Research On Landslide Disaster Situation Assessment And Early Warning Based On Multi-monitoring Field Information Fusion

As one of the most common geological hazards in nature,landslide hazards have the characteristics of great harm,wide area and rapid occurrence,which have caused huge losses to the safety of life and property of the broad masses of the people.Because of the vast territory of our country and the different geological conditions in different areas,the losses caused by landslide disasters are very serious every year.Once a landslide accident occurs on an open-pit slope,it will pose a huge threat to the lives and safety of workers,and bring great losses to mining equipment and other property.Therefore,reasonable emergency measures must be taken to deal with the accident on an open-pit slope to minimize losses.In order to effectively prevent and control landslide risk disasters,corresponding protective measures are taken to minimize the losses caused by disasters.This dissertation attempts to introduce the mature situational awareness theory system in military security and network security into landslide hazard risk assessment and warning.This new theory and method can better deal with landslide prevention and control.Specific research contents include:(1)Multi-field information model of Landslide Evolution Based on spatiotemporal flow data.Aiming at the complex uncertain system of landslide disaster evolution,on the basis of spatio-temporal geological modeling,according to the characteristics of spatial-temporal data changes of landslide disaster evolution,the data integration representation and modeling of stress-strain,temperature,microseismic events and other field information of any unit in space are carried out by using limited monitoring points and spatio-temporal database processing technology.Emphasis is laid on the research of abstract modeling methods for monitoring flow data,such as multi-source heterogeneous,multi-dimensional,etc.in the process of landslide disaster,so as to realize the data access service of internal correlation and space-time unification among various data representations.(2)Landslide disaster situation detection based on information fusion in fog computing environment.Aiming at the characteristics of multi-source heterogeneous flow data of landslide disasters,such as large amount of data,strong mobility and scattered data,a distributed intelligent fog computing platform is proposed.By distributing data processing of local applications to local devices on the edge of the network,the purpose of fast calculation and reducing resource consumption is achieved.In order to analyze multi-source monitoring data more systematically,data preprocessing is carried out by means of stream data time synchronization,noise data cleaning and missing data supplement.Neighborhood rough sets are constructed to process the mixed landslide disaster monitoring data with both nominal and numerical types.The redundant elements are reduced to determine the optimal combination of landslide disaster monitoring fields.Finally,through the information fusion model of SVM-DS,the reduced feature data sets are fused to determine the situation factors of landslide disasters.(3)Landslide disaster situation assessment method based on set pair analysis.The core of situation assessment is the calculation of landslide safety situation value.Based on SVM-DS evidence theory,the data of different monitoring sites at the same monitoring point are fused to obtain the state set of all monitoring points in landslide space.Set pair analysis theory is applied to the situation assessment of landslide disasters.The three-element connection number is expanded to five-element connection number.The state and development trend of landslide are determined by analyzing the change of set pair potential and partial connection degree of landslide disasters.(4)Landslide disaster situation combination forecasting based on variable phase space reconstruction.On the basis of the landslide disaster situation assessment model,the corresponding situation values at the next or more times are calculated.A Bayesian network with m nodes is constructed by using the reconstructed points of phase space as variables,and a single variable chaotic time series prediction model based on Bayesian network is established.Then a multi-variable phase space reconstruction time series situation prediction model based on support vector regression machine is established by using the observation data of multiple variables in the same time period.Finally,two weighted prediction models are synthesized.A combination forecasting model is proposed with high accuracy.(5)Landslide disaster monitoring and early warning system based on situational awareness.From the aspects of system development environment,overall framework and system function modules,the process and main functions of realizing multi-site monitoring landslide hazard situational awareness and early warning system software are introduced,including real-time collection,analysis and prediction of landslide multi-site data,risk index dividing point and index data maintenance of disaster situation assessment and early warning index system,and landslide hazard state.Potential assessment and early warning,landslide disaster data maintenance and reporting,etc.Finally,through the successful application of slope in an open pit mining area,the effectiveness of landslide disaster monitoring and early warning system platform based on situational awareness is illustrated.With the rapid development of Internet of Things technology,more and more heterogeneous multi-source sensor networks are applied in landslide monitoring.They gradually replace the traditional manual monitoring methods and begin to play an important role in monitoring and early warning.Effective establishment of multi-source heterogeneous sensor integrated monitoring system,extraction of landslide multi-field feature information,multi-field information fusion processing and decision analysis are the main methods of landslide prediction.Therefore,in order to achieve the goal of rapid,efficient and accurate assessment and prediction of landslide hazards,and to ensure the intelligent fusion of different monitoring sites for landslide hazards,the research on situation awareness and early warning of landslide hazards based on multi-monitoring field information fusion is carried out.The theory of situation awareness is applied to the field of landslide hazard safety management,and multi-monitoring field information fusion is used as the basis of data collection and data specification.This method then uses neighborhood rough set to reduce a large number of landslide monitoring data sets,uses five-element connection number similarity-difference-inverse evaluation model in set pair analysis to judge the landslide situation in this state,and establishes a combined forecasting model of variable phase space reconstruction to predict the landslide situation value.Finally,the landslide situation value is visually displayed in the early warning platform,which is a landslide.Disaster situational awareness and intelligent early warning management provide reference.