Federated Learning-Enabled Edge Intelligence for Sustainable Smart Cities

Fault detection in industrial process control systems has strong requirements in real-time because detection is required to meet safety, efficiency and continuity requirements. Historically, the methods of monitoring employed tend to be rule-based, dedicated to thresholds, and dependent upon manual inspections, which are inefficient and deficient with respect to scalability and the discovery of minor or incipient faults. Such deficiencies result in more downtimes, inferior quality of the product, and costly operations, particularly in highly intricate and immense industrial scenes. To overcome these problems, this paper introduces an AI-Based Fault Detection Framework that combines deep learning models (CNN-LSTM hybrid) with distributed signal processing at the edge to perform continuous monitoring of sensor, and actuator data streams. The system takes advantage of spatio-temporal feature extraction, probabilistic thresholdbased anomaly detection, and retraining to the changing process conditions. It is a resilient/ fault-resistant orchestration level that enables the system to withstand noisy data, sensor value drop off, and delays in communication channels; hence it is robust enough to be deployed in the real-world industrial environment. Results on benchmark industrial datasets and a pilot-scale process plant testbed show that the proposed AI framework can achieve high performance in fault detection accuracy (97.2%), minimize false positive detections compared to conventional ML approaches (34.6 percent reduction), and decrease average response time by fault detection by 41 percent. These findings verify its effectiveness in enhancing reliability and responsiveness, thus, reducing downtime and optimizing operational resilience. The results indicate that a deep learning based distributed signal processing can advance fault detection technology to be capable of real-world scalability and intelligent decisions that can enable industries to transform into smart, autonomous and resilient process control systems.