An edge-cloud IIoT framework for predictive maintenance in manufacturing systems

Abstract

Despite significant research efforts on Industrial Internet of Things (IIoT) based Predictive Maintenance (PdM) systems, challenges related to the availability of real-time machine operational data, reliable computing-deployment architecture, and implementation in real-time manufacturing environments continue to be major concerns. Hence, this work presents Intelligent PdM (IntelliPdM), an end-to-end IIoT predictive maintenance framework implemented on an edge-cloud platform that processes the real-time heterogeneous data streams (IoT sensors and cameras) and provides intelligent decisions on faults, failures, and maintenance schedules via. endpoints and interactive web user interface (dashboards, alerts/recommendations, and analytics. SmartHome, a synthetic data generation framework was properly configured to generate synthetic data based on limited real-time operational data or open-source benchmark machine health datasets covering all possible industrial fault scenarios. Experimental validations using synthetic data, generated from real-time machine health data collected from a testbed setup at a research center in Western Europe, along with on-site implementation in a large manufacturing unit in Singapore, effectively demonstrate the efficiency of IntelliPdM in delivering accurate and reliable fault diagnostics. Over a 12-months real-time implementation, IntelliPdM demonstrated (i) an accuracy of 93–95%, (ii) 25–30% reduction in maintenance costs, (iii) 70–75% decrease in equipment breakdowns, (iv) 35–45% reduction in downtime, (v) 20–25% increase in production, and (vi) 10x return on investment.