Have you read the
UKAS M3003 ?
it is quite long i know but it explains how to measure the uncertainty.
to sum things up, it goes as simple as this :
Define all sources of errors (Uncertainty from calibration certificate, repeatbility, ....etc.)
find the probability distribution, and based on it then find divisor.
find the coefficient of sensitivity
then find the combined uncertainty.
to understand the probability distribution and coefficient sensitivity, you have to read
the UKAS M3003