Abstract
This research investigates the application of explainable quantum machine learning (QML) for predictive maintenance in the railroad industry. By utilizing Ground-Penetrating Radar (GPR) data to characterize subsurface track conditions (Ballast Fouling Index, Ballast Thickness Index, Layer Roughness Index, and Moisture Likelihood Index), a quantum neural network (QNN) model was developed to predict track geometry (profile and alignment) defects in a Class 3 railroad track. Shapley Additive Explanations (SHAP) were employed to analyze the feature importance and the model's decision-making processes to ensure model interpretability. The QNN model correctly predicted 42 out of 55 test data points. SHAP analysis identified the ballast fouling index and layer roughness index as the most influential parameters, aligning with engineering expectations.
