AI and the World's Hard Problems · Nuclear fusion & energy · arXiv · 2026-10-23
Noise-Robust Quantum-Inspired Synthetic Minority Oversampling for Condition Monitoring and Fault Classification in Industrial and Energy Systems
Class imbalance makes it hard for machine learning to learn minority fault patterns in industrial condition monitoring. The authors extend the Quantum SMOTE oversampling method by adding noise directly into its swap-test-based similarity estimation, so that robustness of the oversampling itself can be tested. Across four public datasets and five classifiers, QSMOTE improved results by up to 170% on an engine-failure dataset, and the effects of noise varied with the noise model and classifier.