Hybrid evidence-and-fuzzy uncertainty propagation under a dual-level analysis framework

With the development of modern computing technology, the uncertainty propagation theory plays an increasing important role in the thermal engineering practice. Under this context, our paper proposes an efficient dual-level framework for hybrid uncertainty propagation analysis. Two kinds of epistemic uncertainties are considered simultaneously in inputs, which are respectively modeled as evidence parameters with basic probability assignment and fuzzy parameters with membership function. The hybrid uncertainty characteristic of output response is interpreted by the interval-type mean value with fuzzy bounds, where the interval mean value is derived by focal element subintervals in the first-level evidence uncertainty analysis, and the membership function of fuzzy bounds is constructed by cut-set operation in the second-level fuzzy uncertainty analysis. In order to enhance the computational efficiency for cross-extreme-value prediction, a dual-level parameter perturbation method (DPPM) with small computational cost is developed, where the subinterval dividing strategy can be adopted in sub-DPPM to further improve the computational precision. By comparing results with the direct optimization method, two examples prove the effectiveness of proposed method in engineering application.