Calibration of a Torque
Wrench as per ISO6789
by
Eddie Tarnow
NLA Test & Measurement Workshop 20 September 2011
Calibration Setup
350,0 N•m
Reference Standard Torque Transducer &
Readout Unit
Unit Under Test Torque Wrench
Force Applied Clockwise Rotation
Calibration Scenario
• The Unit Under Test Torque Wrench is a Type II class A tool (adjustable click type) and has a full scale of 350 N•m.
• It has a setting dial resolution of 2 N•m
• We are to calibrate it according to ISO 6789 which requires a calibration point at full scale (100 % of range) viz. at
350 N•m and the estimation of the measurement uncertainty at this point
GUM Steps
• Model the measurement
• Identify and quantify the sources of uncertainty • Categorize as type A or type B
• Manipulate appropriately to obtain
• Standard uncertainties, u(xi)
• Sensitivity coefficients, ci
• Uncertainty contributor, u(yi)
• Combine to obtain combined standard uncertainty, uc(y)
• Expand to obtain an expanded uncertainty, U, at an appropriate level of confidence
Measurement Model
STD
Ind
STD
UUT
T
Corr
T
=
+
Identifying the sources of uncertainty
T
T
UUT U ResT
STD SCal SRes SCFU
RepQuantifying the sources of uncertainty
• SCAL
• Calibration Results Table from the Calibration Certificate
APPLIED TORQUE
(N•m)
MEAN
UUT CALCULATED TORQUE (N•m) UNCERTAINTY OF MEASUREMENT (± N•m) 0,0 0,0 0,1 99,9 100,0 0,3 199,9 200,2 1,0 299,8 300,5 1,0 399,8 400,8 1,0 499,7 501,1 1,0 599,7 601,5 1,0 699,6 701,8 1,0
Quantifying the sources of uncertainty (2)
• SCAL
• This is the uncertainty due to the accuracy of the Reference Standard Torque Transducer, which is not perfect
• Corrections must first be applied, or the uncertainty increased, to take the error into account (largest error on values either side of the calibration point was +1,0 N•m)
• The Reference Standard Torque Transducer used has a full scale of 700 N•m and was calibrated in 100 N•m steps (See calibration
certificate)
• Therefore we will have to use the polynomial equation to determine the actual torque generated by the UUT at 350 N•m since it is a
measurement point in between 300 N•m and 400 N•m.
• Since we have to interpolate a value we will use the largest reported uncertainty from the calibration certificate for the values on either side of the calibration point which is ± 1 N•m.
Quantifying the sources of uncertainty (3)
• SCAL
• Since we will be using the polynomial to interpolate a value at 350 N•m, we DO NOT need to correct for the + 1 N•m error at 399,8 N•m.
• Therefore total uncertainty for the “accuracy” of the Reference Standard Torque Transducer is ±1 N•m
• This is treated as normal at 95,45% Level of Confidence • The divisor is the coverage factor k which for 95,45%
LOC is 2
• The degrees of freedom are always ∞ or 100 % Reliable
due to the source of traceability being accredited and reputable.
Quantifying the sources of uncertainty (4)
• SRES
• This is due to the resolution of the Reference Standard Torque Transducer Readout Unit
• We must first determine the “effective resolution” • The least significant digit displayed is 0,1 N•m • Resolution is always treated as a Rectangular
distribution source of uncertainty and this is the range. • The semi-range is therefore (0,1 N•m/2)=0,05 N•m
• The divisor is √3
• The degrees of freedom are always ∞ or 100 %
Quantifying the sources of uncertainty (5)
• SCF
• Polynomial Equation Coefficients Table from the Calibration Certificate POLYNOMIAL EQUATION y=a+bx+cx 2 +dx3 POLYNOMIAL COEFFICIENTS NORMAL FUNCTION INVERSE FUNCTION a 2,71846 x10-2 -2,70765 x10-2 b 9,99825 x10-1 1,00017 c -7,89039 x10-6 7,95708 x10-6 d 5,16535 x10-9 -5,22137 x10-9
Standard Error of the polynomial curve fit for a Level of Confidence of 68,27% and 4
degrees of freedom
Quantifying the sources of uncertainty (6)
• SCF
• This is the additional uncertainty which results from the interpolation calculation to determine the torque
generated by the UUT at a point in between the
calibration points of the Reference Standard Torque Transducer
• It is obtained directly from the calibration certificate as the “Standard Error of the polynomial curve Fit” value = ± 0,045 N•m
• This is treated as a normal distribution at a 68,27% Level of Confidence
Quantifying the sources of uncertainty (7)
• URES
• This is due to the resolution of the UUT Torque Wrench scale. (How it influences the setting of the wrench to a specified torque)
• Typically this would be the smallest graduation on the UUT setting dial which for this UUT is 2 N•m
• This is the range of the rectangular distribution • Therefore the semi-range is (2 N•m/2)=1 N•m
• The divisor for Rectangular Distributed uncertainty contributors is √3
• The degrees of freedom for resolution is always ∞ or
Quantifying the sources of uncertainty (8)
• UREP
• This results from the variability in the measurement results obtained after repeating the measurement 5 times.
• It can be dealt with either as “repeatability” or as “reproducibility”
• “Repeatability” – all conditions remained the same during the repeated measurements
• “Reproducibility” – any one or more of the conditions changed during the repeated measurements
• Different approaches can be used to repeat the measurement
Quantifying the sources of uncertainty (9)
Repeatability 20 % Meas 1 20 % Meas 2 20 % Meas 3 20 % Meas 4 20 % Meas 5 20 % Mean 60 % Meas 1 60 % Meas 2 60 % Meas 3 60 % Meas 4 60 % Meas 5 60 % Mean 100 % Meas 1 100 % Meas 2 100 % Meas 3 100 % Meas 4 100 % Meas 5 100 % MeanQuantifying the sources of uncertainty (9)
Reproducibility 20 % Meas 1 20 % Meas 2 20 % Mean 60 % Meas 1 60 % Meas 2 60 % Mean 100 % Meas 1 100 % Meas 2 100 % Mean 20 % Meas 4 60 % Meas 4 100 % Meas 4 20 % Meas 3 60 % Meas 3 100 % Meas 3 20 % Meas 5 60 % Meas 5 100 % Meas 5Quantifying the sources of uncertainty (9)
• UREP
• Treating as “Repeatability” (as per ISO 6789) • We use the ESDM
• ESDM = ESD/SQRT (n) = 0,98/sqrt (5) = 0,436348 N•m
• Treating as “Reproducibility” (preferred option but contrary to ISO 6789)
• We use the ESD • ESD = 0,98 N•m
GUM Steps
• Model the measurement
• Identify and quantify the sources of uncertainty
• Categorize as type A or type B
• Manipulate appropriately to obtain
• Standard uncertainties, u(xi)
• Sensitivity coefficients, ci
• Uncertainty contributor, u(yi)
• Combine to obtain combined standard uncertainty, uc(y)
• Expand to obtain an expanded uncertainty, U, at an appropriate level of confidence
Categorize as type A or type B
• SCAL - type B, not statistically determined
• SRES - type B, not statistically determined
• SCF - type A, statistically determined (standard deviation)
• URES - type B, not statistically determined
GUM Steps
• Model the measurement
• Identify and quantify the sources of uncertainty • Categorize as type A or type B
• Manipulate appropriately to obtain • Standard uncertainties, u(xi)
• Sensitivity coefficients, ci
• Uncertainty contributor, u(yi)
• Combine to obtain combined standard uncertainty, uc(y)
• Expand to obtain an expanded uncertainty, U, at an
appropriate level of confidence
Uncertainty Budget
Torque Wrench Calibration Uncertainty
Budget.xls
GUM Steps
• Model the measurement
• Identify and quantify the sources of uncertainty • Categorize as type A or type B
• Manipulate appropriately to obtain
• Standard uncertainties, u(xi)
• Sensitivity coefficients, ci
• Uncertainty contributor, u(yi)
• Combine to obtain combined standard uncertainty, uc(y)
• Expand to obtain an expanded uncertainty, U, at an appropriate level of confidence
Reporting the result
• The final result is calculated using the “Normal Function” polynomial coefficients
• This is because we want to know the true torque applied to the Reference Standard Torque Transducer when it reads the mean measured value of 350,7 N•m
• The calculated interpolated value was 349,938089 N•m • The calculated measurement uncertainty was
± 1,794160789 N•m
• Rounding the uncertainty to two significant digits gives ± 1,8 N•m
• Rounding the interpolated value to the same number of digits gives 349,9 N•m
• The measurement result is then reported as:
Conclusions
• Both methods in this case prove that the UUT is well within the allowable ± 4% of Maximum (± 14 N•m)
• Using the ESDM (In accordance with ISO 6789) results in the smallest uncertainty (unrealistic??)
• Using the ESD (contrary to ISO 6789) results in the largest uncertainty (realistic??)
• Always use the polynomial for calibrations using the laboratory Reference Standard Torque Transducer
• This will correct for any error on the Reference Standard eliminating the need to apply corrections
• This will solve the problem of the “Applied Torque” not being exactly at the nominal values