A Federated Learning Framework for Quality Inspection and Anomaly Detection in AI-Enabled Manufacturing Systems for Sustainable production

This study introduces a blockchain-enabled federated learning system for intelligent, privacy-preserving quality assessment in dispersed industrial contexts. To handle non-IID and heterogeneous datasets, the suggested system uses multimodal feature extraction from visual, vibration, and temperature data with adaptive federated optimisation. Smart contracts and a blockchain trust layer avoid data tampering and model poisoning by verifying model modifications transparently. Experimental assessments employing casting, electronics, and assembly datasets show 32% accuracy, 14.2% precision, and 16.8% recall advantages over centralised models, coupled with 40% lower communication costs. Results confirm system scalability, energy efficiency, and resilience, making it suitable for real-time industrial implementation. The proposed system provides safe, interpretable, and decentralised Industry 4.0 quality assurance and predictive maintenance.