An epistemic uncertainty, i.e. the uncertainty due to lack of knowledge – e.g., on the side of model, analysis or experiment. The uncertainty is described by a nominal value and an interval. See also BaseUncertainty uncertainty for general information.
Often, it is inconvenient to provide absolute ranges of a value and, instead, relative or absolute deviations are more practical. In the eFMI standard (Functional Mock-up Interface for embedded systems), for example, tolerances for reference results are defined in a similar way as tolerances for numerical integration algorithms. Due to its generality, this description form of eFMI is used here as well:
Utilizing this information, the lower and upper values of the underlaid interval distribution can be computed in the following way:
tol = max(absTol, relTol * |nominal|), lower = nominal − tol, upper = nominal + tol.
For example, the parameter R of a resistor has the nominal value of 200 Ω, the minimum possible value of 190 Ω and the maximum possible value 210 Ω. Then, the following tolerances can be provided:
unitValue ="Ohm"
,
nominal = 200
,
relTol = 0.05
(and absTol = 0
) or
unitValue ="Ohm"
,
nominal = 200
,
absTol = 10
(and relTol = 0
).
Typically, either a description with relTol or with absTol is used, but not a combination of both. On the contrary, there is an important use case where the description with both relTol and absTol is useful: If reference results are provided (e.g. the results of a simulation or an experiment are required to be within some band around a provided reference solution) then a definition with relTol is often practical together with absTol as band for small values of the reference around zero, where relTol makes no sense.