Open DC-DC converter fault-diagnosis benchmark with component-batch shift

Dataset and code accompanying: N.A. Mardi, "Component-batch shift in data-driven fault diagnosis of closed-loop DC-DC converters: an open simulation benchmark with a pre-registered evaluation protocol" (submitted to Electrical Engineering, Springer, 2026). Simulated closed-loop buck, boost and two-phase interleaved boost converters with five classes (healthy, capacitor degradation, inductor degradation, switch open, diode open) at three severities (incipient, moderate, severe). Component values are drawn from datasheet-derived production-lot bands (population P0) and from three in-specification batch-shift populations: D1 second capacitor supplier, D2 different production lot, D3 top-of-window ESR. Every shifted test set has an oracle training set drawn from the same shifted population. Simulation only. No hardware measurements are included. Files: four simulation caches (n7_cache_*.tar) with a SHA-256 manifest, code.zip (simulator, generator and analysis scripts), protocol.zip (pre-registration, amendments and decision log) and results.zip (every median and IQR in the paper). See README.md for the file format and how to reproduce the paper's numbers. Licences: data CC BY 4.0; code MIT (see LICENSE-CODE).