Safety Monitoring And Analysis Of Subway Structure Based On BP Neural Network

As a way of the modern transportation, subway construction project has become one of the hotspot of national infrastructure with the advantage of large volume, high speed, low pollution, etc. Subway structure deformation in the constructions and services is difficult to avoid, which have great impact on the security and stability of subway tunnel and endanger the entire urban mass transit system. So it is very necessary to monitor subway deformation. Regulation schemes can be put forward timely according to deformation prediction and safety assessment based on the results of subway deformation monitoring, which is of great practical significance to guarantee metro safety.Several key techniques in deformation monitoring of subway structure were studied in this paper, the specific contents are as follows:(1) The common methods about deformation monitoring of subway structure were studied, the causes of subway structural deformation, the main content of metro deformation monitoring and control standards of subway structural deformation were summarized. According to the special environment of subway tunnel, the principle, methods and processes of metro automatic monitoring by using measurement robot were elaborated.(2) The method and common models of times series were analyzed as well as the modeling process of stationary time series. Meanwhile, the basic principles of neural network, including mathematical models of neurons, transfer functions, network structure, learning rules and training methods were explained. Also, the model structure and algorithm flow of BP neural network were discussed in detail.(3) Deformation time series of monitoring point on the section of metro tunnels was analyzed and prediction model of subway structural deformation was built. On the basic of AR (q) time series prediction model and BP neural network prediction model, a new prediction model, BP-time series fusion prediction model was proposed, which compensated time series prediction model by BP neural network. Prediction precision of the three models were analyzed and contrasted connecting with an engineering example. Results indicated that the prediction precision of BP-time series fusion model was ±0.18mm, which was the highest prediction precision among the three models, increasing by 76.6% respectively compared with time series models (prediction precision ±0.77mm).(4) The main factors affecting the safety of subway structure were researched within the field of deformation monitoring. The safety evaluation system of subway structure was established including evaluation set and evaluation index. The weight of each evaluation index was analyzed and the safety evaluation model of subway structure was built by using BP neural network, which was proved to be scientific and reasonable with an engineering example.