Abstract
Fluid structure interaction (FSI) problems are becoming highly complex as it has wide range of applications, which assists to model many real-world problems. The finite volume method (FVM) enforces conservation of the governing physics over the designed control volume. Presently, the FVM is the most common discretization technique used within the computational fluid dynamics (CFD) domain. In this day, multiple techniques for simulating the strongly coupled fluid-structure systems numerically is constantly being researched as the CFD analysis approach evolves swiftly. We introduce a segmented neural network-based approach for learning of FSI problems. The FSI simulation domain is discretized into two smaller sub-domains, i.e., fluid (FVM) and solid (FEM) domains and utilize an autonomous neural network for each. A python based scientific library is used to couple the two networks which takes care of boundary data communication, data mapping and equation coupling. The coupled Ansys fluent-transient structural analysis data will be used for training the two neural networks. Changes in the geometrical and material properties of a solid structure such as: bluff/curved - corners/surfaces, physical dimensions, young’s modulus of elasticity and mass moment of inertia will affect the dynamics of the structure.
