ISSN 2996-8223
International Journal of Computer Science and Technology | Vol. 3, No. 8, August 2022 | pp. 57–64
DOI: 10.46882/2022/IJCST/000211
Article Type: Original Research Paper
Title: Real-Time Edge Analytics for Automated Fault Detection in Industrial Cyber-Physical Systems
Names of Authors: Siddharth Mehta¹, Hiroshi Tanaka²
Authors’ Affiliations: ¹Department of Automation and Robotics, Tech-Vanguard Institute, Mumbai, India; ²Industrial IoT Solutions Group, Nippon Automation Corp, Yokohama, Japan
Abstract: High-speed assembly operations within modern manufacturing frameworks require instant anomaly recognition to protect equipment and avoid operational pauses. Traditional architectures transmit raw sensory feeds directly into central databases, inducing substantial network latency delays. This article introduces an adaptive edge-computing framework optimized to run real-time anomaly assessments directly at the device layer. The system relies on a structural One-Class Support Vector Machine (1-SVM) engine designed to flag voltage, thermal, and vibration anomalies. The algorithm monitors machine telemetry patterns continuously and estimates deviation vectors over a moving observation frame. We validated this decentralized architecture on a live multi-axis CNC testbed containing 45 high-frequency sensors. Empirical results prove the system identifies early degradation indicators within a strict latency threshold of ±12 ms. Operating at the device layer reduced the baseline outbound telemetry bandwidth consumption by 74.5%. This shift effectively minimizes cloud storage overhead requirements while preserving edge-layer processing accuracy. A standard t-test confirmed the consistency of the classification precision across multi-day manufacturing runs (p < 0.01).
Keywords: Cyber-Physical Systems, Edge Computing, Fault Detection, Industrial Automation, One-Class SVM, Real-Time Analytics
Manuscript Timeline: Received: May 11, 2022; Revised: June 20, 2022; Accepted: July 15, 2022; Published: August 04, 2022
Subscribe to read the full article: https://internationalscholarsjournals.org/subscribe-to-read