Uncertainty Analysis For Fluid-Structure Interaction Simulations Using Information Grids

ResearchSouth will develop a software product for error estimation and uncertainty analysis of computational mechanics code predictions. We will develop methodologies for at least the following methods: interval analysis, sensitivity derivatives, Monte Carlo methods, moment methods and polynomial chaos. An existing and proven fluid-structure interaction code will be used to model many design points by running cases in parallel on information grids. The design points will be set up automatically by perturbation from nominal conditions. This large database will be reduced through an uncertainty analysis using both probabilistic and deterministic uncertainty methods. We will compute test cases for validation including a fluid-structure interaction case on a generic ship hull and compute error bands on the computations. Based on the computational results, the most promising methodologies will be selected for implementation. The Phase I plan is to continue our on-going literature survey on recent developments in stochastic finite elements, formulate the deterministic and probabilistic models for uncertainty, code a new module to implement these formulations, use our existing fluid-structure interaction software for testing the algorithms, validate the solutions, reach conclusions on feasibility and accuracy, formulate a plan for implementing them into commercial codes, and document the Phase I project.