Cylindrical journal bearings, a class of hydrodynamic bearings, are widely employed in industrial applications due to their ability to support high loads and minimize energy losses from friction. Nonetheless, these bearings are susceptible to issues, such as wear and ovalization, which can alter their circular geometry and affect their vibrational characteristics. Identifying these faults is essential to mitigate their impact on industrial processes. This paper introduces a methodology that utilizes Fourier coefficients to identify non-circular profiles in hydrodynamic bearings. The bearing model and its numerical analysis are implemented using the Finite Volume Method. At the same time, the influence of failures is integrated into a rotating system modeled by Finite Element Method. Based on these models, a dataset is built with common failures in applying hydrodynamic bearings, namely: ovalization, wear, and ovalization with wear. This dataset is used to train a Multi-Layer Perceptron (MLP) neural network to identify the coefficients of a function responsible for describing the different types of non-circular profiles. The bearing profiles are identified indirectly through coefficients rather than discretization points, ensuring that the identification process remains independent of the discretization. This approach is particularly advantageous, since it avoids the complexity to adjust a proper discretization for different the bearing’s diameter. The tests performed in this paper show results with considerably accuracy for the three fault conditions, demonstrating that this approach is promising to fault identification in rotating machines supported by hydrodynamic bearings.