Research On Key Technologies Of Building Deformation Monitoring Based On The Fusion Of GNSS And Accelerometer

In recent decades,the Global navigation satellite system(GNSS)has been widely used in surveying,civil aviation,and navigation as a positioning method.Besides,as a displacement detection method in deformation monitoring,GNSS has been applied to landslide monitoring,subsidence survey,industrial measurement,structural health monitoring,and other fields.Compared with traditional displacement detection technology,GNSS-based displacement detection has many advantages,such as real-time,high precision,and weather independence.Buildings are one of the most active places for human activities,so the safety of buildings is closely related to people’s lives and property.The quality of the building,or external factors such as overload,earthquakes,etc.,have an impact on the structure of the building,which seriously threatens people’s lives and property.Therefore,the monitoring of these structures is of great significance.GNSS-based deformation monitoring technology meets both the long-term monitoring and short-term monitoring requirements of structures.At present,GNSS can provide real-time kinematic and high-precision results,but its application mode in deformation monitoring is mainly static monitoring mode.The main reason is that the positioning accuracy of a single epoch cannot meet the accuracy requirements of dynamic monitoring due to various observation errors.Many studies have explored possible methods to mitigate the effects of these observation errors.However,up to now,there is no clear model that can completely eliminate these errors caused by complex spatiotemporal characteristics.On the other hand,most of the current research focuses on the GNSS positioning algorithm,and the application-level monitoring algorithm is rarely studied.With the modernization of GPS and GLONASS,the development of the Galileo system,and the full operation of Bei Dou-3,it is of great significance to explore application-level displacement detection methods according to application types.Therefore,the purpose of this paper is to study the dynamic monitoring technology based on GNSS,including the site error mitigation technology,dynamic displacement identification and extraction technology,and fusion monitoring technology.For the above purposes,this thesis aims to: 1)Carry out research on error mitigation technology related to GNSS site monitoring to improve positioning accuracy.2)Exploring displacement detection based on GNSS positioning which can be divided into two categories: one is for structural health monitoring that requires displacement and waveforms,such as bridge monitoring or seismic monitoring;the other is permanent displacement detection,which belongs to the sudden displacement detection category.3)Carry out research on the fusion detection technology of GNSS and accelerometer to improve the sensitivity and frequency range of monitoring.The main work and contribution of this thesis are as follows:(1)A method aiming at real-time sidereal filtering based on coordinate time series window matching between two consecutive days has been proposed.Two sets of short baseline data were collected by different types of receivers in harsh environments to validate the method.The results show the feasibility and effectiveness of the proposed method for site-specific real-time multipath mitigation.The multipath suppression effect of the low-cost receiver is better than that of geodetic receivers.The possible reason is that the low-cost receiver lack corresponding multipath suppression strategies at the hardware level or signal processing level.The influence of the parameters of window matching: template window size and similarity measure on the filtering result is discussed.As the template window size increases,the root mean square of the time series after filtering decreases first and then increases.The optimum template window size is around 30 for both cases.Overall,the elastic measures perform better than the lock-step measures in the window matching.(2)The relationship between unmodeled errors and observation features is explored,and a data-driven method based on machine learning is proposed to mitigate the influence of unmodeled errors on GNSS positioning.Convolution neural network,random forest regression,and support vector regression are adopted as the regression model in this method,respectively.The proposed method is verified using historical observation data of three international gnss service(IGS)stations.For the ionospheric free combined precise point positioning model,the overall accuracy of the east coordinate component is improved by more than 60% and 50% respectively in precise point positioning(PPP)static mode and PPP kinematic mode.For the other two coordinate components,the improvement of PPP kinematic mode is about 30%,while the improvement of PPP static mode is different for different stations.(3)A short-term displacement detection method based on GNSS kinematic positioning and time series segmentation has been explored.The segmentation point obtained by segmenting the detection window is regarded as a possible change point,and a test is conducted to determine whether the segmentation point is a change point.The coordinate time series is reconstructed by the Daubechies wavelet to extract the abrupt components.Field experiments were carried out,and two baselines with different lengths were analyzed.The results show that the method can effectively detect the abrupt change point and extract the displacement.The accuracy of the extracted displacement can reach the sub-centimeter level.Finally,the influence of time series segmentation index and segmentation window size on detection results is discussed.(4)A Bayesian inference model based on GNSS kinematic positioning is proposed to identify and extract displacement from coordinate time series.Markov chain Monte Carlo sampling is used for Bayesian estimation.By investigating the posterior distribution of the designed change point parameter,we can identify the change points.Furthermore,we derive the mean value from the posterior distribution of the mean parameter,and further obtain the displacement.The experimental results show that the significant displacement can be identified clearly,and the small displacement can be identified by adding the interval constraint prior.The accuracy of up-displacement extraction from GNSS real-time kinematic positioning can reach within 2 mm in 15 minutes.(5)Interacting multiple model(IMM)Kalman filter is introduced for structural vibration detection and displacement extraction.For the acceleration data from the accelerometer,different model sets are designed,and the effectiveness of the method is verified through four sets of experiments.The findings show that the damping vibration model is suitable for capturing obvious vibration,the constant acceleration model is suitable for capturing slight vibration,and the constant position model is suitable for static state tracking.For obvious vibration,the IMM Kalman filter can be effectively performed for displacement extraction and vibration detection.For slight vibration,the performance of the IMM segmental integration is good.For the displacement data from high-frequency GNSS observations,a stationary state model and a vibration state model are designed to represent the vibration state and static state of the structure.To reduce the influence of trend term caused by unmodeled errors,we preprocess original positioning results by a moving average.To reduce the false alarm rate caused by local disturbance,a hold-on mechanism is proposed.Only when the number of the hold-on epoch is greater than the preset hold-on parameter,is it considered that vibration has occurred.In the field experiments,real-time kinematic(RTK)and PPP kinematic positioning modes were used to process high-frequency observations.Then these kinematic processing results are processed by the proposed method.The standard deviation of the displacement is significantly reduced compared to the original positioning output,and the natural frequency of the structure can be identified through the frequency spectrum analysis of the displacement.(6)The fusion algorithm of GNSS and accelerometer for deformation monitoring is reviewed,and the application scenarios of multi-rate Kalman filter are analyzed.Multi-rate Kalman filtering is suitable for large displacement monitoring,and the GNSS sampling frequency cannot be too low.For small displacement,the advantage of the high sensitivity of the accelerometer in multi-rate Kalman filter has not been brought into play,that is,the detection of weak displacement will be interfered by GNSS displacement.An interactive multiple-model multiple-route Kalman filter for small displacement fusion monitoring is proposed,in which the low-frequency signal or absolute position is maintained by the Kalman filter of GNSS positioning,and the dynamic characteristics or high-frequency signal is extracted by interactive multiple-model Kalman filter of the accelerometer.Experimental results show that the advantages of GNSS and accelerometers are fully utilized in this method,and sub-millimeter-level small displacements can be detected and extracted.