New Proposal for Uncertainty Estimation in Indirect Measurements with Correlated Input Quantities

The paper deals with uncertainty estimation in indirect measurements. The attention is mainly paid to measurement models whose input quantities are modeled as correlated random variates. Starting from a past experience concerning the use of the unscented transform to overcome known limitations of current recommendations in indirect measurement involving uncorrelated input quantities, the authors are going to present a new proposal capable of extending the cited advantages also when correlated input quantities are concerned with. The new proposal is, in particular, addressed to nonlinear and/or nonanalytical measurement models. After describing key features and implementation issues of the proposal, the results obtained in a number of tests on simulated and actual measurement data are given, which assess its reliability and effectiveness