E. Menga, C. López, S. Hernández, A. Baldomir, I. Romero, M. J. Sánchez
DOI Number: N/A
Conference number: IFASD-2017-132
This paper aims to review and explain fundamental tasks of the UQ process, particularly in the context of non-intrusive methods, which sample the deterministic model at points in the multidimensional input parameter space. Low Discrepancy Design (LDD) and Global Sensitivity Analysis (GSA) are shown to be effective analytical tools to define which parameters are relevant, the so-called Design Variables (DVs) and to quantify how the variance of the output is affected by the variance of the inputs. On the other side, especially for computationally expensive models, the direct use of the analytical model, i.e. Finite Element Model (FEM), can be extremely time-consuming. Hence, for complex systems, it is often necessary to generate first a meta-model and then, using it, complete the sensitivity analysis. The reliability and robustness of the method are demonstrated considering two structures with different level of complexity. In the first example the frequency response of a doble pinned beam is studied and discussed. In the second example the uncertain nonlinear vibrations of the RAM Air Inlet (RAI) system are considered. The stochastic reaction forces acting on the rods of the RAI mechanism are studied, considering prescribed density distributions on the interfaces, in terms of free-plays and stiffnesses, on the excitation force and on its dominant frequency. In this case the modal information is computed in NASTRAN and is translated to a MATLAB code, where the process is completed and the results evaluated. In both cases the relevant parameters are defined and the influence of each one on the variance of the output is clearly quantified. At the end of each example the results are discussed with a deep look at the physical coherence of the GSA Indices. Moreover, the second example gives the opportunity to discuss the use of Polynomial Chaos Expansion (PCE) meta-model in the frame of nonlinear vibration analysis.