Intelligent assessment method for risk of overtopping and dam break of tailing pond induced by rainstorm
By constructing a tailings dam failure probability, loss and vulnerability assessment model and combining it with a machine learning algorithm, an intelligent assessment of the tailings dam failure risk is achieved, which solves the problem of inaccurate assessment results in existing technologies and improves the accuracy and efficiency of the assessment.
Patent Information
- Application Number
- CN202510811735.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-18
- Publication Date
- 2025-09-26
AI Technical Summary
Existing tailings dam failure risk assessment methods are mainly qualitative or quantitative, which are greatly affected by human factors and it is difficult to form a unified evaluation standard. Quantitative assessments are also difficult to fully consider the variability of design variables and engineering complexity, resulting in poor accuracy of assessment results.
By using satellite remote sensing images and census data, combined with information such as the geometry and material composition of the tailings dam, a dam failure probability, loss and vulnerability assessment model is constructed, and machine learning algorithms are used to conduct intelligent risk assessment, including data collection, model construction and risk index calculation.
It realizes the intelligent assessment of tailings dam failure risk, improves the accuracy and efficiency of the assessment, provides more advanced and complete evaluation methods, and can quickly assess the risk level of dam failure and visualize it.
Smart Images
Figure CN120706894A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of tailings pond engineering, and in particular to an intelligent assessment method for the risk of tailings pond overtopping and dam failure induced by heavy rain. Background Art
[0002] In recent years, the development of mining resources has generated a large amount of abandoned tailings. To solve the problem of stacking, a large number of tailings ponds have been built, and the safety management of tailings ponds has become a prominent issue. In the past, in order to reduce costs, more attention was paid to the site selection, construction, management and operation of tailings ponds, while insufficient consideration was given to safety prevention and response measures. Tailings ponds have poor seismic performance and low stability, making them susceptible to external factors such as rainfall and earthquakes, resulting in dam failure accidents, posing a serious threat to the safety of life and property of people downstream. At the same time, heavy metals in tailings can also cause significant damage to the environment. Therefore, it is crucial to conduct a risk assessment of tailings dam failure, but current risk assessment methods are still mainly qualitative or quantitative. Qualitative analysis methods do not require a large amount of data and the assessment method is simple, but they are greatly affected by human subjective factors. It is difficult to form a unified evaluation standard applicable to various tailings dams, and the assessment results are less accurate. For example, in 2022, CHEN C (Chen C, Ma B. Safety Assessment of Dam Failure of Tailings Pond Based on Variable Weight Method: A Case Study in China [J]. Mining, Metallurgy & Exploration, 2022, 39 (6): 2401-2413.) subjectively classified indicators such as personnel qualifications, daily management, emergency plans, and geological conditions to construct an assessment system to assess dam failure risks. Quantitative assessment methods can quantify the probability and loss of risks, but it is difficult to fully consider the variability of design variables and the complex characteristics of the project. For example, in 2017, Wang Xunhong (Wang Xunhong, Gu Xiaowei, Xu Xiaochuan, Wang Qing. Tailings dam failure risk assessment based on GA-AHP and cloud matter-element model [J]. Journal of Northeastern University (Natural Science Edition), 2017, 38 (10): 1464-1467) et al. introduced a cloud model to calculate the weights of each risk indicator to construct an assessment system in order to improve the accuracy of qualitative assessment. In 2009, Zheng Xin (Zheng Xin, Xu Kaili. Research on the evaluation model of the severity of tailings dam failure consequences [J]. Industrial Safety and Environmental Protection, 2009, 35(5):30-31.) proposed a dam failure hazard function using a comprehensive factor weighting method. The factors are tailings dam scale, life loss, economic loss, and social and environmental loss. Different impact areas are divided according to the tailings dam failure hazard value. The research proposed in this patent quantifies the probability of tailings dam failure from the perspective of spatial variability of tailings materials and combines losses and vulnerability to conduct intelligent risk assessment. There are relatively few studies. Summary of the Invention
[0003] In order to solve the above problems, the present invention provides an intelligent assessment method for the risk of tailings dam overtopping and dam failure induced by heavy rain. The method uses satellite remote sensing images and the latest census data to statistically analyze the population distribution, GDP, land use types in different areas, geometric parameters such as the geometry of the tailings dam, the material composition of the tailings dam, and the height of the dam body in the area around the tailings dam. Based on the basic information obtained above, an assessment model of risk sub-indicators is constructed, which are a dam failure probability assessment model, a loss assessment model, and a vulnerability assessment model. Through the established model, the risk index of the sub-indicator is obtained, and then the risk assessment model is constructed using the risk index of the sub-indicator, thereby quickly and intelligently assessing the risk of tailings dam overtopping and dam failure.
[0004] The intelligent assessment method for the risk of tailings dam overtopping and failure induced by heavy rain in the present invention comprises the following steps:
[0005] Step S1: Intelligent data collection
[0006] (1) Access to original multi-source data: Use drones and other means to capture and upload point cloud data in real time, and create an intelligent data collection system that can integrate the required tailings dam geometry parameters, population, satellite images, land type, and GDP per capita.
[0007] (2) Query to obtain specific data: By entering a specific year and region, the data collection system is filtered to obtain the raw data that meets the requirements.
[0008] (3) Preprocessing the data: Check the various types of raw data obtained and identify missing values and duplicate values, etc. At the same time, convert the data format and normalize it to ensure the accuracy of the data.
[0009] Step S2: Constructing a dam failure probability model
[0010] (1) The file of geometric parameters of the tailings dam in the acquisition system is called to generate the physical model of the tailings dam. The unsaturated seepage model is constructed through the interface with Geo-Studio to conduct unsaturated seepage stability analysis, and the calculation result files such as the potential seepage surface, critical slip surface and safety factor value are obtained.
[0011] (2) The Monte Carlo method is used to construct a model for tailings dam reliability analysis. Random samples are generated according to the Latin hypercube sampling method, and the KL series expansion method is used to discretize the random field. The random samples and stability calculation result files are used to train the back propagation neural network (BPNN) agent model to perform reliability analysis and calculate the failure probability.
[0012] Step S3: Constructing a dam break loss assessment model
[0013] (1) Constructing a dam-break model: Calling the data in the acquisition system to construct a dam-break model of the tailings pond, and obtaining the changes in the submerged depth and flow velocity of the dam-break tailings flow over time.
[0014] (2) Constructing a loss database: embed the empirical calculation formulas for life loss, economic loss, and ecological environment loss and the program that calls Arcgis into the database, input the result file of the dam break model, output the spatiotemporal distribution map of life, economic and ecological environment loss, and calculate the comprehensive loss index. The loss calculation and comprehensive loss index calculation formulas are as follows:
[0015]
[0016] K i =0.5KK 1i K 2i K 3i K 4i
[0017] Among them, LOL is the life loss value, K i is the fatality rate of the ith residential area, K 1i ~K 4i are the distance coefficient, house instability coefficient, location coefficient and density coefficient of the i-th residential area along the main river to the tailings dam.
[0018]
[0019] Among them, D is the direct economic loss value, β is the economic loss rate, W is the economic value distribution data, and the cultivated land value is 1.95 yuan / m 2 , the value of residential land is 1040 yuan / m 2 , S p is the loss value of the p-th type of property, R is the indirect economic loss value, the total economic loss is the sum of direct economic loss and indirect loss, K i is the indirect coefficient corresponding to the pth type of property, 25% for cultivated land and 15% for residential land.
[0020]
[0021] Among them, F is the ecological environment loss, f i is the ecological loss rate, V si The value of ecosystem services for type i.
[0022]
[0023] Among them, E is the comprehensive loss index.
[0024] Step S4: Constructing a dam break vulnerability assessment model
[0025] Constructing a vulnerability assessment model: Calling the normalized population density distribution, land use type, and GDP per capita data from the acquisition system, and using a method similar to step S3 and the vulnerability calculation formula to construct a vulnerability assessment model and generate a vulnerability index map. The formula is:
[0026] V = weight I × population density + weight II × GDP per capita + weight III × land use type, where V is the vulnerability index.
[0027] Step S5: Constructing an intelligent dam break risk assessment model
[0028] (1) Call the failure probability, comprehensive loss index and vulnerability index calculated in the dam failure probability, loss assessment and vulnerability assessment models established in steps S2, S3 and S4, and use the formula to calculate the risk assessment value to conduct a risk assessment of the tailings dam failure. The formula is:
[0029] R=P f ×E×V
[0030] Among them, R is the risk index, E is the comprehensive loss index, and V is the vulnerability index.
[0031] (2) An intelligent risk assessment model is constructed by combining the dam failure probability, loss assessment and vulnerability assessment models. The natural breakpoint classification method is used to divide the risk levels and generate a risk level map to realize the visualization of risk levels.
[0032] Compared with the existing technology, the present invention has significant innovations. It first starts from the unique perspective of the spatial variability of tailings materials and uses random field theory to accurately describe the spatial variation characteristics of tailings materials. On this basis, the method integrates advanced machine learning algorithms to achieve efficient calculation of slope failure probability. Furthermore, by constructing a tailings dam break model, the method can accurately assess the losses that may be caused by dam break. Ultimately, combined with the failure probability of dam break, actual losses and the vulnerability of the tailings pond, the method can provide an intelligent risk assessment result. This method not only improves the efficiency of tailings dam break risk assessment, but also achieves the comprehensiveness of the assessment, providing a more advanced and complete evaluation method for related fields. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 To build an intelligent risk assessment model diagram.
[0034] Figure 2 To construct a dam break probability model diagram.
[0035] Figure 3 To construct a dam break loss assessment model diagram.
[0036] Figure 4 Schematic diagram of tailings dam grid division and sliding surface.
[0037] Figure 5 Schematic diagram of the dam break 3 minutes after it started.
[0038] Figure 6 Schematic diagram of the dam break 10 minutes after it began.
[0039] Figure 7 Schematic diagram of the dam break 60 minutes after it began.
[0040] Figure 8 Schematic diagram of the dam break 120 minutes after it started.
[0041] Figure 9 Hazard classification chart.
[0042] Figure 10 Vulnerability partitioning diagram. DETAILED DESCRIPTION
[0043] To more clearly illustrate the objectives, technical solutions, and advantages of the present invention, the following will provide an in-depth analysis of the present invention with reference to detailed drawings and specific implementation examples. It should be noted that these specific implementation examples are intended only as an aid to understanding the scope of the present invention and are not intended to limit its scope or limits of application.
[0044] The intelligent assessment method for the risk of tailings dam overtopping and failure induced by heavy rain described in this embodiment includes five parts: intelligent data collection, construction of a dam failure probability model, a dam failure loss model, a vulnerability assessment model and an intelligent risk assessment model.
[0045] This intelligent data collection system uses drones and other means to capture and upload point cloud data in real time, creating an intelligent data collection system that integrates required data such as tailings dam geometry, population, satellite imagery, land type, and GDP per capita. By entering specific years and regions, the data collection system filters and obtains raw data that meets the requirements. Data preprocessing involves inspecting the raw data and identifying missing and duplicate values. Data format conversion and normalization are also performed to ensure accuracy.
[0046] The dam failure probability model is constructed by calling a file on the tailings dam's geometric parameters from the acquisition system to generate a physical model of the tailings dam. An unsaturated seepage model is constructed through an interface with Geo-Studio to perform an unsaturated seepage stability analysis, obtaining calculation result files such as the potential seepage surface, critical slip surface, and safety factor. A model for tailings dam reliability analysis is constructed using the Monte Carlo method. Random samples are generated using the Latin hypercube sampling method, and the KL series expansion method is used to discretize the random field. The random samples and stability calculation result files are then used to train a back propagation neural network (BPNN) proxy model to perform reliability analysis and calculate failure probability.
[0047] The loss assessment model is constructed by calling the data in the acquisition system to construct a tailings dam failure model, and the time-varying process of the flooding depth and flow velocity of the dam failure tailings flow is obtained. The empirical calculation formulas for loss of life, economic loss, and ecological environment loss and the program calling Arcgis are embedded in the database to construct a loss database, input the result file of the dam failure model, output the spatiotemporal distribution map of life, economic and ecological environment loss, and calculate the comprehensive loss index.
[0048] The vulnerability assessment model is constructed by calling the normalized data such as population density distribution, land use type, and GDP per unit area in the collection system, and using a method similar to step S3 and a vulnerability calculation formula to construct the vulnerability assessment model and generate a vulnerability index map.
[0049] The intelligent risk assessment model is constructed by invoking the failure probability, comprehensive loss index, and vulnerability index calculated in the dam failure probability, loss assessment, and vulnerability assessment models established in steps S2, S3, and S4, and using a formula to calculate a risk assessment value to conduct a risk assessment of the tailings dam failure. The intelligent risk assessment model is constructed by combining the dam failure probability, loss assessment, and vulnerability assessment models. The natural breakpoint classification method is used to classify risk levels and generate a risk level map to visualize the risk levels.
[0050] Specifically, this example selects a valley-type tailings pond in Jiangxi Province. Through the point cloud data uploaded by drone photography and data processing, it is known that the tailings pond has a stacking elevation of 150.0m and a total storage capacity of approximately 66 million m 3 Based on the reservoir capacity and dam height, this tailings pond is classified as a Class II pond. The tailings pond consists of five dams: one main dam and four auxiliary dams. The area is less susceptible to earthquakes, and the designed dam slope is approximately 1:5.
[0051] The basic parameters, material parameters, and boundary conditions of the tailings dam in the intelligent data acquisition system were used to establish an unsaturated seepage finite element model in Geo-Studio. The model was divided into 6361 nodes and 6161 quadrilateral and triangular mixed elements with a mesh size of 2 m. The material parameter statistics are shown in Table 1. The heavy rain intensity q was 5.56×10 -6 m / s. The boundary conditions of the calculation model are: set the upstream water level to 64.1m on the right side EF of the dam, set the rainfall boundary on the dam slope surface ABCDE, input the physical and mechanical parameters of the tailings material and the soil-water characteristics and permeability coefficient function curve, and then import the obtained seepage results into the SLOPE / W subdirectory under the SEEP / W module, and use the Morgenstern-Price method for stability analysis. Under heavy rainfall, the infiltration line suddenly drops along the slope at the clay core wall. The stability analysis results of the tailings dam show that the critical safety factor is 1.629, the mesh division and the sliding surface are as follows Figure 4 shown.
[0052] Table 1 Material parameter statistics
[0053]
[0054] The physical parameters of the tailings dam random field are shown in Table 2. The tailings fine sand is divided into two parts, A and B. The horizontal autocorrelation distance is 160m, and the vertical autocorrelation distance is 16m. The Latin hypercube sampling method (LHS) is used for random sampling to generate 2000 groups of samples. The KL series expansion method is used to discretize the random field and calculate the failure probability of the tailings dam.
[0055] Table 2 Random field physical parameter values of tailings dam
[0056]
[0057]
[0058] The neural network model was trained using the generated 2000 sets of random samples. 2 If the accuracy is greater than 90%, the model is considered to meet the requirements. Using the surrogate model combined with the Monte Carlo method for 10,000 calculations, the failure probability is 0.026.
[0059] The grid file and roughness file are obtained from the DEM elevation data and different land use type data obtained through data pre-processing in the intelligent data acquisition system. The breach flow is then calculated using the empirical formula and the flow dfs0 file is generated. The processed grid, flow, and roughness files are imported into the MIKE 21 hydrodynamic model. The tailings flow module is selected, and other parameters are adjusted to construct the dam-break tailings flow evolution model. The main parameters of the model are shown in Table 3.
[0060] Table 3 Dam break simulation parameter selection table
[0061]
[0062] like Figures 5 to 8 As shown in the figure, 10 minutes after the dam burst, the tailings flow reached the first village downstream and spread to the village ahead. 60 minutes after the dam burst, the tailings flow basically stopped flowing and accumulated in the downstream villages, with the maximum flooding depth reaching 2 to 3 meters.
[0063] The risk level is divided according to the calculation results. The division results are as follows: Figure 9 and as shown in Table 4.
[0064] Table 4 Statistics of division of dangerous areas
[0065]
[0066] According to statistics, the cultivated land areas in high, medium and low risk zones are 216,837 m 2 、25525.9m 2 、14921.2m 2 , the residential land area is 639.48m 2 、479.61m 2 、1385.54m 2 The loss rates for agriculture in high-, medium-, and low-risk areas were set at 88.7%, 60.4%, and 51%, respectively, while the loss rates for residential property were 91.6%, 79.9%, and 70.4%. Based on the formula in step three, direct economic losses are equal to the sum of the products of the areas of different risk areas and the corresponding loss rates, and indirect economic losses are equal to the sum of the products of the loss values of different properties and the corresponding indirect coefficients. The total direct economic losses are calculated to be approximately 2.4421 million yuan, and the total indirect economic losses are 408,314.11 yuan, for a total economic loss of 2.8504 million yuan. The main areas of land inundated by the dam tailings flow are cultivated land, forest land, grassland, and wasteland, which will suffer ecological damage, with their ecosystem service values of 11,271.89 yuan per hectare. 2 24265.35 yuan / hm 2 19168.36 yuan / hm 2 160.15 yuan / hm 2The loss rates for high, medium, and low-risk areas are 70%, 50%, and 30%, respectively. According to the formula in Step 3, life loss equals the sum of the product of the population in each residential area and the resident mortality rate, and ecological and environmental loss equals the sum of the product of the value of different ecosystem services and the ecological loss rate. The total life loss is calculated to be 207, and the total ecological and environmental loss is 94,837.97 yuan. Finally, combining the calculated life loss, total economic loss, and total ecological and environmental loss with the comprehensive loss index calculated in Step 3, we obtain a comprehensive loss index of 0.687.
[0067] The population density, GDP per capita, and land use type data from the intelligent data collection system were overlaid and normalized in Arcgis. The weight indexes were 0.0534, 0.1008, and 0.069, respectively. The vulnerability of different regions can be determined by applying them to the vulnerability calculation formula. The higher the vulnerability index, the higher the degree of damage. Finally, the vulnerability distribution map of the study area is obtained as shown below: Figure 10 shown.
[0068] In summary, the risk index can be obtained by using the calculated failure probability, comprehensive loss index and vulnerability index and the formula risk index is equal to the sum of the product of failure probability, comprehensive loss index and vulnerability index.
[0069] The foregoing merely represents preferred embodiments of the present invention, and while the description is relatively specific and detailed, it should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make various improvements and substitutions without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.
Claims
1. An intelligent assessment method for the risk of tailings dam overtopping and failure induced by heavy rain, characterized by: The following steps are involved: Step S1: Intelligent data collection (1) Access to original multi-source data: Use drones and other means to capture and upload point cloud data in real time, and create an intelligent data collection system that can integrate the required tailings dam geometry, population, satellite imagery, land type, and GDP per capita data; (2) Query and obtain specific data: by entering a specific year and region to filter in the data collection system, thus obtaining the raw data that meets the requirements; (3) Preprocessing the data: Check the various types of raw data obtained and identify missing values and duplicate values, and perform data format conversion and normalization to ensure data accuracy; Step S2: Constructing a dam failure probability model (1) Call the file of the geometric parameters of the tailings dam in the acquisition system to generate the physical model of the tailings dam. Through the interface with Geo-Studio, the unsaturated seepage model is constructed to conduct unsaturated seepage stability analysis, and the calculation result files such as the potential seepage surface, critical slip surface and safety factor value are obtained; (2) A model for tailings dam reliability analysis was constructed using the Monte Carlo method. Random samples were generated according to the Latin hypercube sampling method, and the KL series expansion method was used to discretize the random field. The random samples and stability calculation result files were used to train a back propagation neural network (BPNN) proxy model to perform reliability analysis and calculate the failure probability. Step S3: Constructing a dam break loss assessment model (1) Constructing a dam-break model: Using the data in the acquisition system to construct a dam-break model of the tailings pond, the submerged depth and flow velocity of the dam-break tailings flow are obtained over time; (2) Constructing a loss database: embed the empirical calculation formulas for life loss, economic loss, and ecological environment loss and the program that calls Arcgis into the database, input the result file of the dam break model, output the spatiotemporal distribution map of life, economic and ecological environment loss, and calculate the comprehensive loss index. The loss calculation and comprehensive loss index calculation formulas are as follows: K i =0.5KK 1i K 2i K 3i K 4i Among them, LOL is the life loss value, K i is the fatality rate of the ith residential area, K 1i ~K 4i are the distance coefficient, housing fragility coefficient, location coefficient and density coefficient of the ith residential area along the main river to the tailings dam; Among them, D is the direct economic loss value, S p is the loss value of the pth type of property, D is the direct economic loss value, K i is the indirect coefficient corresponding to the pth type of property; Among them, F is the ecological environment loss, f i is the ecological loss rate, V si The value of ecosystem services for type i; Where, E is the comprehensive loss index; Step S4: Constructing a dam break vulnerability assessment model Constructing a vulnerability assessment model: Calling the normalized population density distribution, land use type, and GDP per capita data from the acquisition system, and using a method similar to step S3 and the vulnerability calculation formula to construct a vulnerability assessment model and generate a vulnerability index map. The formula is: V = weight I × population density + weight II × GDP per capita + weight III × land use type, where V is the vulnerability index; Step S5: Constructing an intelligent dam break risk assessment model (1) Call the failure probability, comprehensive loss index and vulnerability index calculated in the dam failure probability, loss assessment and vulnerability assessment models established in steps S2, S3 and S4, and use the formula to calculate the risk assessment value to conduct a risk assessment of the tailings dam failure. The formula is: R=P f ×E×V Among them, R is the risk index, E is the total loss value, and V is the vulnerability index; (2) An intelligent risk assessment model is constructed by combining the dam failure probability, loss assessment and vulnerability assessment models. The natural breakpoint classification method is used to divide the risk levels and generate a risk level map to realize the visualization of risk levels.