Real-time early warning method and system for coordinated deformation of tailings dam based on multi-source fusion

Abstract

A real-time early warning method and system for collaborative deformation of tailings dam bodies based on multi-source fusion. The system comprises: a data acquisition module for collecting multi-source monitoring data and sending it to an edge computing module; an edge computing module for performing spatiotemporal synchronization and rapid feature extraction of multi-source data; a central processing module for calculating DISE values ​​based on target feature data and performing early warning grading of sliding surface generation probabilities based on the range of DISE values; and an early warning execution module for executing corresponding warning actions under the control of the central processing module. Methods: The data acquisition module is used to perform real-time collection of multi-source monitoring data; the edge computing module calculates target feature data based on the received multi-source monitoring data; and the central processing module performs graded early warnings based on the received target feature data. This system and method can efficiently and accurately achieve early and precise prediction and graded early warning of dam instability risks, which is conducive to the active prevention and control of dam deformation risks.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring and safety early warning, and specifically provides a real-time early warning method and system for coordinated deformation of a tailings dam body based on multi-source fusion. Background Art

[0002] Heavy rainfall infiltration causes the seepage line to rise sharply, weakening the soil's shear strength and potentially triggering deep slippage or even dam failure. Sustained rainfall exacerbates pore water pressure in slopes, triggering local landslides that evolve into continuous sliding surfaces. While online seepage line monitoring systems (such as piezometer clusters) are currently widely deployed in the industry, they only capture the water level elevation in the saturated zone and lack the coordinated analysis of multiple factors, including internal displacement, dry beach conditions, and seepage pressure. This makes it impossible to predict structural instability risks, such as the formation of sliding surfaces. Furthermore, existing early warning systems rely on a "threshold alarm → manual verification → decision-making" process, with response delays often exceeding several hours, making it difficult to meet the minute-by-minute safety requirements of heavy rainfall conditions.

[0003] Traditional technologies suffer from four core flaws: First, monitoring elements are too one-sided. Dam instability is the result of a multi-physics coupling involving displacement deformation, seepage field distribution, and dry beach conditions. Existing systems focus solely on the water level at the seepage line, ignoring key indicators such as displacement field changes from internal displacement monitoring units (such as deep inclinometer arrays), the risk of a sharp decrease in dry beach length from dry beach monitoring units (such as radar water level gauges), and the dynamic gradient of seepage pressure. This results in a lack of a collaborative analysis framework for multi-source deformation parameters. For example, the dynamic correlation between the rate of change of the spatial slope of the seepage line and internal displacement acceleration is not quantified, resulting in missed detection of early signs of local slip. Second, risk assessment is rigid. Traditional models rely on fixed threshold alarms, and their parameters (such as critical displacement rate) are often calibrated based on historical experience. However, dam stability is significantly affected by time-varying factors such as rainfall intensity and material aging, making static models unsuitable for complex conditions. For example, during heavy rainfall, changes in soil permeability cause the seepage line to rise at a doubled rate. However, static thresholds cannot dynamically correlate with the intensity of the rainstorm to adjust the warning threshold, resulting in delayed response. Third, data coordination is weak. Multi-source sensors (such as internal displacement units, seepage pressure units, infiltration line slope units, and dry beach monitoring units) experience temporal and spatial offsets in data due to differences in sampling frequency and spatial resolution. Displacement mutations, seepage pressure gradients, and dry beach change rates are not correlated and analyzed on a unified time scale, weakening the ability to identify coordinated deformation risks. Fourth, emergency response is fragmented. Monitoring and response processes are disconnected, requiring manual flood discharge or evacuation after early warning. When the dam body experiences rapid deformation (such as high-risk situations requiring forced flood discharge), delays in manual operations can lead to escalating danger. Furthermore, response effectiveness is not fed back into model optimization, leaving the risk of similar accidents recurring at a high level.

[0004] Therefore, there is an urgent need to provide a real-time early warning system and method for coordinated deformation of tailings dam bodies based on multi-source fusion, so as to realize early and accurate prediction and graded early warning of dam instability risks, thereby facilitating the active prevention and control of dam deformation risks. Summary of the Invention

[0005] In response to the problems existing in the above-mentioned prior art, the present invention provides a real-time early warning system and method for coordinated deformation of tailings dam bodies based on multi-source fusion. The system has a simple structure and a high degree of intelligence. It can efficiently and accurately realize early and accurate prediction and graded warning of the risk of dam body instability, which is conducive to the active prevention and control of dam body deformation risks. The method can realize early and accurate prediction and real-time graded warning of the risk of dam body instability through coordinated analysis of all factors such as displacement-seepage-dry beach-infiltration line.

[0006] In order to achieve the above objectives, the present invention provides a real-time early warning system for collaborative deformation of tailings dam bodies based on multi-source fusion, including a data acquisition module, an edge computing module, a central processing module and an early warning execution module;

[0007] The data acquisition module is used to collect multi-source monitoring data and send it to the edge computing module. The data acquisition module includes an internal displacement monitoring unit, a dry beach monitoring unit, a seepage pressure monitoring unit, and an infiltration line spatial slope monitoring unit. The internal displacement monitoring unit is set at the section with the maximum dam height; the dry beach monitoring unit is set at the section with the shortest dry beach length; the seepage pressure monitoring unit and the infiltration line spatial slope monitoring unit are set at key sections.

[0008] The edge computing module includes a spatiotemporal collaborative correction model and a feature extraction model, which is used to perform spatial collaborative correction processing on multi-source monitoring data through the spatiotemporal collaborative correction model, and calculate target feature data through the feature extraction model. The target feature data includes the shear displacement rate of the dam body. v、 Laplace norm of displacement field space , Beach Chief L , Beach length change rate , pore water pressure gradient , spatial slope change rate , and then send the target feature data to the central processing module;

[0009] The central processing module is used to calculate the DISE value based on the target feature data, and to perform early warning classification of the sliding surface generation probability based on the range of the DISE value, and then control the action of the early warning execution module according to the early warning classification result;

[0010] The early warning execution module is used to execute corresponding warning actions according to the control of the central processing module.

[0011] As a preference, the internal displacement monitoring unit is a deep inclinometer array or a distributed strain optical fiber, the dry beach monitoring unit includes a radar water level meter or a video analysis module; the seepage pressure monitoring unit includes a vibrating string pore water pressure gauge group, and the infiltration line spatial slope monitoring unit includes an array seepage pressure gauge group.

[0012] As a preferred embodiment, the edge computing module is deployed at a monitoring station, and an anti-interference power supply and lightning protection device are installed in the monitoring station. The central processing module is set at the monitoring management center station, which supports the provision of DISE spatiotemporal heat maps and risk tracing reports.

[0013] As a preferred embodiment, the early warning execution module is integrated with an audible and visual alarm device.

[0014] In the present invention, by making the data acquisition module include an internal displacement monitoring unit, a dry beach monitoring unit, a seepage pressure monitoring unit, and an infiltration line spatial slope monitoring unit, the dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data, and infiltration line monitoring data related to the safety of the tailings dam body can be collected simultaneously. By setting an edge computing module with a built-in spatiotemporal collaborative correction model and a feature extraction model, it is convenient to use the spatiotemporal collaborative correction model to perform spatiotemporal synchronization and rapid feature extraction of multi-source data, and then only the target feature data directly related to the tailings dam body safety warning can be sent to the central processing module, effectively reducing the computing power consumption of the central processing module. The central processing module calculates the DISE value based on the target feature data, which can facilitate the rapid and accurate implementation of the early warning classification process of the sliding surface generation probability. Through the setting of the early warning execution module, it is beneficial to perform different graded early warning actions according to the control of the central processing module, and then effectively remind relevant personnel to take necessary emergency measures in a timely manner to maximize the safety of personnel and property.

[0015] The system has a simple structure and a high level of intelligence. It can collect multi-source deformation parameters in real time, use the edge computing module for spatiotemporal coordinated correction and target feature data extraction, and then use the central processing module to calculate the DISE value in real time, thereby achieving graded early warning. The system can efficiently and accurately achieve early and accurate prediction of dam instability risks and graded early warning, which is conducive to the active prevention and control of dam deformation risks.

[0016] The present invention also provides a real-time early warning method for coordinated deformation of a tailings dam body based on multi-source fusion, which adopts a real-time early warning system for coordinated deformation of a tailings dam body based on multi-source fusion, comprising the following steps:

[0017] Step 1: Arrangement of data acquisition modules: Set up 1 to 3 monitoring sections at the section with the maximum dam height, set up 1 to 3 vertical lines at each monitoring section, set up 3 to 15 measuring points on each vertical line, and set up an internal displacement monitoring unit at each measuring point; at the same time, vertically arrange seepage pressure monitoring units and infiltration line spatial slope monitoring units at key sections; and arrange 1 to 3 dry beach monitoring units at the shortest part of the dry beach;

[0018] Step 2: Use the data acquisition module to collect multi-source monitoring data in real time and send the obtained multi-source monitoring data to the edge computing module. The multi-source monitoring data includes dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data, and infiltration line monitoring data;

[0019] Step 3: The edge computing module calculates target feature data based on the received multi-source monitoring data;

[0020] S31: Using the spatiotemporal collaborative correction model to perform spatial collaborative correction processing on the multi-source monitoring data according to formula (1), and sending the processed multi-source monitoring data to the feature extraction model;

[0021] (1);

[0022] Where, is the corrected multi-source fusion data, which is a vector; , is the number of each monitoring unit; is the total number of monitoring units, which is a scalar; is the weight coefficient, ; is the forgetting factor, which is a scalar; is the time variable; is the integral time variable; is a vector function; is an exponential decay function; is the monitoring unit variance;

[0023] S32: Calculate target feature data based on the processed multi-source monitoring data using the feature extraction model. Specifically, calculate the dam body shear displacement rate based on the displacement monitoring data. v and the Laplace norm of the displacement field space , calculate the beach length based on the beach length monitoring data L and beach length change rate , the pore water pressure gradient is calculated based on the seepage pressure monitoring data , calculate the spatial slope change rate based on the infiltration line monitoring data , and then send the target feature data to the central processing module;

[0024] Step 4: The central processing module performs a graded warning based on the received target feature data;

[0025] S41: Model the dynamic instability entropy model according to formula (2) and calculate the DISE value;

[0026] (2);

[0027] Where, is the spatial second-order derivative norm of the displacement field; The material memory factor calibrated by the tailings sand triaxial test; 、 、 is the weight coefficient dynamically optimized by the Bayesian network; is the critical chaos threshold, which is a scalar; is the critical slope change rate;

[0028] S42: Risk level mapping model performs early warning classification of sliding surface generation probability based on DISE value;

[0029] like , it is a safe level and a reminder message is issued to maintain regular monitoring;

[0030] like , it is a blue warning level, the warning execution module will flash a blue light to warn, and issue a reminder to increase the inspection frequency to 3 times / day and check the drainage structure;

[0031] like , it is a yellow warning level, the control warning execution module flashes yellow light to warn, and issues a reminder message to manually clean the drainage well blockage and open the chute cover to drain;

[0032] like , it is the orange warning level, the warning execution module is controlled to flash orange lights to warn, and a reminder message is issued to forcibly open all flood discharge facilities and organize the evacuation of people within 1 km downstream;

[0033] like , it is a red warning level, the warning execution module is controlled to flash red lights for warning, and trigger the whole area sound and light alarm, and issue a surface evacuation order and report the reminder information to the government emergency department;

[0034] in, is the compression function of the rainstorm warning threshold, is the real-time rainfall intensity. When ≥30mm / h, ,when When the speed is less than 30 mm / h, .

[0035] Furthermore, in order to continuously optimize the model parameters through a closed-loop feedback mechanism to ensure early blocking and active prevention and control of the dam instability risk under extreme working conditions, in step five, after each warning, the parameters of the dynamic instability entropy model are reflected and corrected based on the actual treatment effect. At the same time, the monitoring accuracy of each monitoring unit is verified once a quarter, and the critical value database is updated once a year.

[0036] As a preferred embodiment, a dam break evolution model is integrated into the central processing module, and the dam break evolution model is used to predict the impact range based on the real-time DISE value.

[0037] As a preference, in step 2, the sampling period of the data acquisition module is ≤30 minutes.

[0038] The present invention provides a real-time early warning method for coordinated deformation of a tailings dam body based on multi-source fusion. First, during the data acquisition process, the displacement, dry beach, seepage pressure and seepage line slope monitoring units arranged inside the dam body are used to collect dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data and seepage line monitoring data in real time. Secondly, the edge computing module is used to perform spatiotemporal alignment of the multi-source monitoring data, and then calculate multidimensional deformation parameters such as the spatial Laplace norm of the displacement field, the dry beach length change rate, the pore water pressure gradient and the seepage line slope change rate. The target feature data acquisition process can be completed at the edge node, which is beneficial to saving the computing workload and computing power requirements of the central processing module. In particular, under the working condition where there are multiple monitoring stations, the computing workload and computing power requirements of the central processing module can be greatly saved. Then, the central processing module integrates the internal displacement space chaos, seepage time effect, and nonlinear sharp change three-dimensional deformation mechanism of the dry beach to construct a dynamic instability entropy (DISE) model, and dynamically optimizes the weight coefficient through the Bayesian network to obtain the DISE value. The warning is graded according to the range of the DISE value, and then the warning execution module is controlled according to the warning classification structure to execute the corresponding level of warning actions. Specifically, when the risk is low, the inspection frequency is increased and the drainage structures are inspected. When the risk is medium, the drainage well blockage is cleared and the discharge facilities are opened. When the risk is high, the flood discharge facilities are forced to open and the downstream personnel are organized to evacuate. When the risk is high, the full-area sound and light alarm is triggered, a comprehensive evacuation order is issued and reported to the emergency platform.

[0039] This method integrates real-time perception of multi-source deformation, dynamic instability entropy modeling, hierarchical linkage response and closed-loop optimization early warning mechanism. Through the coordinated analysis of all factors of displacement-seepage-dry beach-infiltration line, it can achieve early and accurate prediction of dam instability risk and real-time hierarchical warning, which is conducive to the active prevention and control of dam deformation risk.

[0040] Compared with the prior art, the present invention has the following advantages:

[0041] 1) Multi-source collaborative monitoring has significantly improved its accuracy. By integrating four-dimensional deformation data—dam internal displacement, dry beach status, seepage pressure, and the spatial slope of the seepage line—a dynamic instability entropy (DISE) model was constructed, overcoming the limitations of traditional single-seepage line monitoring. Using the spatiotemporal collaborative correction of the edge computing module and the dynamic optimization of weight coefficients using a Bayesian network, the accuracy of predicting the risk of sliding surface formation was significantly improved, reducing the false alarm rate.

[0042] 2) Real-time risk response capabilities within minutes. Edge computing modules enable rapid synchronization and feature extraction of multi-source data, shortening the sampling cycle to less than 30 minutes and enabling minute-by-minute updates of DISE values. This revolutionizes the traditional manual decision-making process, accelerating dam breach warning response speeds to minutes under heavy rainfall conditions.

[0043] 3) Intelligent hierarchical emergency linkage closed loop. A five-level risk mapping mechanism based on DISE values ​​automatically triggers precise emergency measures: Yellow alerts automatically clean drainage wells and open chutes for discharge; orange alerts forcibly open flood discharge facilities and organize evacuations within 1 km; red alerts link global audible and visual alarms, evacuation command platforms, and government emergency departments, effectively improving emergency response efficiency.

[0044] 4) Dynamic Optimization and Strong Adaptability. A closed-loop feedback correction mechanism for DISE model parameters was established, sensor accuracy was verified quarterly, and the critical value database was updated annually. The warning trigger threshold was adjusted by dynamically linking rainfall intensity with a rainstorm compression function, significantly improving adaptability to extreme conditions and significantly increasing the success rate of dam-break prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of the early warning method of the present invention;

[0046] Figure 2 It is the principle block diagram of the early warning system. DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings.

[0048] like Figure 1 and Figure 2 As shown, the present invention provides a real-time early warning system for collaborative deformation of tailings dam bodies based on multi-source fusion, including a data acquisition module, an edge computing module, a central processing module and an early warning execution module;

[0049] The data acquisition module is used to collect multi-source monitoring data and send it to the edge computing module. The data acquisition module includes an internal displacement monitoring unit, a dry beach monitoring unit, a seepage pressure monitoring unit, and a seepage line spatial slope monitoring unit. The internal displacement monitoring unit is set at the section with the maximum dam height; the dry beach monitoring unit is set at the section with the shortest dry beach length; the seepage pressure monitoring unit and the seepage line spatial slope monitoring unit are set at key sections (generally no less than three representative dam body cross sections that can control the main seepage conditions and cross sections where abnormal seepage is expected to occur).

[0050] The edge computing module includes a spatiotemporal collaborative correction model and a feature extraction model, which is used to perform spatial collaborative correction processing on multi-source monitoring data through the spatiotemporal collaborative correction model, and calculate target feature data through the feature extraction model. The target feature data includes the shear displacement rate of the dam body. v、 Laplace norm of displacement field space , Beach Chief L , Beach length change rate , pore water pressure gradient , spatial slope change rate , and then send the target feature data to the central processing module;

[0051] The central processing module is used to calculate the DISE value based on the target feature data, and to perform early warning classification of the sliding surface generation probability based on the range of the DISE value, and then control the action of the early warning execution module according to the early warning classification result; preferably, the central processing module has a built-in dynamic instability entropy model (DISE) and a risk level mapping model, wherein the dynamic instability entropy model is used to calculate the DISE value, and the risk level mapping model is used to perform early warning classification of the sliding surface generation probability based on the calculated DISE value, and based on the early warning classification;

[0052] The early warning execution module is used to execute corresponding warning actions according to the control of the central processing module.

[0053] As a preference, the internal displacement monitoring unit is a deep inclinometer array or a distributed strain optical fiber, the dry beach monitoring unit includes a radar water level meter or a video analysis module. When a radar water level meter is used, 1 to 3 radar monitoring sections can be set at the shortest length of the dry beach; the seepage pressure monitoring unit includes a vibrating string pore water pressure gauge group, and the infiltration line spatial slope monitoring unit includes an array seepage pressure gauge group. As a preference, the array seepage pressure gauge group is arranged vertically, with a distance between two adjacent measuring points of 20 to 40 m, and non-woven geotextiles are wrapped in the permeable section.

[0054] As a preferred embodiment, the edge computing module is deployed at a monitoring station, and an anti-interference power supply and lightning protection device are installed in the monitoring station. The central processing module is set at the monitoring management center station, which supports the provision of DISE spatiotemporal heat maps and risk tracing reports.

[0055] As a preference, the early warning execution module is integrated with an audible and visual alarm device. Further preferably, the early warning execution module can be connected to the emergency drainage system and the evacuation command platform.

[0056] In the present invention, by making the data acquisition module include an internal displacement monitoring unit, a dry beach monitoring unit, a seepage pressure monitoring unit, and an infiltration line spatial slope monitoring unit, the dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data, and infiltration line monitoring data related to the safety of the tailings dam body can be collected simultaneously. By setting an edge computing module with a built-in spatiotemporal collaborative correction model and a feature extraction model, it is convenient to use the spatiotemporal collaborative correction model to perform spatiotemporal synchronization and rapid feature extraction of multi-source data, and then only the target feature data directly related to the tailings dam body safety warning can be sent to the central processing module, effectively reducing the computing power consumption of the central processing module. The central processing module calculates the DISE value based on the target feature data, which can facilitate the rapid and accurate implementation of the early warning classification process of the sliding surface generation probability. Through the setting of the early warning execution module, it is beneficial to perform different graded early warning actions according to the control of the central processing module, and then effectively remind relevant personnel to take necessary emergency measures in a timely manner to maximize the safety of personnel and property.

[0057] The system has a simple structure and a high level of intelligence. It can collect multi-source deformation parameters in real time, use the edge computing module for spatiotemporal coordinated correction and target feature data extraction, and then use the central processing module to calculate the DISE value in real time, thereby achieving graded early warning. The system can efficiently and accurately achieve early and accurate prediction of dam instability risks and graded early warning, which is conducive to the active prevention and control of dam deformation risks.

[0058] The present invention also provides a real-time early warning method for coordinated deformation of a tailings dam body based on multi-source fusion, which adopts a real-time early warning system for coordinated deformation of a tailings dam body based on multi-source fusion, comprising the following steps:

[0059] Step 1: Arrangement of data acquisition modules: Set up 1 to 3 monitoring sections at the section with the maximum dam height, with 1 to 3 vertical lines on each monitoring section, 3 to 15 measuring points on each vertical line, and an internal displacement monitoring unit at each measuring point. At the same time, vertically arrange seepage pressure monitoring units and infiltration line spatial slope monitoring units at key sections (representative dam cross sections that can control the main seepage conditions, as well as cross sections where abnormal seepage is expected, generally no less than 3), and arrange 1 to 3 dry beach monitoring units at the shortest part of the dry beach.

[0060] Step 2: Use the data acquisition module to collect multi-source monitoring data in real time and send the obtained multi-source monitoring data to the edge computing module. The multi-source monitoring data includes dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data, and infiltration line monitoring data;

[0061] Step 3: The edge computing module calculates target feature data based on the received multi-source monitoring data;

[0062] S31: Using the spatiotemporal collaborative correction model to perform spatial collaborative correction processing on the multi-source monitoring data according to formula (1), and sending the processed multi-source monitoring data to the feature extraction model;

[0063] (1);

[0064] Where, is the corrected multi-source fusion data (spatiotemporal alignment values ​​of displacement, osmotic pressure and other parameters), which is a vector; is the subscript, , is the number of each monitoring unit (such as k =1 is the displacement monitoring unit, k =2 is the osmotic pressure monitoring unit, etc.); is the total number of monitoring units (the number of monitoring units deployed in the system), which is a scalar; is the weight coefficient, , which is the monitoring unit k The fusion weight of is the forgetting factor (controls the decay rate of historical data, the larger the value, the higher the weight of recent data), which is a scalar; is a time variable, defined as the current moment (the reference time point for data fusion); is the integral time variable, defined as the historical moment (in the time window ) Inner sliding; is a vector function, defined as the monitoring unit k At the moment The original monitoring data (such as displacement acceleration, osmotic pressure gradient, etc.); is an exponential decay function, defined as a time weight factor (giving new data a higher weight and old data exponentially decays); To monitor the unit variance, it is defined as sensor k The measurement data variance (characterizing the volatility / reliability of the sensor's historical data);

[0065] S32: Calculate target feature data based on the processed multi-source monitoring data using the feature extraction model. Specifically, calculate the dam body shear displacement rate based on the displacement monitoring data. v and the Laplace norm of the displacement field space , calculate the beach length based on the beach length monitoring data L and beach length change rate , the pore water pressure gradient is calculated based on the seepage pressure monitoring data , calculate the spatial slope change rate based on the infiltration line monitoring data , and then send the target feature data to the central processing module;

[0066] Step 4: The central processing module performs a graded warning based on the received target feature data;

[0067] S41: Model the dynamic instability entropy model according to formula (2) and calculate the DISE value;

[0068] (2);

[0069] Where, is the spatial second-order derivative norm of the displacement field (quantifying the chaotic degree of shear deformation); The material memory factor calibrated by the tailings sand triaxial test; 、 、 is the weight coefficient dynamically optimized by the Bayesian network; is the critical chaos threshold (marking the critical deformation gradient for shear band formation), which is a scalar; is the critical slope change rate; represents the instability threshold of the infiltration line lifting rate; 、 Calibration based on design monitoring indicators and historical operation data;

[0070] S42: Risk level mapping model performs early warning classification of sliding surface generation probability based on DISE value;

[0071] like , it is a safe level and a reminder message is issued to maintain regular monitoring;

[0072] like , it is a blue warning level, the warning execution module will flash a blue light to warn, and issue a reminder to increase the inspection frequency to 3 times / day and check the drainage structure;

[0073] like , it is a yellow warning level, the control warning execution module flashes yellow light to warn, and issues a reminder message to manually clean the drainage well blockage and open the chute cover to drain;

[0074] like , it is the orange warning level, the warning execution module is controlled to flash orange lights to warn, and a reminder message is issued to forcibly open all flood discharge facilities and organize the evacuation of people within 1 km downstream;

[0075] like , it is a red warning level, the warning execution module is controlled to flash red lights for warning, and trigger the whole area sound and light alarm, and issue a surface evacuation order and report the reminder information to the government emergency department;

[0076] in, is the compression function of the rainstorm warning threshold, is the real-time rainfall intensity. When ≥30mm / h, ,when When the speed is less than 30 mm / h, .

[0077] In order to continuously optimize the model parameters through a closed-loop feedback mechanism to ensure early blocking and active prevention and control of the risk of dam instability under extreme working conditions (such as during heavy rain), in step five, after each warning, the parameters of the dynamic instability entropy model are reflected and corrected based on the actual treatment effect. At the same time, the monitoring accuracy of each monitoring unit is verified once a quarter, and the critical value database is updated once a year.

[0078] As a preferred embodiment, the central processing module is internally integrated with a dam break evolution model, which is used to predict the impact range based on the real-time DISE value. When it is a red warning level, an evacuation route map is automatically generated based on the historical route database and pushed to the emergency broadcast system.

[0079] As a preference, in step 2, the sampling period of the data acquisition module is ≤30 minutes.

[0080] The present invention provides a real-time early warning method for coordinated deformation of a tailings dam body based on multi-source fusion. First, during the data acquisition process, the displacement, dry beach, seepage pressure and seepage line slope monitoring units arranged inside the dam body are used to collect dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data and seepage line monitoring data in real time. Secondly, the edge computing module is used to perform spatiotemporal alignment of the multi-source monitoring data, and then calculate multidimensional deformation parameters such as the spatial Laplace norm of the displacement field, the dry beach length change rate, the pore water pressure gradient and the seepage line slope change rate. The target feature data acquisition process can be completed at the edge node, which is beneficial to saving the computing workload and computing power requirements of the central processing module. In particular, under the working condition where there are multiple monitoring stations, the computing workload and computing power requirements of the central processing module can be greatly saved. Then, the central processing module integrates the internal displacement space chaos, seepage time effect, and nonlinear sharp change three-dimensional deformation mechanism of the dry beach to construct a dynamic instability entropy (DISE) model, and dynamically optimizes the weight coefficient through the Bayesian network to obtain the DISE value. The warning is graded according to the range of the DISE value, and then the warning execution module is controlled according to the warning classification structure to execute the corresponding level of warning actions. Specifically, when the risk is low, the inspection frequency is increased and the drainage structures are inspected. When the risk is medium, the drainage well blockage is cleared and the discharge facilities are opened. When the risk is high, the flood discharge facilities are forced to open and the downstream personnel are organized to evacuate. When the risk is high, the full-area sound and light alarm is triggered, a comprehensive evacuation order is issued and reported to the emergency platform.

[0081] This method integrates real-time perception of multi-source deformation, dynamic instability entropy modeling, hierarchical linkage response and closed-loop optimization early warning mechanism. Through the coordinated analysis of all factors of displacement-seepage-dry beach-infiltration line, it can achieve early and accurate prediction of dam instability risk and real-time hierarchical warning, which is conducive to the active prevention and control of dam deformation risk.

[0082] Compared with the prior art, the present invention has the following advantages:

[0083] 1) Multi-source collaborative monitoring has significantly improved its accuracy. By integrating four-dimensional deformation data—dam internal displacement, dry beach status, seepage pressure, and the spatial slope of the seepage line—a dynamic instability entropy (DISE) model was constructed, overcoming the limitations of traditional single-seepage line monitoring. Using the spatiotemporal collaborative correction of the edge computing module and the dynamic optimization of weight coefficients using a Bayesian network, the accuracy of predicting the risk of sliding surface formation was significantly improved, reducing the false alarm rate.

[0084] 2) Real-time risk response capabilities within minutes. Edge computing modules enable rapid synchronization and feature extraction of multi-source data, shortening the sampling cycle to less than 30 minutes and enabling minute-by-minute updates of DISE values. This revolutionizes the traditional manual decision-making process, accelerating dam breach warning response speeds to minutes under heavy rainfall conditions.

[0085] 3) Intelligent hierarchical emergency linkage closed loop. A five-level risk mapping mechanism based on DISE values ​​automatically triggers precise emergency measures: Yellow alerts automatically clean drainage wells and open chutes for discharge; orange alerts forcibly open flood discharge facilities and organize evacuations within 1 km; red alerts link global audible and visual alarms, evacuation command platforms, and government emergency departments, effectively improving emergency response efficiency.

[0086] 4) Dynamic Optimization and Strong Adaptability. A closed-loop feedback correction mechanism for DISE model parameters was established, sensor accuracy was verified quarterly, and the critical value database was updated annually. The warning trigger threshold was adjusted by dynamically linking rainfall intensity with a rainstorm compression function, significantly improving adaptability to extreme conditions and significantly increasing the success rate of dam-break prevention and control.

Claims

1. A real-time early warning system for coordinated deformation of tailings dam based on multi-source fusion, characterized by: It includes data acquisition module, edge computing module, central processing module and early warning execution module; The data acquisition module is used to collect multi-source monitoring data and send it to the edge computing module. The data acquisition module includes an internal displacement monitoring unit, a dry beach monitoring unit, a seepage pressure monitoring unit, and an infiltration line spatial slope monitoring unit. The internal displacement monitoring unit is set at the section with the maximum dam height; the dry beach monitoring unit is set at the section with the shortest dry beach length; the seepage pressure monitoring unit and the infiltration line spatial slope monitoring unit are set at key sections. The edge computing module includes a spatiotemporal collaborative correction model and a feature extraction model, which is used to perform spatial collaborative correction processing on multi-source monitoring data through the spatiotemporal collaborative correction model, and calculate target feature data through the feature extraction model. The target feature data includes the shear displacement rate of the dam body. v、 Laplace norm of displacement field space , Beach Chief L , Beach length change rate , pore water pressure gradient , spatial slope change rate , and then send the target feature data to the central processing module; The spatiotemporal collaborative correction model is used to perform spatial collaborative correction processing on the multi-source monitoring data according to formula (1), and the processed multi-source monitoring data is sent to the feature extraction model; （1）； Where, is the corrected multi-source fusion data, which is a vector; , is the number of each monitoring unit; is the total number of monitoring units, which is a scalar; is the weight coefficient, , is the forgetting factor, which is a scalar; is the time variable; is the integral time variable; is a vector function; is an exponential decay function; is the monitoring unit variance; The feature extraction model is used to calculate the target feature data based on the processed multi-source monitoring data. Specifically, the shear displacement rate of the dam body is calculated based on the displacement monitoring data. v and the spatial Laplace norm of the displacement field , calculate the beach length based on the beach length monitoring data L and beach length change rate , the pore water pressure gradient is calculated based on the seepage pressure monitoring data , calculate the spatial slope change rate based on the infiltration line monitoring data ; The central processing module is used to calculate the DISE value based on the target feature data, and to perform early warning classification of the sliding surface generation probability based on the range of the DISE value, and then control the action of the early warning execution module according to the early warning classification result; According to formula (2), the dynamic instability entropy model is built and the DISE value is calculated; （2）； Where, is the spatial second-order derivative norm of the displacement field; The material memory factor calibrated by the tailings sand triaxial test; 、 、 is the weight coefficient dynamically optimized by the Bayesian network; is the critical chaos threshold, which is a scalar; is the critical slope change rate; The early warning execution module is used to execute corresponding warning actions according to the control of the central processing module.

2. The tailings dam body collaborative deformation real-time early warning system based on multi-source fusion according to claim 1 is characterized in that: The internal displacement monitoring unit is a deep inclinometer array or distributed strain optical fiber, the dry beach monitoring unit includes a radar water level meter or a video analysis module; the seepage pressure monitoring unit includes a vibrating string pore water pressure gauge group, and the infiltration line spatial slope monitoring unit includes an array seepage pressure gauge group.

3. The tailings dam body collaborative deformation real-time early warning system based on multi-source fusion according to claim 1 is characterized in that: The edge computing module is deployed at the monitoring station, and the monitoring station is equipped with an anti-interference power supply and lightning protection device. The central processing module is set at the monitoring management center station, which supports the provision of DISE spatiotemporal heat maps and risk tracing reports.

4. The tailings dam body collaborative deformation real-time early warning system based on multi-source fusion according to claim 1 is characterized in that: The early warning execution module is integrated with an audible and visual alarm device.

5. A tailings dam body collaborative deformation real-time early warning method based on multi-source fusion, using a tailings dam body collaborative deformation real-time early warning system based on multi-source fusion according to any one of claims 1 to 4, characterized in that: The following steps are involved: Step 1: Arrangement of data acquisition module; Set up 1 to 3 monitoring sections at the section with the maximum dam height, set up 1 to 3 vertical lines at each monitoring section, set up 3 to 15 measuring points on each vertical line, and set up an internal displacement monitoring unit at each measuring point; at the same time, vertically set up seepage pressure monitoring units and infiltration line spatial slope monitoring units at key sections; and set up 1 to 3 dry beach monitoring units at the shortest part of the dry beach; Step 2: Use the data acquisition module to collect multi-source monitoring data in real time and send the obtained multi-source monitoring data to the edge computing module. The multi-source monitoring data includes dam displacement monitoring data, beach length monitoring data, seepage pressure monitoring data, and infiltration line monitoring data; Step 3: The edge computing module calculates target feature data based on the received multi-source monitoring data; S31: Using the spatiotemporal collaborative correction model to perform spatial collaborative correction processing on the multi-source monitoring data according to formula (1), and sending the processed multi-source monitoring data to the feature extraction model; （1）； Where, is the corrected multi-source fusion data, which is a vector; , is the number of each monitoring unit; is the total number of monitoring units, which is a scalar; is the weight coefficient, , is the forgetting factor, which is a scalar; is the time variable; is the integral time variable; is a vector function; is an exponential decay function; is the monitoring unit variance; S32: Calculate target feature data based on the processed multi-source monitoring data using the feature extraction model. Specifically, calculate the dam body shear displacement rate based on the displacement monitoring data. v and the spatial Laplace norm of the displacement field , calculate the beach length based on the beach length monitoring data L and beach length change rate , the pore water pressure gradient is calculated based on the seepage pressure monitoring data , calculate the spatial slope change rate based on the infiltration line monitoring data , and then send the target feature data to the central processing module; Step 4: The central processing module performs a graded warning based on the received target feature data; S41: Model the dynamic instability entropy model according to formula (2) and calculate the DISE value; （2）； Where, is the spatial second-order derivative norm of the displacement field; The material memory factor calibrated by the tailings sand triaxial test; 、 、 is the weight coefficient dynamically optimized by the Bayesian network; is the critical chaos threshold, which is a scalar; is the critical slope change rate; S42: Risk level mapping model performs early warning classification of sliding surface generation probability based on DISE value; like , it is a safe level and a reminder message is issued to maintain regular monitoring; like , it is a blue warning level, the warning execution module will flash a blue light to warn, and issue a reminder to increase the inspection frequency to 3 times / day and check the drainage structure; like , it is a yellow warning level, the control warning execution module flashes yellow light to warn, and issues a reminder message to manually clean the drainage well blockage and open the chute cover to drain; like , it is the orange warning level, the warning execution module is controlled to flash orange lights to warn, and a reminder message is issued to forcibly open all flood discharge facilities and organize the evacuation of people within 1 km downstream; like , it is a red warning level, the warning execution module is controlled to flash red lights for warning, and trigger the whole area sound and light alarm, and issue a surface evacuation order and report the reminder information to the government emergency department; in, is the compression function of the rainstorm warning threshold, is the real-time rainfall intensity. When ≥30mm / h, ,when When the speed is less than 30 mm / h, .

6. The method for real-time early warning of tailings dam body coordinated deformation based on multi-source fusion according to claim 5 is characterized in that: In step four, after each warning, the parameters of the dynamic instability entropy model are revised based on the actual treatment effect. At the same time, the monitoring accuracy of each monitoring unit is verified once a quarter, and the critical value database is updated once a year.

7. The method for real-time early warning of tailings dam body coordinated deformation based on multi-source fusion according to claim 5 is characterized in that: The central processing module is internally integrated with a dam break evolution model, which is used to predict the impact range based on the real-time DISE value.

8. The method for real-time early warning of tailings dam body coordinated deformation based on multi-source fusion according to claim 5 is characterized in that: In step 2, the sampling period of the data acquisition module is ≤ 30 minutes.

Images