CRTA Cross-Dataset Fault Diagnosis Artifact

This record contains the code, preprocessed data, result summaries, and reproducibility materials for the paper "CRTA: An Applied Cross-Dataset Benchmark and Consistency-Regularized Alignment Study for Fault Diagnosis". The artifact supports cross-dataset vibration fault diagnosis experiments on CWRU, Paderborn, and an industrial Nangang testbed dataset. It includes five transfer tasks: CWRU to Paderborn, Paderborn to CWRU, CWRU+Paderborn to Nangang, Nangang to CWRU, and Nangang to Paderborn. The code package contains paper source files, training scripts, analysis scripts, plotting scripts, paper figures, summary CSV files, per-run metric files, and reproducibility documentation. The data package contains preprocessed z-score segment data with X.npy, y.npy, and meta.json files for each dataset. The main experiment protocol uses five random seeds, file-disjoint source validation splits, and target-domain labels only for final evaluation. Target-domain samples are used without labels for domain adaptation losses.