Dam risk assessment method, device and equipment based on multi-source data fusion

Abstract

The application relates to the technical field of intelligent evaluation, and discloses a dam risk evaluation method, device and equipment based on multi-source data fusion, which comprises the following steps: determining the monitoring item categories of dam multi-source monitoring data, constructing a risk evaluation index system by selecting key indexes of the monitoring item categories, dividing a dam region in combination with the spatial distribution of measuring points to obtain a dam partition result, processing single-measuring-point measured data in the same region in the dam partition result by using a dam risk evaluation model to obtain a regional basic probability distribution value, and further obtaining a risk evaluation result. By constructing the risk evaluation index system, the application breaks through the limitation of single data dimension, realizes accurate grasping of the risk heterogeneity of different regions in the region division, adopts the dam risk evaluation model combining the cloud model and the evidence theory, solves the dimension unification problem of multi-source data, guarantees the comprehensiveness of risk evaluation, and improves the accuracy of risk evaluation.

Description

Technical Field

[0001] This application relates to the field of intelligent assessment technology, and in particular to a method, apparatus and equipment for dam risk assessment based on multi-source data fusion. Background Technology

[0002] With the rapid development of intelligent assessment technology, dam online monitoring systems can now acquire massive amounts of data in real time. However, existing dam risk assessment methods often rely on single data sources or local indicators, which have significant limitations. For example, assessment models based on a single dimension of deformation or seepage are difficult to fully reflect the complex multi-factor coupled risks of dams. At the same time, the differences in dimensions and uncertainties of different monitoring indicators have not been effectively addressed. Furthermore, while a single data source enables the dynamic response of dams under different environmental conditions, the assessment models are often static and fixed, making it impossible to assess dam risks in real time. This results in one-sided and inaccurate risk assessment results. Summary of the Invention

[0003] The main purpose of this application is to provide a method, apparatus and equipment for dam risk assessment based on multi-source data fusion, which aims to solve the technical problem that existing risk assessment methods cannot effectively integrate multi-source monitoring data due to their reliance on a single data source or static model, resulting in low accuracy of risk assessment.

[0004] To achieve the above objectives, this application proposes a dam risk assessment method based on multi-source data fusion, the method comprising:

[0005] Determine the monitoring item categories for multi-source monitoring data of the dam, and construct a risk assessment indicator system by selecting key indicators for each monitoring item category;

[0006] The dam area is divided according to the risk assessment index system and the spatial distribution of the measuring points to obtain the dam zoning results;

[0007] By using a dam risk assessment model, the measured data of a single measuring point in the same area of ​​the dam zoning results are processed to obtain the basic probability allocation value of the area. The dam risk assessment model is a model built based on cloud model and evidence theory.

[0008] The risk assessment result is obtained based on the aforementioned basic probability allocation value.

[0009] In one embodiment, the step of processing the measured data of a single measuring point in the same area of ​​the dam zoning results using a dam risk assessment model to obtain the basic probability allocation value of the area includes:

[0010] By using the dam risk assessment model, the risk classification of the measured data of single measuring points in the same area of ​​the dam zoning results is carried out, and a risk classification system for single measuring points is obtained.

[0011] Based on the single-point risk classification system, an improved forward cloud emitter is generated by determining the characteristic parameters of the cloud model, wherein the improved forward cloud emitter introduces hyperentropy correction.

[0012] The measured data of the single measurement point is input into the forward cloud transmitter for processing to obtain the membership degree of the risk level corresponding to the safety level value of each measurement point;

[0013] The membership degree is converted into a basic probability assignment value in evidence theory;

[0014] The basic probability allocation values ​​within the same region are merged to obtain the basic probability allocation value for the region.

[0015] In one embodiment, the step of classifying the risk of individual measurement points in the same area of ​​the dam zoning results using a dam risk assessment model to obtain a risk classification system for individual measurement points includes:

[0016] Based on the dam's environmental load, the theoretical risk classification boundary values ​​are determined using a finite element model.

[0017] Based on the statistical distribution of multi-source monitoring data of the dam, kernel density estimation is used to determine the actual risk classification interval;

[0018] The theoretical risk classification boundary value and the actual risk classification interval are weighted and fused to obtain fused data;

[0019] The risk classification results are obtained by using the dam risk assessment model to classify the risk of single measuring points in the same area of ​​the dam zoning results.

[0020] Based on the fused data and the division results, a risk classification system for a single measurement point is constructed.

[0021] In one embodiment, the step of generating an improved forward cloud emitter by determining the characteristic parameters of the cloud model based on the single-measurement-point risk classification system includes:

[0022] Based on the fusion of theoretical risk classification boundary values ​​and actual risk classification intervals, the characteristic parameters of the cloud model corresponding to the risk level of each measurement point are determined.

[0023] Determine the confidence factor based on the accuracy report of the monitoring equipment;

[0024] An improved positive cloud emitter is obtained by correcting the hyperentropy of the feature parameters using the feature parameters and the confidence factor.

[0025] In one embodiment, the step of obtaining the risk assessment result based on the basic probability allocation value includes:

[0026] Based on the aforementioned basic probability allocation values, the initial basic probability allocation conflict coefficients between regions are determined;

[0027] A preset threshold for conflict detection is set by determining the mechanical coupling coefficient between regions;

[0028] If the initial basic probability allocation conflict coefficient exceeds the preset threshold, the inter-regional basic probability allocation conflict coefficients are weighted and fused using the entropy weight method to obtain the inter-regional basic probability allocation conflict coefficient.

[0029] The conflict coefficients of the basic probability allocation between the regions are fused layer by layer to obtain the comprehensive basic probability allocation value;

[0030] The comprehensive basic probability allocation value is converted into a risk score according to the preset quantification rules, and a risk assessment result is obtained based on the risk score.

[0031] In one embodiment, the step of dividing the dam area according to the risk assessment index system and the spatial distribution of measuring points to obtain dam zoning results includes:

[0032] Based on the aforementioned risk assessment index system and spatial distribution of measurement points, potential risk patterns in different areas are determined according to the structural mechanical response of the dam.

[0033] Based on the spatial distribution patterns of key indicators, the physical mechanism and the multi-source monitoring data of the dam are matched to obtain matching results;

[0034] Based on the potential risk patterns and the matching results, the zoning boundaries are adjusted using engineering experience to obtain the dam zoning results.

[0035] Furthermore, to achieve the above objectives, this application also proposes a dam risk assessment device based on multi-source data fusion, the multi-source data fusion-based dam risk assessment device comprising:

[0036] The system construction module is used to determine the monitoring item categories of multi-source monitoring data for dams, and to construct a risk assessment indicator system by selecting key indicators for each monitoring item category.

[0037] The region division module is used to divide the dam area according to the risk assessment index system and the spatial distribution of the measuring points, and obtain the dam zoning results;

[0038] The data conversion module is used to process the measured data of a single measuring point in the same area of ​​the dam zoning results through the dam risk assessment model to obtain the basic probability allocation value of the area. The dam risk assessment model is a model built based on cloud model and evidence theory.

[0039] The result generation module is used to obtain risk assessment results based on the basic probability allocation value.

[0040] Furthermore, to achieve the above objectives, this application also proposes a dam risk assessment device based on multi-source data fusion, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the dam risk assessment method based on multi-source data fusion as described above.

[0041] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the dam risk assessment method based on multi-source data fusion as described above.

[0042] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the dam risk assessment method based on multi-source data fusion as described above.

[0043] The technical solution proposed in this application determines the monitoring item categories of multi-source monitoring data for dams. By selecting key indicators for each monitoring item category, a risk assessment indicator system is constructed. Based on this system and the spatial distribution of monitoring points, the dam area is divided to obtain dam zoning results. Using a dam risk assessment model, the measured data of single monitoring points within the same zoning area are processed to obtain the basic probability allocation value for the area. The dam risk assessment model is based on cloud modeling and evidence theory, and the risk assessment results are obtained based on the basic probability allocation value. This application overcomes the limitations of single-dimensional data by constructing a risk assessment indicator system. The regional division enables accurate understanding of the risk heterogeneity in different areas. The dam risk assessment model, combining cloud modeling and evidence theory, solves the problem of dimensional uniformity for multi-source data, ensuring the comprehensiveness and accuracy of risk assessment. Attached Figure Description

[0044] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1This is a flowchart illustrating an embodiment of the dam risk assessment method based on multi-source data fusion provided in this application.

[0047] Figure 2 This is a flowchart illustrating Embodiment 2 of the dam risk assessment method based on multi-source data fusion provided in this application;

[0048] Figure 3 This is a flowchart illustrating Embodiment 3 of the dam risk assessment method based on multi-source data fusion provided in this application;

[0049] Figure 4 This is a schematic diagram illustrating the framework of a dam risk assessment indicator system based on multi-source data fusion.

[0050] Figure 5 A schematic diagram illustrating the process of dividing the dam area;

[0051] Figure 6 A schematic diagram showing the dam area delineation results;

[0052] Figure 7 This is a schematic diagram of the data fusion process in the dam area.

[0053] Figure 8 A schematic diagram illustrating the calculation process for the comprehensive risk score of a dam.

[0054] Figure 9 This is a schematic diagram of the module structure of the dam risk assessment device based on multi-source data fusion according to an embodiment of this application;

[0055] Figure 10 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the dam risk assessment method based on multi-source data fusion in the embodiments of this application.

[0056] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0057] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0058] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0059] Existing dam risk assessment methods often rely on single data sources or local indicators, which have significant limitations and make it difficult to fully reflect the complex multi-factor coupled risks of dams. At the same time, the differences in the dimensions and uncertainties of different monitoring indicators have not been effectively addressed. Furthermore, while a single data source can realize the dynamic response of the dam under different environmental conditions, the assessment model is often static and fixed, making it impossible to assess dam risks in real time. This results in one-sided and inaccurate risk assessment results.

[0060] Therefore, in order to overcome the above-mentioned shortcomings, this application provides a solution that breaks through the limitation of single data dimension by constructing a risk assessment index system, achieves accurate grasp of the risk heterogeneity of different regions through regional division, and solves the problem of dimension unification of multi-source data by adopting a dam risk assessment model that combines cloud model and evidence theory, thus ensuring the comprehensiveness of risk assessment and improving the accuracy of risk assessment.

[0061] It should be noted that the implementing entity of each embodiment of this application can be a computing service system with data processing, network communication, and program execution functions, such as an electronic system capable of realizing the above functions, a dam risk assessment system based on multi-source data fusion, etc. The following uses a dam risk assessment system based on multi-source data fusion (hereinafter referred to as "the System") as an example to describe the following embodiments.

[0062] Based on this, the embodiments of this application provide a dam risk assessment method based on multi-source data fusion, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the dam risk assessment method based on multi-source data fusion in this application.

[0063] In this embodiment, the dam risk assessment method based on multi-source data fusion includes steps S10 to S40:

[0064] Step S10: Determine the monitoring item categories of the multi-source monitoring data of the dam, and construct a risk assessment indicator system by selecting key indicators of the monitoring item categories.

[0065] With the rapid development of monitoring technology, dam online monitoring systems can now acquire massive amounts of data in real time, covering multi-dimensional indicators such as deformation, seepage, and stress-strain. However, current dam safety management generally suffers from a prominent problem of "emphasizing monitoring while neglecting supervision": on the one hand, while the deployment of monitoring equipment and data acquisition capabilities have significantly improved, the supervision system has not been optimized accordingly, lacking systematic integration and in-depth analysis of multi-source data; on the other hand, there is a disconnect between monitoring data and risk control decisions, failing to achieve a dynamic risk assessment and prevention mechanism.

[0066] The above problems indicate that traditional methods have significant shortcomings in data fusion, dynamic supervision, and multi-dimensional risk linkage analysis. How to break through the limitations of a single data source and build a risk assessment method driven by multi-source data fusion has become an urgent problem to be solved.

[0067] Therefore, this application proposes a dam risk assessment method based on multi-source data fusion. By integrating multi-source monitoring data of the dam, the risk status of the dam can be comprehensively analyzed, thereby improving the accuracy and precision of the risk assessment results.

[0068] First, it needs to be clarified in this step that the monitoring project categories refer to the various observation project systems set up in dam safety monitoring, which may include deformation monitoring (surface displacement, internal displacement), seepage monitoring (seepage flow, uplift pressure), stress and strain monitoring (concrete stress, steel reinforcement stress), environmental quality monitoring (water level, temperature), etc. Key indicators are representative characteristic parameters extracted from various monitoring data, such as indicators for assessing deformation, seepage pressure gradient, stress and strain, displacement change rate, crack propagation rate, etc.

[0069] When constructing a dam risk assessment index system, the first step is to extract features and perform correlation analysis on various multi-source dam monitoring data to identify key parameters sensitive to the dam's safety status. For core monitoring items, typical key indicators are selected, such as dam body horizontal displacement and resisting body horizontal displacement to characterize deformation features; dam foundation uplift pressure and seepage around the dam to reflect seepage status; and concrete stress-strain and anchor cable stress to assess mechanical response. The weights of the indicators are determined using the analytic hierarchy process (AHP) or entropy weight method, and machine learning is employed to dynamically adjust the weight allocation, enabling the system to adapt to the evolution of the dam's operating state and more accurately characterize the time-varying patterns of risk compared to traditional static methods.

[0070] Step S20: Divide the dam area according to the risk assessment index system and the spatial distribution of measurement points to obtain the dam zoning results.

[0071] It should be noted that the spatial distribution of monitoring points refers to the physical arrangement of various monitoring devices (such as sensors, inclinometers, piezometers, strain gauges, etc.) within the dam structure, including their planar location and elevation distribution. Dam zoning involves dividing the dam into several assessment units with similar risk characteristics based on the spatial correlation of monitoring data and the engineering structural characteristics.

[0072] The zoning process can begin by establishing the three-dimensional spatial topological relationships of the measuring points. Then, cluster analysis is performed based on the parameter correlations in the risk assessment index system. A hydraulic-mechanical coupled zoning algorithm is introduced, considering not only geometric distances but also the strength of physical connections between measuring points through seepage field analysis and finite element stress calculations. This makes the zoning results more consistent with the actual working behavior of the dam. For example, measuring points with close hydraulic connections and significant stress transmission can be grouped into the same region to improve the spatial resolution of the risk assessment.

[0073] As one implementation method, step S20 in this embodiment may include: determining the potential risk patterns of different regions based on the risk assessment index system and the spatial distribution of measuring points, according to the mechanical response of the dam structure; matching the physical mechanism and the multi-source monitoring data of the dam based on the spatial distribution law of key indicators to obtain matching results; and adjusting the partition boundaries based on the potential risk patterns and the matching results through engineering experience to obtain dam partitioning results.

[0074] It should be noted that potential risk modes refer to typical failure modes that may occur in different areas of the dam under specific load conditions, such as dam sliding, foundation seepage failure, and structural cracking. Structural mechanical response, on the other hand, refers to the changes in physical quantities such as deformation, stress, and seepage of the dam under internal and external loads, which can be obtained through finite element analysis or measured data.

[0075] When identifying potential risk patterns, a numerical analysis model of the dam is first established by combining key indicators (such as displacement, stress, and seepage pressure) and the spatial distribution of measuring points in the risk assessment indicator system. Based on this model, weak links and possible failure modes in each area under extreme conditions are identified through parameter inversion and working condition simulation. At this point, a risk pattern automatic identification method based on machine learning can be used to establish a risk pattern knowledge graph by training the failure mode features in the historical case library. This enables the system to automatically match the similarity between the current monitoring data and typical risk patterns to obtain matching results.

[0076] Understandably, the spatial distribution patterns of key indicators refer to the characteristics of the variation gradient and abnormal area distribution of each monitoring parameter in the three-dimensional space of the dam. Physical mechanism matching is the process of verifying the correspondence between measured data variation patterns and known mechanical mechanisms (such as seepage-stress coupling effects, temperature-deformation relationships, etc.). The matching process can also employ a mechanism-data dual-driven matching algorithm. On the one hand, it establishes theoretical relationship constraints between parameters based on physical equations; on the other hand, it utilizes deep learning to mine implicit correlation patterns in the monitoring data. The two achieve optimal matching through a collaborative optimization framework, overcoming the problem that purely data-driven methods may violate physical laws and solving the limitation of pure mechanism models in handling complex actual working conditions.

[0077] Furthermore, zoning boundary adjustment refers to the process of optimizing and correcting the initially defined regional boundaries based on professional judgment. Engineering experience here specifically refers to the accumulated knowledge of dam structural behavior, failure mechanisms, and monitoring data characteristics among experts, including knowledge of typical failure cases and experience with parameter thresholds. The adjustment method can employ interactive optimization: the system first recommends zoning schemes based on algorithms, then displays key information such as risk pattern distribution and mechanism matching degree through a visual interface. Experts can interactively adjust the boundaries and view the evaluation results in real time. Alternatively, a digital system of expert knowledge can be built to transform engineering experience into quantifiable adjustment rules and constraints, such as setting the minimum zoning area for areas of abrupt hydraulic gradient changes and the boundary smoothness requirements for stress concentration zones. This makes the experience-based adjustment process more standardized and traceable, retaining the advantages of expert judgment while avoiding subjective arbitrariness.

[0078] Step S30: Using the dam risk assessment model, the measured data of single measurement points in the same area of ​​the dam zoning results are processed to obtain the basic probability allocation value of the area. The dam risk assessment model is a model constructed based on cloud model and evidence theory.

[0079] It should be noted that single-point measured data refers to time-series observations collected by specific monitoring equipment, such as the displacement of a certain inclinometer on a certain day. Basic probability assignment (BPA) is a fundamental quantity used in evidence theory to represent the credibility of a proposition; here, it specifically refers to the probability distribution of a certain partition under different risk states.

[0080] In its implementation, the model processing involves two stages: First, the quantitative monitoring data of each measuring point is transformed into a probability distribution of qualitative concepts using a cloud model. The basic probability allocation of each measuring point is calculated through an inverse cloud algorithm. Then, evidence synthesis is performed on the BPA of multiple measuring points within the same partition. Furthermore, an improved cloud model parameter estimation method can be introduced. Considering the characteristics of dam monitoring data, a sliding time window is used to dynamically update cloud digital features (expectation, entropy, and hyperentropy), enabling the model to adapt to changes in data distribution and overcoming the limitations of fixed parameters in traditional cloud models.

[0081] This application presents an innovative architecture for dam risk assessment, integrating cloud models, evidence theory, and quantum computing. The construction steps include: First, establishing a multi-scale sensing system based on multi-source monitoring data of the dam. At the macroscopic level, cloud models are used to process conventional monitoring parameters such as deformation and seepage, while at the microscopic level, quantum computing units are introduced to analyze the evolution characteristics of the material's microstructure. Second, a quantum-enhanced evidence theory framework is constructed, improving traditional evidence fusion algorithms through quantum state encoding technology and designing quantum evidence fusion rules capable of simultaneously processing macroscopic monitoring data and microscopic feature parameters. Third, a dynamic risk boundary generation mechanism is introduced, cross-scale correlation between the risk membership degree output by the cloud model and the microscopic degradation index obtained from quantum computing, establishing an adaptive risk assessment matrix that considers the time-varying characteristics of materials. Finally, the above components are integrated to form a complete model architecture. A digital twin algorithm is used to virtually verify the model and fine-tune parameters to form an optimized dam risk assessment model, ensuring the collaborative performance of each module. This construction process breaks through the single-scale limitations of traditional risk assessment models, forming a model capable of simultaneously analyzing macroscopic responses and microscopic mechanisms.

[0082] Step S40: Obtain the risk assessment result based on the basic probability allocation value.

[0083] Understandably, risk assessment results refer to a quantitative evaluation of the safety status of each zone and the dam as a whole. This typically includes outputs such as risk level classification, risk probability distribution, and risk evolution trends. The key to obtaining these results lies in the selection and improvement of the evidence synthesis algorithm. Based on traditional Dempster-Shafer (DS) evidence theory, a conflict evidence identification mechanism and an adaptive weight allocation strategy can be introduced: first, the degree of conflict between pieces of evidence is measured using Jousselme distance; then, the synthesis rules are dynamically adjusted according to the magnitude of the conflict. A locally weighted synthesis method is used for highly conflicting evidence, effectively solving the decision paradox problem that may arise with traditional methods when evidence is highly conflicting, making the assessment results more reliable. The final output not only includes the conventional risk level but also generates a risk spatial distribution cloud map and a time evolution curve, providing multi-dimensional reference for engineering decisions.

[0084] This embodiment overcomes the limitations of traditional methods with their single data dimension by establishing a risk assessment index system. Furthermore, it partitions the data based on the spatial distribution of monitoring points, achieving a precise characterization of risk heterogeneity in different regions. This effectively avoids the masking of local risks by global assessments. The dam risk assessment model, combining cloud models and evidence theory, solves the problem of dimensional uniformity in multi-source data, overcomes the impact of uncertainty in monitoring equipment data, ensures the comprehensiveness of risk assessment, and improves its accuracy. Moreover, by identifying potential risk patterns and verifying the rationality of partitioning through spatial clustering of monitoring point data, and finally using engineering experience to correct the boundaries of special geological areas, the data within each region not only conforms to mechanical laws but also maintains statistical similarities, thereby improving the detection rate of local risks.

[0085] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 Step S30 may include steps S301 to S305:

[0086] Step S301: Using the dam risk assessment model, risk classification is performed on the measured data of single measuring points in the same area of ​​the dam zoning results to obtain a risk classification system for single measuring points.

[0087] It should be noted that a single-point risk classification system refers to a multi-level risk classification standard established based on measured data (such as displacement values, seepage pressure values, etc.) of a single monitoring point, typically including different levels such as safe, warning, and dangerous. The risk classification process needs to comprehensively consider the characteristics of the monitoring parameters, historical variation patterns, and engineering experience thresholds.

[0088] In practical implementation, firstly, long-term monitoring data from each monitoring point are statistically analyzed to determine its normal fluctuation range and abnormal threshold. Then, combining dam design standards, specifications, and expert experience, a grading standard applicable to different types of monitoring points is established. A dynamic grading method can be adopted, using a time series prediction model to update the grading threshold in real time, enabling risk grading to adapt to changes in the dam's operating status. Simultaneously, a fuzzy boundary processing algorithm based on image processing is introduced to set transition intervals at grade boundaries, avoiding potential misjudgments caused by traditional rigid grading.

[0089] As one implementation method, step S301 in this embodiment may include: determining the theoretical risk classification boundary value based on the dam environmental load using a finite element model; determining the actual risk classification interval using kernel density estimation based on the statistical distribution of multi-source monitoring data of the dam; weighted fusion of the theoretical risk classification boundary value and the actual risk classification interval to obtain fused data; risk classification of the measured data of single measuring points in the same area of ​​the dam zoning results using a dam risk assessment model to obtain classification results; and constructing a single measuring point risk classification system based on the fused data and the classification results.

[0090] It should be noted that the theoretical risk classification boundary value refers to the theoretical dividing point between each risk level obtained through numerical calculation, reflecting the theoretical response characteristics of the dam under specific environmental loads (such as water pressure, temperature load, etc.). The finite element model is used here as a theoretical analysis tool to simulate the mechanical behavior of the dam under various working conditions through discretization calculation.

[0091] In practical implementation, a water-thermal-mechanical coupled finite element model including the dam body, foundation, and reservoir area is first established. Then, various load combinations specified in the design code are input for calculation. Based on the calculation results, extreme values ​​of parameters such as stress and deformation in key parts are extracted, and theoretical risk classification boundary values ​​are determined by combining material strength and safety factors. This process can incorporate multi-condition automatic coupling analysis algorithms to perform parallel calculations of hundreds of load combinations. Machine learning algorithms can automatically identify the most unfavorable load combination, improving the comprehensiveness and reliability of theoretical boundary determination. Simultaneously, uncertainty analysis methods are introduced to provide the probability distribution characteristics of the theoretical boundary rather than a single deterministic value.

[0092] The actual risk grading range is the actual distribution range of each risk level obtained from long-term monitoring data, reflecting the dam's behavioral characteristics in a real operating environment. Kernel density estimation is a nonparametric statistical method that can estimate the probability density function from a finite sample, making it particularly suitable for handling complex distributions in dam monitoring data.

[0093] In the specific implementation, historical monitoring data is first cleaned and features are extracted. Then, for each key indicator, an adaptive bandwidth kernel density estimation method is used to calculate its probability distribution. The actual risk interval boundary is determined based on the characteristic points (such as inflection points and extreme points) of the probability density curve in the probability distribution. This process introduces a spatiotemporal correlation kernel density estimation method, which considers not only the magnitude distribution of data values ​​but also spatial location and temporal evolution factors. A more accurate density estimation is achieved through a three-dimensional kernel function, effectively identifying spatial heterogeneity and temporal non-stationarity characteristics in the monitoring data. Subsequently, the theoretical risk grading boundary values ​​and the actual risk grading intervals are weighted and fused. The resulting fused data is an optimized risk grading standard obtained by combining theoretical calculations and actual observations. The fusion algorithm uses an adaptive weighting method based on information entropy. First, the reliability indicators of the theoretical risk grading boundary values ​​and the actual risk grading intervals are evaluated separately, and then the fusion weights are dynamically adjusted based on the reliability.

[0094] Risk classification results refer to the output of classifying real-time monitoring data from each measuring point into corresponding risk levels according to predetermined standards. The classification method employs a multi-threshold fuzzy discrimination algorithm, setting a transition interval near strict classification boundaries, achieving smooth classification through membership functions, and also considering a collaborative classification strategy based on parameter correlation. It not only relies on single indicator values ​​but also comprehensively analyzes the changing trends and spatial distribution characteristics of relevant parameters. For example, the classification of displacement measuring points considers not only the absolute value but also the differences and acceleration of change between them and adjacent measuring points to improve the accuracy of risk identification and early warning capabilities.

[0095] In addition, during the construction of the single-monitoring-point risk classification system, a dynamic update mechanism for the classification system was established. The system is regularly optimized and adjusted based on new monitoring data and engineering feedback to ensure that the classification system can maintain its accuracy and applicability and effectively track the performance evolution during the aging process of the dam.

[0096] Step S302: Based on the single-point risk classification system, an improved forward cloud emitter is generated by determining the characteristic parameters of the cloud model, wherein the improved forward cloud emitter introduces hyperentropy correction.

[0097] It should be noted that the forward cloud emitter is an algorithmic structure that can convert quantitative values ​​into qualitative conceptual membership degrees. Its core is the three characteristic parameters of the cloud model: expected value ( ),entropy( ) and hyperentropy ( The over-entropy correction is an improvement measure proposed to address the shortcomings of traditional cloud models in processing dam monitoring data. It aims to balance the model's expression of uncertainty: too high an entropy will lead to excessive dispersion of membership, while too low an entropy may ignore reasonable fluctuations.

[0098] In its implementation, the improved forward cloud transmitter construction process first determines the characteristic parameters corresponding to each level based on the single measurement point risk classification system, and then optimizes the model's ability to express uncertainty.

[0099] In one implementation, step S302 in this embodiment may include: determining the characteristic parameters of the cloud model corresponding to the risk level of each measuring point based on the fused data of the theoretical risk classification boundary value and the actual risk classification interval; determining the confidence factor according to the accuracy report of the monitoring equipment; and correcting the over-entropy of the characteristic parameters through the characteristic parameters and the confidence factor to obtain the improved positive cloud emitter.

[0100] It should be noted that the expected value, entropy, and hyperentropy of the cloud model's characteristic parameters respectively represent the central value, fuzziness, and randomness of the risk level. The fused data serves as a classification benchmark, providing numerical basis for the risk level of each monitoring point, ensuring that the cloud model's parameter settings conform to both theoretical calculations and actual monitoring patterns.

[0101] When determining the feature parameters, the boundaries of each level are first calculated based on the fused risk grading intervals. The expected value is set as the median of the level interval. Entropy reflects the interval span, while hyperentropy expresses the dispersion of the data. A dynamic parameter optimization algorithm is introduced to adaptively adjust the values ​​of entropy and hyperentropy by combining the time-varying characteristics of historical data from the measuring points. For example, for long-term stable seepage pressure measuring points, hyperentropy is reduced to decrease random interference; while for displacement measuring points affected by seasonality, hyperentropy is appropriately increased to enhance the model's adaptability to fluctuations.

[0102] In addition, the confidence factor is a weighted coefficient used to quantify the reliability of monitoring data. Its value depends on the measurement accuracy, stability and maintenance records of the equipment. Equipment accuracy reports usually include key indicators such as instrument calibration error, long-term drift characteristics and environmental adaptability.

[0103] When determining the confidence factor, the equipment technical documentation is first analyzed to extract parameters such as accuracy level and measurement error range. This is then combined with on-site maintenance records (such as calibration frequency and fault history) for a comprehensive score. A multi-dimensional confidence assessment model can be used, considering not only the equipment's inherent accuracy but also factors such as installation quality and environmental interference (e.g., electromagnetic fields and temperature fluctuations). A more reasonable confidence weight is output using fuzzy comprehensive evaluation. For example, the confidence factor for strain gauges installed in areas with significant vibration may be appropriately lowered to reflect potential data noise.

[0104] The improved positive cloud emitter is an optimized cloud model generation algorithm. Its core principle is to dynamically adjust the over-entropy through a confidence factor, making the membership degree of the model output more closely reflect the reliability of the data. The over-entropy correction process employs a weighted adjustment strategy, using the confidence factor as the adjustment coefficient for the over-entropy. For example, for high-confidence measurement points (such as newly calibrated vibrating wire instruments), the over-entropy is compressed to reduce randomness; while for low-confidence measurement points (such as resistive instruments subject to electromagnetic interference), the over-entropy is appropriately relaxed to accommodate greater uncertainty. In this process, a nonlinear correction function can be used to set differentiated over-entropy adjustment rules for different risk levels. For example, the over-entropy correction for high-risk levels is more conservative (using a first adjustment threshold) to avoid misjudgments; while low-risk levels allow for more lenient adjustments (using a second adjustment threshold less than the first adjustment threshold) to improve the model's sensitivity to minor anomalies. The resulting cloud emitter can more accurately map the relationship between monitoring data and risk levels, improving the reliability of the assessment results.

[0105] Step S303: Input the measured data of the single measurement point into the forward cloud transmitter for processing to obtain the membership degree of the risk level corresponding to the safety level value of each measurement point.

[0106] It should be noted that membership degree refers to a fuzzy measure of whether measured data belongs to a certain risk level, with a value between 0 and 1, indicating the degree of matching between the data and the level; the safety level value is a quantitative indicator that can be used for risk assessment, obtained by standardizing the original monitoring data.

[0107] The process of inputting single-point measured data into the forward cloud transmitter for processing first involves preprocessing the raw single-point measured data (such as denoising and normalization), and then inputting it into the improved forward cloud transmitter for calculation. Through a multi-scale membership calculation method, not only is the membership degree of the data at the current moment calculated, but its short-term changing trend membership characteristics are also analyzed. By fusing information across the time dimension, a more comprehensive risk assessment result is obtained to identify situations where the current value is within a safe range but the changing trend is abnormal, thus achieving early risk warning.

[0108] Step S304: Convert the membership degree into the basic probability assignment value of the evidence theory.

[0109] Understandably, the transformation process in this step requires addressing the issue of how to reasonably convert fuzzy membership degrees into probability assignments. The transformation method employs an improved membership-probability mapping algorithm. First, the membership degrees are normalized, and then a confidence factor is introduced to adjust the transformation weights. An information entropy-based transformation optimization strategy is used, which determines the optimal basic probability assignment values ​​by maximizing information content and minimizing conflict.

[0110] Step S305: Merge the basic probability allocation values ​​within the same region to obtain the basic probability allocation value for the region.

[0111] It should be noted that the basic probability allocation value for a region is the overall risk assessment result obtained by integrating the evaluation information of all measuring points within a certain partition. The fusion process needs to address the information conflict and redundancy issues between different measuring points. The fusion algorithm adopts an improved evidence synthesis rule and introduces three innovations on the basis of traditional DS theory: 1) a weighted fusion strategy based on the reliability of measuring points; 2) a mechanism for identifying and processing conflicting evidence; and 3) an adaptive adjustment factor for regional characteristics.

[0112] Specifically, the step of fusing basic probability allocation values ​​within the same region to obtain regional basic probability allocation values ​​may include: comprehensively considering the accuracy of monitoring equipment, the stability of historical data, and the quality of real-time signals; performing a reliability score for each measuring point; assigning a larger fusion weight to high-reliability measuring points based on the score results; and simultaneously introducing regional historical benchmark data to compensate and correct low-quality measuring points. By calculating the difference matrix between measuring point evidence (processed BPA data, i.e., the results obtained after analyzing the characteristics of each measuring point's BPA, such as the determination of the main focal elements, evidence uncertainty, and evidence specificity), a subset of evidence with significant conflicts is automatically identified. For conflicting evidence, an information entropy assessment method is used to intelligently reconstruct its risk level distribution, retaining effective information features while eliminating abnormal fluctuation interference, providing a consistent source of evidence for subsequent fusion. Combining the finite element analysis results of the dam structure, and considering engineering factors such as material properties and load transfer paths, a spatial correlation adjustment factor is dynamically generated. During the final evidence synthesis, the fusion weights are automatically optimized based on the spatial distribution characteristics of the measuring points, ensuring that the risk assessment results conform to both statistical data patterns and engineering mechanics principles.

[0113] For ease of understanding, the following example is provided, but it is not intended to limit the dam risk assessment method based on multi-source data fusion in this application. In the dam risk assessment scenario, based on the risk assessment index system and spatial distribution of monitoring points established in the above steps, and comprehensively considering the spatial distribution characteristics of all monitoring points, a three-dimensional fusion of the arch dam structural mechanical response, the spatial distribution patterns of monitoring indicators, and engineering experience is used to establish a zoning mapping relationship strongly correlated with the physical mechanism. This divides the dam into multiple regions, each of which can include multi-dimensional monitoring items such as deformation and seepage, and the data variation patterns within each region are the same. Based on the single-monitoring-point regional division, a dam risk assessment model based on the cloud model-DS evidence theory is established. The single-monitoring-point data of the same region are fused to obtain the regional risk assessment results. Specific steps may include:

[0114] A1. Based on the risk classification interval and risk classification boundary value of a single measurement point, a single measurement point risk classification system is constructed as shown in Table 1.

[0115] Table 1 Risk Classification System for Single Monitoring Points of Dams

[0116]

[0117] A2. Based on the single-point risk classification system established in step S1, calculate the characteristic parameters of the cloud model and generate a positive cloud emitter.

[0118] A3. Import the measured values ​​of the single measurement point data into the positive cloud transmitter generated in step S2, calculate the membership degree of the risk level corresponding to the safety level value of each measurement point, and convert the membership degree of the indicator into BPA in DS evidence theory.

[0119] The membership degree expression is:

[0120] (1)

[0121] In the formula, This represents the membership degree of each risk level corresponding to the fusion indicator. The selected number of cloud droplets.

[0122] The BPA expression is:

[0123] (2)

[0124] In the formula, The value represents the uncertainty.

[0125] A4. Use the DS evidence theory to fuse the BPA values ​​of different types of single measurement points in the same area obtained in step S3.

[0126] The synthesis rules are as follows:

[0127] (3)

[0128] (4)

[0129] In the formula, To identify n pieces of evidence within the framework, their BPAs are respectively The corresponding focal elements are Ai (i=1, 2, ..., n), and K is the conflict coefficient.

[0130] A5. Based on the principle of maximizing membership, the risk level corresponding to the maximum value of BPA after merging each region is the risk level of that region, thus realizing the risk assessment of the region.

[0131] Furthermore, by employing the DS evidence theory to further integrate the BPA values ​​from different regions to obtain a comprehensive BPA, the specific steps for quantifying the comprehensive BPA to calculate the comprehensive risk score may include:

[0132] B1. Treat the composite BPA values ​​calculated for each region as separate pieces of evidence and construct an evidence set. Then, the conflict coefficient K between the comprehensive BPA of each pair of regions is calculated sequentially, and a threshold is set. When the conflict coefficient of two pieces of evidence reaches a threshold, the BPA values ​​of the two pieces of evidence are weighted and averaged before being fused based on the DS fusion rule. If the conflict coefficient is below the threshold, BPA fusion is performed directly according to the DS fusion rule to construct comprehensive evidence. .

[0133] B2. Calculation of comprehensive evidence The conflict coefficient between the evidence body of the next partition is calculated, and the steps are repeated until the BPA of all partitions is merged to obtain the overall BPA value of the dam.

[0134] B3. Calculate the overall risk score of the dam on the day based on the overall BPA value of the dam.

[0135] Specifically, in B3, the membership values ​​of the comprehensive BPA corresponding to different risk levels are weighted, a fixed weight is assigned to the membership under different risk levels, and the overall score is calculated based on the weight assigned to the membership values, thereby realizing the conversion of the comprehensive BPA to the comprehensive risk score of the dam.

[0136] Based on the comprehensive risk score of the dam mentioned above, a four-level risk classification system is established (as shown in Table 2) to achieve the overall risk level assessment of the dam.

[0137] Table 2 Comprehensive Risk Classification System for Dams

[0138]

[0139] This embodiment is based on an improved forward cloud emitter with dynamic overentropy correction, which enables the membership calculation of data to have error tolerance capabilities. It overcomes the long-term monitoring distortion problem caused by the fixed parameters of traditional cloud models, and transforms the membership output of the cloud model into the basic probability assignment of evidence theory. This enables probabilistic decision-making based on multi-point data in the region, avoiding misjudgment of a single indicator. Furthermore, it calculates the theoretical boundary through a finite element model and extracts the actual distribution from historical data by combining kernel density estimation, reflecting the true behavioral characteristics of the dam. The weighted fusion of the two avoids the idealization bias of pure theoretical models and overcomes the overfitting risk of pure data methods. The confidence factor is calculated in real time through the accuracy report of monitoring equipment, and the overentropy is adjusted accordingly, which improves the robustness of the model under complex working conditions such as equipment aging and environmental interference.

[0140] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 Step S40 may include steps S401 to S405:

[0141] Step S401: Based on the basic probability allocation value, determine the initial basic probability allocation conflict coefficient between regions.

[0142] It should be noted that the initial basic probability allocation conflict coefficient is used to quantify the inconsistency between risk assessment results in different regions, reflecting the degree of conflict between different sources of evidence in evidence theory. This coefficient can be calculated based on Jousselme distance or similarity measure to characterize the differences in BPA vectors in different regions.

[0143] In the specific implementation, the basic probability allocation values ​​of each region for different risk levels are first extracted to construct a probability distribution vector. Then, the degree of conflict between pairs of regions is calculated using distance measurement formulas (such as cosine similarity or Euclidean distance). A spatiotemporal correlation correction factor is introduced, which considers not only the BPA difference at the current moment when calculating the conflict coefficient, but also the historical conflict patterns and spatial proximity relationships to avoid misjudgment caused by instantaneous data fluctuations. For example, if adjacent regions have shown low conflict over a long period of time, even if the current conflict coefficient increases briefly, it can still be judged to be within an acceptable range.

[0144] Step S402: Set a preset threshold for conflict detection by determining the mechanical coupling coefficient between regions.

[0145] It should be noted that the mechanical coupling coefficient is a physical indicator reflecting the intensity of interaction between different areas of the dam, including parameters such as stress transfer rate and seepage connectivity; the preset threshold is the upper limit of conflict tolerance determined based on engineering experience and numerical simulation, used to distinguish between normal differences and abnormal conflicts.

[0146] The threshold setting process first obtains the mechanical correlation strength between regions (such as the constraint stiffness of adjacent dam sections and the deformation coordination of the foundation rock mass) through finite element analysis or field tests, and then quantifies it into a mechanical coupling coefficient between regions of 0 to 1. During this process, the threshold standard can be automatically adjusted according to the dam's operating conditions (such as high water level period, after earthquake). For example, under the condition of sudden drop in reservoir water level, the redistribution of dam stress may lead to temporary conflict aggravation. At this time, the system will appropriately increase the threshold to avoid excessive alarms.

[0147] Step S403: If the initial basic probability allocation conflict coefficient exceeds the preset threshold, the inter-regional basic probability allocation conflict coefficient is weighted and fused using the entropy weight method to obtain the inter-regional basic probability allocation conflict coefficient.

[0148] It should be noted that the entropy weighting method is an objective weighting method based on information entropy, used to determine the weight allocation of conflict evidence in each region. Weighted fusion aims to reduce the interference of high-conflict areas on the overall assessment while retaining effective risk signals. The fusion process first calculates the information entropy of the BPA (Balance of Perceptual Aspect) for each region; regions with lower entropy values ​​(higher certainty) have greater weights. Then, a weighted average is applied to the conflict coefficients exceeding a threshold to weaken the contribution of abnormal regions. During this process, the root cause regions of conflict (such as sensor failure areas or true risk areas) can be identified by analyzing the weight distribution. For example, if a region consistently receives a low weight, it may indicate that the monitoring equipment there needs maintenance and calibration.

[0149] Step S404: The conflict coefficients of the basic probability allocation between the regions are fused layer by layer to obtain the comprehensive basic probability allocation value.

[0150] Layer-by-layer fusion refers to the recursive synthesis of conflict coefficients from bottom to top according to the structural hierarchy of the dam (such as dam section-dam block-whole). The comprehensive BPA value is the global risk assessment result obtained through the Dempster synthesis rule or its improved version.

[0151] In its implementation, the fusion algorithm employs a hierarchical evidence network. It first synthesizes evidence within homogeneous sub-regions (such as multiple monitoring sections of the same dam section), and then integrates it across regions. For low-conflict levels, it uses classic DS synthesis, while for high-conflict levels, it switches to weighted synthesis or the PCR6 rule to avoid the "one-vote veto" problem. For example, the conflict between the seepage zone of the dam foundation and the deformation zone of the dam body may reflect real mechanical phenomena. In this case, the system will retain some conflict information rather than forcibly unify it.

[0152] Step S405: Convert the comprehensive basic probability allocation value into a risk score according to a preset quantification rule, and obtain a risk assessment result based on the risk score.

[0153] It should be noted that the risk score is the standard quantitative output of BPA, which is usually designed as a continuous value or discrete level from 0 to 100; the preset quantitative rules define the mapping relationship between the probability distribution and the risk value.

[0154] The comprehensive basic probability allocation value conversion process employs a fuzzy quantification strategy, setting score ranges for the core and transition zones for each risk level. This constructs a multi-dimensional risk matrix, outputting local risk indices (such as deformation risk and seepage risk sub-items) in addition to the total score, and generating a visual heatmap. For example, a dam section might have an overall score of 70 (yellow alert), but a seepage sub-item score of 90 (red alert), indicating a need for focused inspection of the seepage prevention system. The final result provides disposal recommendations through a decision tree model, forming a closed-loop risk management system.

[0155] For ease of understanding, please refer to Figures 4-8This explanation is provided, but it is not intended to limit the dam risk assessment method based on multi-source data fusion proposed in this application. In dam management, a dam risk assessment method based on multi-source data fusion is applied to first organize online monitoring data of the dam, then rationally divide the dam into zones, and finally construct a risk assessment model to achieve risk assessment of the dam area and the entire dam. Specific steps may include:

[0156] Step 1: Divide the single monitoring point of the dam into three categories: deformation, seepage, and stress-strain, and determine the specific monitoring content and measuring points corresponding to each category.

[0157] Specifically, the monitoring categories and monitoring items are set as shown in Table 3.

[0158] Table 3 Monitoring Categories and Monitoring Item Settings

[0159]

[0160] Continued table

[0161]

[0162] Step Two: Select the measuring points corresponding to five monitoring items—dam horizontal displacement X1, resisting body horizontal displacement X2, dam vertical displacement X3, dam foundation deep deformation X4, and dam joint and crack deformation X5—as indicators for evaluating deformation B1; select the measuring points corresponding to four monitoring items—dam foundation uplift pressure X6, dam foundation and dam body seepage pressure X7, seepage around the dam X8, and dam seepage flow X9—as indicators for evaluating seepage B2; select the dam concrete stress-strain X... 10 Stress X of anchor cables in dam body and abutments 11 The measuring points corresponding to the two monitoring items are used as indicators to evaluate stress-strain B3, thereby establishing a risk assessment indicator system. The schematic diagram of the risk assessment indicator system framework is shown below. Figure 4 As shown.

[0163] Step 3: Combining the different patterns of deformation, seepage, and stress-strain at different locations on the dam, the single measuring points of the dam are divided into regions. A schematic diagram of the dam region division process is shown below. Figure 5 As shown, based on the dam's geometric morphology analysis, load transfer path analysis, and identification of geometric boundaries in key response areas, the analysis is conducted from the perspective of structural characteristics and mechanical response. Single-index partitions are divided, such as seepage-dominant partitions (arch crown area, transition area, dam abutment area), stress-dominant partitions (dam foundation seepage area, arch crown seepage area), and deformation-dominant partitions (low stress area, high stress area). Then, through multi-index coupling and comprehensive adjustment, the measurement points are mapped and the partitions are verified to complete the construction of the dam's comprehensive partitions.

[0164] Specifically, the dam can be divided into 43 sections from the left bank to the right bank, further dividing it into six regions: the arch crown beam area, the dam crest area, the dam foundation area, the left / right transition area, and the left / right abutment area. A schematic diagram of the dam's regional division is shown below. Figure 6 As shown, the specific area divisions are as follows:

[0165] The area extending downwards from the design elevation of the dam crest to 1 / 5 of the dam height in dam sections 10-34 is designated as the dam crest zone, denoted as Zone A.

[0166] It is proposed that dam sections 1-9 be designated as the right bank shoulder area, denoted as area B2, and dam sections 35-43 as the left bank shoulder area, denoted as area B1.

[0167] It is proposed that dam sections 10-20 be designated as the right bank transition zone, denoted as Zone C2, and dam sections 26-34 as the left bank transition zone, denoted as Zone C1.

[0168] Sections 21-25 of the dam are designated as the arched beam area (excluding the dam foundation area and the dam crest area). This section of the dam is located in the river channel and is designated as area D.

[0169] The lower third of dam sections 10-34 is designated as the dam foundation area, denoted as Zone E.

[0170] Step 4: Introduce a dam risk assessment model based on cloud modeling and evidence theory. Convert the safety level values ​​of individual measurement points within the dam area into membership degrees under different risk levels, and further map the membership degrees to BPA values ​​under the DS evidence theory framework. A schematic diagram of the dam area data fusion process is shown below. Figure 7 As shown, firstly, based on the monitoring data of each measuring point (measuring point 1 to measuring point n), the digital characteristics (expected value) of the cloud model are calculated according to the risk classification boundary value. ,entropy hyperentropy The process involves generating a forward cloud transmitter (an improved forward cloud transmitter), inputting the safety level values ​​of the measuring points into the cloud transmitter, calculating the membership degree of each risk level, and converting it into a BPA. Then, the BPA of similar monitoring items (such as deformation, seepage, stress and strain) is fused using DS evidence, and the fusion results of dissimilar monitoring items are fused a second time. Finally, the comprehensive BPA is converted into a risk score, and the regional risk level is determined based on the score.

[0171] Specifically, if and Let be the minimum and maximum values ​​of the i-th fusion index at the j-th risk level, respectively. Then, the three characteristic parameters of the cloud model are calculated using the following formula:

[0172] (5)

[0173] In the formula, , , These represent the expected value, entropy, and hyperentropy of the cloud model's features, respectively. It is a constant, and in this example it is taken as 10.

[0174] Furthermore, an improved forward cloud generator is generated based on cloud feature parameters. The algorithm steps for the one-dimensional forward cloud generator are as follows:

[0175] S1, with input For the expectation, Generate a normal random number for the standard deviation. ;

[0176] S2, generate an expected value. The standard deviation is Normal random numbers ;

[0177] S3, Calculation Generate a cloud droplet ;

[0178] S4. Repeat the above steps until the specified number of cloud droplets are generated.

[0179] Based on the positive cloud generator generated in steps S1 to S4, the measured values ​​of the monitoring points are imported to generate BPA values. The DS fusion rules are applied to each monitoring point under each monitoring item within the region to gradually fuse its BPA values. Then, the BPA values ​​of different monitoring items within the same region are further fused, thereby achieving data fusion for the region and completing the determination of the region's risk level.

[0180] In this example, the risk levels of areas A, B1, B2, C1, C2, D, and E are respectively A, A, A, A-, A, and A-.

[0181] Step 5: Using the DS evidence theory, the BPA values ​​from different regions are further integrated to obtain the comprehensive BPA. The comprehensive BPA is then used to quantify and calculate the comprehensive risk score. A schematic diagram of the dam comprehensive risk score calculation process is shown below. Figure 8 As shown, firstly, the BPA values ​​of each partition are used as independent evidence to construct an evidence set M; then, the conflict coefficient K between adjacent pieces of evidence is calculated iteratively (i starts from 1), and based on a preset threshold Q, the evidence is processed either by pairwise fusion using the DS fusion rule or by a weighted average method, generating comprehensive evidence e and updating the evidence set e. i+1 The process is repeated until all n-1 fusions are completed. Finally, the risk score is calculated by quantifying the comprehensive evidence, and the overall risk level of the dam is determined accordingly.

[0182] Specific steps may include:

[0183] S1. In the process of integrating BPA values ​​from different regions using DS evidence theory, a conflict coefficient threshold is set. The BPA value is 0.8. The BPA values ​​of the regions are merged to obtain the comprehensive BPA value.

[0184] S2, if the combined BPA value after fusion is Then the comprehensive risk score F of the dam is:

[0185] (6)

[0186] In the formula, , , , The comprehensive risk score is calculated by taking values ​​of 90, 70, 50, and 30 respectively.

[0187] In the monitoring data of a certain 4 days, the minimum value of the comprehensive risk score F of the dam was 82 and the maximum value was 95. The safety level of the dam was A or A- for all 4 days. Thus, the dam risk assessment was achieved.

[0188] The beneficial effects of the dam risk assessment method based on multi-source data fusion proposed in this application include the following:

[0189] 1. Traditional methods often employ a global assessment, neglecting the risk heterogeneity across different dam regions due to structural characteristics or environmental differences. This application, by combining physical mechanisms and spatial distribution characteristics, divides the dam into multiple risk zones. The data variation patterns within each zone are consistent, and data fusion is performed independently, achieving accurate local risk assessment. Furthermore, through comprehensive BPA quantitative calculation, the overall dam risk level is refined into more intuitive score intervals, providing decision-makers with a more operational basis for risk management.

[0190] 2. By systematically integrating multi-source online monitoring data, a risk assessment indicator system covering multiple indicators was constructed. At the same time, a dam risk assessment model based on cloud model-evidence theory was built, which solved the problems of differences in the dimensions and uncertainties of different monitoring indicators. It also overcame the shortcomings of traditional methods that rely on a single data source and are difficult to cover the multi-dimensional influencing factors of dam safety, and greatly improved the accuracy and reliability of risk assessment.

[0191] 3. The dam risk assessment method based on multi-source data fusion proposed in this application possesses good dynamic adaptability and engineering practicality. It can adapt to the real-time updates and dynamic changes of dam monitoring data, overcoming the limitations of traditional static models in dealing with complex environmental conditions. Furthermore, the design of the risk assessment index system and grading standards closely integrates engineering experience, balancing theoretical rigor with practical application needs. It can be widely applied to risk assessment scenarios for dams of different scales and types, and has high promotional value.

[0192] This embodiment employs multi-scale window segmentation and multi-dimensional feature fusion to comprehensively capture the distribution characteristics of high-dimensional time-series data. By driving the sparsity weight adjustment of the random projection matrix through data distribution characteristics, adaptive optimization of the dimensionality reduction process is achieved. The sparsity mode is dynamically adjusted according to the contribution of each dimension in the principal component direction, so that the low-dimensional data after dimensionality reduction can retain key feature information to the greatest extent while significantly reducing computational complexity. The multi-resolution hash table construction method based on local volatility balances detection accuracy and computational efficiency. By quantifying the local volatility features of each data point, the data space is divided into different regions and configured with differentiated resolution parameters, so that high volatility regions can obtain fine detection capabilities, and low volatility regions can maintain efficient screening, thereby improving the real-time performance and accuracy of the system as a whole.

[0193] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the dam risk assessment method based on multi-source data fusion in this application. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0194] This application also provides a dam risk assessment device based on multi-source data fusion. Please refer to [link / reference]. Figure 9 The dam risk assessment device based on multi-source data fusion includes:

[0195] System construction module 10 is used to determine the monitoring item categories of multi-source monitoring data of the dam, and to construct a risk assessment indicator system by selecting key indicators of the monitoring item categories;

[0196] The area division module 20 is used to divide the dam area according to the risk assessment index system and the spatial distribution of the measuring points, and obtain the dam zoning results;

[0197] The data conversion module 30 is used to process the measured data of a single measuring point in the same area of ​​the dam zoning results through the dam risk assessment model to obtain the basic probability allocation value of the area. The dam risk assessment model is a model built based on cloud model and evidence theory.

[0198] The result generation module 40 is used to obtain risk assessment results based on the basic probability allocation value.

[0199] This application provides a dam risk assessment device based on multi-source data fusion. The dam risk assessment device based on multi-source data fusion includes: at least one processor; and a memory communicatively connected to at least one processor.

[0200] The following is for reference. Figure 10This diagram illustrates a structural schematic of a dam risk assessment device based on multi-source data fusion suitable for implementing embodiments of this application. The device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the dam risk assessment device based on multi-source data fusion. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the abnormal data identification device to communicate wirelessly or wiredly with other devices to exchange data.

[0201] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0202] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the dam risk assessment method based on multi-source data fusion in the above embodiments.

[0203] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the dam risk assessment method based on multi-source data fusion as described above.

[0204] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A dam risk assessment method based on multi-source data fusion, characterized in that, The method includes the following steps: Determine the monitoring item categories for multi-source monitoring data of the dam, and construct a risk assessment indicator system by selecting key indicators for each monitoring item category; The dam area is divided according to the risk assessment index system and the spatial distribution of the measuring points to obtain the dam zoning results; By using a dam risk assessment model, the measured data of a single measuring point in the same area of ​​the dam zoning results are processed to obtain the basic probability allocation value of the area. The dam risk assessment model is a model built based on cloud model and evidence theory. The risk assessment result is obtained based on the aforementioned basic probability allocation value; The step of processing measured data from single measuring points in the same area of ​​the dam zoning results using a dam risk assessment model to obtain the basic probability allocation value for the area includes: By using the dam risk assessment model, the risk classification of the measured data of single measuring points in the same area of ​​the dam zoning results is carried out, and a risk classification system for single measuring points is obtained. Based on the single-point risk classification system, an improved forward cloud emitter is generated by determining the characteristic parameters of the cloud model, wherein the improved forward cloud emitter introduces hyperentropy correction. The measured data of the single measurement point is input into the forward cloud transmitter for processing to obtain the membership degree of the risk level corresponding to the safety level value of each measurement point; The membership degree is converted into a basic probability assignment value in evidence theory; The basic probability allocation values ​​within the same region are merged to obtain the basic probability allocation value for the region.

2. The dam risk assessment method based on multi-source data fusion as described in claim 1, characterized in that, The steps of classifying the risk of individual measurement points in the same area of ​​the dam zoning results using a dam risk assessment model to obtain a risk classification system for individual measurement points include: Based on the dam's environmental load, the theoretical risk classification boundary values ​​are determined using a finite element model. Based on the statistical distribution of multi-source monitoring data of the dam, kernel density estimation is used to determine the actual risk classification interval; The theoretical risk classification boundary value and the actual risk classification interval are weighted and fused to obtain fused data; The risk classification results are obtained by using the dam risk assessment model to classify the risk of single measuring points in the same area of ​​the dam zoning results. Based on the fused data and the division results, a risk classification system for a single measurement point is constructed.

3. The dam risk assessment method based on multi-source data fusion as described in claim 1, characterized in that, The step of generating an improved forward cloud emitter by determining the characteristic parameters of the cloud model based on the single-measurement-point risk classification system includes: Based on the fusion of theoretical risk classification boundary values ​​and actual risk classification intervals, the characteristic parameters of the cloud model corresponding to the risk level of each measurement point are determined. Determine the confidence factor based on the accuracy report of the monitoring equipment; An improved positive cloud emitter is obtained by correcting the hyperentropy of the feature parameters using the feature parameters and the confidence factor.

4. The dam risk assessment method based on multi-source data fusion as described in any one of claims 1 to 3, characterized in that, The step of obtaining the risk assessment result based on the basic probability allocation value includes: Based on the aforementioned basic probability allocation values, the initial basic probability allocation conflict coefficients between regions are determined; A preset threshold for conflict detection is set by determining the mechanical coupling coefficient between regions; If the initial basic probability allocation conflict coefficient exceeds the preset threshold, the inter-regional basic probability allocation conflict coefficients are weighted and fused using the entropy weight method to obtain the inter-regional basic probability allocation conflict coefficient. The conflict coefficients of the basic probability allocation between the regions are fused layer by layer to obtain the comprehensive basic probability allocation value; The comprehensive basic probability allocation value is converted into a risk score according to the preset quantification rules, and a risk assessment result is obtained based on the risk score.

5. The dam risk assessment method based on multi-source data fusion as described in any one of claims 1 to 3, characterized in that, The step of dividing the dam area according to the risk assessment index system and the spatial distribution of measuring points to obtain the dam zoning results includes: Based on the aforementioned risk assessment index system and spatial distribution of measurement points, potential risk patterns in different areas are determined according to the structural mechanical response of the dam. Based on the spatial distribution patterns of key indicators, the physical mechanism and the multi-source monitoring data of the dam are matched to obtain matching results; Based on the potential risk patterns and the matching results, the zoning boundaries are adjusted using engineering experience to obtain the dam zoning results.

6. A dam risk assessment device based on multi-source data fusion, characterized in that, The dam risk assessment device based on multi-source data fusion includes: The system construction module is used to determine the monitoring item categories of multi-source monitoring data for dams, and to construct a risk assessment indicator system by selecting key indicators for each monitoring item category. The region division module is used to divide the dam area according to the risk assessment index system and the spatial distribution of the measuring points, and obtain the dam zoning results; The data conversion module is used to process the measured data of a single measuring point in the same area of ​​the dam zoning results through the dam risk assessment model to obtain the basic probability allocation value of the area. The dam risk assessment model is a model built based on cloud model and evidence theory. The result generation module is used to obtain risk assessment results based on the basic probability allocation values; The data conversion module is further used to classify the risk of single-point measured data in the same area of ​​the dam zoning results using a dam risk assessment model, thereby obtaining a single-point risk classification system; based on the single-point risk classification system, an improved forward cloud emitter is generated by determining the characteristic parameters of the cloud model, wherein the improved forward cloud emitter introduces hyperentropy correction; the single-point measured data is input into the forward cloud emitter for processing to obtain the membership degree of the risk level corresponding to the safety level value of each measuring point; the membership degree is converted into the basic probability allocation value of the evidence theory; and the basic probability allocation values ​​in the same area are fused to obtain the regional basic probability allocation value.

7. A dam risk assessment device based on multi-source data fusion, characterized in that, The dam risk assessment device based on multi-source data fusion includes: a memory, a processor, and a dam risk assessment program based on multi-source data fusion stored in the memory and executable on the processor. When the dam risk assessment program based on multi-source data fusion is executed by the processor, it implements the dam risk assessment method based on multi-source data fusion as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores a dam risk assessment program based on multi-source data fusion, which, when executed by a processor, implements the dam risk assessment method based on multi-source data fusion as described in any one of claims 1 to 5.

9. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the dam risk assessment method based on multi-source data fusion as described in any one of claims 1 to 5.

Images