Data-Driven Approach for Detection of Physical Faults and Cyber Attacks in Manufacturing Motor Drives

Abstract

In recent decades, the utilization of digital control units and communication networks in modern manufacturing systems has increased significantly. These valuable and safety-critical systems are facing new threats from physical and cyber domains. There is still a gap between studies of physical faults and cyber-attacks on motor drives. it is critical for future motor drives in manufacturing systems to develop attack and fault detection and diagnostic solutions to guarantee system safety and security. To narrow this gap, this paper proposes a data-driven method for detecting and distinguishing cyber-attacks and some common physical faults for manufacturing motor drives. The proposed method integrates the PCC line current spectra and four widely used data-driven classifiers to detect and distinguish cyber-attacks and physical faults. We form a comprehensive case study to validate the proposed methods, including sophisticated false data injection attacks and the two most common physical faults, inter-turn short circuit faults and bearing faults. The final testing results suggest that the proposed method could achieve 95% or higher detection accuracy.