Research On Structure Deformation Prediction And Security Assessment Model Of Underground Engineering

With the development of China’s economy,the development and utilization of underground space is also becoming more huge and extensive,and the scale of the project is also more and more large and complex,which puts forward higher requirements for the construction and operation safety of underground space.With the rapid development of information technology and control technology,automatic monitoring system has become an important means to ensure the safety of underground space structure construction and operation.Through the analysis and processing of the monitoring data of the monitoring system,the hidden safety hazards in the project can be discovered in advance so that corresponding countermeasures can be taken,which is an important basis for the subsequent construction engineering design,and has also great reference value for engineering construction and operation safety.Based on the national key research and development project ‘Urban Underground Large Space Construction Safety Visualization Automatic Monitoring System’,this paper establishes an underground engineering structure deformation prediction and safety evaluation model based on the research on the current situation of the structural deformation prediction and safety evaluation model.Through the analysis of the actual monitoring data,it is found that there are some abnormal,noise and data loss in the original monitoring data,which will have a certain impact on the data analysis and modeling.In order to solve these problems,the3s anomaly handling criteria,EMD-wavelet denoising method,and support vector machine are used to preprocess the monitoring data.Considering the internal characteristics of massive monitoring data and different data components,a TCN-LSTM structure deformation prediction model based on wavelet decomposition is proposed using the advantages of time convolution network and long short term memory in processing time series.The preprocessed data is decomposed by wavelet transform,and the detail component and trend component are determined by component correlation analysis.According to the characteristics of different components and the deep learning model,the detail component and trend component are predicted by TCN and LSTM models respectively.Orthogonal design is used to optimize the model parameters,and then the prediction results of the components are fused to obtain the final prediction results.Finally,aiming at the existing problems of structural safety assessment based on monitoring data,a structural safety assessment model is proposed based on fuzzy neural network.The method proposed in this paper is verified by the actual monitoring data in a construction site,and the main conclusions are as follows:(1)The data preprocessing is realized effectively by using 3s anomaly handling criteria,EMD-wavelet denoising method,and support vector machine,which can provide effective data for the subsequent data analysis.(2)The Temporal convolutional network model can effectively obtain the long-term memory of the sequence,and then extract the internal features of the sequence in depth,which has great advantages in predicting the detail components.(3)Compared with the existing WNN,DBN-SVR,SVR,LSTM and TCN models,the RMSE,MAE and MAPE of the proposed WA-TCN-LSTM model are reduced by 84.55%,77.65% and 54.62%,respectively.(4)The proposed safety assessment model can accurately reflect the safety status of the structure and provide guarantee for the safety of underground engineering structures.