基于机器学习的双系统强震仪监测信号识别方法  

Machine Learning-Based Dual-System Strong Earthquake Instrument Monitoring Signal Identification Method

作　　者：孟蔓菁 李博 Meng Manjing;Li Bo(Research Institute of Building Sciences(Wuhan)Investigation and Design Co.,Ltd,Wuhan,Hubei 430071;Shanghai Research institute of Building Science Co.,Ltd)

机构地区：[1]建科(武汉)勘测设计有限公司,湖北武汉430071 [2]上海市建筑科学研究院有限公司

出　　处：《计算机时代》2025年第4期35-40,共6页Computer Era

摘　　要：针对地震监测中存在的误报率持续偏高问题,本研究提出了一种可接收双源信号的强震监测仪,并在此基础上给出一种双源信号地震波类型的识别方法。运用机器学习方法进行双源信号的一致性判断与提取,并构建基于分解的多目标进化算法(MOEA/D)与一维卷积神经网络(1D-CNN)相结合的多目标优化模型。该模型克服了以往传统强震仪因采用单类型加速度计传感器采集地震信号,而导致误触发率较高、识别精度欠佳的问题。经实践验证,采用本文方案进行地震监测的准确率能够达到92.38%。In response to the high false alarm rate existing in earthquake monitoring,this study proposes a strong earthquake monitoring instrument capable of receiving dual-source signals is proposed.Based on this,a method for identifying the types of earthquake wave from dual-source signals is introduced.This method utilizes machine learning for consistency judgment and extraction of dual-source signals,and employs a multi-objective optimization model combining the decomposition-based multiobjective evolutionary algorithm(MOEA/D)and one-dimensional convolutional neural network(1D-CNN).This approach addresses the drawbacks of traditional strong seismic instruments,which typically use a single-type accelerometer sensor that often results in high false triggers and inaccurate identification of seismic signals.Through practical verification,the accuracy of earthquake monitoring using the scheme proposed in this paper can reach 92.38%.

关 键 词：强震监测 双源信号识别 MOEA/D 1D-CNN 

分 类 号：P315.61[天文地球—地震学]