Central AsianResearch Index

DESIGN AND IMPLEMENTATION STRUCTURE OF PERIPHERAL COMPUTING DEEP NEURON NETWORKS FOR EARTHQUAKE WARNING SYSTEM IN KYRGYZSTAN

This research proposes a cutting-edge hierarchical Edge Computing framework leveraging Deep Neural Networks (DNN) to enhance the efficacy of Earthquake Early Warning (EEW) systems in the seismically active regions of Kyrgyzstan. Conventional cloud-centric approaches suffer from prohibitive data transmission latency and bandwidth bottlenecks, which are detrimental to real-time disaster mitigation. To address these challenges, this study shifts the computational burden of seismic phase identification from centralized clouds to the network edge. We developed a highly optimized, lightweight Convolutional Recurrent Neural Network (CRNN) architecture specifically designed for resource-constrained IoT gateways. By employing advanced model compression techniques, including INT8 quantization and knowledge distillation, the framework achieves high-speed inference on ARM-based hardware without compromising classification accuracy. Furthermore, a collaborative "Spatio-Temporal Voting" mechanism is integrated to cross-validate triggers among adjacent nodes, effectively mitigating false alarms caused by urban noise. Experimental evaluations using the STEAD dataset and historical seismic records from the National Academy of Sciences of the KR demonstrate that the proposed framework reduces end-to-end warning latency by an average of 3.5 to 5 seconds. This advancement provides a vital window for critical infrastructure protection and public safety response, aligning with the "Digital Kyrgyzstan" national development strategy for disaster resilience.

Сведения

Опубликовано
2026 · Т. 614, № 1005, С. 717-725
Язык
Английский
ISSN
1694-7851 · 1694-8505
Издатель
Kyrgyz State University named after I. Arabaev
Аффилиации
Kyrgyz State University named after I. Arabaev