טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
M.Sc Thesis
M.Sc StudentTeper Lina
SubjectEstimation of Uncertainty in a Model for Calculation
Uncertainty of ISO GUM
DepartmentDepartment of Quality Assurance and Reliability
Supervisor Professor Emeritus Amos Notea


Abstract

The importance of uncertainty in measurements is reflected in the international definition of Traceability and in the international Quality Assurance Standard ISO 9000 and the standard for laboratories ISO 17025. ISO published a document entitled "Guide to the Expression of Uncertainty in Measurement" (GUM for short) and accepted throughout the world. According to the GUM, the measured value is obtained by substituting the values of the input quantities in the model function. The present work was undertaken with a view to comparing the above technique with estimation of uncertainties, by means of mathematical modeling based on known parameters distributions.Were developed a model based on the distributions of the uncertainty of the contributing sources, yielding the combined uncertainty, and a method for calculating the ranges of expanded uncertainty for a prescribed confidence level, directly from the combined distribution. This approach  avoids the problematicity in estimating the number of effective degrees of freedom and in a use of the t-distribution in the GUM.The model programmed in MathCAD, was used in analyzing several test examples published by ISO and different accreditation bodies, as well as, measurements carried out under laboratory conditions. Comparison of the model results and those from the GUM shows that the latter one suitable for cases in which the uncertainty contribution are small relative to the measured value and where the expanded uncertainty distribution is symmetric or every close to symmetric.. The proposed model is easy to understand, and based mainly, on the knowledge of the uncertainty source distributions, thereby dispensing with the assumptions and approximations required in the GUM approach. It thus sets the limits for the applicability of the latter.