UCL Discovery

Abstract

This paper considers the treatment of errors in metrology. After a classification of the sources of error into systematic, random, and gross, the treatment of these is divided between management systems and statistical techniques. In the treatment of systematic errors, the emphasis is shown to be on minimising their impact by setting up appropriate documented procedures for calibration. Random errors are seen to be an inevitable occurrence, and the emphasis is on quantifying their effect on the quality of the final results through the use of redundant observations and rigorous least squares techniques. The final classification, the gross error, is potentially the most serious in terms of the effect that this can have on the final solution, and hence the importance is stressed of not simply making check observations, but of quantifying the probability of gross errors remaining undetected.