AI-supported Condition Monitoring in Practice: Insights from ProKInect

Digitalization in production enables efficient machine monitoring through sensors and IoT. The BMBF-funded ProKInect project aims to implement a manufacturer-independent condition monitoring system. Challenges include data protection and lack of standardization. Condition monitoring detects machine deviations at an early stage and optimizes maintenance. AI methods such as machine learning and fuzzy logic analyze sensor data and predict conditions. ProKInect uses adaptive neuro-fuzzy systems and possibilistic logic to manage uncertainty. A TruLaser laser cutting machine is used as a test platform. AI agents analyze motor current, speed, and sensor data. PySyft ensures secure data transmission and processing. An FMEA identifies error sources and exclusion criteria for accurate diagnosis. ProKInect demonstrates the effectiveness of possibilistic symptom evaluation and AI agents. Challenges such as sensor variability and data interpretation remain. Future research focuses on real-time analysis and federated learning to improve privacy. Ethics and transparency remain key issues.
ProKInect has made progress in AI-based condition monitoring by combining physical models, neural networks, and fuzzy logic. Secure data processing using PySyft and accurate fault diagnosis using FMEA were critical. Future integration of additional data sources and ethical AI design promise further improvements.