This thesis investigates the application of Physics-Informed Neural Networks (PINNs) to the Euler-Bernoulli beam equation, with a focus modelling structural dynamics problems. PINNs offer a powerful framework for solving partial differential equations (PDEs) by embedding physical laws directly into the loss function of a neural network. The study begins by reproducing results from a recent benchmark paper on beam dynamics using PINNs, validating the methodology and implementation. A series of experiments are then conducted to explore the effects of different network architectures, optimiser choice, learning rate strategies, and loss weighting on model accuracy and convergence. Particular attention is given to the challenges of overfitting, training instability, and generalisation across the spatio-temporal domain. The thesis also extends the baseline model to simulate more complex loading scenarios, including point loads and spatially varying forces, to better reflect structural dynamics problems. Results demonstrate that while PINNs can accurately approximate beam deflections under idealised conditions, their performance is sensitive to hyper-parameter tuning and problem formulation. The findings highlight both the potential and limitations of PINNs in structural dynamics and suggest directions for future research in structural dynamics problems and physics-informed machine learning.