In this project, we are developing a modular simulation-based framework starting with a physics framework based on cylinder-on-flat contact model to first emulate the wheel-rail interaction behaviour and prove applicability. This model runs on GPU using PhysicNeMo or PyTorch DeepXDE frameworks with increasing input parameters and more articulated physical structure. We aim with PINNs to VPINNs to further enhance accuracy, computational efficiency (meshless), robustness while maintaining physical interpretability of the degradation behaviour under data scarcity. Last, compare PINNs framework to classical numerical model (when available) or finite difference-based methods (FEM) for final validation. Final, expand and validate the project objective on different computing paradigm shifts such as PINNs but also novel ones like QML, QNN or QPINNs to convey which paradigm is the most suited or effective based on our specific use case via a set of visualisations, results and benchmarks when it is possible. The final goal is also to show pro and contro of each model and paradigm from the classical Physics-based and hybrid one (PINNs) to quantum machine learning. PhD advisor: Prof. Uday Kumar. I have experience with HPC cluster like Kebnekaise at HPC2N using both PhysicsNeNo framework and Apptainer container, and I have got acceptance by journal and conference in 2025 & 2026: three papers related to PINN in Transportation using GPU cluster. I started in summer 2022 working with Uppsala Physics group (Prof. Olle Eriksson) in Quantum Physics (Advanced Materials simulation), Density functional theorem calculation using both Dardel and Tetralith. Start date: Monday 16 Feb 2026