Flow sensing method for fluid-structure interaction systems via multilayer proper orthogonal decomposition

Flow sensing is widely used to forecast flow field in fluid-structure interaction (FSI) systems. The FSI system with multiple flexible structures usually involves complex unsteady flow and large number of sensors, which makes it difficult to perform flow sensing. In this study, a flow sensing method for FSI systems via multilayer proper orthogonal decomposition (POD) is proposed to achieve real-time forecast of flow field using structural deformation. First, we establish the POD model of structural deformation. To improve model accuracy for flow field, we propose the multilayer POD model, which mainly focus on the local modeling accuracy in the region with complex flow structures. Then, we establish the multilayer model of flow field. Furthermore, the deep neural network model is employed to map the mode coefficients of the structure to all the mode coefficients of the multilayer POD. The proposed method is applied in two FSI systems, including the flow past a flexible filament clamped behind a cylinder and the flow past flexible filament set. Both constant inflow and transient flow conditions are considered. The results indicate that the proposed flow sensing method exhibits excellent spatial-temporal performance, performs accurately in flow properties forecasting, and is suitable for FSI systems with complex flow structures such as coherent vortices.