This paper focuses on the use of the possibility theory and of the fuzzy su
bset theory to deal with the uncertainty of the measurements handled in ins
trumentation systems, Methods are described for building fuzzy subsets from
numerical data coming from imprecise "physical" sensors on the one hand, a
nd handling approximate estimations provided by "human" sensors on the othe
r hand. The propagation of the fuzzy representation of acquired measures in
further treatments, such as those involved in performance indicators aimed
at controlling manufacturing production, is also considered.