Surface Deformation Prediction of Expressway Combined with PS-InSAR Technology and LSTM Model
To achieve large-scale monitoring and high-precision prediction for time series deformation of expre-ssway,PS-InSAR technology was utilized to gather settlement data of a research segment,and its deformation characteristics was analyzed.A highway surface deformation prediction model based on long short-term mem-ory(LSTM)network was developed for deformation prediction at selected feature points,and its performance was compared with support vector machine(SVM),convolutional neural network(CNN),and back propaga-tion neural network(BPNN)models.Results demonstrate that the PS-InSAR monitoring aligns closely with leveling measurements,with errors under 5 mm.The study section showed a cumulative settlement ranging from 76.615 5 mm to 33.122 4 mm,with a rising trend around the section and a major settlement location in high-fill and deep-cut sections.All models achieved prediction errors within 2 mm,with the LSTM model's root mean square error(RMSE)and mean absolute error(MAE)being less than 1 mm,indicating its superi-or suitability for settlement prediction.The proposed deformation prediction method for highways with the combination of PS-InSAR technology and LSTM model can provide a promising framework for large-scale ex-pressway deformation monitoring and early road safety warnings.
万方数据
维普
