Tailing pond grading early warning method and system based on intelligent disaster hidden danger identification
Through sensing data, the characteristics and types of disaster hazards of tailings ponds are identified, and risk assessment is carried out in combination with static and dynamic early warning modules, which solves the problems of rough risk assessment and lagging early warning of tailings ponds, and achieves accurate safety monitoring and early warning of tailings ponds.
Patent Information
- Application Number
- CN202510407075.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-02
- Publication Date
- 2025-08-05
AI Technical Summary
The existing tailings pond has rough risk assessment, lagging monitoring and early warning, and poor early warning accuracy. Traditional monitoring methods rely on manual inspection efficiency and are greatly affected by human factors. The existing early warning system lacks the ability to comprehensive multi-source data analysis and scientific grading methods.
The tailings pond hierarchical early warning method and system based on intelligent identification of disaster hazards, uses sensing data to identify the characteristics and types of disaster hazards, combines static and dynamic early warning modules to conduct risk assessment, and uses machine learning and time series analysis to generate hierarchical early warning information.
The accurate assessment of the disaster risk of tailings ponds has been achieved, the timeliness and accuracy of safety monitoring and early warning have been improved, and the safe operation of tailings ponds has been ensured.
Smart Images
Figure CN120431702A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of disaster early warning, and in particular to a tailings pond graded early warning method and system based on intelligent identification of disaster hazards. Background Art
[0002] Currently, numerous technical challenges remain to be addressed in the field of tailings dam safety monitoring and early warning. Traditional methods for monitoring tailings dam disasters rely primarily on regular manual inspections and simple instrumental measurements. This approach is not only inefficient but also subject to significant human influence, making it difficult to provide a comprehensive, real-time understanding of the dam's operational status. For example, manual inspections fail to detect subtle changes within the dam, and by the time a problem surfaces, it may have already developed to a more serious level. Furthermore, existing early warning systems are mostly based on a single monitoring indicator and lack the ability to comprehensively analyze multi-source data. For example, warnings based solely on water level or displacement data fail to fully consider the impact of environmental factors surrounding the tailings dam, such as rainfall, wind speed and direction, and temperature, on its stability. Furthermore, a scientific and rational approach to early warning classification is lacking, resulting in inaccurate early warning information and a failure to effectively support relevant decision-making.
[0003] The existing tailings ponds have technical problems such as rough risk assessment, delayed monitoring and early warning, and poor early warning accuracy. Summary of the Invention
[0004] This application provides a tailings pond classification warning method and system based on intelligent identification of disaster hazards, which is used to solve the technical problems of rough risk assessment, delayed monitoring and warning, and poor warning accuracy of existing tailings ponds.
[0005] In view of the above problems, this application provides a tailings pond classification warning method and system based on intelligent identification of disaster hazards.
[0006] The first aspect of the present application provides a tailings pond graded early warning method based on intelligent identification of disaster hazards, the method comprising:
[0007] Disaster hidden dangers are identified based on the tailings pond sensor data to obtain disaster hidden danger characteristics and disaster identification types; the disaster identification type and the disaster hidden danger characteristics are used as input data, and warning identification is performed through a static warning grading module to obtain a static warning identification level; the environmental time series parameters of the tailings pond are obtained; the sensor data timestamps of the disaster identification type and the disaster hidden danger characteristics are time-aligned with the environmental time series parameters; the disaster identification type is used to activate the pre-trained disaster type dynamic warning module, and the disaster identification type and disaster hidden danger characteristics are input into the disaster type dynamic warning module in combination with the time series alignment relationship of the environmental time series parameters to perform dynamic warning level identification and obtain the dynamic warning identification level and corresponding time series; the static warning identification level and the dynamic warning identification level are integrated for warning grading to generate graded warning information, and the graded warning information is used to make warnings of different levels and stages according to the warning level and warning time node.
[0008] The second aspect of the present application provides a tailings pond classification early warning system based on intelligent identification of disaster hazards, the system comprising:
[0009] A disaster identification type acquisition module is used to identify disaster hazards based on the tailings pond sensor data, and obtain disaster hazard characteristics and disaster identification types; a static warning identification level acquisition module is used to use the disaster identification type and the disaster hazard characteristics as input data, perform warning identification through the static warning grading module, and obtain a static warning identification level; an environmental timing parameter acquisition module is used to obtain the environmental timing parameters of the tailings pond; a timing alignment module is used to perform timing alignment with the environmental timing parameters based on the sensor data timestamps of the disaster identification type and the disaster hazard characteristics; a dynamic warning identification level acquisition module is used to activate the pre-trained disaster type dynamic warning module using the disaster identification type, and input the disaster identification type and disaster hazard characteristics combined with the timing alignment relationship of the environmental timing parameters into the disaster type dynamic warning module for dynamic warning level identification, to obtain the dynamic warning identification level and corresponding timing; a graded warning information generation module is used to integrate the static warning identification level with the dynamic warning identification level for warning grading, and generate graded warning information, which is used to perform warnings of different levels and stages according to the warning level and warning time node.
[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0011] Disaster hazard identification is performed based on tailings pond sensor data to obtain disaster hazard characteristics and disaster identification types; the disaster identification type and the disaster hazard characteristics are used as input data to perform warning identification through a static warning grading module to obtain a static warning identification level; the environmental time series parameters of the tailings pond are obtained; time series alignment is performed; the disaster identification type is used to activate a pre-trained disaster type dynamic warning module, and the disaster identification type and disaster hazard characteristics are combined with the time series alignment relationship of the environmental time series parameters into the disaster type dynamic warning module to perform dynamic warning level identification, obtaining a dynamic warning identification level and corresponding time series; the static warning identification level and the dynamic warning identification level are integrated for warning grading to generate graded warning information. This achieves the technical effect of achieving accurate assessment of tailings pond disaster risks, improving the timeliness and accuracy of tailings pond safety monitoring and warning, and ensuring the safe operation of the tailings pond. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0013] Figure 1 A schematic flow chart of a tailings pond classification warning method based on intelligent identification of disaster hazards provided in an embodiment of the present application;
[0014] Figure 2 Schematic diagram of the structure of the tailings pond classification warning system based on intelligent identification of disaster hazards provided in the embodiment of the present application.
[0015] Explanation of the reference numerals: disaster identification type acquisition module 10 , static warning identification level acquisition module 20 , environmental time series parameter acquisition module 30 , time series alignment module 40 , dynamic warning identification level acquisition module 50 , graded warning information generation module 60 . DETAILED DESCRIPTION
[0016] This application provides a tailings pond classification warning method and system based on intelligent identification of disaster hazards, which is used to solve the technical problems of rough risk assessment, delayed monitoring and warning, and poor warning accuracy of existing tailings ponds.
[0017] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative work are within the scope of protection of this application.
[0018] Example 1, as Figure 1 As shown, the present application provides a tailings pond classification early warning method based on intelligent identification of disaster hazards, the method comprising:
[0019] Step S100: Identify potential disaster risks based on tailings pond sensor data to obtain potential disaster risk characteristics and disaster identification types.
[0020] Specifically, tailings ponds are equipped with numerous sensors to collect various sensor data. These data cover key information such as stack displacement, pore water pressure, water level, and vibration. After the data is collected, it is first input into a pre-built disaster hazard identification model. When building this model, it is necessary to construct a multi-category tailings pond disaster database based on a large amount of historical data. Then, training and test data sets for each disaster type are extracted from this database. These data sets contain stack displacement, pore water pressure, water level, vibration, and the corresponding disaster hazard characteristics and disaster types. Machine learning algorithms are used to train and test the identification models for each disaster type, and ultimately a reliable disaster hazard identification model is obtained. After pre-processing the sensor data collected in real time and inputting it into this model, the disaster hazard characteristics of the tailings pond and the corresponding disaster identification type can be output, thereby determining the potential safety hazards and hazard types that may currently exist in the tailings pond.
[0021] Step S200: using the disaster identification type and the disaster hidden danger characteristics as input data, performing warning identification through a static warning grading module to obtain a static warning identification level.
[0022] Specifically, the obtained disaster identification type and disaster hidden danger characteristics are used as input data and passed into the static early warning classification module. Before building this module, it is necessary to collect full-level samples of each disaster type. These samples cover all state data of the disaster type, including normal state and dangerous state at different levels, as well as the corresponding disaster hidden danger characteristics and disaster loss data under each state. Risk classification of all-level samples is carried out according to the disaster loss data, and the risk classification threshold is determined. Then, based on this threshold, the full-level samples of each disaster type are divided separately to establish a risk classification feature comparison list for each disaster type. The static early warning classification module constructed based on these lists will use the input data as an index to search and match in the risk classification feature comparison list, and then obtain the static early warning identification level, and preliminarily evaluate the safety risk level of the tailings pond from a static level.
[0023] Step S300: Obtaining environmental time series parameters of the tailings pond.
[0024] Specifically, various specialized monitoring devices, such as rain sensors, wind speed and direction meters, temperature sensors, and water level monitors, are installed in the area surrounding the tailings pond. Rain sensors monitor rainfall in real time, recording rainfall intensity and duration in detail; wind speed and direction meters accurately measure wind speed and direction; temperature sensors capture changes in ambient temperature; and water level monitors closely monitor fluctuations in the water level within the tailings pond and the surrounding groundwater level. These devices continuously collect data, generating environmental time series data including rainfall, wind speed and direction, temperature, and water level changes. These data, with precise time stamps, provide a key basis for subsequent time series alignment with disaster risk sensor data and for risk assessment using the dynamic early warning module, thus enabling more accurate tiered early warnings for tailings ponds.
[0025] Step S400: performing time sequence alignment with the environmental time sequence parameters according to the timestamp of the sensor data of the disaster identification type and the disaster hidden danger characteristics.
[0026] Specifically, after completing disaster hazard identification, obtaining sensor data indicating the hazard type and characteristics, and labeling them with corresponding timestamps, and also acquiring the tailings pond's environmental time series parameters, a time series alignment operation is performed. This operation, based on the time dimension, closely correlates the sensor data reflecting the tailings pond's status with the surrounding environmental data. Based on the sensor data's timestamps, the time stamps in the environmental time series parameters are compared one by one, and the hazard hazard sensor data from the same or similar moments are matched and integrated with the corresponding environmental parameters. For example, if the stack displacement sensor data at a certain moment shows an abnormal change, the corresponding environmental parameters such as rainfall, wind speed and direction, temperature, and water level change are identified and combined. This time series alignment ensures that the input data is synchronized and correlated when subsequently performing risk assessment using the disaster type dynamic warning module. This ensures that the assessment results more accurately reflect the actual risk status of the tailings pond at a specific point in time and under specific environmental conditions, providing a reliable data foundation for generating graded warning information.
[0027] Step S500: Use the disaster identification type to activate the pre-trained disaster type dynamic warning module, and input the disaster identification type and disaster hazard characteristics combined with the timing alignment relationship of the environmental timing parameters into the disaster type dynamic warning module to perform dynamic warning level identification, and obtain the dynamic warning identification level and corresponding timing.
[0028] Specifically, the pre-trained dynamic disaster warning module, using a long short-term memory (LSTM) network, is activated based on the disaster type. After acquiring time-aligned data on the disaster type, hazard characteristics, and environmental time series parameters, the LSTM network leverages its unique advantages. Through its gating mechanism, the LSTM effectively handles long-term dependencies in the input data. Its forget gate determines which information from the previous moment is retained or discarded, its input gate controls which parts of the current input data are added to the memory cell, and its output gate determines which information is output. When processing tailings pond data, it effectively integrates and analyzes hazard characteristic data such as pile displacement, pore water pressure, water level, and vibration at different times, as well as corresponding environmental time series data such as rainfall, wind speed and direction, temperature, and water level changes. By training on a large amount of historical data, the LSTM network constructs a complex model of the time-varying state of the tailings pond and environmental factors. Then, when fed with current data, it accurately identifies the dynamic warning level and generates the corresponding time series, providing critical dynamic risk assessment information for tailings pond safety management.
[0029] Step S600: The static warning identification level and the dynamic warning identification level are integrated into warning levels to generate graded warning information. The graded warning information is used to provide warnings at different levels and stages according to the warning level and warning time node.
[0030] Specifically, the static and dynamic warning identification levels are deeply integrated to produce graded warning information that effectively guides safety precautions. First, based on the static warning identification level, real-time warning information is generated, providing an intuitive assessment of the current risk level of the tailings pond. Next, focusing on the dynamic warning identification level, the time difference between it and the static warning identification level and the risk deterioration are analyzed in depth. The time difference reflects the time span of risk change, while the risk deterioration reflects the degree of risk aggravation. These two key indicators are used to determine the warning time constraint for the dynamic risk and the corresponding warning level information. For example, if the dynamic warning identification level indicates a rapid increase in risk over a short period of time and the risk deterioration is significantly greater than the static warning level, the warning time constraint will be shortened accordingly, and the warning level will be increased accordingly. Finally, based on the dynamic risk warning time constraint and the corresponding warning level information, post-level warning information is generated, detailing the time interval between the real-time warning information and the specific warning level information. After this integration process, the graded warning information is ultimately generated. This information will be strictly distributed to relevant personnel in a phased and targeted manner according to different warning levels and precise warning time nodes, ensuring that staff can take corresponding preventive and response measures at the most appropriate time based on different levels of risk, and maximize the safe operation of the tailings pond.
[0031] In one possible implementation, step S100 further includes:
[0032] Step S110: The tailings pond sensor data includes stack displacement, pore water pressure, water level, and vibration. The stack displacement, pore water pressure, water level, and vibration are input into a disaster hazard identification model to obtain the disaster hazard characteristics and disaster identification type.
[0033] Specifically, the random forest algorithm is used as the core algorithm of the disaster hazard identification model. Sensors deployed in the tailings pond continuously collect data such as pile displacement, pore water pressure, water level, and vibration. The random forest algorithm processes this data. First, during the training phase, it randomly extracts multiple subsets from the original training data set with replacement, and each subset is used to build a decision tree. In the process of building the decision tree, for each node split, the algorithm randomly selects a part of the features to find the best split point, which can increase the independence between decision trees. After many decision trees form a forest, each decision tree makes an independent prediction for the input pile displacement, pore water pressure, water level, and vibration data. Then, the prediction results of all decision trees are combined to obtain the final prediction through voting or averaging. For example, when identifying the characteristics of disaster hazards and the types of disaster identification, the data features are mined according to different decision trees to determine whether the tailings pond has the risk of dam instability suggested by abnormal displacement patterns, or to identify disaster types such as pipe bursts based on the characteristics of pore water pressure and water level changes. Ultimately, the current disaster hazard characteristics of the tailings pond and the corresponding disaster identification types are given, providing a reliable basis for subsequent early warning work.
[0034] In one possible implementation, step S110 further includes:
[0035] Step S111: Construct a multi-category tailings dam disaster database based on historical data.
[0036] Step S112: constructing a training data set and a test data set for each disaster type based on the multi-category tailings pond disaster database, wherein the training data set and the test data set both include stack displacement, pore water pressure, water level, vibration and corresponding disaster hazard characteristics and disaster types.
[0037] Step S113: Based on the training data set and the test data set of each disaster type, the identification model of each disaster type is trained and tested by a machine learning algorithm to obtain a disaster hazard identification model.
[0038] Step S114: After pre-processing the stack displacement, pore water pressure, water level, and vibration, the pre-processing steps are input into the disaster hazard identification model to identify the characteristics of the disaster hazard and the type of disaster. The disaster hazard identification model then outputs the characteristics of the disaster hazard and the type of disaster identified.
[0039] Specifically, a wide range of historical data was collected to construct a multi-category tailings dam disaster database. This historical data includes monitoring records of pile displacement, pore water pressure, water level, vibration, and other aspects of different tailings dams over many years of operation. It also covers detailed information on various disasters such as dam collapses, piping, and landslides. This includes the corresponding monitoring data at the time of the disaster, the characteristics of the potential hazards, and the specific type of disaster. This data was systematically organized and categorized to form a comprehensive and multi-dimensional database, laying a solid foundation for subsequent model training and testing.
[0040] Based on a multi-category tailings dam disaster database, we carefully divided training and test datasets for each disaster type. Each dataset contains key data such as stack displacement, pore water pressure, water level, and vibration, as well as the corresponding disaster hazard characteristics and disaster type. The training dataset allows the random forest algorithm to learn the complex relationship between various monitoring data and disaster situations under different disaster scenarios, while the test dataset is used to evaluate the performance of the trained model.
[0041] The model was trained and tested using a random forest algorithm using training and test datasets for each disaster type. The algorithm randomly extracts multiple subsets from the training dataset with replacement and constructs a decision tree for each subset. During the decision tree construction process, for each node, a subset of features is randomly selected to determine the optimal split point, thereby increasing independence between decision trees. After the numerous decision trees form the random forest model, the training data is continuously learned and parameters are adjusted. The model is then validated using a test dataset to examine its predictive accuracy and generalization ability on unknown data. Based on these test results, the model is further optimized, ultimately resulting in a high-performance disaster hazard identification model.
[0042] The real-time data on stack displacement, pore water pressure, water level, and vibration is preprocessed. This includes removing noise and outliers from the data and normalizing the data to make it comparable across different ranges. The preprocessed data is then fed into a trained disaster risk identification model based on a random forest algorithm. Each decision tree in the model independently analyzes and predicts the input data. A voting mechanism then synthesizes the predictions from all decision trees to identify the corresponding disaster risk characteristics and types. Accurate results are then output, effectively identifying and warning of tailings pond disaster risks.
[0043] In one possible implementation, step S200 further includes:
[0044] Step S210: Collect full-level samples of each disaster type. The full-level samples are all state data of the disaster type, including normal state and dangerous state at different levels, as well as corresponding disaster hazard characteristics and disaster loss data under each state.
[0045] Step S220: Risk grading of the samples of all levels is performed according to the disaster loss data, and a risk grading threshold is determined.
[0046] Step S230: Divide all level samples of each disaster type according to the risk grading threshold, and establish a risk grading feature comparison list for each disaster type.
[0047] Step S240: Based on the risk grading feature comparison list of each disaster type, a static warning grading module is constructed, which is used to search the risk grading feature comparison list using the input data as an index to obtain the static warning identification level.
[0048] Specifically, comprehensive sampling of all disaster types and levels requires in-depth collection of detailed data on tailings pond hazards under different conditions, from normal operation to various hazard levels. For each condition, not only must corresponding data on potential hazards such as pile displacement, pore water pressure, water level, and vibration be recorded, but also accurate statistics on the corresponding disaster losses must be compiled. This loss data includes information on potential economic losses, environmental damage, and risk of casualties. This extensive and meticulous collection provides a rich and comprehensive data foundation for subsequent risk assessment and early warning grading.
[0049] The K-Means clustering algorithm is used to classify all risk categories based on disaster loss data and determine risk classification thresholds. First, the collected disaster loss data for all disaster categories is organized into multidimensional vectors. These vectors contain information on economic losses, potential casualties, and the degree of environmental damage. The K-Means clustering algorithm randomly selects K initial cluster centers (K is set based on the tailings pond risk level; for example, 4 represents low, medium, high, and very high risk levels). The distance (Euclidean distance is commonly used) between each sample vector and each cluster center is calculated, and the sample is assigned to the cluster with the closest cluster center. Next, the center of each cluster is recalculated, and the mean of all sample vectors within the cluster is used as the new cluster center. This process of sample assignment and cluster center updating is repeated until the cluster center remains stable or changes minimally, at which point the clustering results become stable. Each cluster represents a risk level, and the boundary between clusters is determined as the risk classification threshold based on statistical data analysis and practical business considerations. For example, after clustering a large number of samples, the numerical range of certain indicators such as economic loss and environmental damage degree is determined to be the boundary between low risk and medium risk. This numerical range is a key risk classification threshold, which can be used to achieve scientific classification of tailings pond disaster risks.
[0050] Based on the established risk grading thresholds, all samples of each disaster type are meticulously divided. Samples of each disaster type are categorized and organized according to different risk levels. For each risk level, the unique disaster hazard characteristics of the samples at that level are identified, thereby establishing a comparison list of risk grading characteristics for each disaster type. This comparison list clearly demonstrates the correspondence between different risk levels and corresponding disaster hazard characteristics, facilitating quick and accurate risk assessment.
[0051] Based on the established risk classification feature comparison list for each disaster type, a static early warning classification module was built. This module uses the input disaster identification type and potential hazard feature data as an index to efficiently search and match within the risk classification feature comparison list. This method allows for rapid identification of the corresponding risk level, thereby obtaining a static early warning identification level. This provides a timely and intuitive assessment of the current static risk level of the tailings pond, enabling the implementation of appropriate preventative measures.
[0052] In one possible implementation, step S500 further includes:
[0053] Step S510: Establish a full-cycle time series chain according to the level stage of the full-level samples of each disaster type, collect environmental parameters to establish a corresponding relationship between the environmental parameters and the full-level samples, and fit them into the full-cycle time series chain.
[0054] Step S520: Analyze the impact relationship of environmental parameters on each level stage according to the full-cycle time series chain, perform time series risk prediction training through the time series prediction model architecture based on the impact relationship, and obtain a dynamic risk prediction model for each disaster type.
[0055] Step S530: Combine the dynamic risk prediction models of all disaster types to construct the dynamic early warning module.
[0056] Specifically, for all-level samples of each disaster type, a full-cycle time series chain is established based on its level stage. The full-level samples cover complete data from normal status to different risk level stages. When constructing the time series chain, the data of each level stage are arranged in order with time as the axis. At the same time, environmental parameters closely related to the tailings pond, such as rainfall, wind speed, temperature, etc., are continuously collected. Through in-depth mining and analysis of historical data, an accurate correspondence between these environmental parameters and full-level samples is established. For example, when the rainfall reaches a certain numerical range, it corresponds to the tailings pond sample being at a specific risk level stage. Subsequently, these corresponding relationships are accurately fitted into the full-cycle time series chain, thereby constructing a complete time series data structure that includes the influence of environmental factors.
[0057] Based on the constructed full-cycle time series chain, we conduct an in-depth analysis of the impact of environmental parameters on each level stage. The LSTM algorithm has unique advantages in processing time series data, capturing long-term dependencies in the data. The data in the full-cycle time series chain is divided into an input sequence and a corresponding target sequence. The input sequence contains environmental parameters and historical disaster level information, while the target sequence represents future disaster levels. An LSTM network is trained for time series risk prediction. The forget gate in the network determines which information from the previous moment is forgotten, the input gate controls which parts of the current input data are added to the cell state, and the output gate determines the final output information. By continuously adjusting the network weights and biases, the error between the predicted results and the actual disaster level is minimized, ultimately obtaining a dynamic risk prediction model for each disaster type.
[0058] The dynamic risk prediction models for all disaster types are combined to construct a dynamic early warning module. During their respective training, the dynamic risk prediction models for each disaster type have learned the risk evolution patterns of that disaster type under different environmental conditions. By integrating these models, the dynamic early warning module comprehensively considers multiple disaster types. Based on real-time input environmental parameters and the current disaster status, it accurately predicts the risk level of each disaster type at different points in the future, providing timely and effective early warning information for tailings pond safety management.
[0059] In one possible implementation, step S500 further includes:
[0060] Step S510: The environmental parameters include: rainfall, wind speed and direction, temperature, and water level changes.
[0061] Specifically, for each disaster type, full-level sample, and the construction of a full-cycle time series chain based on its level stage, it is necessary to comprehensively collect specific environmental parameters—namely, rainfall, wind speed and direction, temperature, and water level changes—and establish a close correspondence between these parameters and the full-level sample, ultimately fitting them into the full-cycle time series chain. Around the tailings pond, various specialized monitoring devices play a key role. Rain sensors continuously record rainfall data, accurately recording rainfall intensity and cumulative rainfall down to the minute. Anemometers monitor wind speed and direction in real time, providing a basis for analyzing wind impacts on the tailings pond. Temperature sensors stably measure ambient temperature, capturing its fluctuations over time. Water level monitors closely monitor the dynamics of the water level within the tailings pond and the surrounding groundwater level. After collecting these environmental parameter data, in-depth research was conducted to examine the correlation between environmental parameter changes in historical data and the various levels of tailings pond disasters. For example, it was found that when rainfall exceeds a certain value and continues to increase, the probability of abnormal tailings pond pile displacement increases significantly, indicating a corresponding relationship between this rainfall amount and a specific disaster level stage. Similarly, the impact of wind speed and direction, temperature and water level changes on the stability of tailings ponds is analyzed, and their connection with different states in all-level samples is found. Then, these corresponding relationships are accurately fitted into the full-cycle time series chain constructed according to the level stages of all-level samples in chronological order, laying a solid foundation for the subsequent in-depth analysis of the impact of environmental factors on the disaster risk of tailings ponds and the establishment of a dynamic risk prediction model.
[0062] In one possible implementation, step S600 further includes:
[0063] Step S610: Generate real-time level warning information according to the static warning recognition level.
[0064] Step S620: Determine the warning time constraint of the dynamic risk and the corresponding level warning information according to the dynamic warning identification level, the time difference between the dynamic warning identification level and the static warning identification level, and the risk deterioration amount.
[0065] Step S630: generating post-level warning information according to the warning time constraint of the dynamic risk and the corresponding level warning information, including the interval time with the real-time level warning information and warning level information.
[0066] Specifically, real-time warning information is generated based on the static warning identification level. This level is an assessment of the current risk level based on tailings dam sensor data and historical disaster sample data. Once this level is obtained, it is immediately converted into intuitive real-time warning information. For example, the current risk status of the tailings dam is displayed to relevant personnel using eye-catching color codes and concise text descriptions, such as "Low risk, currently operating basically stable" or "High risk, urgent measures required", so that personnel can immediately understand the immediate safety status of the tailings dam.
[0067] Based on the dynamic warning identification level, the time difference between it and the static warning identification level, and the risk deterioration, the system determines the warning time constraint for the dynamic risk and the corresponding warning message. The dynamic warning identification level reflects the changing trend of the tailings dam risk over time and environmental factors. The time difference reflects the time span of the risk change, for example, the time it takes for the dynamic risk to progress from a low risk at the static warning level to the current higher risk. The risk deterioration quantifies the degree of risk escalation, for example, from a minor displacement change to a severe displacement change that could lead to dam collapse. Taking these factors into consideration, the system uses a complex algorithm to determine the warning time constraint for the dynamic risk—the time it will take for the risk to reach its expected level—and generates a warning message of the corresponding level. If the dynamic risk indicates that the risk will increase from the current medium risk to a high risk within the next six hours, the warning time constraint will be six hours, and the corresponding warning message will be "High risk warning, risk escalation expected in six hours, precautionary measures required."
[0068] Based on the warning time constraints of dynamic risks and the corresponding level warning information, post-level warning information is generated to further refine the warning content and clarify the interval time with the real-time level warning information and the warning level information. Assuming that the real-time level warning is medium risk, and the dynamic risk warning shows that it will be upgraded to high risk after 6 hours, the post-level warning information may be "A high-risk warning will be issued 6 hours after the real-time warning. During this period, pay close attention to changes in the tailings pond." This allows staff to clearly understand the time nodes and risk level changes of warnings at different stages, so that they can plan and arrange response measures in advance to ensure the safe operation of the tailings pond.
[0069] Example 2 is based on the same inventive concept as the tailings pond classification warning method based on intelligent identification of disaster hazards in the previous embodiment. Figure 2 As shown, this application provides a tailings pond classification warning system based on intelligent identification of disaster hazards. The system and method embodiments in the embodiments of this application are based on the same inventive concept. The system includes:
[0070] The disaster identification type acquisition module 10 is used to identify disaster hazards based on tailings pond sensor data and obtain disaster hazard characteristics and disaster identification types.
[0071] The static warning identification level acquisition module 20 is used to take the disaster identification type and the disaster hidden danger characteristics as input data, perform warning identification through the static warning classification module, and obtain a static warning identification level.
[0072] The environmental time series parameter acquisition module 30 is used to obtain the environmental time series parameters of the tailings pond.
[0073] The timing alignment module 40 is configured to perform timing alignment with the environmental timing parameters according to the timestamp of the sensor data of the disaster identification type and the disaster hidden danger characteristics.
[0074] The dynamic warning identification level acquisition module 50 is used to activate the pre-trained disaster type dynamic warning module using the disaster identification type, and input the disaster identification type and disaster hazard characteristics combined with the timing alignment relationship of the environmental timing parameters into the disaster type dynamic warning module to perform dynamic warning level identification, and obtain the dynamic warning identification level and corresponding timing.
[0075] The graded warning information generation module 60 is used to integrate the static warning identification level and the dynamic warning identification level into warning levels to generate graded warning information. The graded warning information is used to provide warnings at different levels and stages according to the warning level and warning time node.
[0076] Furthermore, the system is also used to implement the following functions:
[0077] The tailings pond sensor data includes pile displacement, pore water pressure, water level, and vibration. The pile displacement, pore water pressure, water level, and vibration are input into a disaster hazard identification model to obtain the disaster hazard characteristics and disaster identification type.
[0078] Furthermore, the system is also used to implement the following functions:
[0079] A multi-category tailings pond disaster database is constructed based on historical data; a training data set and a test data set for each disaster type are constructed based on the multi-category tailings pond disaster database, and the training data set and the test data set both include stack displacement, pore water pressure, water level, vibration and corresponding disaster hazard characteristics and disaster types; based on the training data set and the test data set for each disaster type, an identification model for each disaster type is trained and tested through a machine learning algorithm to obtain a disaster hazard identification model; after pre-processing the stack displacement, pore water pressure, water level and vibration, the data are input into the disaster hazard identification model to identify the disaster hazard characteristics and disaster types, and the disaster hazard characteristics and disaster identification type are output through the disaster hazard identification model.
[0080] Furthermore, the system is also used to implement the following functions:
[0081] Collect full-level samples of each disaster type, which are all status data of the disaster type, including normal status and dangerous status at different levels, as well as corresponding disaster hidden danger characteristics and disaster loss data under each state; perform risk classification on the full-level samples according to the disaster loss data, and determine the risk classification threshold; divide the full-level samples of each disaster type according to the risk classification threshold, and establish a risk classification feature comparison list for each disaster type; based on the risk classification feature comparison list for each disaster type, construct a static warning classification module for searching the risk classification feature comparison list using input data as an index to obtain the static warning identification level.
[0082] Furthermore, the system is also used to implement the following functions:
[0083] According to the level stages of the full-level samples of each disaster type, a full-cycle time series chain is established, and environmental parameters are collected to establish the corresponding relationship between the environmental parameters and the full-level samples, and fitted into the full-cycle time series chain; the influence relationship of environmental parameters on each level stage is analyzed according to the full-cycle time series chain, and time series risk prediction training is performed through the time series prediction model architecture based on the influence relationship to obtain a dynamic risk prediction model for each disaster type; the dynamic risk prediction models of all disaster types are combined to construct the dynamic warning module.
[0084] Furthermore, the system is also used to implement the following functions:
[0085] The environmental parameters include: rainfall, wind speed and direction, temperature, and water level changes.
[0086] Furthermore, the system is also used to implement the following functions:
[0087] Generate real-time level warning information based on the static warning identification level; determine the warning time constraint of the dynamic risk and the corresponding level warning information based on the dynamic warning identification level, and the time difference and risk deterioration amount between the dynamic risk and the static warning identification level; generate post-level warning information based on the warning time constraint of the dynamic risk and the corresponding level warning information, including the interval time with the real-time level warning information and warning level information.
[0088] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0089] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.
[0090] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.
Claims
1. A tailings pond classification early warning method based on intelligent identification of disaster hazards, characterized by: include: Identify potential disaster hazards based on tailings pond sensor data to obtain potential disaster characteristics and disaster identification types; Using the disaster identification type and the disaster hidden danger characteristics as input data, a static warning classification module is used to perform warning identification to obtain a static warning identification level; Obtain environmental time series parameters of the tailings pond; Performing time sequence alignment with the environmental time sequence parameters based on the timestamp of the sensor data of the disaster identification type and the disaster hidden danger characteristics; Utilizing the disaster identification type to activate the pre-trained disaster type dynamic warning module, the disaster identification type and disaster hidden danger characteristics are combined with the time sequence alignment relationship of the environmental time series parameters to be input into the disaster type dynamic warning module for dynamic warning level identification, thereby obtaining the dynamic warning identification level and the corresponding time series; The static warning identification level and the dynamic warning identification level are integrated into warning levels to generate graded warning information. The graded warning information is used to issue warnings at different levels and stages according to the warning level and warning time node.
2. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 1 is characterized in that: The method of identifying potential disaster risks based on tailings pond sensor data includes: The tailings pond sensor data includes pile displacement, pore water pressure, water level, and vibration. The pile displacement, pore water pressure, water level, and vibration are input into a disaster hazard identification model to obtain the disaster hazard characteristics and disaster identification type.
3. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 2 is characterized in that: Obtaining the disaster hazard characteristics and disaster identification types, including: Construct a multi-category tailings dam disaster database based on historical data; Constructing a training data set and a test data set for each disaster type based on the multi-category tailings dam disaster database, wherein the training data set and the test data set both include stack displacement, pore water pressure, water level, vibration, and corresponding disaster hazard characteristics and disaster types; Based on the training data set and test data set of each disaster type, a recognition model of each disaster type is trained and tested by a machine learning algorithm to obtain a disaster hazard recognition model; After pre-processing, the stack displacement, pore water pressure, water level and vibration are input into the disaster hazard identification model to identify the disaster hazard characteristics and disaster type, and the disaster hazard characteristics and disaster identification type are output through the disaster hazard identification model.
4. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 1 is characterized in that: The static warning classification module is used to perform warning identification to obtain a static warning identification level, including: Collect full-level samples of each disaster type. The full-level samples are all state data of the disaster type, including normal state and dangerous state at different levels, as well as the corresponding disaster hazard characteristics and disaster loss data under each state; Perform risk grading on the samples of all levels according to the disaster loss data, and determine a risk grading threshold; Divide all level samples of each disaster type according to the risk grading threshold, and establish a risk grading feature comparison list for each disaster type; Based on the risk grading feature comparison list of each disaster type, a static warning grading module is constructed to search the risk grading feature comparison list using input data as an index to obtain the static warning identification level.
5. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 4 is characterized in that: Before activating the pre-trained disaster type dynamic warning module using the disaster identification type, the method further includes: According to the level stage of the full-level samples of each disaster type, a full-cycle time series chain is established, and environmental parameters are collected to establish a corresponding relationship between the environmental parameters and the full-level samples, and fitted into the full-cycle time series chain; Analyze the impact of environmental parameters on each level stage according to the full-cycle time series chain, conduct time series risk prediction training based on the impact relationship through the time series prediction model architecture, and obtain a dynamic risk prediction model for each disaster type; The dynamic risk prediction models of all disaster types are combined to construct the dynamic early warning module.
6. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 5 is characterized in that: The environmental parameters include: rainfall, wind speed and direction, temperature, and water level changes.
7. The tailings pond classification early warning method based on intelligent identification of disaster hidden dangers according to claim 1 is characterized in that: The static warning identification level and the dynamic warning identification level are integrated into warning levels to generate graded warning information, including: Generate real-time warning information based on the static warning recognition level; Determine the warning time constraint of the dynamic risk and the corresponding level warning information based on the dynamic warning identification level, the time difference between the dynamic warning identification level and the static warning identification level, and the risk deterioration amount; According to the warning time constraint of the dynamic risk and the corresponding level warning information, post-level warning information is generated, including the interval time with the real-time level warning information and the warning level information.
8. Tailings pond classification early warning system based on intelligent identification of disaster hazards, characterized by: The system is used to implement the tailings pond graded early warning method based on intelligent identification of disaster hazards according to any one of claims 1 to 7, and the system includes: The disaster identification type acquisition module is used to identify disaster hazards based on tailings pond sensor data and obtain disaster hazard characteristics and disaster identification types; A static warning identification level acquisition module is used to use the disaster identification type and the disaster hidden danger characteristics as input data, perform warning identification through the static warning classification module, and obtain a static warning identification level; Environmental time series parameter acquisition module, used to obtain environmental time series parameters of the tailings pond; A timing alignment module, configured to perform timing alignment with the environmental timing parameters based on the timestamp of the sensor data of the disaster identification type and the disaster hidden danger characteristics; A dynamic warning recognition level acquisition module is used to activate a pre-trained disaster type dynamic warning module using the disaster recognition type, input the disaster recognition type and disaster hidden danger characteristics into the disaster type dynamic warning module in combination with the time sequence alignment relationship of the environmental time sequence parameters to perform dynamic warning level recognition, and obtain a dynamic warning recognition level and corresponding time sequence; The graded warning information generation module is used to integrate the static warning identification level and the dynamic warning identification level into warning levels to generate graded warning information. The graded warning information is used to provide warnings at different levels and stages according to the warning level and warning time node.