A multi-agent system framework for real-time self-diagnostic condition monitoring of induction furnace equipment in developing industrial economies

Unplanned equipment failures in developing industrial economies result in substantial production losses, elevated maintenance costs, and reduced plant availability. This research presents the design and implementation of a multi-agent system (MAS) based self-diagnostic framework for real-time condition monitoring of induction furnace equipment, demonstrated through a case study at a copper cable manufacturing company in Zimbabwe. The proposed system integrates a heterogeneous sensor network comprising infrared temperature sensors, electromagnetic flow meters, conductivity transducers, and a linear float sensor with a Siemens S7-1200 PLC, OPC server, and a Foundation for Intelligent Physical Agents (FIPA)-compliant JADE (Java Agent Development Framework) multi-agent platform to achieve autonomous fault detection, diagnosis, and notification. Four specialized agents: Diagnosis, Decision, Database, and Notification collaborate via FIPA ACL messaging to process real-time data, execute JESS-based rule inference, persist operational records in MySQL, and deliver automated email alerts to maintenance personnel. Experimental validation on a hardware-in-the-loop testbed demonstrated successful detection of critical fault conditions, including refractory overheating, cooling water flow degradation, water conductivity anomalies, and inductor power deviations. Connecting the contribution within the growing Industry 4.0 and predictive maintenance literature, this work advances the application of agent-based architectures for condition-based maintenance (CBM) in resource-constrained African industrial environments where IoT-enabled solutions remain nascent.