Research on a Landslide Stability Prediction Method

滑坡工程的稳定性关乎施工工程的安全健康状况，分析滑坡的变形监测数据，并进行准确的滑坡变形预测，对可能出现的安全隐患做出预判有很重要的意义。本文利用奇异谱分析(SSA, Singular Spectrum Analysis)方法分离出滑坡变形监测数据中的趋势项与周期项；利用神经网络与小波方法对趋势项与周期项进行预测与重构，二者结合得到重构后的预测值。最后通过施工工程中滑坡变形监测数据进行分析，预测结果表明经过奇异谱分析之后的小波–神经网络预测模型效果更加稳定，优于单独的小波–神经网络模型预测结果。The stability of landslide engineering is related to the safety and health status of construction projects. Analyzing the deformation monitoring data of landslides and making accurate predictions of landslide deformation are of great significance for predicting potential safety hazards. This article uses Singular Spectrum Analysis (SSA) to separate trend and period terms from landslide deformation monitoring data; Using neural networks and wavelet methods to predict and reconstruct trend and period terms, and combining the two to obtain the reconstructed predicted values. Finally, through the analysis of landslide deformation monitoring data in construction projects, the prediction results show that the wavelet neural network prediction model after singular spectrum analysis is more stable and superior to the prediction results of the standalone wavelet neural network model.