Identification of Viscoelastic Material Properties Using a Physics-Informed Neural Network Framework with Non-Uniform Complex Modulus
Please login to view abstract download link
In this work, the physics-informed neural network (PINN) framework applied to the identification of viscoelastic material properties in a cantilever beam is presented, considering both the real and complex parts of the elastic modulus. The model leverages the Euler-Bernoulli beam theory to characterize the behavior of a structure with an embedded viscoelastic layer. By formulating the problem in the frequency domain, the approach integrates experimental data and the governing partial differential equations (PDEs) into the loss function, allowing the estimation of both the spatial variation of the elastic modulus and the damping properties of the material. The proposed methodology is validated experimentally, where displacement measurements are used to reconstruct the modulus profile of the material. The use of a non-uniform complex modulus allows for accurate modeling of viscoelastic effects, capturing both stiffness and energy dissipation across the structure. The results demonstrate the capability of the PINN approach to provide an accurate surrogate model for the identification of material properties in continuum mechanics, offering promising applications in non-destructive testing and structural health monitoring.
