Development and Validation of an Integrated Test Platform for Earthquake Early Warning Systems in Major Engineering Infrastructure

Earthquake early warning (EEW) algorithms intended for high-speed railways, reservoir dams, converter stations, and other major engineering infrastructure must be tested under realistic signal and interference conditions before field deployment. Conventional offline simulations can evaluate individual algorithms, but they do not reproduce the complete operational chain from multichannel waveform input to embedded processing, communication, warning output, and performance assessment. This study develops an integrated EEW test platform based on a National Instruments (NI) CompactRIO (cRIO) embedded architecture. The platform combines deterministic Field-Programmable Gate Array (FPGA)-based acquisition of up to five three-component stations, real-time processing and alarm control, Transmission Control Protocol/Internet Protocol (TCP/IP) communication, and server-side software for waveform display, event management, storage, logging, and automated test reporting. A progressive warning workflow is implemented in which the initial single-station estimate is updated by two-station constraints and then refined using three to five triggered stations. Validation combines MATLAB-based algorithm simulation, waveform replay through an NI signal generator, embedded execution, network transmission, and server-side supervision. The proposed platform bridges offline algorithm testing and complete functional testing by integrating acquisition, embedded processing, communication, warning generation, and server-side supervision within a unified cRIO-based framework, enabling complete end-to-end evaluation of the EEW operational workflow. The test dataset contains 100 field-relevant cases, including 79 natural earthquakes with local magnitudes M_L 3—7 and 21 non-seismic interference cases. Within this dataset, all earthquake cases completed the warning chain without missed detections, and none of the interference cases generated a false alarm.