Measurement uncertainty and measurement deviation

Practice

Dipl.-Ing. Susanne Nisch

WZL/Fraunhofer IPT

Practice Mesurement uncertainty and mesurement deviation

Content Definitions and Terms Influence Faktors on measurement uncertainty Temperature Messurement object Messurement device

Method for estimation of measuremnet uncertainty Examples

WZL/Fraunhofer IPT

Seite 2

Definition and Terms

Measurement uncertainty is a parameter associated with the result of a measurement that characterises the dispersion of values which could reasonably be attributed to the measurand.

Mesurement error is the deviation from the true value of a value gained from measurements and assigned to the measurand, or the measurement results minus the true value of the measurand.

The limit of error is the maximum amount of measurement deviation of a measuring device.

WZL/Fraunhofer IPT

Seite 3

Influence Factors on Measurement deviation

Value Proceeding
Defining the measurand (specifying the quantity to be measured, depending on the required measurement uncertainty)

Example

Value of the measurand (non-realisable true value) Values of the measurand due to incomplete definition

Determining the thickness of a given sheet of material at 20 C Measuring the thickness of the sheet at 25 C with a micrometer and measuring the applied pressure

Measuring the realised quantity

Unadjusted observed values

Unadjusted arithmetic mean of the observed values

Correcting all known systematic influences

Correcting the influences of temperature and pressure

Measurement result (best estimate)

Residual error

Complete measurement result

WZL/Fraunhofer IPT

Seite 4

Cause and effect diagram of production metrology

Human Measured oject
Surface structure e.g. roughness Form deviations Commitment to perform Efficiency Mass Dimensions

Method
Measurement strategy Measur. procedure

Measur. principle Evaluation

Complete measurement result

Contamination Humidity Vibrations Temperature Strucutre Statistic behaviour Resolution Dynamic behaviour Internal evaluation of the device

Environment
WZL/Fraunhofer IPT

Measuring device
Seite 5

Types of measurement errors

Measurement errors

systematic measurement errors

coincidental measurement errors

known systematic errors

unknown systematic errors

correction

residual errors

measurement result
WZL/Fraunhofer IPT

Measurement uncertainty

Seite 6

Influence Factor Temperature

Temperature measurement is not everything in length metrology, but it cannot be done without Three types of temperature influences: Deviation of the temperature level from the reference temperature Temporal temperature fluctuations Spatial temperature fluctuations Types of heat transfer: Thermal conduction Convection Heat radiation Linear expansion characteristic:
WZL/Fraunhofer IPT

DL=L**Dt
Seite 7

Coefficient of thermal expansion for solid bodies

Aluminium alloy = 23..24 [10-6/K] Glass = 8..10 [10-6/K] Steel Zerodur a = 10..12 [10-6/K] a = 0..0.05 [10-6/K]

Example: A 100 mm in length Steelruler will stretch by more than 1m with a temperature difference of 1K! The effect of change in length are negligible if: ~0 Werk = M und tWerk = tM tWerk = tM = 20 C

WZL/Fraunhofer IPT

Seite 8

Coefficient of thermal expansion for solid bodies

When measuring, the change in length of scale must beconsidered as well as the change in length of the measured objekt: L=L*(Work*tWork-M*tM) The calculable influence of temperature contains an uncertainty which results from the uncertainty of the temperature measurement and the uncertainty of the coefficients of expansion. This is computed by partially deriving and squared addition of the individual parts, where ua represents the uncertainty of the coefficient of expansion and ut the uncertainty of temperature measurement:
u =L * (u Werk * t Werk ) + (u tWerk * Werk ) + (u M * t M ) + (u tM * M )
WZL/Fraunhofer IPT Seite 9

Length measurement error conditional upon temperature

Deviation [m/m]
25 20 15 10 3 m 5 0 18 19 -5
-10 -15 -20 -25

Boundaries of the area of uncertainty with correction thermal correction

3m/m

20

21

22

23

24

25

Temperature ] tWork=tM [C

Boundaries of the area of uncertainty without correction

Seite 10

WZL/Fraunhofer IPT

Influence factor: measured object

Measurement deviation due to features of the measured object: Surface Structure Reflection (optical measurement technology) Roughness, Waviness (taktile) Hardness 40 40
a [m] 10

Measurement error through deformation

1 0,1 0,01

F = 10 N F=5N F=1N

r [mm]

Example: ball-plane (steel)

Flattening (Hertz Pressure) through measurement force
WZL/Fraunhofer IPT Seite 11

Influence factor: slim measured object e.g. deflection of slim measured objects
F=mg

For the deflection of uniform measured objects (pipes, rulers, plates) under dead weight:

a) Support at the ends: maximum deflection

d max = 100%

b) Support at each 0,22 l: minimum deflection

l
d max = 2%

0,22 l
WZL/Fraunhofer IPT

0,22 l
Seite 12

Influence Factor: measurement device

Measurement errors due to inaccurate guides play an important part in measurement devices with built-in measure embodiment. The clearance, which is technically required in guides for measuring pins, touch probes or eyepieces, causes tilting. The influence of these on the measurement result are large (first order) or small (second order), depending on how the measure embodiment and the measured object have been positioned. Abbe Principle: In order to avoid errors of the first order, the scale of the measuring device must bepositioned such that the distance to be measured forms a straight-line continuation of the scale. (Ernst Abbe, 1893)

WZL/Fraunhofer IPT

Seite 13

Influence factor: measurement device Caliper Error of first order by Abbe

2 0

10

11

12

13

14

15

16

L
L'

L
Meabweichung Measurement Error: L =L L ~p p = L L -L ~

Meobjekt parallel versetzt Measured object aligned to measure embodiment zur Maverkrperung => => complying to des the Abbe'schen Abbe principle Verletzung Prinzips
WZL/Fraunhofer IPT

L': abgelesener Mewert L : measured L : wahre Lnge value p : Parallelversatz L : true value

P : parallel offset
Seite 14

Influence factor: measurement device Capiler Breach of Law of Abbe

With a parallel offset of the measuring distance and the reference distance, a small tilt already causes measurement errors which are no longer negligible! Measurement deviation: L = L-L = p*tan() for << 1 with described in angle minutes: L=p*

Example: Tilt = 2 parallel offset p = 30 mm -> L = 17 m

L L'

L p

WZL/Fraunhofer IPT

Seite 15

Influence factor: measurement device micrometer Abidance of law of Abbe

Measurement object, measurement surface and lead screw form a straight line. The measurement deviation for <<1 can be described as:
1 L = A 1 cos( ) = A( 1 + tan( ) 1) A( 1 + 1)
A

L A j A

Measured object aligned to measure embodiment => complying to the Abbe principle

L
Error of measurement Measurement Error: 2 2 L = L L Dj ~A L /2 = L' - L ~ A

L : L': measured read-out measurement value value : true length L :L true value
WZL/Fraunhofer IPT Seite 16

Methods for the estimation of measurement uncertainty

The basis of every procedure for estimating measurement uncertainty ist the Guide to the Expression of Uncertainty in Measurement (GUM) The evaluation of a measurement can be carried out in four steps:
i. Setting up a model which mathematically describes the relationships of the measurands (y1,y2,...yn) to all other quantities involved (x1,x2,...xn) y=f(x1,x2,...xn) Preparation of the given measurement values and other available data Calculation of the measurement result and measurment uncertainty of the measurand from the prepared data Specification of a complete measurement result and determination of the extended uncertainty.

i. ii. iii.

WZL/Fraunhofer IPT

Seite 17

Procedures for estimating measurement uncertainty ii. Determining the standard uncertainty
Method Form of distribution
Normal distribution

Calculation

A
Normal distribution

ss x

u=

s n a 4 a 3 a 6

Standard uncertainty of mean value s: standard deviation n: number of observed values Assumption:

ss x

u=

The estimated value lies within the boundaries a+ and a- with a confidence level of 95 %. Assumption: The estimated value lies within the boundaries a+ and a- with a confidence level of 100 %. Assumption:

Uniform distribution

B
aa+

u=
Tringular distribution

u=
a WZL/Fraunhofer IPT

a+

The estimated value lies within the boundaries a+ and a- with a confidence level of 100 %.
Seite 18

Procedures for estimating measurement uncertainty iv. Specification of the complete measurement result
The measurement uncertainty of the measurand, the combined standart uncertainty uc, is determined through squared addition of the individual uncertainty components. As long as all input variables are independent from each other: df df df uc = u + u + ... + u dx x1 dx x2 dx xn 1 2 n or simply:

uc = u x1 + u x 2 + ... + u xn

Multiplication with the extension factor k, shows the extended measurement uncertainty: U=k*uc k = 2 equals a range of confidence of 95%
WZL/Fraunhofer IPT Seite 19

Evaluation of measurement uncertainty e.g. Dial Gauge

Example: Dial Gauge for the determination of the hight of a gauge block
10 0 10 20 20 30 30 40 40

Dial gauge

Tripod Gauge block

Plane table
WZL/Fraunhofer IPT Seite 20

Evaluation of measurement uncertainty e.g. Dial Gauge

Main influence factors Surroundings:
The measurement is carried out at 20 C (error limit of the temperature

definition 2 K)
Thermal expansion coefficient of the workpiece 12*10-6 /K (uncertainty

1*10-6 /K) Measuring device:

Measurement deviations for a temperature range of 18-22 C are in an area

of 0.02 mm (95% plausibility)

Values are distributed normally Systematic deviation of b = -0.06 mm Levelness of the plane table, the support face of the tripod and the

formation of the tripod are not known

WZL/Fraunhofer IPT Seite 21

Evaluation of measurement uncertainty e.g. Dial Gauge Random influences

20 measurements are taken from different points of the plane table: Mean value of the observation: Standard deviation: x = 100.02 mm s = 0.09 mm

The systematic deviation necessitates a correction of the mean value for the real value: y = 100.02 mm 0.06 mm = 99.96 mm Uncertainty components: Repeated Measurements

u1 =
WZL/Fraunhofer IPT

0.09 mm = 0.02 mm 20
Seite 22

Evaluation of measurement uncertainty e.g. Dial Gauge Lenth deviation caused by temperature
The error limit from the temperature definition is 2K When quoting error limits without stating the distribution it is useful to assume a uniform distribution of the values
Uniform distribution

a-

a+

Assumption: The estimated value lies within the boundaries a+ and a- with a confidence level of 100 %.

u=

a 3

u t =

2K 3

u2 = L * (uWerk * t Werk )2 + (u tWerk * Werk )2 + (uM * t M )2 + (utM * M )2 (from page 9) 2K 1 * 12 10 6 ) + 0 + 0 = 1.4 m K 3

Seite 23

u2 = 99.96 mm * 0 + (

WZL/Fraunhofer IPT

Evaluation of measurement uncertainty e.g. Dial Gauge Deviation of dial gauge

normal distribution 95% of the values are within a range of 0.02mm (Methode B)
Normal distribution

Assumption: the estimated value lies within the boundaries a+and a-with a confidence level of 95%

a u= 4

a-

a+

u3 =

0.02 mm = 0.01 mm 2

WZL/Fraunhofer IPT

Seite 24

Evaluation of measurement uncertainty e.g. Dial Gauge evaluation of the complete measurement result
Combined standard uncertainty:

uc = u1 + u2 + u3 = 0.022 mm
Extended measurement uncertainty: extension factor k=2 equals a range of confidence of 95%

U = k* uc = 0.044 mm
Complete result:

Y = (x + b) U = 99.960 mm 0.044 mm

WZL/Fraunhofer IPT

Seite 25

Practice: Evaluation of measurement uncertainty for using a ruler

Take your ruler and measure the distance between the two parallel lines. The nominal size is 100 mm 1 mm. Estimate the first decimal place (0.1 mm). Calculate the mean value and the standard deviation by using the next page. Which influences can appear? Evaluate the complete measurement result (mean value extended value). Consider influences of temperature caused length deviation (error limit a = 2K; Lineal = 120 * 10-6 /K; uniform distributed) and the deviation of the ruler (error limit a = 0.1 mm; nominal distributed).

Reference value: 100 mm

WZL/Fraunhofer IPT

Seite 26

Practice: series of measurements

Mean value:

1 n x = xi n i=1

Standard deviation:

1 n (xi x)2 n 1 i=1

n 1 2 3 4 5 6 7 8 9 10

xi

xi x

Influences:
Human:__________________________________________

Environment:______________________________________

Measured object:___________________________________

Device:___________________________________________
WZL/Fraunhofer IPT Seite 27

bungsaufgabe: Berechnung der Messunsicherheit

Influences
Random variance Temperature deviation Error of ruler

Error limit or standard deviation

Method and distribution

calculation

Standard uncertainty

Combined standard uncertainty: Extended uncertainty:

____________________________

____________________________

Complete measurement result:

WZL/Fraunhofer IPT

____________________________
Seite 28