Machine Learning-Based Fault Detection for UR3 Collaborative Robot: A Multimodal Data Analysis Approach

Abstract

Collaborative robots (cobots) are increasingly deployed in industrial environments, necessitating robust fault detection systems to ensure operational safety and efficiency. This study presents a comprehensive machine learning framework for fault detection in UR3 collaborative robots using multimodal sensor data. We analyzed a dataset comprising 7,409 samples with 20 features, including joint currents, temperatures, and velocities. Five machine learning algorithms were evaluated: Logistic Regression, Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), and Decision Tree. Our results demonstrate that KNN achieved the best overall performance with an accuracy of 94.42%, F1-score of 0.446, and AUC of 0.812. Feature importance analysis revealed that joint currents (J3, J2) and joint velocities (J5) are the most critical indicators for fault prediction. Dimensionality reduction techniques (PCA and t-SNE) confirmed distinct separability between normal and fault conditions. This work provides valuable insights for developing predictive maintenance systems in collaborative robotics applications.