Edge AI Powers Industrial Predictive Maintenance Demo

The development of a groundbreaking demonstrator has been a significant milestone in the realm of industrial maintenance, drawing inspiration from the successful outcomes of the iCampus project ForTune. This innovative creation seamlessly integrates sensor technology, data acquisition, and AI-driven data evaluation to revolutionize condition monitoring and predictive maintenance practices. The brainchild of Fraunhofer IPMS, the demonstrator represents a leap forward in preventive maintenance strategies for plants and machinery.

Dr. Marcel Jongmanns, the esteemed project leader at Fraunhofer IPMS, sheds light on the transformative nature of this solution, stating, "Our cutting-edge approach empowers precise monitoring of machine conditions through the utilization of sensors and intelligent data analysis. By embedding AI capabilities within the sensors, we can proactively identify potential damages, optimize maintenance schedules, and minimize operational downtime."

The demonstrator itself serves as a tangible embodiment of technological prowess, featuring a scaled-down conveyor belt that serves as a canvas for showcasing a state-of-the-art toolbox designed for industrial plant monitoring. Leveraging multimodal sensors, the system captures accelerations in various spatial dimensions alongside corresponding rotation rates. Furthermore, the inclusion of magnetic field sensors and acoustic or ultrasonic sensors enhances the monitoring capabilities of industrial equipment, with a specific focus on belt tension and jam detection.

Central to the demonstrator's functionality are AI models that have been meticulously crafted through extensive data analysis, enabling precise predictions of potential damages. Real-time calibrations further enhance the accuracy of these models, ensuring adaptability to diverse operational environments. The integration of in-house sensors with an edge computing unit based on RISCV architecture empowers efficient data processing at the point of use, facilitating complex AI operations and real-time analysis.

By overcoming traditional limitations in computing power for real-time modeling within embedded systems, the Fraunhofer IPMS system sets a new standard for predictive maintenance technologies. The seamless incorporation of changing environmental factors into the analysis process enhances the system's predictive capabilities, allowing for the integration of a multitude of sensors and significantly improving the accuracy of equipment condition forecasts. Collaborations with industry leaders like Vetter Kleinförderbänder GmbH underscore the growing interest in such cutting-edge systems.