|
|
Uncertainty Quantification Methods
Uncertainty quantification can be done using probabistic methods or
deterministic methods. In probabilistic methods, the uncertainty is
represented as a probability distribution function (PDF) or cummulative
distribution function (CDF). See thesis by Mantis
(ASDL, Gatech) where a strong case has been made for the routine use of
CDFs and their propagation using Bayesian statistics.
Deterministic quantification of uncertainty can be
done using interval analysis, Dempster-Shafer theory, convex modelling
and fuzzy computation method. It is only advisable to use deterministic
model in case of inadequate data or computation power. Presently we are
focusing on the probabilistic methods.
The next issue in the quantification of uncertainty is the processing
of raw data available from the measurements. Darmofal et. al. have
suggested Principal Component Analysis (PCA) for this. We are
investigating this matter presently. For a more indepth introduction of
this approach, please refer to the PhD thesis by Garzon here.
|
|
|
|
