Hysteresis & Linearity
100 200 300 400 500
100
200
300
400
500
0
Hysteresis
Linearity
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
-15 -5 5 15 25 35 45 55
Temperature °C
Cor
rect
ion
to r
efer
ence
hP
a
PA11a Increasing Temperature
PA11a Decreasing Temperature
PTB220B Increasing Temperature
PTB220B Decreasing Temperature
PA11a & PTB220B Hysterisis
GoodGoodClean Mercury Rising Pressure
BadBadPossibly Dirty MercuryFalling Pressure
Very BadVery BadDirty MercuryFalling Pressure
Mercury Barometers
Response Versus Temperature
Opposite is a plot of the corrections required for HMP45D probes versus TempNote – response is quite consistent – but not linear
5 10 15 20 25 30 35 40 45 50
0.0
0.5
1.0
1.5
2.0
2.5
3.0Response versus TemperatureVaisala HMP45D Humidity Probes
Cor
rect
ion
(o C)
Temperature (oC)
With-in run precision.
Variability on an occasion
Reproducibility
Variability on different occasions
Between-run precision
Repeatability
Contd.
Opposite is a plot of the RH reached by the humidity generator versus timeSystem was cycled between 3 RH levelsRepeatability is the closeness of the match in RH achieved
2000 4000 6000 8000 100000
20
40
60
80
100
% R
H
Time (sec)
Reproducibility
Reproducibility is the “between trial” variabilityOpposite is a plot of the long term error for a barometer
Calibration Errors
Co
rrec
tio
n (
hP
a)Jun-65 May-70 May-75 May-80 May-85 May-90
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Populations
0 20 40 60 80 100
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Original probes tested Probes from other batches
Temp = 25oC
Err
or
(Re
f -
Pro
be
) %
RH
Reference % RH
Opposite is a plot of the offset errors for a batch of humidity probes.The error for any particular probe for any measurement will be approximately normally distributedThe offset or bias of the all probes is also expected to be normally distributed!
Resolution
Resolution is the smallest increment in value the instrument can returnResolution will affect the precision of the instrumentResolution will not ordinarily affect the accuracy of an instrument
Resolution = Uncertainty
Half of the least significant digit on an analogue instrument
The least significant digit on a digital instrument.
The uncertainty of this thermometer is ± 2°C.
If the scale has 10°C divisionsThe resolution is 5°C
If the scale has 2°C divisions
The resolution is 1°C
20
30
10
0
40
50
Contd
Both probes have the same resolutionRed probes has approx four times the uncertainty
200 300 40021.90
21.95
22.00
22.05
22.10
22.15
22.20
Tem
per
atur
e (o C
)
Time (sec)
Probe 1 Probe 2
Confidence
100%95%65%<1%
± 0.5°C± 5°C
± 10°C± 30°C
20
30
10
0
40
50
Errors Vs Blunders
By definition most measurements will not be exactly “right” they will be in error to some degreeA blunder is when a human is in the loop and produces a mistakeIe. Misreads a thermometer as 35.25oC instead of 25.25oC
Calibration
Comparing the reading of an instrument when it is exposed to a known artifact or conditionEither the instrument is adjusted to read “correctly” orA table of corrections is produced so that the operator can “correct” the instrument reading to the true readingMay need to interpolate
VerificationMost of the work of the RIC involves verifying that an instrument/probe etc is in specificationThis is not a calibration since corrections etc are not suppliedHence equipment sent to the field is within spec but may lie anywhere within the specification - two humidity probes could differ in readings by 4% RH and both could still be in spec
0 20 40 60 80 100
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
HMP45D HMP45A
Temp = 23oC
Co
rre
ctio
n (
Re
f - P
rob
e)
% R
H
Reference % RH
Field TolerancesSensor
ComparisonMethod
UncertaintyField
Tolerance
Pressure Standard 0.3hPa 0.5hPa
TemperatureWithin ScreenPsychrometer
0.3°C0.4°C
0.5°C0.6°C
RelativeHumidity
Within ScreenPsychrometer
4%5%6%
Wind Speed ? 10% N/A
Wind Direction Compass 5% 10%
RainfallWith Syphon
Without Syphon
3% (<250mm/h)
4% (250 – 350mm/h)
8%8%
Exercise 1Currently the inspection handbook “checks” an AWS RTD with an Inspection grade Mercury in glass thermometerRTD accurate to 0.2oC – MIG accurate to?Single measurement after 1 hour of stabilisationWhat is are the flaws in this procedure?Come up with some alternatives alternatives
Exercise 2
Currently the inspectors check an AWS humidity probe with an wet/dry bulb thermometersOne wet/dry measurement after 1 hour of stabilisationWhat is are the flaws in this procedureSuggest alternatives
Exercise 3
NCC alerted RIC to anomalous readings from manual sites (red) and AWS humidity probes (black)The manual obs (wet/dry bulb) appear to over-estimate the dew point
0 20 40 60 80 100 120
-6
-4
-2
0
2
4
6
8
10
12
14
De
w P
oint
(o C
)
Time
AWS Data Manual Obs
Contd.Plotted opposite is the DP from manual obs (x-axis) versus the AWS derived DP (y-axis)In a perfect world the data would lie along the line y = xPostulate a model as to what has gone wrongAssume humidity probe was checked and found to be in-spec within previous 6 months.
-6 -4 -2 0 2 4 6 8 10 12-4
-2
0
2
4
6
8
10
12
14
Slope = 0.75
Man
ual O
bs D
ew P
oint
(o C
)
AWS Dew Point (oC)
Scatter Plot Linear Fit of Data1_B
Hypotheses
1 – The humidity probe is stuft!2 – The manual observers were drunk!3 – Both 1 & 24 – Both sets of data are correct!Come up with some others –Also assume all measurements made were correct!
Best GuessIt is troubling that the line of best fit does not have a slope of 1 and this suggests there may be a problem with the algorithms used to calculate DP.Having said that, it is most likely that both sets of data are essentially “correct”.RH probes (currently in use) measure RH Wet/dry bulb measurements really measure evaporation rate – not really the same thingWet/dry measurements over-estimate humidity by up to 20% in still air conditions.A useful comparison would be RH from each technique after selecting data obtained when the wind speed was greater than 2 m/s
Data Quality
Site Selection and Installation
Measurements
Quality Assurance
Quality Control
Data
Instrument Selection
Final Product
Meteorologists & Climatologists
PMAs
Observers
ROMs & Engineers
Lab
Training Double checkUse Calibrated instrumentsMinimises the number of variablesUse standard test procedures“If it is not broken don’t fix it”Document, document, document
How to improve the Data Quality
Field Adjustment
Don’tJust DON’T!An adjustment in the field will remove all traceabilityIf it is out of spec – remove and return
Questions?