Reduced Order Level Modeling of Structure-Based Uncertainty on Fluid Forces for the Dynamics of Nearly-Straight Pipes

Abstract

This investigation is focused on the consideration of structural uncertainties in nearly-straight pipes conveying fluid and on the effects of these uncertainties on the dynamic behavior of the pipes. Of interest more specifically are the structural uncertainties which affect directly the fluid flow and its feedback on the structural response, i.e., uncertainties on/variations of the inner cross-section and curvature of the pipe. A finite element-based discovery effort is first carried out on randomly tapered straight pipes to understand how the uncertainty in inner cross-section affects the behavior of the pipes. It is found that the dominant effect originates from the variations of the exit flow speed, induced by the change in inner cross-section at the pipe end, with the uncertainty on the cross-section at other locations playing a secondary role. The development of a generic model of the uncertainty in fluid forces is next considered by proceeding directly at the level of modal models by randomizing simultaneously the appropriate mass, stiffness, and damping matrices. The maximum entropy framework is adopted to carry out the stochastic modeling of these matrices with appropriate symmetry constraints guaranteeing that the nature, e.g., divergence or flutter, of the bifurcation is preserved when introducing uncertainty. To achieve this property, it is proposed that the fluid related mass, damping, and stiffness matrices of the stochastic reduced order model (ROM) all be determined from a single random matrix and a random variable. The predictions from this stochastic ROM are found to closely match the corresponding results obtained with the randomized finite element model.