Experimental Data Compression for GPU-Based Solution of Inverse Coefficient Problem for Vibrational Testing Data

The paper is devoted to an inverse problem of acquiring elastic properties of a thin plate from vibrational testing data. A direct problem is solved as a boundary value problem for transverse movement of a plate with known geometry. The direct problem computes an amplitude-frequency characteristic as a result. The inverse problem of acquiring the elastic parameters from an experimentally obtained frequency response is solved as a nonlinear optimization problem. The computations are performed on GPU. The present work investigates how the loss function depends on the parameters of the material. Based on this study the paper proposes the approach to optimize memory requirements and computation time for the target inverse problem. The convergence study is performed to ensure that the proposed data compression approach provides a stable solution.