Reliable condition monitoring is essential for ensuring the safe and stable operation of industrial robots. However, deep learning methods often require substantial computational resources, while conventional machine learning methods may have limited capability in capturing complex sensor patterns. To address these issues, this paper proposes a lightweight industrial robot anomaly diagnosis framework based on multi-domain feature fusion and XGBoost classification. Time-domain statistical features and frequency-domain FFT features are extracted from multichannel sensor signals and fused into a compact feature representation for binary anomaly classification. Experiments on the Voraus-AD dataset demonstrate that the proposed framework achieves an Accuracy of 93.02%, an F1-score of 90.76%, and an AUC of 0.9845. Comparative experiments with Logistic Regression, Support Vector Machine, and Random Forest further validate the effectiveness of the proposed feature fusion strategy. The results indicate that the proposed framework provides an efficient solution for industrial robot condition monitoring in resource-constrained scenarios.

Industrial robot fault diagnosis,anomaly detection,multi-domain feature fusion,sensor signal processing,Xgboost