,
Assistant Professor, Department of Electrical and Electronics Engineering, Knowledge Institute of Technology, Salem, Tamil Nadu, India, Research Scholar, Division of Robotics Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
,
Associate Professor and Head, Division of Robotics Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
Assistant Professor, Department of Electrical and Electronics Engineering, Knowledge Institute of Technology, Salem, Tamil Nadu, India, Research Scholar, Division of Robotics Engineering, Karunya Institute of Technology and Sciences, Coimbatore, Tamil Nadu, India
The robotic arms deployed in contemporary Industry 4.0 systems are becoming more and more dependent on interconnected sensors, controllers, and information-driven smartness to accomplish more intricate manufacturing activities. Although predictive techniques of maintenance have enhanced the ability to detect mechanical failures, the systems are susceptible to cyber-physical vulnerabilities like spoofing by sensors, replay attacks, and manipulation of signals. Current solutions normally solve the problem of mechanical fault diagnosis and cyber intrusion detection separately, which is not able to offer overall protection to smart robotic systems. To address this shortcoming, this paper presents HT-GP-SecureFusion, a combined Hierarchical Temporal-Graph Physics-Guided Fusion Network that will be used to simultaneously forecast mechanical failures and detect cyber-physical intrusions in robotic arms. The new framework is based on the previous architecture of HT-GP-FusionNet and includes a Cyber Anomaly Branch (CAB) and hierarchical temporal encoders along with a kinematic graph transformer. Latent regularization in the form of physics-guided is included to ensure that the actions of robot systems are consistent with their mechanical behavior. Moreover, adversarial generative augmentation is used to generate real-world mechanical fault and cyber-attack conditions, which enhances the model's robustness and its generalization. The jointly optimized multi-objective loss function is used to guarantee the performance of fault classification, cyber anomaly detection, and physical consistency. The effectiveness of the proposed framework is proved by experimental assessment on multi-sensor robotic arm datasets with normal operations, mechanical faults, and with the simulation of cyber-attack scenarios. The framework achieves robust dual-task performance: fault detection with 95.9% accuracy, 99.2% precision, and 96.1% recall (F1=91.6), and cyber-attack detection with 97.3% accuracy, 98.6% precision, and 95.9% recall (F1=97.3). The model substantially outperforms baseline deep learning approaches (LSTM, CNN-LSTM, Transformer), demonstrating +45.8% improvement in fault F1 and +35.4% improvement in cyber F1 over the best baseline. Expected Calibration Error values below 0.070 confirm well-calibrated uncertainty estimates critical for safety-critical industrial deployment. These findings substantiate the fact that HT-GP-SecureFusion is an effective tool to detect both mechanical and cyber intrusions. The suggested framework is thus a reliable-resilience proposal to the safe and dependable robotic systems in the next-generation smart manufacturing premises.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.
