Railway transportation demand is rapidly increasing, which requires enhanced prognostics and health management (PHM) approaches for effective predictive maintenance planning. There has been a significant evolution in prognostics and prediction techniques, e.g. reliability-, damage accumulation-, data analytics- and condition-based prediction. The ability to predict damage progression requires an understanding of initiation criteria or root causes through accurate stress, displacement, and deformation field analysis. Existing commercial multiphysics simulation tools lack the source code accessibility, customisation flexibility, and computing platform portability essential for railway prognostic applications. These limitations are addressed by implementing a physics-informed neural network (PINN) using PhysicsNeMo framework, as a computational instrument for 3D wheel-rail contact analysis. This PhysicsNeMo framework integrates classical Hertzian contact theory with neural network capabilities, enabling accurate prediction of stress and displacement fields critical for damage accumulation assessment. Validation demonstrates appreciable accuracy in predicting contact mechanics parameters essential for prognostic health management. The framework provides railway maintenance practitioners with detailed visualisation of stress distributions and displacement fields, enabling identification of fault root causes and damage propagation mechanisms. Results establish a foundation for intelligent predictive maintenance strategies, supporting the railway industry’s transition toward physics-informed prognostics for improved asset health management and operational reliability.