基于深度学习技术的地震实时流数据自动处理辅助工具开发

Development of auxiliary tools for automatic processing of seismic real-time stream data based on deep learning technology

以JOPENS系统实时流接收为基础,应用Redis共享内存技术和近年来发展较快的深度学习震相自动识别技术,设计一套可7×24小时不间断稳定接收并实时识别连续地震流数据中P、S震相的系统,为地震台网实时数据处理提供一套辅助工具,并在福建省地震局测震台网128个台站的实时数据流上进行测试。该工具由Redis实时数据流共享模块与深度学习震相到时自动拾取、MSDP震相格式转换3个模块组成,可以实时接收并自动识别台网地震连续波形,生成P、S震相报告,并可导入MSDP人机交互工具进一步处理,在一定程度上可以减轻人工处理工作量。

We provided a set of auxiliary tools for real-time stream data processing of the seismic networks. The tool consists of three modules: the Redis real-time data stream sharing module, the automatic phase picking module by deep learning, and the phase format conversion tool for MSDP. This tool has been tested on the real-time data streams of 128 stations in the Seismic Network of Fujian Province. Its biggest advantage is that it can receive continuous waveforms of the network in real-time and automatically identify and generate P and S phase reports. Phase reports can also be imported into MSDP human-computer interaction software for further processing, which can reduce the manual processing workload to a certain extent.