COMBINING MULTIPLE REFERENCE MEASUREMENT COLLECTIONS INTO A WEIGHTED REFERENCE MEASUREMENT COLLECTION

• US 20060184334A1
• Filed: 02/14/2005
• Published: 08/17/2006
• Est. Priority Date: 02/14/2005
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Sendelbach, Matthew J.
• Granted Patent: US 7,107,177 B2
• Time in Patent Office: Days
• US Class Current: 702/127
• CPC Class Codes: G01D 1/02 giving mean values, e.g. root means square values measuring root mean square values of currents or voltages G01R19/02

Abstract

Combining of reference measurement collections (RMCs) of at least three reference measurement systems into a weighted reference measurement collection (wRMC) is disclosed. Each RMC includes a plurality of corresponding sample measurements, each of which has a measurement value of the same sample. The invention plots corresponding measurement values to generate a plurality of data pairs for each possible RMC pairing. A best-fit line of the plurality of data pairs for each RMC pairing is then generated, and a residual for each data pair is calculated. A weight is then assigned to each sample measurement for each RMC based on the residuals associated with a respective RMC to which the sample measurement belongs, favoring a smaller residual more than a larger residual. A weighted reference measurement is then generated based on the weights, and the measurement value for the respective sample measurement for each RMC.

Claims

1. A method for determining a reliability of a target measurement by combining reference measurement collections of at least three reference measurement systems into a weighted reference measurement collection, each reference measurement collection (RMC) including a plurality of corresponding sample measurements, each corresponding sample measurement having a measurement value of the same sample, the method comprising the steps of:

for each possible RMC pairing, plotting corresponding measurement values to generate a plurality of data pairs;

generating a best-fit line of the plurality of data pairs for each RMC pairing;

calculating a residual for each data pair, each residual associated with each RMC of the RMC pairing to which the respective residual belongs;

assigning a weight to each sample measurement for each RMC, wherein each weight is based on the residuals associated with a respective RMC to which the sample measurement belongs and favors a smaller residual more than a larger residual;

for each respective sample measurement, generating a weighted reference measurement based on the measurement value for the respective sample measurement for each RMC and the weights of the respective sample measurement for each RMC; and

comparing the weighted reference measurement with the target measurement to determine the reliability of the target measurement.

2. The method of claim 1, wherein the best-fit line generating step includes using a fit analysis chosen from one of:

a total measurement uncertainty (TMU) analysis and an ordinary-least squares (OLS) analysis.

3. The method of claim 2, wherein the residual calculating step includes determining the residual for each data pair based on the measurement values of each data pair, a slope and y-intercept of the best-fit line and a residual angle, wherein the residual angle depends on the fit analysis.

4. The method of claim 1, wherein the weight assigning step includes for a respective sample measurement:

calculating an inverse coefficient of the residual for each data pair associated with the RMC to which the sample measurement belongs; and

calculating the weight by summing the inverse coefficients.

5. The method of claim 1, wherein the weight assigning step includes for a respective sample measurement calculating a residual sum inverse coefficient for each data pair associated with the RMC to which the sample measurement belongs.

6. The method of claim 1, wherein the weighted reference measurement generating step includes for a respective sample measurement:

calculating the weighted reference measurement by dividing a first sum of multiples of each measurement value and a respective weight of the respective sample measurement by a second sum of all weights for the respective sample measurement.

7. The method of claim 1, wherein the weighted reference measurement generating step includes for a respective sample measurement:

calculating the weighted reference measurement by dividing a first sum of inverses of multiples of each measurement value and a respective weight of the respective sample measurement by a second sum of inverses of all weights for the respective sample measurement.

8. A system for determining a reliability of a target measurement by combining reference measurement collections of at least three reference measurement systems into a weighted reference measurement collection, each reference measurement collection (RMC) including a plurality of corresponding sample measurements, each corresponding sample measurement having a measurement value of the same sample, the system comprising:

means for, for each possible RMC pairing, plotting corresponding measurement values to generate a plurality of data pairs;

means for generating a best-fit line of the plurality of data pairs for each RMC pairing;

means for calculating a residual for each data pair, each residual associated with each RMC of the RMC pairing to which the respective residual belongs;

means for assigning a weight to each sample measurement for each RMC, wherein each weight is based on the residuals associated with a respective RMC to which the sample measurement belongs and favors a smaller residual more than a larger residual; and

means for, for each respective sample measurement, generating a weighted reference measurement based on the measurement value for the respective sample measurement for each RMC and the weights of the respective sample measurement for each RMC, the weighted reference measurement being compared with the target measurement to determine the reliability of the target measurement.

9. The system of claim 8, wherein the best-fit line generating means includes means for using a fit analysis chosen from one of:

a total measurement uncertainty (TMU) analysis and an ordinary-least squares (OLS) analysis.

10. The system of claim 8, wherein the residual calculating means includes means for determining the residual for each data pair based on the measurement values of each data pair, a slope and y-intercept of the best-fit line and a residual angle, wherein the residual angle depends on the fit analysis.

11. The system of claim 8, wherein the weight assigning means includes means for, for a respective sample measurement:

calculating an inverse of the residual for each data pair associated with the RMC to which the sample measurement belongs; and

calculating the weight by summing the inverses.

12. The system of claim 8, wherein the weighted reference measurement generating means includes means for, for a respective sample measurement:

calculating the weighted reference measurement by dividing a first sum of multiples of each measurement value and a respective weight of the respective sample measurement by a second sum of all weights for the respective sample measurement.

13. A program product stored on a computer readable medium, which when executed, enables a computer infrastructure to determine a reliability of a target measurement by combining reference measurement collections of at least three reference measurement systems into a weighted reference measurement collection, each reference measurement collection (RMC) including a plurality of corresponding sample measurements, each corresponding sample measurement having a measurement value of the same sample, the program product comprising computer program code for enabling the computer infrastructure to perform the following steps:

for each possible RMC pairing, plotting corresponding measurement values to generate a plurality of data pairs;

generating a best-fit line of the plurality of data pairs for each RMC pairing;

calculating a residual for each data pair, each residual associated with each RMC of the RMC pairing to which the respective residual belongs;

assigning a weight to each sample measurement for each RMC, wherein each weight is based on the residuals associated with a respective RMC to which the sample measurement belongs and favors a smaller residual more than a larger residual; and

for each respective sample measurement, generating a weighted reference measurement based on the measurement value for the respective sample measurement for each RMC and the weights of the respective sample measurement for each RMC, the weighted reference measurement being compared with the target measurement to determine the reliability of the target measurement.

14. The program product of claim 13, wherein the best-fit line generating step includes using a fit analysis chosen from one of:

a total measurement uncertainty (TMU) analysis and an ordinary-least squares (OLS) analysis; and

wherein the residual calculating step includes determining the residual for each data pair based on the measurement values of each data pair, a slope and y-intercept of the best-fit line and a residual angle, wherein the residual angle depends on the fit analysis.

15. The program product of claim 13, wherein the weight assigning step includes for a respective sample measurement:

calculating an inverse of the residual for each data pair associated with the RMC to which the sample measurement belongs; and

calculating the weight by summing the inverses.

16. A method for determining a reliability of a target measurement by combining reference measurement collections of at least three reference measurement systems into a weighted reference measurement collection, each reference measurement collection (RMC) including a plurality of corresponding sample measurements, each corresponding sample measurement having a measurement value of the same sample, the method comprising the steps of:

assigning a weight to each sample measurement of each reference measurement collection indicative of an ability of the respective reference measurement system to measure that sample compared to other reference measurement systems;

generating a weighted reference measurement collection (wRMC) by calculating, for each respective sample measurement, a weighted reference measurement based on the measurement value for the respective sample measurement for each RMC and the weights of the respective sample measurement for each RMC; and

determining the reliability of the target measurement based on the weighted reference measurement collection.

17. The method of claim 16, wherein the weighted reference measurement collection generating step includes for each respective sample measurement:

calculating the weighted reference measurement by dividing a first sum of multiples of each measurement value and a respective weight of the respective sample measurement by a second sum of all weights for the respective sample measurement.

18. The method of claim 16, wherein the assigning step includes:

for each possible RMC pairing, plotting corresponding measurement values to generate a plurality of data pairs;

generating a best-fit line of the plurality of data pairs for each RMC pairing;

calculating a residual for each data pair, each residual associated with each RMC of the RMC pairing to which the respective residual belongs; and

setting a weight to each sample measurement for each RMC, wherein each weight is based on the residuals associated with a respective RMC to which the sample measurement belongs and favors a smaller residual more than a larger residual.

19. The method of claim 18, wherein the best-fit line generating step includes using a fit analysis chosen from one of:

a total measurement uncertainty (TMU) analysis and an ordinary-least squares (OLS) analysis.

20. The method of claim 18, wherein the weight setting step includes for a respective sample measurement:

calculating an inverse of the residual for each data pair associated with the RMC to which the sample measurement belongs; and

calculating the weight by summing the inverses.