A method and system for identifying collapse hazards based on geological environment multi-factor data
By collecting multi-factor data and constructing multi-level analysis and neural network models, and dynamically adjusting the early warning threshold, the problem of poor identification effect of collapse hazards in traditional methods is solved, and high-precision and intelligent collapse early warning is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANJIN GEOLOGICAL RES & MARINE GEOLOGY CENT
- Filing Date
- 2026-03-11
- Publication Date
- 2026-05-12
AI Technical Summary
Traditional geological disaster monitoring and early warning methods lack comprehensive consideration of multi-dimensional factors, making it difficult to capture the entire evolution process of landslide disasters. They have weak early identification capabilities, and the early warning thresholds are statically set, which cannot adapt to the multi-scale nonlinear landslide evolution in complex mountainous areas, resulting in poor identification performance.
Multi-factor geological environmental data are collected, and the potential index is calculated using the analytic hierarchy process and the information content method. A three-dimensional geomechanical model is constructed, and a physical information neural network is used to predict the risk of landslides. The warning threshold is then optimized and adjusted using Bayesian methods to achieve dynamic adaptive early warning.
It enables intelligent early warning systems that scientifically quantify and zonate potential landslide hazards, provide high-precision predictions, and ensure timely responses, thereby improving the effectiveness and practicality of landslide hazard identification.
Smart Images
Figure CN121808571B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geological disaster monitoring and early warning technology, specifically to a method and system for identifying potential landslide hazards based on multi-factor geological environmental data. Background Technology
[0002] Traditional geological disaster monitoring and early warning methods mainly rely on single surface deformation measurements (such as GNSS point monitoring) or periodic remote sensing surveys. They lack comprehensive consideration of multi-dimensional factors such as geological structure, meteorology, hydrology, and human activities, making it difficult to capture the entire chain evolution process of landslide disasters. Furthermore, existing technologies mostly focus on pre-disaster warnings when disasters are about to occur, with weak early identification capabilities for potential hazards. This makes it impossible to achieve "early detection of hazards and early warning of risks," leading to frequent occurrences of "long-term stability at registered sites and sudden disasters at unregistered sites." At the same time, the warning thresholds of existing technologies are mostly set statically and fail to be dynamically adjusted according to geological environmental evolution and real-time monitoring data. This makes it difficult to adapt to the multi-scale and nonlinear landslide evolution characteristics of complex mountainous areas, resulting in poor identification effects of existing methods and systems for landslide hazards.
[0003] Based on the above, this invention proposes a method and system for identifying landslide hazards based on multi-factor geological environmental data, which has a good effect on landslide hazard identification. Summary of the Invention
[0004] To overcome the shortcomings of existing geological disaster monitoring and early warning methods, which mainly rely on single surface deformation measurements (such as GNSS point monitoring) or periodic remote sensing surveys, lacking comprehensive consideration of multi-dimensional factors such as geological structure, meteorology, hydrology, and human activities, it is difficult to capture the entire chain evolution process of landslide disasters. Furthermore, existing technologies mostly focus on pre-disaster warnings when disasters are about to occur, with weak early identification capabilities for potential hazards, failing to achieve "early detection of hazards and early warning of risks." This leads to frequent phenomena of "long-term stability at registered sites and sudden disasters at unregistered sites." At the same time, the warning thresholds of existing technologies are mostly statically set and fail to be dynamically adjusted according to geological environmental evolution and real-time monitoring data, making it difficult to adapt to the multi-scale and nonlinear landslide evolution characteristics of complex mountainous areas. As a result, existing methods and systems have poor identification effects on landslide hazards. This invention proposes a method and system for identifying landslide hazards based on multi-factor geological environmental data, which has a better landslide hazard identification effect.
[0005] A method for identifying potential landslide hazards based on multi-factor geological environmental data includes the following steps:
[0006] Collect multi-factor geological environment data of the monitoring area. The multi-factor geological environment data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. Preprocess the multi-factor geological environment data to obtain preprocessed multi-factor geological environment data.
[0007] Based on the preprocessed multi-factor geological environment data, the potential index of each location in the monitoring area is calculated by using the analytic hierarchy process and the information content method. The monitoring area is then divided into target areas according to the potential index and the preset division rules, resulting in high potential target areas, medium potential target areas and low potential target areas.
[0008] A multi-scale detailed survey of high-potential target areas in the monitoring area was conducted, and a quantifiable three-dimensional geomechanical model was constructed. Based on the preprocessed multi-factor geological environment data and the three-dimensional geomechanical model, key early warning factor data of high-potential target areas were obtained. The key early warning factor data included development factor data, basic factor data, and triggering factor data.
[0009] A target area collapse risk probability prediction model based on physical information neural network is constructed. The key early warning factor data is preprocessed and input into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target areas.
[0010] The initial classification threshold for the predicted probability of collapse risk is set based on the cumulative distribution function of historical data. The Bayesian optimization algorithm is used to adaptively and dynamically adjust the initial classification threshold to obtain the current classification threshold for the predicted probability of collapse risk. Based on the current classification threshold and the predicted probability of collapse risk, the high potential target area is classified into collapse risk levels and corresponding alarms are issued.
[0011] As a preferred aspect of the invention, the topographic data includes DEM data, slope data, aspect data, and elevation data; the geological structure data includes stratigraphic lithology data, fault distance data, and joint and fissure development density data; the meteorological and hydrological data includes rainfall intensity data, cumulative rainfall data, and soil and rock saturation data; the human activity data includes slope cutting rate change data, road construction progress data, and slope cutting housing construction progress data; the deformation monitoring data includes GNSS displacement data, dip angle change rate data, and crack change data; the development factor data includes unstable rock mass stability index data and unfavorable structural surface combination index data; the basic factor data includes tectonic stress coefficient data, rock mass integrity coefficient data, and hydrogeological condition index data; and the triggering factor data includes rainfall intensity data and human activity intensity data.
[0012] As a preferred aspect of the invention, the specific steps for preprocessing multi-factor geological environment data to obtain preprocessed multi-factor geological environment data are as follows:
[0013] Using dates as an index, multi-factor geological environmental data are converted into time series format, the starting reference time points of all time series are unified, the time frequencies of all time series are unified, and missing values of time series with low time frequencies at new time points are filled by linear interpolation or polynomial interpolation methods.
[0014] To handle null or missing values in a time series, interpolation methods such as linear interpolation or polynomial interpolation can be used to fill in the missing values, or the mean or median of the time series can be used directly to fill in the missing values.
[0015] Outliers in a time series that do not conform to the expected pattern can be identified using the Z-Score or IQR method. Outliers can be removed and replaced with the mean or median of the time series, or interpolation methods can be used to repair outliers.
[0016] Min-max standardization is used to standardize data of different dimensions. The formula for calculating min-max standardization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. Represents the maximum value in a time series. This represents the standardized data value.
[0017] As a preferred aspect of the invention, the specific steps for comprehensively calculating the potential index of each location in the monitoring area based on preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method are as follows:
[0018] Each data point in the preprocessed multi-factor geological environment data is considered as a factor, and the weight coefficients of each factor are obtained through the analytic hierarchy process.
[0019] Historical landslide disaster data were acquired for various locations within the monitoring area, and the information content method was used to calculate the information content values of various factors related to landslide disasters. The specific calculation formula is as follows:
[0020]
[0021] in Indicates the first The information content value of the class factor for landslide disasters Indicates the condition under which the collapse occurs. The probability of class factors appearing, Indicates the first The probability of class factors appearing in the monitored area;
[0022] Based on the weight coefficients of various factors, the information value of each factor on the collapse disaster is weighted and summed to obtain the potential index of each location in the monitoring area.
[0023] As a preferred aspect of the invention, the target area collapse risk probability prediction model based on a physical information neural network includes an input layer, three fully connected layers, a physical constraint layer, and an output layer. The input layer receives preprocessed key early warning factor data. The first fully connected layer receives the output of the input layer and performs feature extraction to capture nonlinear relationships. The second fully connected layer receives the output of the first fully connected layer and performs feature dimensionality reduction to extract key patterns. The third fully connected layer receives the output of the second fully connected layer and performs high-level semantic mapping. The physical constraint layer receives the output of the third fully connected layer and embeds physical equation residuals to constrain the network output to conform to mechanical mechanisms. The output layer receives the output of the physical constraint layer and outputs the collapse risk prediction probability through a Sigmoid activation function.
[0024] As a preferred aspect of the invention, the specific steps of setting an initial classification threshold for the collapse risk prediction probability based on the cumulative distribution function of historical data, and adaptively and dynamically adjusting the initial classification threshold using a Bayesian optimization algorithm to obtain the current classification threshold for the collapse risk prediction probability are as follows:
[0025] The predicted probability of landslide risk before a landslide occurs is obtained from historical landslide disaster data in various locations within the monitoring area. The 25th, 50th, and 75th percentiles of the predicted landslide risk probability are calculated using the cumulative distribution function of historical data and used as the initial classification thresholds for the predicted landslide risk probability.
[0026] Historical landslide disaster data of the monitoring area is obtained and early warning event samples are extracted from them. The early warning accuracy rate, early warning underreporting rate and early warning false alarm rate are calculated based on the early warning event samples, and an objective function is constructed accordingly. Based on the objective function, the initial classification threshold is iteratively optimized using a Bayesian optimization algorithm until the objective function reaches a preset value or the change in the objective function is less than the preset threshold multiple times, thus obtaining the current classification threshold for the landslide risk prediction probability.
[0027] A system for identifying potential landslide hazards based on multi-factor geological environmental data includes:
[0028] The data acquisition and processing module is used to collect multi-factor geological environmental data of the monitoring area. The multi-factor geological environmental data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. The multi-factor geological environmental data is preprocessed to obtain preprocessed multi-factor geological environmental data.
[0029] The hidden danger target area identification module is used to comprehensively calculate the potential index of each location in the monitoring area based on the preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method. According to the potential index and the preset division rules, the monitoring area is divided into target areas, resulting in high potential target areas, medium potential target areas and low potential target areas.
[0030] The 3D geological modeling module is used to conduct multi-scale detailed investigations of high-potential target areas in the monitoring area and construct quantifiable 3D geomechanical models. Based on preprocessed multi-factor geological environment data and 3D geomechanical models, key early warning factor data of high-potential target areas are obtained. Key early warning factor data includes development factor data, basic factor data and triggering factor data.
[0031] The model building and prediction module is used to build a target area collapse risk probability prediction model based on physical information neural network. It preprocesses the key early warning factor data and inputs it into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target areas.
[0032] The risk classification and early warning module is used to set an initial classification threshold for the predicted probability of collapse risk based on the cumulative distribution function of historical data, and to adaptively and dynamically adjust the initial classification threshold using a Bayesian optimization algorithm to obtain the current classification threshold for the predicted probability of collapse risk. Based on the current classification threshold and the predicted probability of collapse risk, the module classifies the high potential target area for collapse risk and issues an alarm of the corresponding level.
[0033] The present invention has the following advantages:
[0034] 1. This invention comprehensively calculates the potential index of each location in the monitoring area based on preprocessed multi-factor geological environment data and using the analytic hierarchy process (AHP) and information content method. It not only integrates expert experience and historical data statistical patterns, but also quantitatively calculates the weight coefficients and information content contribution values of each factor. This enables the scientific quantification and precise zoning of the potential of the monitoring area, thereby improving the landslide hazard identification effect of this method and system based on multi-factor geological environment data.
[0035] 2. This invention constructs a target area collapse risk probability prediction model based on a physical information neural network and uses it to predict the collapse risk probability of high potential target areas. It can not only construct a mechanism-constrained and data-driven collaborative prediction model and utilize the powerful nonlinear fitting ability of deep learning, but also force the network output to conform to physical laws, thereby achieving high-precision, interpretable and strong generalization of collapse risk probability prediction. This improves the collapse risk identification effect of this method and system based on multi-factor geological environment data for identifying collapse hazards.
[0036] 3. This invention sets an initial grading threshold for the predicted probability of landslide risk based on the cumulative distribution function of historical data, and uses a Bayesian optimization algorithm to adaptively and dynamically adjust the initial grading threshold. It can continuously optimize the threshold parameters based on real-time monitoring data feedback and early warning effect evaluation, thereby achieving intelligent early warning with accurate grading, timely response and self-evolution, and improving the practicality of this method and system for identifying landslide hazards based on multi-factor geological environmental data. Attached Figure Description
[0037] Figure 1 This is a flowchart illustrating a method for identifying potential landslide hazards based on multi-factor geological environmental data, as used in an embodiment of the present invention.
[0038] Figure 2 This is a schematic diagram of the structure of a system for identifying potential landslide hazards based on multi-factor geological environmental data, as used in an embodiment of the present invention. Detailed Implementation
[0039] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.
[0040] Example 1: A method for identifying potential landslide hazards based on multi-factor geological environmental data, such as... Figure 1 As shown, it includes the following steps:
[0041] Step S1: Collect multi-factor geological environment data of the monitoring area. The multi-factor geological environment data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. Preprocess the multi-factor geological environment data to obtain preprocessed multi-factor geological environment data.
[0042] Step S2: Based on the preprocessed multi-factor geological environment data, the potential index of each location in the monitoring area is calculated using the analytic hierarchy process (AHP) and the information content method. The monitoring area is then divided into target areas according to the potential index and the preset division rules, resulting in high potential target areas, medium potential target areas, and low potential target areas.
[0043] Step S3: Conduct a multi-scale detailed survey of high potential target areas in the monitoring area and construct a quantifiable three-dimensional geomechanical model. Based on the preprocessed multi-factor geological environment data and the three-dimensional geomechanical model, obtain key early warning factor data for high potential target areas. Key early warning factor data includes development factor data, basic factor data, and triggering factor data.
[0044] Step S4: Construct a target area collapse risk probability prediction model based on physical information neural network, preprocess the key early warning factor data and input it into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target area.
[0045] Step S5: Set the initial classification threshold for the collapse risk prediction probability based on the cumulative distribution function of historical data, and use the Bayesian optimization algorithm to adaptively and dynamically adjust the initial classification threshold to obtain the current classification threshold for the collapse risk prediction probability. Based on the current classification threshold and the collapse risk prediction probability, classify the collapse risk of the high potential target area and issue an alarm of the corresponding level.
[0046] The topographic data includes DEM data, slope data, aspect data, and elevation data; the geological structure data includes stratigraphic lithology data, fault distance data, and joint and fissure development density data; the meteorological and hydrological data includes rainfall intensity data, cumulative rainfall data, and soil and rock saturation data; the human activity data includes slope cutting rate change data, road construction progress data, and slope cutting housing construction progress data; the deformation monitoring data includes GNSS displacement data, dip angle change rate data, and crack change data; the development factor data includes unstable rock mass stability index data and unfavorable structural surface combination index data; the basic factor data includes tectonic stress coefficient data, rock mass integrity coefficient data, and hydrogeological condition index data; and the triggering factor data includes rainfall intensity data and human activity intensity data.
[0047] The specific steps for preprocessing multi-factor geological environment data to obtain preprocessed multi-factor geological environment data are as follows:
[0048] Using dates as an index, multi-factor geological environmental data are converted into time series format, the starting reference time points of all time series are unified, the time frequencies of all time series are unified, and missing values of time series with low time frequencies at new time points are filled by linear interpolation or polynomial interpolation methods.
[0049] To handle null or missing values in a time series, interpolation methods such as linear interpolation or polynomial interpolation can be used to fill in the missing values, or the mean or median of the time series can be used directly to fill in the missing values.
[0050] Outliers in a time series that do not conform to the expected pattern can be identified using the Z-Score or IQR method. Outliers can be removed and replaced with the mean or median of the time series, or interpolation methods can be used to repair outliers.
[0051] Min-max standardization is used to standardize data of different dimensions. The formula for calculating min-max standardization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. Represents the maximum value in a time series. This represents the standardized data value.
[0052] The specific steps for comprehensively calculating the potential index of each location in the monitoring area based on preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method are as follows:
[0053] Each data point in the preprocessed multi-factor geological environment data is considered as a factor, and the weight coefficients of each factor are obtained through the analytic hierarchy process.
[0054] Historical landslide disaster data were acquired for various locations within the monitoring area, and the information content method was used to calculate the information content values of various factors related to landslide disasters. The specific calculation formula is as follows:
[0055]
[0056] in Indicates the first The information content value of the class factor for landslide disasters Indicates the conditions under which the collapse occurs. The probability of class factors appearing, Indicates the first The probability of class factors appearing in the monitored area;
[0057] Based on the weight coefficients of various factors, the information value of each factor on the collapse disaster is weighted and summed to obtain the potential index of each location in the monitoring area.
[0058] It should be noted that the above-mentioned analytic hierarchy process is a conventional existing technique, so it will not be described in detail here.
[0059] The above steps, based on preprocessed multi-factor geological environmental data and employing the analytic hierarchy process (AHP) and information content method, comprehensively calculate the potential index of each location in the monitoring area. This not only integrates expert experience and historical data statistical patterns but also quantitatively calculates the weight coefficients and information content contribution values of each factor. As a result, the potential of the monitoring area is scientifically quantified and accurately zoned, improving the landslide hazard identification effect of this method and system based on multi-factor geological environmental data.
[0060] The target area collapse risk probability prediction model based on a physical information neural network includes an input layer, three fully connected layers, a physical constraint layer, and an output layer. The input layer receives preprocessed key early warning factor data. The first fully connected layer receives the output of the input layer and performs feature extraction to capture nonlinear relationships. The second fully connected layer receives the output of the first fully connected layer and performs feature dimensionality reduction to extract key patterns. The third fully connected layer receives the output of the second fully connected layer and performs high-level semantic mapping. The physical constraint layer receives the output of the third fully connected layer and embeds physical equation residuals to constrain the network output to conform to mechanical mechanisms. The output layer receives the output of the physical constraint layer and outputs the collapse risk prediction probability through a sigmoid activation function.
[0061] The above steps construct a target area collapse risk probability prediction model based on a physical information neural network and predict the collapse risk probability of high potential target areas accordingly. This not only enables the construction of a mechanism-constrained and data-driven collaborative prediction model and utilizes the powerful nonlinear fitting capability of deep learning, but also forces the network output to conform to physical laws. This achieves high-precision, interpretable, and highly generalizable collapse risk probability prediction, improving the collapse hazard identification effect of this method and system based on multi-factor geological environment data.
[0062] The specific steps for setting an initial grading threshold for the predicted collapse risk probability based on the cumulative distribution function of historical data, and then using a Bayesian optimization algorithm to adaptively and dynamically adjust the initial grading threshold to obtain the current grading threshold for the predicted collapse risk probability are as follows:
[0063] The predicted probability of landslide risk before a landslide occurs is obtained from historical landslide disaster data in various locations within the monitoring area. The 25th, 50th, and 75th percentiles of the predicted landslide risk probability are calculated using the cumulative distribution function of historical data and used as the initial classification thresholds for the predicted landslide risk probability.
[0064] Historical landslide disaster data of the monitoring area is obtained and early warning event samples are extracted from them. The early warning accuracy rate, early warning underreporting rate and early warning false alarm rate are calculated based on the early warning event samples, and an objective function is constructed accordingly. Based on the objective function, the initial classification threshold is iteratively optimized using a Bayesian optimization algorithm until the objective function reaches a preset value or the change in the objective function is less than the preset threshold multiple times, thus obtaining the current classification threshold for the landslide risk prediction probability.
[0065] It should be noted that the Bayesian optimization algorithm mentioned above is a conventional existing technology, so it will not be described in detail here.
[0066] The above steps set an initial classification threshold for the probability of landslide risk prediction based on the cumulative distribution function of historical data, and use a Bayesian optimization algorithm to adaptively and dynamically adjust the initial classification threshold. This allows for continuous optimization of the threshold parameters based on real-time monitoring data feedback and early warning effect evaluation, thereby achieving intelligent early warning with accurate classification, timely response, and self-evolution. This enhances the practicality of the method and system for identifying landslide hazards based on multi-factor geological environmental data.
[0067] Example 2: A system for identifying potential landslide hazards based on multi-factor geological environmental data, such as... Figure 2 As shown, it includes:
[0068] The data acquisition and processing module is used to collect multi-factor geological environmental data of the monitoring area. The multi-factor geological environmental data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. The multi-factor geological environmental data is preprocessed to obtain preprocessed multi-factor geological environmental data.
[0069] The hidden danger target area identification module is used to comprehensively calculate the potential index of each location in the monitoring area based on the preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method. According to the potential index and the preset division rules, the monitoring area is divided into target areas, resulting in high potential target areas, medium potential target areas and low potential target areas.
[0070] The 3D geological modeling module is used to conduct multi-scale detailed investigations of high-potential target areas in the monitoring area and construct quantifiable 3D geomechanical models. Based on preprocessed multi-factor geological environment data and 3D geomechanical models, key early warning factor data of high-potential target areas are obtained. Key early warning factor data includes development factor data, basic factor data and triggering factor data.
[0071] The model building and prediction module is used to build a target area collapse risk probability prediction model based on physical information neural network. It preprocesses the key early warning factor data and inputs it into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target areas.
[0072] The risk classification and early warning module is used to set an initial classification threshold for the predicted probability of collapse risk based on the cumulative distribution function of historical data, and to adaptively and dynamically adjust the initial classification threshold using a Bayesian optimization algorithm to obtain the current classification threshold for the predicted probability of collapse risk. Based on the current classification threshold and the predicted probability of collapse risk, the module classifies the high potential target area for collapse risk and issues an alarm of the corresponding level.
[0073] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.
Claims
1. A method for identifying potential landslide hazards based on multi-factor geological environmental data, characterized in that, Includes the following steps: Collect multi-factor geological environmental data of the monitoring area. The multi-factor geological environmental data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. Preprocess the multi-factor geological environmental data to obtain preprocessed multi-factor geological environmental data. Based on the preprocessed multi-factor geological environment data, the potential index of each location in the monitoring area is calculated by using the analytic hierarchy process and the information content method. The monitoring area is then divided into target areas according to the potential index and the preset division rules, resulting in high potential target areas, medium potential target areas and low potential target areas. A multi-scale detailed survey of high-potential target areas in the monitoring area was conducted, and a quantifiable three-dimensional geomechanical model was constructed. Based on the preprocessed multi-factor geological environment data and the three-dimensional geomechanical model, key early warning factor data of high-potential target areas were obtained. The key early warning factor data included development factor data, basic factor data, and triggering factor data. A target area collapse risk probability prediction model based on physical information neural network is constructed. The key early warning factor data is preprocessed and input into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target areas. The initial classification threshold for the predicted probability of collapse risk is set based on the cumulative distribution function of historical data. The Bayesian optimization algorithm is used to adaptively and dynamically adjust the initial classification threshold to obtain the current classification threshold for the predicted probability of collapse risk. Based on the current classification threshold and the predicted probability of collapse risk, the high potential target area is classified into collapse risk levels and corresponding alarms are issued.
2. The method for identifying potential landslide hazards based on multi-factor geological environmental data according to claim 1, characterized in that, The topographic data includes DEM data, slope data, aspect data, and elevation data; the geological structure data includes stratigraphic lithology data, fault distance data, and joint and fissure development density data; the meteorological and hydrological data includes rainfall intensity data, cumulative rainfall data, and soil and rock saturation data; the human activity data includes slope cutting rate change data, road construction progress data, and slope cutting housing construction progress data; the deformation monitoring data includes GNSS displacement data, dip angle change rate data, and crack change data; the development factor data includes unstable rock mass stability index data and unfavorable structural surface combination index data; the basic factor data includes tectonic stress coefficient data, rock mass integrity coefficient data, and hydrogeological condition index data; and the triggering factor data includes rainfall intensity data and human activity intensity data.
3. The method for identifying potential landslide hazards based on multi-factor geological environmental data according to claim 2, characterized in that, The specific steps for preprocessing multi-factor geological environment data to obtain preprocessed multi-factor geological environment data are as follows: Using dates as an index, multi-factor geological environmental data are converted into time series format, the starting reference time points of all time series are unified, the time frequencies of all time series are unified, and missing values of time series with low time frequencies at new time points are filled by linear interpolation or polynomial interpolation methods. To handle null or missing values in a time series, interpolation methods such as linear interpolation or polynomial interpolation can be used to fill in the missing values, or the mean or median of the time series can be used directly to fill in the missing values. Outliers in a time series that do not conform to the expected pattern can be identified using the Z-Score or IQR method. Outliers can be removed and replaced with the mean or median of the time series, or interpolation methods can be used to repair outliers. Min-max standardization is used to standardize data of different dimensions. The formula for calculating min-max standardization is as follows: ,in Represents the original data value. This represents the minimum value in a time series. Represents the maximum value in a time series. This represents the standardized data value.
4. The method for identifying potential landslide hazards based on multi-factor geological environmental data according to claim 3, characterized in that, The specific steps for comprehensively calculating the potential index of each location in the monitoring area based on preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method are as follows: Each data point in the preprocessed multi-factor geological environment data is considered as a factor, and the weight coefficients of each factor are obtained through the analytic hierarchy process. Historical landslide disaster data were acquired for various locations within the monitoring area, and the information content method was used to calculate the information content values of various factors related to landslide disasters. The specific calculation formula is as follows: in Indicates the first The information content value of the class factor for landslide disasters Indicates the conditions under which the collapse occurs. The probability of class factors appearing, Indicates the first The probability of class factors appearing in the monitored area; Based on the weight coefficients of various factors, the information value of each factor on the collapse disaster is weighted and summed to obtain the potential index of each location in the monitoring area.
5. The method for identifying potential landslide hazards based on multi-factor geological environmental data according to claim 4, characterized in that, The target area collapse risk probability prediction model based on physical information neural network includes an input layer, three fully connected layers, a physical constraint layer and an output layer, wherein the input layer is used to receive preprocessed key early warning factor data; The first fully connected layer receives the output of the input layer and performs feature extraction to capture nonlinear relationships. The second fully connected layer receives the output of the first fully connected layer and performs feature dimensionality reduction to extract key patterns. The third fully connected layer receives the output of the second fully connected layer and performs high-level semantic mapping. The physical constraint layer receives the output of the third fully connected layer and embeds physical equation residuals to constrain the network output to conform to mechanical mechanisms. The output layer receives the output of the physical constraint layer and outputs the collapse risk prediction probability through the Sigmoid activation function.
6. The method for identifying potential landslide hazards based on multi-factor geological environmental data according to claim 5, characterized in that, The specific steps for setting an initial grading threshold for the predicted collapse risk probability based on the cumulative distribution function of historical data, and then using a Bayesian optimization algorithm to adaptively and dynamically adjust the initial grading threshold to obtain the current grading threshold for the predicted collapse risk probability are as follows: The predicted probability of landslide risk before a landslide occurs is obtained from historical landslide disaster data in various locations within the monitoring area. The 25th, 50th, and 75th percentiles of the predicted landslide risk probability are calculated using the cumulative distribution function of historical data and used as the initial classification thresholds for the predicted landslide risk probability. Historical landslide disaster data of the monitoring area is obtained and early warning event samples are extracted from them. The early warning accuracy rate, early warning underreporting rate and early warning false alarm rate are calculated based on the early warning event samples, and an objective function is constructed accordingly. Based on the objective function, the initial classification threshold is iteratively optimized using a Bayesian optimization algorithm until the objective function reaches a preset value or the change in the objective function is less than the preset threshold multiple times, thus obtaining the current classification threshold for the landslide risk prediction probability.
7. A system for identifying landslide hazards based on multi-factor geological environmental data, applied to the method for identifying landslide hazards based on multi-factor geological environmental data as described in any one of claims 1-6, characterized in that, Including: The data acquisition and processing module is used to collect multi-factor geological environmental data of the monitoring area. The multi-factor geological environmental data includes topographic data, geological structure data, meteorological and hydrological data, human activity data and deformation monitoring data. The multi-factor geological environmental data is preprocessed to obtain preprocessed multi-factor geological environmental data. The hidden danger target area identification module is used to comprehensively calculate the potential index of each location in the monitoring area based on the preprocessed multi-factor geological environment data and using the analytic hierarchy process and information content method. According to the potential index and the preset division rules, the monitoring area is divided into target areas, resulting in high potential target areas, medium potential target areas and low potential target areas. The 3D geological modeling module is used to conduct multi-scale detailed investigations of high-potential target areas in the monitoring area and construct quantifiable 3D geomechanical models. Based on preprocessed multi-factor geological environment data and 3D geomechanical models, key early warning factor data of high-potential target areas are obtained. Key early warning factor data includes development factor data, basic factor data and triggering factor data. The model building and prediction module is used to build a target area collapse risk probability prediction model based on physical information neural network. It preprocesses the key early warning factor data and inputs it into the trained target area collapse risk probability prediction model to obtain the collapse risk prediction probability of high potential target areas. The risk classification and early warning module is used to set an initial classification threshold for the predicted probability of collapse risk based on the cumulative distribution function of historical data, and to adaptively and dynamically adjust the initial classification threshold using a Bayesian optimization algorithm to obtain the current classification threshold for the predicted probability of collapse risk. Based on the current classification threshold and the predicted probability of collapse risk, the module classifies the high potential target area for collapse risk and issues an alarm of the corresponding level.