Physics-informed quantum neural network for wheel-rail adhesion and optimal slip control

Reliable utilisation of wheel-rail adhesion is critical for heavy-haul locomotives operating near traction limits under variable rail conditions. Existing physics-based adhesion models provide physical interpretability but often require calibration and can be difficult to adapt for real-time optimal slip estimation, whereas purely data-driven approaches may capture nonlinear adhesion behaviour but can produce physically inadmissible predictions outside the training domain. This study develops a physics-informed quantum neural network (PIQNN) for adhesion regime classification and optimal slip prediction using longitudinal creepage, traction response, and rolling speed. The PIQNN is physics-informed by constraining the learned response through a differentiable reduced-order traction surrogate inspired by FASTSIM-Polach wheel-rail contact behaviour, including bounded adhesion mobilisation, speed-dependent friction limits, and peak-slip consistency. The trained model is then embedded as an estimation module within a locomotive traction-control co-simulation framework coupling a GENSYS multibody locomotive model with a Python-based quantum circuit emulation. Representative heavy-haul operating scenarios show that the proposed framework provides stable and physically consistent optimal slip estimates across speed and adhesion regime variations, with improved adhesion utilisation relative to a conventional slip controller within the tested simulation cases.