Structural defect detection is an important problem in civil engineering. Full Waveform Inversion (FWI) recently has been further developed to address this problem by emitting waves to the building attached with sensors and reconstructing sensor signals-much like a CT scan. However, decoding the defects from sensor signals is much time consuming and mathematically impossible, since the corresponding inverse problems are difficult to solve and usually are ill-posed in real engineering applications. To solve these issues, data-driven approaches from deep learning have been investigated by researchers. Data-driven surrogate models like DeepONets, Fourier Neural Operators, and PINNs show strong strengths in computational efficiency than the classical wave equation solvers. Additionally, a well-designed regularization network is also able to address the ill-posedness of wave inversion. Thus, it will be promising to solve the wave equation by combining the surrogate models and different Machine Learning approaches. In this master thesis, we focus on applying DeepONets architecture to wave equations and analyzing the results acquired by it. Initially, various information about wave equations and traditional solvers is introduced. Moreover, we get our data from simulations executed at high-capacity GPU servers. Different DeepONets architectures (Stacked and Unstacked) with various subnetworks (FCNN, CNN) are afterwards implemented to solve the equation. In the end, each approach is evaluated and subsequently, the best one is emphasized.     «

Structural defect detection is an important problem in civil engineering. Full Waveform Inversion (FWI) recently has been further developed to address this problem by emitting waves to the building attached with sensors and reconstructing sensor signals-much like a CT scan. However, decoding the defects from sensor signals is much time consuming and mathematically impossible, since the corresponding inverse problems are difficult to solve and usually are ill-posed in real engineering applic...     »