基于CNN-LSTM-CBAM模型的地震前兆重力異常檢測研究

打開文本圖片集

中圖分類號：TP183；TP39；P315. 7 2 + 6 文獻標識碼：A

文章編號：2096-4706（2025）08-0041-05

Abstract： This research proposes an anomaly detection method in earthquake precursor gravity data based on the CNNLSTM-CBAMmodel.The anomalydetection inearthquake precursor gravitydata is crucial for improving the timelinessof earthquakepredictions.Itextracts spatial features ofthegravitydata using CNN，anduses theLSTMtocapture long-term dependencyrelationships inthe time series.The CBAMisintroduced toenhance the model's abilityto focusonimportant features，thereby improving anomaly detection performance.Experimental comparisons with the anomaly detection methods suchas AutoEncoder，CNN，LST，andCNN-LSTmethodsshowthattheproposedmodelinthispaperoutperformsotrsin metrics such as MAE，MSE，RMSE，and .This model effectively identifies potential and abnormal dataand providesa reliable foundation forearthquakeriskmanagementandearly waming.Thisresearchofersnewinsights into theanalysisofearthquake precursor data.

Keywords： earthquake precursor anomaly; gravity data; time series; LSTM; Atention Mechanism

0 引言

在印度板塊與歐亞板塊相互作用及太平洋板塊影響下，中國是板塊內地震活動最強烈、頻率最高的地區(qū)之一{I]。（剩余6966字）