SpectONet: Physics-Guided Neural Network for Beam Dynamics

A new framework called SpectONet has been created by researchers to address vibration issues in Euler-Bernoulli beams through a physics-guided spectral deep operator network. This innovative approach combines the operator-learning features of DeepONet with physics-informed constraints and Chebyshev-Gauss-Lobatto (CGL) sensor positioning. In contrast to traditional DeepONet, which utilizes uniformly spaced sensors, SpectONet strategically places nonuniform spectral sensors, concentrating more points near the boundaries of the domain. This enhances the finite-dimensional representation of boundary-sensitive structural responses while minimizing the necessary inputs for the branch network. The training objective incorporates the governing beam equation along with initial and boundary conditions to ensure predictions are physically consistent. Numerical tests on three synthetic EBB vibration scenarios validate the method's effectiveness.