Optimal configuration method and system of tailings safety monitoring system based on causal role classification and adversarial explanatory structure model

Abstract

The application discloses a tailing pond safety monitoring system optimization configuration method and system based on a causal role classification and an adversarial explanation structure model, and comprises the following steps: obtaining historical monitoring data of a tailing pond safety monitoring system; taking monitoring indexes as influencing factors, dividing into five causal roles in a disaster-causing causal chain and establishing a factor set; establishing an adjacency matrix and calculating a reachable matrix; adopting two extraction rules of result priority and cause priority to obtain UP type and DOWN type multi-level hierarchical structures respectively; comparing the two structures to identify pivotal factors; generating a monitoring index priority configuration table according to the pivotal factors, marking corresponding monitoring indexes as the highest priority monitoring objects and outputting early warning configuration parameters; and adjusting the collection frequency and / or the early warning threshold according to the early warning configuration parameters by a data collection and access module. The application realizes systematic screening of factors through causal role classification, eliminates hierarchical disputes through a bidirectional adversarial extraction mechanism, realizes a closed loop from data analysis to system optimization by directly mapping analysis results into system configuration parameters, and can be widely applied to the field of tailing pond safety monitoring.

Description

Technical Field

[0001] This invention belongs to the field of tailings dam safety monitoring and disaster early warning technology, specifically involving an optimized configuration method and system for a tailings dam safety monitoring system based on a causal role classification and adversarial explanation structure model. Background Technology

[0002] Tailings dams are dams built around rivers, valley mouths, or depressions to store tailings or solid waste generated after mineral processing in metal or non-metal mines for sedimentation and accumulation. While serving as a disposal site for mining waste, they also possess the dual attributes of being a high-risk source and a source of environmental pollution. Statistical analysis shows that flooding and seepage damage are the two main causes of tailings dam failures, both closely related to the rise in the seepage line caused by heavy rainfall in the reservoir area. Of 161 tailings dam failures domestically and internationally, flooding accounted for 32.3%, and seepage damage accounted for 28.6%, totaling 60.9%. In my country, flooding accounted for 36.9% of tailings dam failures, making it the leading cause. Therefore, the scientific configuration and optimization of tailings dam safety monitoring systems, enabling them to accurately identify and prioritize the monitoring of key risk indicators, is of great significance for ensuring the safe operation of tailings dams.

[0003] Currently, many mining companies have deployed automated safety monitoring systems in tailings ponds, collecting dozens of monitoring indicators in real time, such as pond water level, seepage line, rainfall, and dry beach slope. However, the computing power, communication bandwidth, and storage resources of these safety monitoring systems are limited. On the one hand, tailings pond safety monitoring involves multiple monitoring items and points, requiring the system to process massive amounts of data from different sensors simultaneously. On the other hand, resource allocation within the monitoring system faces bottlenecks—the energy consumption of wireless sensor networks and the length of data transmission paths restrict the overall system efficiency. Under these resource constraints, identifying the key monitoring indicators from the numerous monitoring metrics of the safety monitoring system to optimize resource allocation is a pressing technical problem that needs to be solved in this field.

[0004] To address the aforementioned issues, some scholars have employed interpretive structural models (ESMs) to conduct hierarchical analysis of the risk influencing factors of tailings dams. For example, some studies have combined bibliometrics and expert decision-making to screen key influencing indicators of tailings dam flood risk, and then used ESMs to classify these indicators into hierarchical levels. However, these existing technologies have the following problems in engineering applications.

[0005] First, existing explanatory structural models only employ a result-oriented, one-way hierarchical extraction method—that is, tracing the cause upwards layer by layer from the final result. This one-way extraction method leads to the hierarchical classification results being highly dependent on the initial causal relationship judgment. For the same tailings dam, different technicians may give completely different hierarchical classification results of influencing factors. When faced with contradictory analytical conclusions, managers cannot determine which indicators should be focused on monitoring, and are forced to monitor all indicators comprehensively, resulting in a serious waste of monitoring resources and funds.

[0006] Second, existing methods lack a systematic classification framework when screening influencing factors. Methods relying on bibliometrics or direct screening based on expert experience mix different types of factors, such as driving factors, impedance factors, and state factors, together for hierarchical classification. This approach is logically inconsistent and easily leads to distorted hierarchical analysis results.

[0007] Third, existing methods lack a systematic framework for comparing and verifying the results of fault tree analysis with those of interpretive structural models. When the two methods yield inconsistent conclusions, managers cannot determine which result to accept, making it difficult to directly use the analysis findings to guide engineering decisions.

[0008] To address the aforementioned problems, this invention provides an optimized configuration method for a tailings dam safety monitoring system based on causal role classification and an adversarial interpretation structure model. This method achieves systematic screening of factors through causal role classification, eliminates hierarchical disputes through the UP and DOWN bidirectional extraction mechanism of the adversarial interpretation structure model, performs independent verification through fault tree analysis, and directly maps the analysis results to the configuration parameters of the monitoring system, providing managers with a definitive decision-making basis. Summary of the Invention

[0009] This invention aims to solve the technical problems in existing tailings dam safety monitoring system resource allocation methods, such as unidirectional hierarchical extraction leading to inconsistent analysis results, lack of a systematic classification framework for factor screening, lack of multi-source independent verification mechanisms, and inability to directly convert analysis conclusions into system configuration parameters. It provides a tailings dam safety monitoring system optimization configuration method and system based on causal role classification and adversarial interpretation structure model.

[0010] To solve the above-mentioned technical problems, the technical solution proposed in this application is as follows:

[0011] This invention provides an optimized configuration method for a tailings dam safety monitoring system based on a causal role classification and adversarial explanation structure model, characterized by the following steps: S1: Obtain historical monitoring data from the tailings dam safety monitoring system. The historical monitoring data includes time series data of multiple monitoring indicators, including at least the reservoir water level, the depth of the seepage line, rainfall, the slope of the dry beach, and the operating status data of the flood discharge facilities. S2: Using the multiple monitoring indicators as influencing factors, the multiple monitoring indicators are divided into five causal roles according to their roles in the disaster causal chain. Based on the five causal roles, specific influencing factors corresponding to each causal role are screened to establish a factor set. S3: Establish an adjacency matrix based on the causal relationship between various influencing factors, and perform an exponentiation operation on the adjacency matrix to obtain the reachability matrix; S4: The reachability matrix is ​​hierarchically divided using a result-first extraction rule to obtain a UP-type multi-level hierarchical structure; the reachability matrix is ​​hierarchically divided using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure. S5: Compare the UP-type multi-level hierarchical structure with the DOWN-type multi-level hierarchical structure. If the same influencing factor has a different hierarchical position in the two structures, then the factor is identified as the pivot factor in the causal network. S6: Based on the key factors, generate a monitoring indicator priority configuration table in the tailings dam safety monitoring system, mark the monitoring indicators corresponding to the key factors as the highest priority monitoring objects in the safety monitoring system, and output the corresponding early warning configuration parameters. The early warning configuration parameters include the collection frequency adjustment value and early warning threshold adjustment value of the highest priority monitoring objects.

[0012] Furthermore, the five causal roles include driving factors, impedance factors, inherent factors, state factors, and control factors; the driving factors include peak flow and total flood volume; the impedance factors include flood discharge capacity, flood control reservoir capacity, and seepage discharge capacity; the inherent factors include dam height, outer slope ratio, dry beach slope, and tailings particle size; the state factors include reservoir water level and phreatic line depth; and the control factors include monitoring and early warning reliability and emergency response timeliness.

[0013] Furthermore, when selecting specific influencing factors based on the five types of causal roles, the indicators within the same criterion layer belong to the same causal role based on the principle of causal homogeneity, and the indicators are reduced based on the principle of information non-redundancy: if the safety status of the first indicator can be fully inferred from the monitoring value of the second indicator, then the second indicator is retained and the first indicator is discarded.

[0014] Furthermore, the result-priority extraction rule in step S4 is as follows: if the reachable set R(Si) of the influencing factor Si is a subset of the antecedent set A(Si), then Si is extracted as the current highest level; the cause-priority extraction rule is as follows: if the antecedent set A(Si) of the influencing factor Si is a subset of the reachable set R(Si), then Si is extracted as the current highest level.

[0015] Furthermore, it also includes step S7: constructing a tailings dam flood disaster accident tree, calculating the minimum cut set and minimum path set of the accident tree, and calculating the structural importance of each basic event; comparing and verifying the ranking of the structural importance with the pivotal factors determined in step S5; if a factor is identified as a pivotal factor and has the highest structural importance, then the factor is marked as the highest priority monitoring object in the safety monitoring system.

[0016] Furthermore, the ranking rule for structural importance is as follows: the more times a basic event appears in the minimum cut set, the greater its structural importance; for basic events that appear the same number of times, the lower the order of the minimum cut set, the greater its structural importance.

[0017] On the other hand, this application also claims protection for a tailings dam safety monitoring system employing any of the methods described above, comprising: The data acquisition and access module is used to collect time series data of multiple monitoring indicators of the tailings dam. The monitoring indicators include at least the reservoir water level, the depth of the seepage line, the rainfall, the slope of the dry beach, and the operating status data of the flood discharge facilities. The factor set construction module is used to establish a factor set with the multiple monitoring indicators as influencing factors. The matrix operation module is used to establish an adjacency matrix based on the causal relationship between various influencing factors, and to perform exponentiation on the adjacency matrix to obtain a reachability matrix; The adversarial explanation structure model construction module is used to hierarchically divide the reachable matrix using a result-first extraction rule to obtain a UP-type multi-level hierarchical structure, and to hierarchically divide the reachable matrix using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure. The pivot factor identification module is used to compare the UP-type multi-level hierarchical structure with the DOWN-type multi-level hierarchical structure. If the same influencing factor has a different hierarchical position in the two structures, then the factor is identified as a pivot factor in the causal network. The configuration optimization module is used to generate a monitoring indicator priority configuration table based on the hub factors, mark the monitoring indicators corresponding to the hub factors as the highest priority monitoring objects, and output the early warning configuration parameters, including the collection frequency adjustment value and the early warning threshold adjustment value, to the data acquisition and access module. The data acquisition and access module increases the acquisition frequency of the highest priority monitoring object and / or lowers its early warning trigger threshold according to the early warning configuration parameters.

[0018] Furthermore, the data acquisition and access module is configured with a file parsing interface for receiving monitoring data files exported from external monitoring systems. By parsing the encoding information in the filename of the monitoring data file, it automatically identifies the tailings dam and monitoring point to which it belongs, and loads the corresponding parameter configuration from the configuration file directory accordingly. The filename format is "tailings dam code_point code.extension". The data acquisition and access module extracts the tailings dam code and monitoring point code from the filename using regular expressions.

[0019] Furthermore, it also includes a data preprocessing module for anomaly detection and missing value imputation preprocessing of the monitoring data; the anomaly detection adopts a dual-threshold physical constraint mechanism: anomaly decrease thresholds for reservoir water level change and seepage line change are set respectively. When the reservoir water level decrease exceeds the reservoir water level abnormal decrease threshold or the seepage line decrease exceeds the seepage line abnormal decrease threshold at adjacent collection times, the corresponding data point is marked as anomaly and removed; the missing value imputation adopts a cubic spline interpolation method.

[0020] Furthermore, the configuration file directory adopts a hierarchical structure, with each tailings pond corresponding to an independent tailings pond parameter configuration file, and each monitoring point corresponding to an independent monitoring point parameter configuration file; the early warning configuration parameters are stored in the corresponding monitoring point parameter configuration file, and the data acquisition and access module automatically loads the configuration file when the platform starts.

[0021] Compared with the prior art, the present invention achieves the following beneficial technical effects: This invention achieves systematic screening of monitoring indicators through causal role classification, making the analysis logic clear and discernible; it eliminates the difference in conclusions caused by a single rule through the bidirectional extraction mechanism of the adversarial interpretation structure model, making the analysis results consistent and reproducible; by directly mapping the analysis results to the configuration parameters of the collection frequency and early warning threshold, the security monitoring system can automatically concentrate computing and communication resources on key monitoring indicators under limited resources, avoiding the omission or delay of key data due to resource dispersion, and realizing a complete closed loop from data analysis to system optimization. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the tailings dam safety monitoring system optimization configuration method based on the causal role classification and adversarial explanation structure model of the present invention.

[0024] Figure 2 This is a schematic diagram of the UP-type multi-level hierarchical directed topology graph of the present invention.

[0025] Figure 3 This is a schematic diagram of the DOWN-type multi-level hierarchical directed topology graph of the present invention.

[0026] Figure 4 This is a schematic diagram of the tailings dam flood disaster accident tree according to the present invention.

[0027] Figure 5 This is a schematic diagram of the module structure of the tailings dam safety monitoring system of the present invention.

[0028] Figure 6 This is a schematic diagram of the operation page of an embodiment of the tailings dam safety monitoring system of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1 As shown in one embodiment of this application, an optimized configuration method for a tailings dam safety monitoring system based on causal role classification and an adversarial interpretation structure model is provided. The method includes the following main steps: acquiring historical monitoring data of the tailings dam safety monitoring system, the historical monitoring data including time-series data of multiple monitoring indicators; classifying the multiple monitoring indicators into five causal roles according to their roles in the disaster-causing causal chain, and establishing a factor set; establishing an adjacency matrix and calculating a reachability matrix; performing UP-type and DOWN-type bidirectional hierarchical partitioning using an adversarial interpretation structure model; comparing the two structures to identify pivotal factors; generating a monitoring indicator priority configuration table based on the pivotal factors, marking the corresponding monitoring indicators as the highest priority monitoring objects, and outputting early warning configuration parameters. This method, by simultaneously employing both result-first and cause-first extraction rules, ensures that the analysis conclusions no longer depend on the selection of a single rule, eliminating the differences in conclusions caused by different rules.

[0031] In one embodiment of this application, step S1 first acquires historical monitoring data from the tailings dam safety monitoring system. This historical monitoring data includes time-series data for multiple monitoring indicators. These indicators include at least the reservoir water level, phreatic line depth, rainfall, dry beach slope, and the operational status of the drainage facilities. This data originates from a sensor network deployed at the tailings dam site—reservoir water level sensors are deployed near drainage wells to collect real-time reservoir water level elevation data; phreatic line sensors are buried inside the dam body to monitor changes in phreatic line depth; rain gauges are deployed around the reservoir area to record rainfall; and dry beach monitoring devices are used to measure the dry beach slope. The acquisition cycle of historical monitoring data can be determined according to the tailings dam's monitoring specifications and actual operational needs. These monitoring indicators cover a complete range from external flood loads to the inherent properties of the dam body, and from real-time safety status to the operational status of the drainage facilities, providing a comprehensive data foundation for subsequent causal relationship analysis and identification of key factors.

[0032] In one embodiment of this application, step S2 uses the multiple monitoring indicators as influencing factors, classifying them into five causal roles according to their roles in the disaster-causing causal chain. Based on these five causal roles, specific influencing factors corresponding to each causal role are selected to establish a factor set. The five causal roles include: driving factors, which are flood loads applied externally to the tailings dam system and are at the starting point of the disaster-causing causal chain; resistance factors, which are the ability of the engineering system to resist or dissipate flood loads and are in the middle of the disaster-causing causal chain, used to counteract the effects of driving factors; inherent factors, which are dam body properties that remain basically unchanged during operation after the design is determined and are in the middle of the disaster-causing causal chain, used to regulate the transmission of driving factors; state factors, which are dynamic variables that reflect the current safety level of the system in real time and are at the end of the disaster-causing causal chain, directly representing the current risk state; and control factors, which are the ability of human management behavior to intervene in the risk evolution and are outside the disaster-causing causal chain, indirectly affecting the risk evolution process. By classifying the causal roles as described above, the originally mixed influencing factors are systematically categorized according to their position in the disaster causal chain, avoiding the distortion of hierarchical analysis caused by mixing different types of factors such as driving factors and state factors in the screening.

[0033] In one embodiment of this application, the driving factors include peak flow and total flood volume. Peak flow determines the rate of reservoir water level rise, and total flood volume determines the cumulative rise in reservoir water level. The impedance factors include flood discharge capacity, flood regulation capacity, and seepage capacity. Flood discharge capacity determines the rate of reservoir water level fall, flood regulation capacity buffers the response of reservoir water level to total flood volume, and seepage capacity controls the depth of the phreatic line. The inherent factors include dam height, outer slope ratio, dry beach slope, and tailings particle size. Dam height determines the geometric boundary of the seepage field, outer slope ratio affects the morphology of the phreatic line within the dam body, dry beach slope determines the starting position of the phreatic line, and tailings particle size controls the permeability of the dam body. The state factors include reservoir water level and phreatic line depth. Reservoir water level directly determines the risk of flood overtopping, and phreatic line depth directly determines the risk of dam instability. The control factors include the reliability of monitoring and early warning and the timeliness of emergency response. The reliability of monitoring and early warning is quantified by the rate of missing monitoring data, while the timeliness of emergency response is quantified by the average time difference between system alarm and alarm deactivation. The above 13 indicators constitute the complete monitoring indicator system that the tailings dam safety monitoring system needs to collect and process.

[0034] In one embodiment of this application, when screening specific influencing factors based on the five categories of causal roles, the principle of causal homogeneity is used to ensure that indicators within the same criterion layer belong to the same causal role. The principle of causal homogeneity means that indicators of different types of causal roles should not be placed in the same criterion layer for equal weighting. For example, "total flood volume," which belongs to the driving factor, and "reservoir water level," which belongs to the state factor, cannot be directly compared in the same criterion layer because they occupy different positions in the causal chain, and conflation would lead to logical confusion. Simultaneously, indicator reduction is performed based on the principle of information non-redundancy: if the safety status of the first indicator can be fully inferred from the monitoring value of the second indicator, then the second indicator is retained and the first indicator is discarded. For example, "safety superelevation" is the difference between "dam crest elevation" and "reservoir water level elevation," which can be fully inferred from the dam height and reservoir water level; therefore, it is not retained as an independent indicator to avoid duplicate inclusion of relevant information. "Dam construction materials" and "tailings physical parameters" refer to the same physical object in tailings dams using tailings and are merged into a single indicator, "tailings particle size." The actual flood discharge capacity is determined by three factors: "designed flood discharge capacity," "reliability of flood discharge facilities," and "operation and maintenance of flood discharge facilities." These factors are combined into a single indicator: "flood discharge capacity." Through the selection based on the principles of causal homogeneity and information non-redundancy, the established set of factors has a clear causal logical hierarchy.

[0035] In one embodiment of this application, step S3 establishes an adjacency matrix based on the causal relationships between the influencing factors. Let the factor set be S = {S1, S2, ..., S...} n}, where n is the number of influencing factors. Based on professional knowledge and practical engineering experience in tailings dam engineering, the direct causal relationships between each influencing factor are compared pairwise. If factor S i For factor S j If there is a direct impact, then the corresponding element a in the adjacency matrix A is... ij =1; if there is no direct impact, then a ij =0. The diagonal element a of the adjacency matrix. ii =0. The adjacency matrix A is an n×n 0-1 square matrix that fully records the direct causal relationships between various influencing factors. The reachability matrix is ​​obtained by exponentiation of the adjacency matrix. The formula for calculating the reachability matrix M is: M = (A + I) k Where I is the identity matrix and k is an integer. The elements m in the reachable matrix M... ij =1 indicates that from factor S i To factor S j There exists a path with a length greater than or equal to 0, i.e., factor S. i Factor S can be reached through direct or indirect causal relationships. j The reachability matrix reflects all causal relationships between influencing factors and serves as the data foundation for subsequent hierarchical division.

[0036] like Figure 2 and Figure 3 As shown, in one embodiment of this application, step S4 employs a result-first extraction rule to hierarchically divide the reachability matrix, obtaining a UP-type multi-level hierarchical structure. Regarding influencing factors S... i Its reachable set R(S) i Let A(S) be the set of factors corresponding to the column with a value of 1 in the i-th row of the reachability matrix M, and let its antecedent set be S(S). i Let R(S) be the set of factors corresponding to the row with a value of 1 in the i-th column of the reachability matrix M. The result-priority extraction rule is: if R(S) i ) A(S i ), then S i Extracting from the current highest level. This rule means that when all factors reachable by a factor are covered by its antecedents, that factor is at the end of the current hierarchical system; that is, it is the result of other factors rather than the cause of further results. Extracting layer by layer according to this rule, each extracted set of elements is placed at the top of the structure diagram, arranged sequentially from top to bottom, ultimately resulting in a UP-type multi-level hierarchical directed topology graph, such as... Figure 2 As shown. The reachability matrix is ​​hierarchically partitioned using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure. The cause-first extraction rule is: if A(S i ) R(S i), then S i Extracting from the current highest level. This rule means that when all antecedents of a factor are covered by its reachable set, that factor is at the beginning of the current hierarchical system; that is, it is the cause of other results, not the result of other factors. Extracting layer by layer according to this rule, each extracted set of elements is placed below the structure graph, arranged sequentially from bottom to top, ultimately resulting in a DOWN-type multi-level hierarchical directed topology graph, such as... Figure 3 As shown.

[0037] In one embodiment of this application, step S5 compares the UP-type multilevel hierarchical structure with the DOWN-type multilevel hierarchical structure. Since the UP-type structure traces the cause upwards from the result, while the DOWN-type structure derives the result downwards from the cause, the two methods may produce different results for the hierarchical division of the same system. If the same influencing factor has different hierarchical positions in the UP-type and DOWN-type structures, then that factor is identified as a pivotal factor in the causal network. The existence of hierarchical disputes indicates that the factor has a multifaceted role in the causal network—it may be both the result of some factors and the cause of others. Such factors are usually located in the middle transmission layer of the causal network, possessing high pivotal nature and significantly influencing both upstream and downstream factors. Through this bidirectional adversarial extraction method, hierarchical division no longer depends on a single-direction extraction rule, eliminating the differences in conclusions caused by different rule choices, and making the analysis results reproducible.

[0038] In one embodiment of this application, step S6 generates a monitoring indicator priority configuration table in the tailings dam safety monitoring system based on the pivotal factors. The monitoring indicators corresponding to the pivotal factors are marked as the highest priority monitoring objects in the safety monitoring system, and corresponding early warning configuration parameters are output. The early warning configuration parameters include the collection frequency adjustment value and the early warning threshold adjustment value for the highest priority monitoring objects. Specifically, for indicators marked as highest priority monitoring objects, the system increases their collection frequency from the default value to a preset high-frequency collection value to ensure that the changing trend of the indicator can be captured in a timely manner during risk evolution. Simultaneously, based on the historical data distribution characteristics and engineering experience of the indicator, the system lowers its early warning trigger threshold, enabling it to trigger an early warning at the early stage of abnormal fluctuations in the indicator. These configuration parameters are output to the data acquisition and access module in the form of executable instructions, which are then automatically executed by the data acquisition and access module. This step transforms the analysis conclusions from a purely cognitive level into executable configuration instructions for the safety monitoring system, achieving a complete closed loop from "data analysis" to "system optimization."

[0039] like Figure 4As shown, in one embodiment of this application, the method further includes constructing a tailings dam flood disaster fault tree, calculating the minimum cut set and minimum path set of the fault tree, and calculating the structural importance of each basic event. The tailings dam flood disaster fault tree is decomposed layer by layer downwards with "tailings dam flood disaster" as the top event. Flood disasters mainly include two modes: flood overtopping and dam instability. Flood overtopping is caused by insufficient freeboard and management factors, and insufficient freeboard is further caused by rising reservoir water level; dam instability is caused by reduced dam stability and management factors, and reduced dam stability is further caused by the rise of the phreatic line and inherent dam factors. Basic events in the fault tree include large peak flow, insufficient flood discharge capacity, large total flood volume, insufficient flood control reservoir capacity, unreliable monitoring and early warning, insufficient emergency response capacity, poor seepage discharge capacity, excessive dam height, excessively steep outer slope ratio, excessively steep dry beach slope, and excessively fine tailings particle size. The minimum cut set and minimum path set are calculated based on the fault tree. The minimum cut set is the minimum combination of basic events that sufficiently lead to the occurrence of the top event, and each minimum cut set represents a disaster-causing path. The structural importance of each basic event is calculated, and the ranking rule for structural importance is as follows: the more frequently a basic event appears in the minimum cut set, the greater its structural importance; for basic events with the same frequency of occurrence, the lower the order of the minimum cut set, the greater its structural importance. This ranking of structural importance is then compared and verified with the previously determined pivotal factors. If a factor is simultaneously identified as a pivotal factor and has the highest structural importance, then that factor is marked as the highest priority monitoring object in the safety monitoring system. This comparison and verification mechanism provides independent support for the analysis conclusions from the failure logic side, enhancing the reliability of the decision-making basis.

[0040] like Figure 5 and Figure 6 As shown, in one embodiment of this application, a tailings dam safety monitoring system for implementing the above method is also provided. This system includes a data acquisition and access module, a factor set construction module, a matrix operation module, an adversarial interpretation structure model construction module, a pivot factor identification module, and a configuration optimization module.

[0041] The data acquisition and access module is used to collect time-series data of multiple monitoring indicators of the tailings dam. These monitoring indicators include at least the dam water level, seepage line depth, rainfall, dry beach slope, and the operational status of drainage facilities. This module connects to the sensor network deployed at the tailings dam site via wired or wireless communication, receiving real-time monitoring data from dam water level sensors, seepage line sensors, rain gauges, and dry beach monitoring devices. The acquisition frequency of the data acquisition and access module can be adjusted according to the early warning configuration parameters output by the configuration optimization module—the acquisition frequency is automatically increased for indicators marked as the highest priority monitoring objects; for non-priority indicators, the acquisition frequency can be appropriately reduced to save system resources.

[0042] In one embodiment of this application, the data acquisition and access module is configured with a file parsing interface for receiving monitoring data files exported from an external monitoring system. By parsing the encoding information in the filename of the monitoring data file, the module automatically identifies the tailings dam and monitoring point, and loads the corresponding parameter configuration from the configuration file directory accordingly. The filename format is "tailings dam code_point code.extension". The data acquisition and access module extracts the tailings dam code and monitoring point code from the filename using regular expressions. This mechanism achieves automated identification and configuration loading with zero manual intervention.

[0043] In one embodiment of this application, the system further includes a data preprocessing module for anomaly detection and missing value imputation preprocessing of the monitoring data. The anomaly detection employs a dual-threshold physical constraint mechanism: anomaly drop thresholds are set for both the reservoir water level change and the phreatic line change. When the reservoir water level drop exceeds the threshold for anomaly drop in either the reservoir water level or the phreatic line drop exceeds the threshold for anomaly drop in the phreatic line at adjacent acquisition times, the corresponding data point is marked as an anomaly and removed. The design basis of this dual-threshold physical constraint mechanism is that—under normal operating conditions, the tailings dam water level and phreatic line will not experience a drastic drop beyond the physically possible range in the short term. Such a jump usually indicates sensor failure, communication interruption, or external interference, and should be removed. The missing value imputation employs a cubic spline interpolation method: using effective monitoring data points as nodes, a piecewise cubic polynomial is constructed. While ensuring the continuity of the first and second derivatives at each node, the monitoring values ​​at missing time points are interpolated and imputed. The cubic spline interpolation method, by constructing a piecewise cubic polynomial and ensuring continuous first and second derivatives at each node, can generate a curve with high smoothness that passes through all known data points, thus better reflecting the natural trend of reservoir water level and seepage line changes.

[0044] In one embodiment of this application, the configuration file directory adopts a hierarchical structure, with each tailings pond corresponding to an independent tailings pond parameter configuration file, and each monitoring point corresponding to an independent monitoring point parameter configuration file. The tailings pond parameter configuration file includes at least the following fields: tailings pond code, tailings pond name, current dam height, current dry beach slope, geographical coordinates (longitude), and geographical coordinates (latitude). The monitoring point parameter configuration file includes at least the following fields: monitoring point code, monitoring point name, and the code of the tailings pond to which it belongs. The early warning configuration parameters are stored in the corresponding monitoring point parameter configuration file. When the platform starts, the data acquisition and access module automatically scans the configuration file directory, identifies all configured tailings ponds and monitoring points, and loads them into memory. Upon receiving a monitoring data file, it uses the extracted tailings pond code and monitoring point code as indexes to match and apply the corresponding parameter configuration from memory.

[0045] The factor set construction module is used to establish a factor set based on the multiple monitoring indicators as influencing factors. This module classifies the multiple monitoring indicators into five causal roles according to their roles in the disaster causal chain. The five causal roles include driving factors, resistance factors, inherent factors, state factors, and control factors. The module also selects indicators based on the principles of causal homogeneity and / or information non-redundancy.

[0046] The matrix operation module is used to build an adjacency matrix based on the causal relationships between various influencing factors, and then perform exponentiation on the adjacency matrix to obtain a reachability matrix. This module receives the factor set output by the factor set construction module, determines the direct causal relationships between the factors according to the pre-defined causal relationships in the engineering knowledge base, constructs the adjacency matrix, and then performs matrix exponentiation until convergence to obtain the reachability matrix.

[0047] The adversarial explanation structure model construction module is used to perform hierarchical partitioning of the reachability matrix using a result-first extraction rule to obtain a UP-type multi-level hierarchical structure, and to perform hierarchical partitioning of the reachability matrix using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure.

[0048] The pivot factor identification module is used to compare the UP-type multi-level hierarchical structure with the DOWN-type multi-level hierarchical structure. If the same influencing factor has a different hierarchical position in the two structures, then the factor is identified as a pivot factor in the causal network.

[0049] The configuration optimization module generates a monitoring indicator priority configuration table based on the hub factors, marks the monitoring indicators corresponding to the hub factors as the highest priority monitoring objects, and outputs early warning configuration parameters, including collection frequency adjustment values ​​and early warning threshold adjustment values, to the data acquisition and access module. The data acquisition and access module increases the collection frequency of the highest priority monitoring object and / or lowers its early warning trigger threshold according to the early warning configuration parameters. Specifically, the data acquisition and access module increases the data collection frequency of the highest priority monitoring object according to the collection frequency adjustment value and sets more sensitive early warning trigger conditions according to the early warning threshold adjustment value. When the real-time data of the monitoring indicator exceeds the adjusted early warning threshold, the system immediately issues an early warning signal. Through the above configuration optimization, the security monitoring system can allocate resources to one or a few of the most critical monitoring indicators under limited computing and communication resources, avoiding the omission or delay of critical data due to the dispersion of monitoring resources.

[0050] In one embodiment of this application, the system further includes a fault tree analysis module for constructing a fault tree for tailings dam flood disasters, calculating the minimum cut set and minimum path set of the fault tree, calculating the structural importance of each basic event, and verifying the ranking of the structural importance against the pivot factors determined by the pivot factor identification module. If a factor is simultaneously identified as a pivot factor by the pivot factor identification module and determined to have the highest structural importance by the fault tree analysis module, the configuration optimization module marks that factor as the highest priority monitoring object in the safety monitoring system.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing the configuration of a tailings dam safety monitoring system based on a causal role classification and adversarial explanation structure model, characterized in that, Includes the following steps: S1: Obtain historical monitoring data from the tailings dam safety monitoring system. The historical monitoring data includes time series data of multiple monitoring indicators, including at least the reservoir water level, the depth of the seepage line, rainfall, the slope of the dry beach, and the operating status data of the flood discharge facilities. S2: Using the multiple monitoring indicators as influencing factors, the multiple monitoring indicators are divided into five causal roles according to their roles in the disaster causal chain. Based on the five causal roles, specific influencing factors corresponding to each causal role are screened to establish a factor set. S3: Establish an adjacency matrix based on the causal relationship between various influencing factors, and perform an exponentiation operation on the adjacency matrix to obtain the reachability matrix; S4: The reachability matrix is ​​hierarchically divided using a result-first extraction rule to obtain a UP-type multi-level hierarchical structure; the reachability matrix is ​​hierarchically divided using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure. S5: Compare the UP-type multi-level hierarchical structure with the DOWN-type multi-level hierarchical structure. If the same influencing factor has a different hierarchical position in the two structures, then the factor is identified as the pivot factor in the causal network. S6: Based on the key factors, generate a monitoring indicator priority configuration table in the tailings dam safety monitoring system, mark the monitoring indicators corresponding to the key factors as the highest priority monitoring objects in the safety monitoring system, and output the corresponding early warning configuration parameters. The early warning configuration parameters include the collection frequency adjustment value and early warning threshold adjustment value of the highest priority monitoring objects.

2. The method according to claim 1, characterized in that, The five causal roles include driving factors, impedance factors, inherent factors, state factors, and control factors. The driving factors include peak flow and total flood volume; the impedance factors include flood discharge capacity, flood control reservoir capacity, and seepage capacity; the inherent factors include dam height, outer slope ratio, dry beach slope, and tailings particle size; the state factors include reservoir water level and phreatic line depth; and the control factors include monitoring and early warning reliability and emergency response timeliness.

3. The method according to claim 1, characterized in that, When selecting specific influencing factors based on the five categories of causal roles, the principle of causal homogeneity is used to ensure that indicators within the same criterion layer belong to the same causal role, and the principle of information non-redundancy is used to simplify the indicators: if the safety status of the first indicator can be fully inferred from the monitoring value of the second indicator, then the second indicator is retained and the first indicator is discarded.

4. The method according to claim 1, characterized in that, The result priority extraction rule in step S4 is: if the influencing factor S i The reachable set R(S) i ) is the antecedent set A(S) i If Si is a subset of ), then Si will be extracted as the current highest level; The reason-priority extraction rule is as follows: if the influencing factor S i Antecedent set A(S) i R(S) is a reachable set. i If Si is a subset of ), then Si will be extracted as the current highest level.

5. The method according to claim 1, characterized in that, The method also includes step S7: constructing a tailings dam flood disaster accident tree, calculating the minimum cut set and minimum path set of the accident tree, and calculating the structural importance of each basic event; comparing and verifying the ranking of the structural importance with the pivotal factors determined in step S5; if a factor is identified as a pivotal factor and has the highest structural importance, then the factor is marked as the highest priority monitoring object in the safety monitoring system.

6. The method according to claim 5, characterized in that, The ranking rule for structural importance is as follows: the more times a basic event appears in the minimum cut set, the greater its structural importance; for basic events that appear the same number of times, the lower the order of the minimum cut set, the greater its structural importance.

7. A tailings dam safety monitoring system employing the method described in any one of claims 1 to 6, characterized in that, include: The data acquisition and access module is used to collect time series data of multiple monitoring indicators of the tailings dam. The monitoring indicators include at least the reservoir water level, the depth of the seepage line, the rainfall, the slope of the dry beach, and the operating status data of the flood discharge facilities. The factor set construction module is used to establish a factor set with the multiple monitoring indicators as influencing factors. The matrix operation module is used to establish an adjacency matrix based on the causal relationship between various influencing factors, and to perform exponentiation on the adjacency matrix to obtain a reachability matrix; The adversarial explanation structure model construction module is used to hierarchically divide the reachable matrix using a result-first extraction rule to obtain a UP-type multi-level hierarchical structure, and to hierarchically divide the reachable matrix using a cause-first extraction rule to obtain a DOWN-type multi-level hierarchical structure. The pivot factor identification module is used to compare the UP-type multi-level hierarchical structure with the DOWN-type multi-level hierarchical structure. If the same influencing factor has a different hierarchical position in the two structures, then the factor is identified as a pivot factor in the causal network. The configuration optimization module is used to generate a monitoring indicator priority configuration table based on the hub factors, mark the monitoring indicators corresponding to the hub factors as the highest priority monitoring objects, and output the early warning configuration parameters, including the collection frequency adjustment value and the early warning threshold adjustment value, to the data acquisition and access module. The data acquisition and access module increases the acquisition frequency of the highest priority monitoring object and / or lowers its early warning trigger threshold according to the early warning configuration parameters.

8. The system according to claim 7, characterized in that, The data acquisition and access module is equipped with a file parsing interface for receiving monitoring data files exported from external monitoring systems. By parsing the encoding information in the filename of the monitoring data file, it automatically identifies the tailings dam and monitoring point to which it belongs, and loads the corresponding parameter configuration from the configuration file directory accordingly. The filename format is "tailings dam code_point code.extension". The data acquisition and access module extracts the tailings dam code and monitoring point code from the filename using regular expressions.

9. The system according to claim 7, characterized in that, It also includes a data preprocessing module for anomaly detection and missing value imputation preprocessing of the monitoring data; the anomaly detection adopts a dual threshold physical constraint mechanism: an abnormal decrease threshold for the change in reservoir water level and an abnormal decrease threshold for the change in the seepage line are set respectively. When the decrease in reservoir water level at adjacent collection times exceeds the abnormal decrease threshold for reservoir water level or the decrease in seepage line exceeds the abnormal decrease threshold for seepage line, the corresponding data point is marked as an anomaly and removed. The missing value filling method uses cubic spline interpolation.

10. The system according to claim 8, characterized in that, The configuration file directory adopts a hierarchical structure, with each tailings pond corresponding to an independent tailings pond parameter configuration file and each monitoring point corresponding to an independent monitoring point parameter configuration file; the early warning configuration parameters are stored in the corresponding monitoring point parameter configuration file, and the data acquisition and access module automatically loads the configuration file when the platform starts.

Images