In this work, a physics-informed neural network is employed to run fatigue optimization for an Euler-Bernoulli beam under dynamic loading. The performance of the network is then compared to the optimization carried out via a finite-element solver. Initially, the physics- informed neural network capabilities were explored through solving steady state and transient Euler-Bernoulli beams, including free vibration and forced vibration under dynamic loading. The results delivered by the network were then assessed by those obtained by the finite element solver. The network was expanded to allow for the optimization of the cross sectional parameters of the beam to minimise fatigue accumulation. The network was found to be able to deliver reliable results with a significant performance improvement (up to 600 times faster) compared to the finite-element-based optimization. This highlights the potential capabilities of neural networks as a tool for solving optimization-based problems.     «

In this work, a physics-informed neural network is employed to run fatigue optimization for an Euler-Bernoulli beam under dynamic loading. The performance of the network is then compared to the optimization carried out via a finite-element solver. Initially, the physics- informed neural network capabilities were explored through solving steady state and transient Euler-Bernoulli beams, including free vibration and forced vibration under dynamic loading. The results delivered by the network...     »