KR-20260063178-A - System and method for predicting road slope stability using machine learning techniques

KR20260063178AKR 20260063178 AKR20260063178 AKR 20260063178AKR-20260063178-A

Abstract

The objective of the present invention, which is to solve the aforementioned problem, is to provide a system and method for predicting the stability or potential for road slope collapse of a road slope to be constructed in the future by performing computational learning on the collapse and non-collapse situations of a slope based on the collapse history of existing road slopes, by adding data such as topographic elevation, geological structure analysis methods, and weathering grade maps, as well as data related to the physical properties (cohesion, internal friction angle, unit weight) and specifications (height, slope, groundwater level) of each slope and unit slope data, and by utilizing the learning results. A road slope stability prediction system using a machine learning technique according to the present invention for solving the aforementioned problem comprises: a data collection unit that collects road slope information including field survey data, wide-area drawing information, projection network analysis information, and weathering grade map information regarding the road slope; a prediction modeling unit that generates and learns a slope collapse prediction model using machine learning with the collected road slope information; and a slope collapse prediction unit that predicts the collapse probability of a new road slope input using the learned prediction model.

Inventors

정석열
황상기

Assignees

정석열
(주)에스알이앤씨

Dates

Publication Date: 20260507
Application Date: 20241030

Claims (13)

A data collection unit that collects road slope information including field survey data, wide-area map information, projection network analysis information, and weathering grade map information regarding road slopes; A prediction modeling unit that generates and trains a slope collapse prediction model using machine learning based on collected road slope information; and Characterized by including a slope collapse prediction unit that predicts the probability of collapse of a new road slope input using the learned prediction model. Road slope stability prediction system using machine learning techniques.
In paragraph 1, The above data collection unit is, A road slope DB that receives and stores the above road slope information; and Characterized by including a preprocessing unit that preprocesses data stored in the above road slope DB into data suitable for a machine learning model. Road slope stability prediction system using machine learning techniques.
In paragraph 1, The above slope collapse prediction model is, Characterized by including a step-by-step prediction model classified according to step-by-step road slope information prior to design, during the design process, and after construction of the slope. Road slope stability prediction system using machine learning techniques.
In paragraph 1, The above data collection unit is, Characterized by including a weathering grade prediction unit that generates and trains a machine learning-based weathering grade prediction model based on data from a geochemistry map DB and weathering grade data from a road slope DB to predict the weathering grade for an arbitrary location. Road slope stability prediction system using machine learning techniques.
(a) A step in which a data collection unit collects road slope information including field survey data, wide-area map information, projection network analysis information, and weathering grade information regarding the road slope; (b) a step of generating and training a slope collapse prediction model using machine learning based on road slope information collected by the prediction modeling unit; and (c) characterized by including the step of predicting the probability of collapse of a new road slope input using the learned prediction model, Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above field survey data is, Characterized by including information on lateral shape, rock mass shape, soil layer depth, slope type, upper slope, height, groundwater, slope shape, berm, valley water, rockfall, loose stone, and rock mass shape. Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above wide-area map information is, Characterized by including drawing information produced using geological maps and Digital Elevation Models (DEMs), Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above projection network analysis information is, Characterized by including data on the arrangement and location of geological structures, and a quantitative stability index of the slope evaluated by extracting the slope face structure near the location of the slope from a database and analyzing it using a projection network. Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above weathering grade information is, Information generated by a weathering grade prediction system that predicts the weathering grade for an arbitrary location by generating and training a machine learning-based weathering grade prediction model based on data from a geochemistry map DB and weathering grade data from a road slope DB. Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above slope collapse prediction model is, Characterized by including a step-by-step prediction model classified according to step-by-step road slope information prior to design, during the design process, and after construction of the slope. Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above step (a) is, (a1) A step in which a data collection unit collects road slope information including field survey data, wide-area map information, projection network analysis information, and weathering grade information regarding road slopes, and weather and environmental data; and (a2) Characterized by including a step of preprocessing collected data, Road slope stability prediction method using machine learning techniques.
In paragraph 5, The above slope collapse prediction model is, Characterized by using at least one of a decision model, an ensemble model, and a neural network analysis model, Road slope stability prediction method using machine learning techniques.
A computer program stored on a medium to execute a method for predicting road slope stability using a machine learning technique according to any one of paragraphs 5 through 12 on a computer.

Description

System and method for predicting road slope stability using machine learning techniques The present invention relates to a road slope stability prediction system and method, and more specifically, to a machine learning-based system and method for predicting the stability or potential for collapse of a road slope to be constructed in the future using various road slope data. Factors affecting slope stability include internal factors such as slope specifications (height, slope angle), soil cohesion, internal friction angle, soil unit weight, geological structure, and groundwater, as well as external factors such as rainfall, earthquakes, and incidental loads caused by anthropogenic development. Furthermore, slope stability analysis can be classified into cases involving the analysis of the stability of pure soil slopes, slopes containing bedrock, and natural slope collapses such as landslides. In all cases, the reality is that it is impossible to obtain perfect stability analysis techniques or results because the interrelationships between the aforementioned internal and external factors are highly complex and non-uniform. Generally, the correlation of collapse factors becomes more complex in the order of soil, rock, and natural slopes, and as the uncertainty of natural conditions increases, analysis techniques also show significant differences. For example, in the case of soil slopes, mechanical analyses such as limit equilibrium theory or finite element methods are commonly performed using verifiable factors such as slope specifications, soil physical properties (cohesion, internal friction angle, unit weight, pore water pressure), and groundwater. While such analysis is sometimes utilized in cases involving bedrock, rough statistical analysis techniques, such as much simplified limit equilibrium theory or projection network analysis, are used due to the excessive complexity of physical properties and environmental factors. Meanwhile, for the collapse of natural slopes such as landslides, the area where collapse can occur is very wide, and since the pattern and scale of the collapse are not at a level where detailed field data can be obtained, techniques such as the overlay analysis of wide-area data (e.g., topographic orientation, elevation, slope, curvature, land mobility, rainfall distribution map, etc.) using GIS are utilized. As such, while there are significant differences in stability analysis methods depending on the type of slope, they share a commonality in that the interrelationships of physical properties, external factors, and the natural environment, as well as the importance of their contribution to slope failure, cannot be easily defined mechanically. Therefore, in the assessment of slope stability, the establishment of interrelationships based on empirical formulas or statistics has been commonly performed. However, given that quantitative analysis of various factors affecting slope stability has not been commonly performed in the stability assessment of artificial slopes such as road slopes, there is a need for the development of systems or methods to predict slope stability or the possibility of collapse based on such analysis. FIG. 1 is a diagram showing the block configuration of a road slope stability prediction system using a machine learning technique according to an embodiment of the present invention. Figure 2 is a diagram showing the detailed flow of a method for predicting road slope stability using machine learning techniques. Figure 3 is a diagram showing the predicted failure grade using a road slope stability prediction system according to an embodiment of the present invention. Figure 4 is a diagram showing the distribution pattern of failure grades applied to a road slope stability prediction system and method using machine learning techniques according to an embodiment of the present invention. FIG. 5 is a schematic diagram of a decision tree created for collapse and non-collapse using altitude, plane, and wedge failure data applied to an embodiment of the present invention. Figure 6 is a diagram showing the weathering grade map of the database and the weathering grade map predicted by the system according to an embodiment of the present invention. Figure 7 is a drawing produced from the difference between the weathering grade map and the predicted grade map applied to an embodiment of the present invention. FIG. 8 is a drawing showing the average of the pixel values of drawings produced by interpolating the predicted value of collapse according to an embodiment of the present invention, after subtracting the pixel value of the interpolated collapse value drawing of the original data from the pixel value of the drawing of the interpolated collapse value of the original data. FIG. 10 is a diagram showing the ranking order of items used in analysis in an embodiment of the present invention. The difference in applicable items is due to the difference in items that can be inve