Hybrid interval model for uncertainty analysis of imprecise or conflicting information

For many engineering problems with epistemic uncertainty, evidence theory provides a flexible modeling framework to tackle with imprecise and conflicting information. However, the traditional model established under evidence theory framework can solely address uncorrelated or correlated evidence variables, which curtails its application in engineering practices. This paper aims to develop a unified model for evidence theory-based uncertainty analysis with consideration of parameter independence and dependence coexist scenarios. With the aid of subspace decomposition, a hybrid ellipsoid-interval model integrating evidence theory is firstly developed to quantify uncorrelated and correlated evidence variables within a unified framework. To make full use of sample information, a data-driven strategy is further proposed to determine the basic probability assignment of the above focal elements. Subsequently, the belief and plausibility measures are calculated using proposed hybrid ellipsoid-interval model integrating evidence theory. To improve the efficiency and accuracy of belief degree calculation, a region optimalbased ensemble metamodel is further introduced by applying component metamodels together. Eventually, a test example and an aeroengine blade engineering application are investigated to substantiate the effectiveness of the proposed model and method.