Gas circuit breakers (GCBs) are power equipment that play a critical role in protecting transmission systems by interrupting abnormal currents caused by lightning strikes. Traditionally, GCBs have been maintained through time based maintenance (TBM), which involves periodic inspections conducted by maintenance personnel. However, due to growing concerns over the shortage of skilled maintenance workers, there is an increasing demand for a shift toward condition based maintenance (CBM), which requires technologies capable of diagnosing the operational status of GCBs to improve maintenance efficiency. In this study, to secure a sufficient volume of training data for the development of diagnostic technologies for GCBs, we utilized data generated by a one-dimensional computer-aided engineering (1D-CAE) model. Furthermore, to enhance the interpretability of the diagnostic results produced by black-box machine learning models, we developed a counterfactual explanation generation technique. This technique enables the presentation of diagnostic rationales in a form that is sufficiently understandable to human operators.
SAITO et al. (Wed,) studied this question.