Design and implementation of landslide comprehensive prediction system in Three Gorges reservoir area

This research is supported by the Science and Technology Department of the Ministry of Land and Resources and the special research plan of the National Natural Science Foundation (project number: 40 172096). This paper was first published in the No.2 issue of China Geological Disasters and Prevention 14.

Li Dongshan Li

(National Professional Laboratory of Geological Disaster Prevention of Chengdu University of Technology, Chengdu, Sichuan, 6 1005 1)

Because of the complexity, randomness and uncertainty of landslide deformation evolution process, the difficulty of landslide prediction is greatly increased. The forecasting system proposed in this paper integrates more than 20 kinds of landslide forecasting models into the landslide forecasting database, organically combines quantitative and qualitative forecasting with expert system, and modifies and perfects the forecasting parameters and results according to the updated monitoring data. With the extension of time and the continuous updating of monitoring data, the prediction accuracy is getting higher and higher, and the dynamic tracking prediction is realized. The whole system is developed on the basis of GIS platform, which uses the powerful spatial data and graphic management functions of GIS itself, and also has the functions of data maintenance and data processing.

GIS expert system for landslide prediction; Comprehensive forecast

People's understanding of landslides began in the Alps in the late19th century. So far, landslide research has a history of 100 years. However, the research on landslide prediction started late. Most scholars believe that the empirical formula of landslide prediction put forward by Japanese scholar M.Satio in 1960s can be used as the starting point of landslide prediction research. After that, through the painstaking exploration of scholars, the theory and method of landslide prediction have made great progress, which has gone through the process from phenomenon prediction, empirical prediction to statistical prediction, grey prediction to nonlinear prediction, and has now entered the stage of system comprehensive prediction and real-time tracking dynamic prediction. The comprehensive forecasting system proposed in this paper is the integration of the above forecasting methods. Using GIS system to manage all kinds of landslide information to realize graphic visualization, using expert system method to organically combine quantitative prediction based on theoretical model with qualitative prediction based on macro phenomenon, and finally realize comprehensive prediction and real-time tracking prediction.

1 Overall design of the system [4]

1. 1 system structure

The system is realized by component programming, and based on MapGIS, combined with GIS and m is, a landslide expert knowledge base is established to adapt to the prediction of various landslides. The system structure is shown in figure 1.

Figure 1 system structure diagram

The functions of each module are as follows: ①MapGIS platform: the basic operating platform, which realizes graphic visualization, spatial query and analysis of data, and three-dimensional dynamic display of models. ② Prediction model component: More than 20 prediction models are integrated. ③ Expert system component: comprehensive analysis of qualitative prediction and quantitative prediction. ④ Database management system: comprehensively manage all kinds of original information and achievement information.

Technical route of 1.2 system

The system takes GIS as the core and embeds professional modules, and the data flow between modules is controlled by a specially developed database management module. The system has clear structure, portability, visibility and practicability. The technical route is shown in Figure 2.

Figure 2 System technical route block diagram

2 the specific design of the system

2. 1 forecasting model component

This component integrates most existing forecasting models. Including short-term and impending slip prediction: Verhulst model, cooperative model, grey system, catastrophe model, Verhulst inverse function model, Saito model, Kawamura model, exponential smoothing model, second-order regression model, Pearl model, third-order regression model, neural network model, creep strip joint model and sliding deformation power model. Medium-term prediction: golden section method, nonlinear dynamic model, fractal theory, time series analysis theory, Kalman filter method, catastrophe model. Long-term prediction: stability coefficient method, comprehensive evaluation of stability model, etc.

The flow of this module is: firstly, test the applicability of the model, and the applicable model obtains data from the database, then preprocesses the data, and returns the results to the database after calculation. The calculation results of each module will be comprehensively evaluated in the expert system. This process is shown in Figure 3. 2.2 expert system module

Fig. 3 Structure diagram of prediction model

The establishment of expert system [5] organically combines the quantitative prediction of theoretical model with the qualitative prediction and judgment based on macro phenomena, and finally realizes the functions of comprehensive prediction and real-time tracking prediction.

The expert system itself also includes a qualitative prediction model, which organizes all the information that can be collected into a knowledge base, including macro qualitative information and monitoring data, information provided by model calculation and so on. Make predictions according to certain rules. The prediction results are landslide failure probability and landslide stage, which are the same as those of some qualitative prediction models. At the top of the expert system, the qualitative prediction results of the expert system should be combined with the qualitative prediction results and comprehensive prediction criteria provided by the prediction model base according to the corresponding rules of the rule base to get the final results of the system. At the same time, because the prediction model base will provide many quantitative prediction results (landslide time prediction), the expert system should also get the final landslide time according to the corresponding rules.

The final results provided by the expert system are: landslide failure probability; Landslide stage (creep stage, uniform deformation stage, accelerated deformation stage and imminent sliding stage); Time of landslide failure.

2.2. 1 composition of expert system

As shown in Figure 4, the expert system basically consists of the following parts:

(1) knowledge base: The knowledge of the system is obtained through comprehensive analysis and arrangement of the knowledge of many landslide experts, model calculation results, a large number of landslide examples and a large number of domestic and foreign documents. It is used to store all the knowledge and experience that affect landslide prediction. The establishment of knowledge base is actually to solve the problem of expressing expert knowledge and experience. The inference engine in the expert system uses the knowledge and information in the knowledge base to reason, make judgments and draw conclusions.

Fig. 4 Basic structure and principle of expert system

(2) Inference machine: Inference machine can also be called intelligent control system. In expert system, it is a super knowledge base, which controls the operation of other knowledge bases. The inference engine uses the knowledge in the knowledge base and the information provided by users to search and reason and find the answer to the question. Its reasoning mechanism is a combination of positive and negative reasoning.

(3) Comprehensive database: used to store the random information provided by users and the intermediate results and final conclusions obtained during the system operation.

(4) User interface: it is used for man-machine dialogue during the system operation, including the input of original information and data, the output of intermediate and final conclusions, etc.

(5) Auxiliary explanation function: it is used to provide help information to users at any time and answer related questions raised by users, such as explaining the reasoning process and asking the system to explain the questions raised. This function provides a friendly user interface for the system and increases users' confidence in using the system.

The process of expert system

Establishment of (1) knowledge base. Through the analysis of a large number of examples, the results show that displacement, deformation rate, acoustic emission and stress in the process of slope evolution are all essential characteristic parameters reflecting the slope deformation evolution, and rainfall is the main factor inducing landslide. Other factors include: groundwater dynamics, distribution and evolution of cracks, deformation and evolution of the surface, changes in surface vegetation, changes in surface vegetation, deformation and destruction of ground buildings and various landslide precursor phenomena. Earthquake is the decisive factor to induce collapse and landslide. When the earthquake intensity is greater than 7 degrees, the landslide can be revived and immediately enter the critical sliding stage.

(2) the expression of knowledge

The system adopts production representation in the knowledge base:

If p 1 and p2 ... and pnTHEN e[jd] [tm]

Here P 1, p2, ..., p. is the premise of the rule, e is the conclusion of the rule, p is the failure probability of the landslide, pn∈[0, 1], jd is the deformation stage of the landslide, which is divided into four stages: creep stage, uniform deformation stage, accelerated deformation stage and imminent sliding stage. Tm is the time of landslide failure.

(3) Establishment of rule base. For the uncertain factors in the system, the system adopts the deterministic factor method, that is, the uncertain information is given a deterministic factor value. The value range of this factor is [- 1,+1], which is usually determined by the way of Figure 5 and the experience of many landslide experts, and then the correction factor value is adjusted through the comparative analysis of the prediction results and monitoring data.

Fig. 5 the relationship between decisive factors and their languages

Each parameter should consider the possible influence on landslide prediction in different degrees, and these influences are attributed to the rules in the knowledge base of expert system. At the same time, the expert system should also deal with the quantitative and qualitative prediction results of the prediction models, and give each prediction model a certain factor value, so as to make a comprehensive prediction. The system has established a 1 1 knowledge base with more than 200 rules. Its overall structure is shown in Figure 6.

Fig. 6 The overall structure of system knowledge and the process diagram of reasoning and judgment.

In the actual forecasting process, according to the latest monitoring data, the forecasting results are revised and improved at any time, so that with the extension of time and the continuous updating of monitoring data, the forecasting accuracy is getting higher and higher, and the dynamic tracking forecasting is realized. At the same time, the numerical simulation model should be adjusted and revised at any time according to the updated monitoring data (field monitoring data or geomechanics simulation monitoring data), so that the calculation results and monitoring results can achieve the best fitting in each time period. On the other hand, by comparing and analyzing the monitoring data of each stage with the prediction results of the deformation trend prediction model, the deformation prediction model can be modified and improved at any time, so that the deformation prediction results are as close to the actual situation as possible.

2.3 Database Management Module

Due to the limited management function of MAPGIS to the database, the external database method is adopted to comprehensively manage all kinds of information. Including the input/output, editing, modification and query functions of survey data, monitoring data, forecast criterion data and macro-indication database. At the same time, it is necessary to establish the knowledge base, rule base and result base of expert system, which is convenient for users and experts to query and modify.

2.4 GIS working platform

GIS has two remarkable characteristics: ① it can manage spatial data information; ② Various spatial analysis methods can comprehensively analyze different information, seek the relationship between spatial entities, and analyze and deal with phenomena and processes distributed in a certain range. Its research object and content are spatial multi-source data. Landslide prediction data are typical spatial multi-source data; The analysis and evaluation of landslide spatial prediction is fundamentally a spatial analysis problem, so the application of GIS in landslide prediction has practical significance.

The main functions provided by GIS platform include graphic processing, data space query, data space analysis and dynamic three-dimensional display. The structure is shown in Figure 7.

Fig. 7 Structure diagram of GIS platform

① The graphics processing function can be realized directly through the control provided by MAPGIS.

(2) The data query function provides the functions of searching maps by attributes and searching maps by attributes according to conditions. Because time is an important factor in the landslide prediction system, it also provides the function of historical query.

(3) Digital Terrain Model (DTM) is realized by using the control provided by MapGIS, and Digital Elevation Model (DEM) is generated by using contour lines or TIN, which can be used for elevation analysis and ground parameter calculation (slope, slope radiance, ground roughness, etc.). ) and multi-angle azimuth display of three-dimensional model.

(4) Dynamic display of morphological changes of each surface. MapGIS does not provide three-dimensional dynamic display. This system uses Nurbs function to fit the surface, and Nurbs function is calculated from enough sampling points (monitoring points) data. When the values of sampling points change, new Nurbs surfaces are generated. Opengl provides the function of drawing Nurbs surfaces, which can be dynamically displayed by frame animation or real-time animation technology [6].

2.5 system integration

The secondary development of MAPGIS is based on components, which greatly reduces the intensity of programming and greatly enhances the portability, visibility and practicability of the system. For the embedding of professional modules, seamless connection is realized.

3 Conclusion

This paper introduces in detail the technical route, design and implementation of the single landslide comprehensive prediction system. Based on GIS platform, the system adopts object-oriented software development technology and visual computer language, which is a real-time landslide prediction system with friendly development interface, convenient operation, intuitive graphic display and high integration. Because the system makes full use of the powerful spatial data and graphics management functions of GIS itself, the system has powerful functions of data and graphics management, data maintenance, data processing and so on. At the same time, the system will combine quantitative prediction with qualitative prediction and judgment by using expert system method, and finally realize the functions of comprehensive prediction and real-time tracking prediction.

refer to

[1] Huang Runqiu Shen Fang. Geographic Information System and Geological Environment Assessment [J]. Geological Disasters and Environmental Protection, 2000, 1 1( 1).

Xu Qiang, Huang Runqiu. Temporal and Spatial Prediction of Geological Disasters [J]. Journal of Mountainous Areas, 2000, 18(B02)

Li, Chen Mingdong. Several basic problems of landslide prediction [J]. Journal of Engineering Geology,1999,7 (3)

Wu Xincai. Design and implementation of geographic information system [M]. Beijing: Electronic Industry Press, 2002.

Fu Ronghua and Li. Application of expert system in determining rock mass structure type [J]. Geological disasters and environmental protection,1993,4 (2)

Sun, Yang Changgui. Computer graphics (new edition) [M]. Beijing: Tsinghua University Publishing House, 1995.