Federated Edge Intelligence: IoT-Enabled Earthquake Detection with Real-Time Learning

Earthquakes pose a significant threat to life and infrastructure, necessitating efficient early warning systems. Traditional earthquake early warning systems (EEWS) rely on centralized cloud-based processing, which suffers from high latency, scalability issues, and data privacy concerns. This paper proposes a Federated Edge Intelligence (FEI)-based IoT-enabled earthquake detection framework that integrates edge computing, federated learning (FL), and deep learning models for real-time seismic monitoring. The IoT sensor layer consists of ADXL335 accelerometers and Arduino Uno microcontrollers, enabling real-time seismic data collection and localized preprocessing. At the edge layer, Long Short-Term Memory (LSTM) networks perform anomaly detection on seismic signals, reducing response time and enhancing earthquake detection accuracy. FL is employed to enable collaborative model training across multiple edge nodes without sharing raw data, ensuring privacy preservation and minimizing communication overhead. The proposed system was evaluated on key performance metrics, including detection accuracy, false alarm rate (FAR), response time, model convergence, and communication overhead. Experimental results demonstrate that the FEI framework achieves a detection accuracy of 96.3%, significantly reduces false alarms, and minimizes response time by 37% compared to traditional cloud-based methods. Additionally, the federated learning model stabilizes within 30 communication rounds, reducing bandwidth consumption while maintaining high model performance. These results highlight the effectiveness of the FEI approach in enabling fast, accurate, and privacy-preserving earthquake detection and intensity forecasting.