Time Series Modeling Method And Application Research Of GNSS Deformation Monitoring Based On Deep Learning

With the development of the national economy,some large-scale engineering projects have emerged.Due to the diversity of the composition of large-scale engineering projects,the difficulty of management and the complexity of the environment,it often leads to some safety accidents.For the safety of engineering,deformation monitoring of large-scale engineering projects is imperative.Traditional deformation monitoring methods include static leveling instruments and fully automatic measuring robots.These tooling methods play a certain role in practical application,but they also have limitations.Based on GNSS technology,it has a series of advantages such as 24/7 uninterrupted,high precision and automation.This technology has also been applied to displacement deformation monitoring of large-scale engineering and disaster warning.Whether it is a globally monitored IGS station or a large-scale GNSS deformation monitoring,it is a long-term and continuous process that accumulates a large amount of time series data propelled along the time axis.These data contain the true deformation information of the monitoring points,so it is especially important for engineering researchers to analyze the true deformation information and trends of the monitoring body from these monitoring data.In the process of analyzing time series data,scholars have proposed many method models,such as time series analysis model,artificial neural network,wavelet analysis and so on.However,these methods are based on small data volume analysis,and can not solve the timing analysis problem of large data volume.Based on the time series data of deformation monitoring acquired by GNSS technology,this paper introduces the deep learning method for time series modeling analysis and discusses the deformation information of monitoring points.The main research contents are as follows:(1)Summarize the common tools and methods of deformation monitoring,and focus on the deformation monitoring system based on GNSS technology.In addition,time series data features generated by the monitoring system are described,and a preprocessing method for GNSS time series data is proposed.(2)Summarize the development process of deep learning,as well as the similarities and differences between deep learning and artificial neural networks.Then,the structure and operation principle of the cyclic neural network used for time series analysis in depth learning are described in detail,and then the current mainstream deep learning framework is simply combed.(3)Through the engineering case,the GNSS deformation monitoring time series of the IGS monitoring station is modeled and applied.It mainly includes the verification of the accuracy and adaptability of the deep learning model,as well as the application of regional settlement based on data from multiple monitoring stations.The results show that before modeling the GNSS deformation monitoring time series,it is necessary to perform data preprocessing on its own characteristics to improve the quality of the data set.In addition,the feasibility and adaptability of the cyclic neural network in the field of GNSS deformation monitoring are verified by experiments on the engineering case data of the IGS monitoring station.Finally,it is verified that this model can provide certain data support in the estimation analysis of regional settlement.