SAE-SVM AI Hybrid System for Fault Diagnosis of Industrial 6 DOF Robot based on Acceleration of Wrist

Abstract

Industrial robots with six degrees of freedom (6 DOF) have become indispensable in manufacturing due to their precision, flexibility, and efficiency. However, their reliability is challenged by faults arising from dynamic operational stresses, particularly in joints and wrists, which directly affect performance and safety. Traditional fault diagnosis techniques often struggle with the complexity and nonlinearity of acceleration signals, highlighting the need for advanced artificial intelligence (AI)-based solutions. This review focuses on the potential of a hybrid Sparse Autoencoder–Support Vector Machine (SAE-SVM) system for effective fault diagnosis using wrist acceleration data. SAEs enable automated feature extraction by learning hierarchical representations, while SVMs provide robust classification of fault states, making the hybrid framework suitable for handling high-dimensional and noisy signals. The paper critically examines existing approaches, highlighting their strengths in accuracy and robustness while identifying persistent challenges such as data scarcity, real-time implementation, and the lack of model interpretability. Special attention is given to the research gap in explainability, where current systems operate as black boxes, limiting their adoption in safety-critical environments. Future opportunities include integrating explainable AI, expanding benchmark datasets, leveraging edge computing for real-time deployment, and developing scalable solutions adaptable to varied robotic systems. Overall, the SAE-SVM hybrid approach offers a promising pathway toward reliable, intelligent, and transparent fault diagnosis in industrial robotics.