Uncertainty of Measurement

Prof. Vernon Buchanan

School of Engineering
 Learning Outcomes
P Understand the meaning of uncertainty of
measurement
P Follow standard procedures to evaluate
uncertainty of measurement
P State good measurement practices that reduce
uncertainty of measurement
 The Fact

P Measurement involves comparison of quantities.

P Comparison always involves some uncertainty.
P The measured result is only an estimate of the true
value.
 Definition of Uncertainty
P Dictionary usage
< Doubt: inability to make a decision
< Skepticism: not willing to believe without
proof
< Suspicion: lack of faith
P Measurement
< The probability that the dispersion of the
result of measurement will lie in a certain
range
– How far from the true value can the result be?
Uncertainty of measurement is quantification of
the doubt that exists about the measurement result.
 What is the True Value?
P Since there is doubt in
measurement, we need to
state:
< How big is the margin of error
(interval)?
< How sure are you that the true
value is within the margin
(level of confidence)?
P When we give the interval and the level of
confidence of a measurement, we are stating the
uncertainty of a measurement.
< Example, the length of a line is 50 mm ± 2.5 mm, at
90% level of confidence.
Interrelations between true value, error,
measured value, and uncertainty.

U Expanded
U Re lative 
Measured value
 Typical Sources of Uncertainty
P Equipment P Environment
< Wear < Temperature
< Hysteresis < Relative humidity
< Backlash < Atmospheric
< Incorrect scale graduation pressure
< Electrical and magnetic < Corrosion
noises < Radiation
< Misalignment P Operator
< Calibration < Wrong calculations
< Friction < Misreading
< Linearity < Misalignment
< Drift < Parallax effect
Example: possible errors in the
uncertainty in the length of a piece of
string using a tape measure
P Possible errors of the measuring tape
< Calibration, extension in the tape, tape discrimination
P Possible errors due to the item being measured
< Is the string straight? Is it affected by temperature and
humidity? Are the ends of the string well defined?
P Possible errors due to the measuring process
< Is the string parallel with the tape?
< Is the end of the string and tape line up?
< How repeatable and reproducible is the measurement?
 What is not a Measurement
Uncertainty?
P Mistakes made by operators
< Avoid them by working carefully and by checking work.
P Tolerances
< They are acceptance limits which are chosen for a process or a
product.
P Specifications
< A specification tells you what you can expect from a product.
P Accuracy (or inaccuracy)
< Accuracy is a qualitative term; uncertainty is quantitative.
P Error
< Error is the difference between the measured value and the
‘true value’ of the thing being measured.
 Evaluation of Uncertainty
There are two ways to evaluate uncertainties.
P Type A Evaluations
< Calculating standard uncertainty by statistical methods
using a series of repeated independent measurements
P Type B Evaluations
< Calculating standard uncertainty by other means
– Previous measurement data
– Manufacturer’s specification; Calibration certificates
– Data taken from handbooks and historical records
– Experience
 Type A Evaluation
P The standard uncertainty is the estimated standard
deviation (u) of a measurand.
< Normal distribution
– Measurements are more likely to be near the average than
farther away.

 xi  x
2


n
i 1
uA =
n 1
 Type B Evaluation
P Rectangular distribution
– Measurements are evenly spread between the highest and the
lowest.

a b
b a 1
uB  
2 3
 Steps to Calculate Uncertainty
P Select the measurand and the method/equipment.
P Decide what elements (sources) of the measuring
system contribute to the uncertainty of measurement.
P Express these elements quantitatively in terms of a
standard uncertainty (deviation) having similar units
and similar coverage factor (confidence level).
P Combine the standard uncertainty to determine the
overall standard uncertainty.
P Expand the standard uncertainty according to the
desired level of confidence.
P Report the value together with the size of the
uncertainty interval and the level of confidence.
 Combination of Standard Uncertainty
Method of Partial Differentiation
Given a = f(x,y,z), then the combined uncertainty is:
2
 a   a   a 
2 2

Ua   ux   uy   uz 
 x   y   z 
For example, the area, A, of a rectangle is length, L,
times the width, W, that is A = L ×W.
2 2
 A   A 
U A   u L    uW   W  u L    L  uW 
2 2

 L   W 
Example - calculating the uncertainty
in the radius of a component

 M  d 2

R
8d
Nominal Radius =

38166.  5000.  2

 274996
. mm
8  5000
.
Given: M = 381.66 ± 0.09 mm
d = 50.00 ± 0.01 mm
 What Elements Contribute to the
Measurement Uncertainty?
P Cylinder diameter
< ud = 0.01/%3 = 0.006 mm, k = 1 (68.4%)
P Micrometer calibration and scale uncertainty
< uM = 0.09/%3 = 0.052 mm, k = 1
P Temperature difference, T
< uT = 0.02/%3 = 0.012 mm, k = 1
P Surface characteristics of sector and surface plate
P Elastic deformation
P Misalignment
P Clamping arrangement
 Contd.
P What elements have negligible effects?
P For this exercise all other elements, including the
last four on the previous slide, are assumed
negligible. Thus, we will only consider errors
due to:
< Cylinder diameter
< Micrometer calibration and scale uncertainty
< Temperature
 Estimating Measurement Uncertainty
 M  d 2 2
 R   R 
2
Given R  Then U a   u M    ud 
8d  M   d 
P Derivation of M with respect to R
R M  d 38166
.  50.00
   1658
.
M 4d 4  50.00
P Derivation of d with respect to R
R

 M  d   2d  M  d 
2

 7.158
d 8d 2
 Estimation of Uncertainty
R
Uncertainty from M = u M  1658
.  0.052  0.086 mm
M
R
Uncertainty from d = u M  7.158  0.006  0.043 mm
d
Uncertainty from T = uT  0.012 mm

Combined uncertainty  0.086 2  0.0432  0.012 2  0.097 mm

P Expanded uncertainy (k = 2) = 2 × 0.097 = ± 0.195 mm

P Uncertainty Statement
< Radius of sector is 274.996 ± 0.195 mm, 95.4%
confidence
 Note!!!
P Coverage factor is a numerical value or multiple of σ
denoting the level of confidence.
< Manufacturers generally use k = 3.89 (99.99%
confidence), for a failure rate of 1 in ten thousand.
P Expanded uncertainty is the value that defines the
interval which can be expected to include a large
portion of the distribution of values of the
measurand.
P It is sometimes more useful to express uncertainty as
relative uncertainty, Urel, which is the ratio of the
expanded uncertainty, Uex and the measured value y:
U ex
U rel 
y
 How to reduce uncertainty
P Calibrate measuring instruments.
P Make corrections to compensate for known
errors.
P Make your measurements traceable.
P Choose the best measuring instruments.
P Check measurements by repeating them.
P Check calculations.
P Use rounding correctly in your calculations.