Dam safety online monitoring system and method based on multi-source information fusion

Abstract

The invention relates to the field of dam safety monitoring, in particular to a dam safety online monitoring system and method based on multi-source information fusion, a digital twinborn center module is used for constructing a physical dam and digital model real-time mapping system, serving as a hub to push data and reference values, receiving feedback and performing iterative optimization; the data acquisition module is used for acquiring environment quantity, measured values and patrol inspection data and constructing an effective data set; the monitoring identification module is used for monitoring risk defects based on the effective data set and a central real-time simulation reference value, and judging an abnormal level according to a four-level rule; the structure rechecking module is used for establishing a structure model by combining a finite difference method and a PINN algorithm based on the central real-time digital twinborn body, and outputting a rechecking result; and the comprehensive early warning module is used for optimizing an early warning threshold value through a GA algorithm in combination with the abnormal level, the rechecking result and central mapping state evaluation, performing real-time early warning and triggering a response mechanism. Therefore, the problems of data fragmentation, single anomaly identification and the like in the prior art are solved.

Description

Technical Field

[0001] This invention relates to the field of dam safety monitoring, specifically to an online dam safety monitoring system and method based on multi-source information fusion. Background Technology

[0002] Driven by both the continuous improvement of the water conservancy project safety management system and the increasing complexity of dam operation risks, online dam safety monitoring has become a core component in ensuring basin flood control safety, rational water resource allocation, and the stability of people's livelihoods. During dam operation, multiple safety hazards must be addressed simultaneously. At the environmental level, real-time monitoring of upstream water level fluctuations, rainfall changes, abnormal inflows, and external risks such as earthquakes is necessary. At the structural level, continuous monitoring of key indicators such as dam displacement, dam foundation seepage pressure, joint deformation, and dam shoulder stability is required. Any abnormality in any of these aspects could lead to dam crack expansion, dam foundation piping, or even more serious safety accidents. Meanwhile, current dam monitoring is gradually moving towards automation and high-frequency operation. Data from measurement points such as displacement, seepage pressure, and stress are rapidly collected, and multi-source information such as environmental parameters and inspection data is generated simultaneously. Traditional decentralized monitoring models can no longer meet the core requirements of multi-source information fusion evaluation and timely monitoring and early warning. Therefore, constructing an online monitoring system that can integrate multi-source data, identify anomalies in real time, and provide precise hierarchical early warnings has become an inevitable choice for improving dam safety management capabilities.

[0003] However, the existing dam safety monitoring system still has many shortcomings and is difficult to adapt to the standardized and refined monitoring needs. On the one hand, data processing is fragmented. Traditional monitoring stores environmental quantities, measurement data, and inspection information in different systems, lacking a unified data validity identification mechanism. It cannot automatically trigger error judgment and mathematical model evaluation to filter valid data, nor can it quickly correlate other monitoring quantities at the same location to determine the scope of the anomaly when an anomaly occurs at a certain measuring point. Moreover, the mode of relying on manual review of suspected invalid data is inefficient and seriously affects the smooth progress of the monitoring process. On the other hand, the anomaly identification and early warning system has limitations. Traditional single mathematical models cannot adapt to the needs of multi-model collaborative identification and are prone to misjudgment in some scenarios where the patterns of monitoring quantities are not significant. At the same time, the early warning does not achieve multi-layer integration of the monitoring layer, the inspection layer, and the structural safety layer. It only relies on a single type of data to carry out early warning, ignoring important structural information such as deformation of key parts of the dam foundation and cracks in the corridor, resulting in insufficient accuracy of early warning. It cannot meet the progressive reasoning and evaluation requirements from monitoring measuring points to monitoring projects, monitoring locations, and then to the dam as a whole, and cannot provide comprehensive and reliable technical support for dam safety decision-making. Summary of the Invention

[0004] This application provides an online monitoring system and method for dam safety based on multi-source information fusion to solve problems such as data fragmentation, single anomaly identification, and lack of early warning levels in the prior art.

[0005] The first aspect of this application provides an online monitoring system for dam safety based on multi-source information fusion, comprising: a digital twin central module, a data acquisition module, a monitoring and identification module, a structural verification module, and a comprehensive early warning module; wherein, the digital twin central module is used to construct a real-time mapping system between the physical dam and the digital model, and to fuse multi-source real-time data to correct the model through a data assimilation algorithm, serving as a hub to push data and benchmark values, receive feedback, and perform iterative optimization; the data acquisition module is used to collect environmental quantity data, measured data, and inspection data, transmit them to the digital twin central module for standardized storage, and identify invalid data through a two-layer identification model and adaptive verification. The data, combined with suspected data recall and manual review, constructs an effective dataset and sends it back; the monitoring and identification module, based on the effective dataset and the central real-time simulation benchmark, monitors risk defects using the distribution law method and the improved envelope domain method, and determines the anomaly level according to a four-level rule; the structure review module, based on the central real-time digital twin, establishes a structure model using the finite difference method and the PINN algorithm, outputs the review results and feeds them back to the central system to update risk labels; the comprehensive early warning module, combined with the anomaly level, review results, and central mapping status assessment, optimizes the early warning threshold using the GA algorithm, provides real-time early warning, and clarifies the reporting level according to a five-level rule to trigger the response mechanism.

[0006] Preferably, the digital twin central module includes a physical digital mapping unit, a data assimilation and correction unit, a data hub unit, and an iterative optimization unit. The physical digital mapping unit is used to construct a digital model of the monitored object, matching the geometric and mechanical characteristics of the physical dam. The data assimilation and correction unit is used to fuse multi-source real-time data to correct model parameters. The data hub unit is used to receive standardized data, push real-time simulation benchmark values, and receive module feedback. The iterative optimization unit is used to update the model by combining structural verification deviations and early warning feedback.

[0007] Preferably, the data acquisition module includes a multi-source data acquisition unit, a data standardization unit, a two-layer invalid identification unit, and a valid dataset construction unit. The multi-source data acquisition unit acquires environmental quantity data, measurement data, and inspection data. The data standardization unit converts the data into a unified format and stores it. The two-layer invalid identification unit identifies invalid data through error judgment, statistical regression, and low-probability methods. The valid dataset construction unit combines suspected data recall with manual review to filter valid data.

[0008] Preferably, the monitoring and identification module includes a benchmark comparison unit, a multi-method anomaly identification unit, a four-level anomaly determination unit, and a result push unit. The benchmark comparison unit calculates the difference between the measured value and the centrally pushed benchmark value. The multi-method anomaly identification unit identifies risks and defects using distribution pattern methods and improved envelope domain methods. The four-level anomaly determination unit determines the status as normal, minor anomaly, general anomaly, and severe anomaly. The result push unit pushes anomaly results and triggers a preliminary alarm.

[0009] Preferably, the structural verification module includes a model calling unit, a multi-method structural calculation unit, a calculation result verification unit, and a result feedback unit. The model calling unit is used to acquire the central digital model and optimize the mesh. The multi-method structural calculation unit is used to calculate the dam stress and stability safety factor using the finite difference method, limit equilibrium method, arch-beam load distribution method, and PINN algorithm. The calculation result verification unit is used to compare the calculated values ​​with the measured values ​​and adjust the parameters. The result feedback unit is used to provide feedback on the verification results and update the risk labels.

[0010] Preferably, the comprehensive early warning module includes a multi-dimensional evaluation unit, a five-level early warning determination unit, and a graded reporting unit. The multi-dimensional evaluation unit integrates anomaly levels, verification results, and central mapping status, combining digital mapping status to evaluate the overall performance of the dam. The five-level early warning determination unit determines normal, Level IV, Level III, Level II, and Level I early warning statuses based on environmental risk and dam safety early warning standards. The graded reporting unit triggers corresponding measures such as encrypted observation, on-site inspection, and emergency response according to early warning information reporting rules.

[0011] The second aspect of this application provides a method for online monitoring of dam safety based on multi-source information fusion, comprising: acquiring environmental quantity data, measured data, and inspection data; simultaneously constructing a digital twin model of the physical dam; identifying invalid data through a two-layer identification model and adaptive verification based on the environmental quantity data, measured data, and inspection data; constructing an effective dataset by combining suspected data recall and manual verification; driving initial calibration of the twin model; monitoring risks and defects based on the effective dataset and real-time simulated benchmark values ​​pushed by the digital twin center using the distribution law method and improved envelope domain method; calculating the difference between the measured value and the benchmark value, determining the anomaly level according to a four-level rule; simultaneously, calling the real-time model of the digital twin center, calculating the dam stress and stability safety factor by combining the finite difference method and PINN algorithm, generating structural verification results and updating risk labels; and optimizing the early warning threshold through the GA algorithm based on the anomaly level and verification results, combined with the real-time mapping state assessment of the digital twin model, issuing real-time early warnings, clarifying the reporting level according to a five-level rule, and triggering a response mechanism.

[0012] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for online monitoring of dam safety based on multi-source information fusion as described in the above embodiments.

[0013] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for online monitoring of dam safety based on multi-source information fusion as described in the above embodiments.

[0014] The fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method for online monitoring of dam safety based on multi-source information fusion as described in the above embodiments.

[0015] Therefore, this application has the following beneficial effects: This application embodiment constructs a real-time mapping system between a physical dam and a digital model through a digital twin central module. It combines a data assimilation algorithm with multi-source real-time data from environmental quantities, measurements, and inspections to correct the model. This, along with the data acquisition module's two-layer invalid identification and valid dataset construction, effectively addresses issues such as scattered data storage and invalid data interference, improving the reliability and correlation analysis capabilities of monitoring data in complex operating environments. Furthermore, relying on the monitoring identification module's distribution pattern method, improved envelope domain method, and other methods for collaborative identification, combined with an attention mechanism to enhance key measurement point feature extraction, it improves anomaly identification accuracy in scenarios with weak measurement patterns and local structural anomalies. Simultaneously, a structural verification module... The finite difference method, limit equilibrium method, and PINN algorithm are integrated for calculation, combined with real-time state mapping using a digital twin model, to perform spatiotemporal correlation analysis on parameters such as dam stress and stability safety factor, providing multi-dimensional data support for dam safety assessment. The comprehensive early warning module optimizes early warning thresholds through the GA algorithm, and, in conjunction with a five-level reporting and response mechanism, can respond in real-time to anomaly levels and structural verification results, triggering corresponding response measures. This effectively avoids the escalation of risks due to delayed early warnings, controls the misjudgment and omission rates in dam safety monitoring, improves the timeliness and accuracy of dam operation safety management, reduces manual monitoring costs, and ensures the stability of watershed flood control and water resource allocation. This addresses the problems of data fragmentation, single anomaly identification, and lack of early warning levels in existing technologies.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the structure of an online monitoring system for dam safety based on multi-source information fusion, according to an embodiment of this application. Figure 2 This is a schematic diagram of a digital twin central module provided according to an embodiment of this application; Figure 3 This is a schematic diagram of a data acquisition module provided according to an embodiment of this application; Figure 4 This is a schematic diagram of a monitoring and identification module provided according to an embodiment of this application; Figure 5 This is a schematic diagram of a structure verification module provided according to an embodiment of this application; Figure 6 This is a schematic diagram of a comprehensive early warning module provided according to an embodiment of this application; Figure 7 This is a schematic diagram of an online monitoring system for dam safety based on multi-source information fusion, according to an embodiment of this application. Figure 8 This is a flowchart illustrating an online monitoring method for dam safety based on multi-source information fusion, according to an embodiment of this application. Figure 9 This is a schematic diagram of an online monitoring method for dam safety based on multi-source information fusion according to an embodiment of this application; Figure 10 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The following description, with reference to the accompanying drawings, illustrates an online dam safety monitoring system and method based on multi-source information fusion, according to an embodiment of this application. Addressing the issue of singular anomaly identification mentioned in the background section, this application provides an online dam safety monitoring system based on multi-source information fusion. In this system, a real-time mapping system between the physical dam and its digital model is constructed through a digital twin central module. A data assimilation algorithm is used to fuse environmental quantities, measurements, and real-time data from inspections to correct the model. This, combined with a two-layer invalid identification and valid dataset construction by the data acquisition module, effectively addresses issues such as scattered data storage and invalid data interference, improving the reliability and correlation analysis capabilities of monitoring data under complex operating environments. Furthermore, relying on the distribution pattern method, improved envelope domain method, and other methods of collaborative identification in the monitoring identification module, and combining an attention mechanism to strengthen the extraction of key measurement point features, this system addresses issues such as weak patterns in measurements and local... In scenarios involving structural anomalies, the accuracy of anomaly identification is improved. Simultaneously, the structural verification module integrates the finite difference method, limit equilibrium method, and PINN algorithm for calculation, combined with real-time state mapping using a digital twin model. This enables spatiotemporal correlation analysis of parameters such as dam stress and stability safety factor, providing multi-dimensional data support for dam safety assessment. The comprehensive early warning module optimizes early warning thresholds through the GA algorithm, and, in conjunction with a five-level reporting and response mechanism, can respond in real-time to anomaly levels and structural verification results, triggering corresponding response measures. This effectively avoids the escalation of risks due to delayed early warnings, controls the misjudgment and omission rates in dam safety monitoring, improves the timeliness and accuracy of dam operation safety management, reduces manual monitoring costs, and ensures the stability of watershed flood control and water resource allocation. Therefore, it solves the problems of data fragmentation, single anomaly identification, and lack of early warning levels in existing technologies.

[0020] Figure 1 This is a schematic diagram of the structure of an online monitoring system for dam safety based on multi-source information fusion, provided as an embodiment of this application.

[0021] This application provides an online monitoring system for dam safety based on multi-source information fusion. The system 10 includes: The digital twin central module 100, data acquisition module 200, monitoring and identification module 300, structure verification module 400, and comprehensive early warning module 500 are included.

[0022] The digital twin central module 100 is used to construct a real-time mapping system between the physical dam and the digital model. It integrates multi-source real-time data through a data assimilation algorithm to correct the model, and serves as the hub to push data and benchmark values, receive feedback, and perform iterative optimization. The data acquisition module 200 is used to collect environmental quantity data, measurement data, and inspection data, and transmit them to the digital twin central storage for standardization. It identifies invalid data through a two-layer identification model and adaptive review, and constructs a valid dataset by combining suspected data recall and manual review, and then sends it back. The monitoring and identification module 300 monitors risks and defects based on the valid dataset and the central real-time simulated benchmark value, and uses the distribution law method and improved envelope domain method to determine the anomaly level according to the four-level rules. The structural review module 400 establishes a structural model based on the central real-time digital twin, combined with the finite difference method and PINN algorithm, and outputs the review results to the central to update the risk label. The comprehensive early warning module 500 combines the anomaly level, review results, and central mapping status assessment, optimizes the early warning threshold through the GA algorithm, provides real-time early warning, and clarifies the reporting level according to the five-level rules to trigger the response mechanism.

[0023] It is understood that in this embodiment, a real-time mapping system between the physical dam and the digital model is constructed through a digital twin central module. This is combined with a data assimilation algorithm to fuse environmental quantities, measurements, and multi-source real-time data from inspections to correct the model. This, along with the data acquisition module's dual-layer invalid identification and valid dataset construction, effectively addresses issues such as scattered data storage and invalid data interference, improving the reliability and correlation analysis capabilities of monitoring data in complex operating environments. Furthermore, relying on the monitoring identification module's distribution pattern method, improved envelope domain method, and other methods for collaborative identification, combined with an attention mechanism to strengthen the extraction of key measurement point features, the accuracy of anomaly identification is improved in scenarios with weak measurement patterns and local structural anomalies. Simultaneously, [the following text is incomplete and requires further context: "...combined with..."] The structural verification module integrates the finite difference method, limit equilibrium method, and PINN algorithm for calculation, combined with real-time state mapping using a digital twin model, to perform spatiotemporal correlation analysis on parameters such as dam stress and stability safety factor, providing multi-dimensional data support for dam safety assessment. The comprehensive early warning module optimizes early warning thresholds through the GA algorithm, and, in conjunction with a five-level reporting and response mechanism, can respond in real-time to anomaly levels and structural verification results, triggering corresponding response measures. This effectively avoids the escalation of risks due to delayed early warnings, controls the misjudgment and omission rates in dam safety monitoring, improves the timeliness and accuracy of dam operation safety management, reduces manual monitoring costs, and ensures the stability of watershed flood control and water resource allocation. Therefore, it solves the problems of data fragmentation, single anomaly identification, and lack of early warning levels in existing technologies.

[0024] In this embodiment of the application, the digital twin central module 100 further includes: Figure 2 As shown, there are physical digital mapping unit, data assimilation correction unit, data hub unit and iterative optimization unit.

[0025] Among them, the physical digital mapping unit is used to construct a digital model of the monitored object and match the geometric and mechanical characteristics of the physical dam; the data assimilation and correction unit is used to integrate multi-source real-time data to correct model parameters; the data hub unit is used to receive standardized data, push real-time simulation benchmark values ​​and receive module feedback; and the iterative optimization unit is used to update the model by combining structural verification deviations and early warning feedback.

[0026] It should be noted that the formula for the digital model is:

[0027]

[0028] in, The dam's state in the digital model; for The measured data vector of the physical dam at any given time; It is a multiphysics mapping model; This is the model parameter vector; This is a dynamic correction item; For prediction The state of the digital twin at any given moment; Let be the state evolution function; for The state quantity of the twin at any given moment; for The rate of change of the twin's state at any given time; for The external excitation vector at time t; For the differential time element.

[0029] It is understood that the digital mapping unit in this application embodiment constructs a digital model of the monitored object, accurately matching the geometric and mechanical characteristics of the physical dam. This allows for the complete reproduction of the dam's structural morphology, foundation geological conditions, and material mechanical parameters, providing a high-precision model foundation for subsequent multi-source data correlation analysis and structural calculations. This ensures the consistency between the digital model and the physical dam, avoiding monitoring deviations caused by model distortion. The data assimilation and correction unit corrects model parameters by fusing multi-source real-time data, dynamically integrating environmental quantities, measurements, and inspection data. It calibrates key variables such as load conditions and material parameters in the digital model in real time, ensuring the model always closely matches the actual operating state of the dam and improving structural integrity. The accuracy of structural calculations and anomaly assessment is enhanced. The data hub unit, by receiving standardized data, pushing real-time simulation benchmark values, and receiving feedback from modules, can achieve unified storage, flow, and interaction of multi-source data. It also provides a unified benchmark reference for modules such as monitoring and identification, and structural verification, ensuring data consistency across modules and improving the continuity of the monitoring process. The iterative optimization unit, by combining structural verification deviations and early warning feedback to update the model, can continuously optimize model parameters and calculation logic based on actual monitoring results, gradually reducing model calculation deviations, enhancing the digital model's ability to predict dam safety status, promoting the dynamic upgrading of dam safety monitoring, and ensuring the reliability and effectiveness of long-term monitoring.

[0030] For example, in the online safety monitoring practice of a concrete double-curvature arch dam (maximum dam height 210m, dam crest elevation 1135m, dam crest arch width 12m, dam base arch width 63m) at a hydropower station, the specific applications and data are as follows: First, based on the dam construction drawings, geological survey report, and material mechanical parameters, a digital twin model matching the physical dam at a 1:1 scale was built. The model accurately reproduced the cross-sectional curve parameters of the arch crown beam (radius of curvature 810m, central angle 85°) and the dam segment joint spacing (15m / segment). Simultaneously, C18036 concrete (compressive strength 12.14MPa, elastic modulus 3.5×10⁻⁶) was imported. 4 C18030 concrete (compressive strength 11.32 MPa, elastic modulus 3.2 × 10⁻⁶ MPa), C18030 concrete (compressive strength 11.32 MPa, elastic modulus 3.2 × 10⁻⁶ MPa). 4Mechanical data, including MPa) and fresh granite deformation modulus of the dam foundation (50 GPa), were collected daily to ensure high consistency between the model and the physical dam in terms of structure and mechanical properties. During the flood season, multi-source data were collected daily, including upstream water level (fluctuation range 1080-1120 m, maximum daily increase 5 m), dam displacement (daily average radial displacement of PL14-1 measuring point of the arch crown 0.2 mm, maximum daily change 0.5 mm), dam foundation seepage pressure (uplift pressure of TC4-8 measuring point 0.35 MPa, fluctuation ±0.02 MPa), and ambient temperature (daily average 25-32℃, diurnal temperature difference 8℃). The model parameters were corrected using a data assimilation algorithm, reducing the initial displacement calculation deviation of 1.5 mm to within 0.3 mm. The unified data hub received 31 horizontal displacement, 76 vertical displacement, and 42 seepage pressure measurements daily. The system processes standardized data from various points (12,000 records per day), pushes simulated values ​​such as the radial displacement benchmark value of 116.15 mm for PL14-1 measuring point and the uplift pressure benchmark value of 0.3 MPa for TC4-8 measuring point, and receives anomaly feedback in real time (e.g., PL14-3 measuring point exceeds the benchmark value by 2 mm, and the calculated value of the dam abutment stability safety factor of 3.3 deviates from the standard allowable value of 3.5 by 0.15). Combining structural verification deviations and Level IV early warnings (slight seepage at the dam abutment, flow rate of 0.02 L / min), the internal friction angle of the dam foundation rock mass shear strength is adjusted from 45° to 44.2°, the safety factor is corrected to 3.52, and the temperature load influence coefficient is optimized (from 0.05 to 0.045), reducing the frequency of similar early warnings by 60%, fully meeting the technical requirements of multi-source information fusion, structural verification, and iterative optimization in the report.

[0031] In this embodiment of the application, the data acquisition module 200 includes: Figure 3 As shown, there are multi-source data acquisition unit, data standardization unit, two-layer invalid identification unit, and effective dataset construction unit.

[0032] The multi-source data acquisition unit is used to acquire environmental quantity data, measurement data, and inspection data; the data standardization unit is used to convert the data into a unified format and store it; the two-layer invalid identification unit is used to identify invalid data through error judgment, statistical regression, and low probability methods; and the effective dataset construction unit is used to combine suspected data recall and manual review to screen effective data.

[0033] It is understood that the multi-source data acquisition unit in this application embodiment, by acquiring environmental quantity data, measurement data, and inspection data, can comprehensively cover the external operating conditions, structural response, and on-site defects of the dam, providing complete data support for safety analysis and avoiding one-sided safety assessments. The data standardization unit converts the three types of data into a unified format for storage, eliminating format differences and ensuring that data can be analyzed across types, improving flow and calculation efficiency. The dual-layer invalid identification unit identifies invalid data by combining error judgment with statistical regression and low-probability methods, screening abnormal data from two dimensions, reducing misjudgments by a single method, and improving data credibility. The effective dataset construction unit combines suspected data recall and manual review to screen effective data, performing secondary verification on suspected invalid data, avoiding algorithmic misjudgments that remove effective data, ensuring the accuracy and completeness of the dataset used for safety monitoring, and laying a reliable foundation for subsequent anomaly identification and structural review.

[0034] For example, in the online safety monitoring of a concrete double-curvature arch dam, the multi-source data acquisition unit simultaneously collects three types of core data: environmental data covering upstream water level (daily average monitoring value 1080-1120m, collection frequency 1 time / 10 minutes) and rainfall (daily maximum monitoring value 95mm, collection frequency 1 time / hour); measurement data including 31 horizontal displacement measuring points (e.g., the daily average radial displacement change of 0.2mm at the PL14-1 measuring point of the arch crown), 76 vertical displacement measuring points (e.g., the vertical displacement of the TC4-8 measuring point is stable at -19.9mm), and 42 seepage pressure measuring points (e.g., the daily average uplift pressure of the dam foundation measuring point is 0.35MPa); and inspection data recording the on-site conditions such as dam surface cracks (3 locations with a length ≤5cm) and dam shoulder seepage (flow rate ≤0.02L / min), comprehensively covering the key data dimensions of dam operation. The data standardization unit converts three types of data into a unified format for storage, standardizing displacement data to millimeters, pressure data to megapascals, and timestamps to "year-month-day hour:minute:second" format. It processes an average of 12,000 data entries daily, eliminating format differences between different devices and ensuring data can be directly used for cross-type correlation analysis, improving data flow efficiency by 40%. The dual-layer invalidity identification unit filters out two invalid data entries through error judgment (one displacement data entry exceeding the instrument's range by 5mm, and one seepage pressure data entry violating the physical logic of "increasing seepage pressure with rising water level"), and then uses statistical regression and low-probability methods to identify three abnormal data entries (displacement residual exceeding 3σ, and seepage pressure measurement values ​​exhibiting low-probability deviations). This dual-dimensional elimination of invalid data improves data reliability to 98.5%. The effective dataset construction unit conducted a recall test on 5 suspected invalid data points (re-collecting displacement measuring points 3 times and seepage pressure measuring points 2 times), combined with manual on-site verification (confirming 1 data point as an instrument malfunction and 4 data points as temporary interference), and finally screened out 11,987 valid data points, ensuring that the dataset used for subsequent anomaly identification and structural verification is accurate and complete, and providing a reliable data foundation for dam safety monitoring.

[0035] In this embodiment of the application, the monitoring and identification module 300 includes: Figure 4 As shown, the system includes a benchmark comparison unit, a multi-method anomaly identification unit, a four-level anomaly judgment unit, and a result push unit.

[0036] The benchmark comparison unit is used to calculate the difference between the measured value and the benchmark value pushed by the central system; the multi-method anomaly identification unit is used to identify risk defects through the distribution law method and the improved envelope domain method; the four-level anomaly judgment unit is used to judge the status according to normal, minor anomaly, general anomaly and serious anomaly; and the result push unit is used to push the abnormal results and trigger the initial alarm.

[0037] It is understood that, through the benchmark comparison unit, the deviation of the monitoring data from the theoretically reasonable range can be intuitively quantified by calculating the difference between the measured value and the benchmark value pushed by the central system. This allows for the rapid location of measurement points that exceed normal fluctuations, avoiding omissions of anomalies due to a lack of quantitative comparison. The multi-method anomaly identification unit identifies risks and defects using the distribution law method and the improved envelope domain method. This adapts to different monitoring data characteristics, reduces the limitations of a single method in adapting to complex working conditions, and improves the comprehensiveness and accuracy of risk and defect identification. The four-level anomaly judgment unit determines the status according to normal, minor anomaly, general anomaly, and severe anomaly. It can classify and define the dam safety status based on the degree of deviation of the monitoring data and the scope of the defect's impact, avoiding risk misjudgment or over-response caused by ambiguous judgment. The result push unit pushes the anomaly results and triggers a preliminary alarm, which can promptly transmit the anomaly information to the corresponding control level, quickly initiate the subsequent verification process, shorten the anomaly response lag time, and buy a window of opportunity for the early handling of dam safety risks.

[0038] For example, in the online safety monitoring of a concrete arch dam, the benchmark comparison unit receives benchmark values ​​from various measuring points pushed by the central control (such as the benchmark value of horizontal displacement at the dam crest being 116.15 mm and the benchmark value of seepage pressure at the dam foundation being 0.3 MPa), and calculates the difference between the measured value and the benchmark value in real time. For instance, the measured radial displacement at measuring point PL14-3 of the arch crown on a certain day was 118.3 mm, which differed from the benchmark value by 2.15 mm; the measured seepage pressure at measuring point TC4-8 of the dam foundation was 0.38 MPa, which differed from the benchmark value by 0.08 MPa, thus intuitively quantifying the degree of data deviation. The multi-method anomaly identification unit selects corresponding methods for different data characteristics: For the horizontal displacement measuring point group of dam section 10#-19#, the correlation between measuring points is calculated using the distribution law method. It was found that the correlation between PL14-3 and its adjacent measuring point decreased from 0.92 to 0.65, indicating the existence of a local anomaly; For the uplift pressure data of the dam foundation, the normal range (0.28-0.32MPa) is constructed using the improved envelope domain method, identifying that the seepage pressure of measuring point TC4-8 exceeds the envelope domain, accurately locating two risk defects. The level four anomaly judgment unit combines the difference and the defect impact: The displacement difference of PL14-3 is 2.15mm and the trend is converging, which is judged as a slight anomaly; the seepage pressure of TC4-8 exceeds the envelope domain by 0.06MPa and continues to rise, which is judged as a general anomaly; The differences of the remaining measuring points are all within 0.5mm or 0.02MPa, which are judged as normal, with no serious anomalies. The result push unit immediately pushes the abnormal results to the production technology department management personnel, simultaneously triggering a blue alert (minor abnormality) and a yellow alert (general abnormality), along with the measurement point location, difference data, and the basis for abnormality judgment, prompting maintenance personnel to initiate on-site verification within 2 hours, effectively shortening the abnormality response time.

[0039] In this embodiment of the application, the structural verification module 400 includes, as follows: Figure 5 The model calling unit, multi-method structure calculation unit, calculation result verification unit, and result feedback unit are shown.

[0040] The model call unit is used to acquire the central digital model and optimize the mesh; the multi-method structure calculation unit is used to calculate the dam stress and stability safety factor through the finite difference method, limit equilibrium method, arch beam load distribution method and PINN algorithm; the calculation result verification unit is used to compare the calculated values ​​with the measured values ​​and adjust the parameters; and the result feedback unit is used to provide feedback on the verification results and update the risk labels.

[0041] It is understood that the embodiments of this application, through the model calling unit, can accurately reproduce the structural details of the dam by acquiring the central digital model and optimizing the mesh, thereby improving the model's calculation accuracy of dam stress and deformation, avoiding structural calculation deviations caused by coarse meshes, and providing a high-quality model foundation for subsequent multi-method structural calculations. The multi-method structural calculation unit calculates dam stress and stability safety factors through the finite difference method, limit equilibrium method, arch-beam load distribution method, and PINN algorithm, which can cover key safety indicators of dam structures from different mechanical analysis dimensions, adapt to the calculation needs of different dam types and working conditions, and reduce the reliance on single methods. The calculation limitations of complex structures are addressed to improve the comprehensiveness of structural safety assessments. The calculation result verification unit adjusts parameters by comparing calculated and measured values, calibrating the material mechanics parameters in the calculation model based on actual dam monitoring data. This reduces the deviation between theoretical calculations and actual working conditions, ensuring the reliability of structural calculation results. The result feedback unit, by providing feedback on verification results and updating risk labels, can promptly synchronize structural verification conclusions to the safety monitoring center, dynamically adjusting the dam's risk level. This provides accurate structural safety basis for subsequent graded early warning and emergency response, ensuring the timeliness and targeted nature of dam operation safety management.

[0042] For example, in the structural verification of a concrete hyperbolic arch dam, the model calling unit obtained a 1:1 digital twin model of the dam from the monitoring center, optimized the mesh for the key areas of the dam abutment (left and right bank arch seats), adjusted the original mesh size from 5m×5m to 2m×2m, refined the rock mass fissures and dam body joint structure, and made the mesh fit the actual geological and construction conditions, providing high-precision model support for subsequent structural calculations, and effectively reducing the stress calculation deviation caused by the coarse mesh (the deviation range was reduced from the original 8% to less than 3%). The multi-method structural calculation unit selects corresponding algorithms for different structural safety indicators of the dam: the finite difference method is used to calculate the dam stress, and the maximum compressive stress of the dam body at the 1135m elevation of the arch crown beam is found to be 12.0MPa; the limit equilibrium method is used to calculate the stability safety factor of the dam abutment, and the result under the basic combination is 3.45; the load distribution ratio of the arch dam is calculated using the arch-beam load distribution method, with the arch bearing 75% of the load and the beam bearing 25% of the load; the stress calculation results are corrected by combining the PINN algorithm, and the calculated value of the local stress concentration area is optimized to 11.8MPa, comprehensively covering the key indicators of dam stress and stability safety. The calculation result verification unit compared the calculated values ​​with the measured values ​​using different methods: the dam stress calculated by the finite difference method (12.0 MPa) was compared with the measured value (11.7 MPa) at the dam stress-strain measurement point, with a difference of 0.3 MPa; the abutment stability safety factor calculated by the limit equilibrium method (3.45) was compared with the allowable value (3.5) in the specification, with a deviation of 0.05; accordingly, the elastic modulus of the dam foundation rock mass in the model was adjusted (from 50 GPa to 48 GPa) and the internal friction angle (slightly adjusted from 45° to 44.8°), reducing the deviation between the calculated and measured values ​​to within 0.1 MPa, and the safety factor was corrected to 3.52, which meets the specification requirements. The result feedback unit fed back the structural verification results (dam stress qualified, abutment stability safety factor met the standard) to the safety monitoring center, and simultaneously updated the original "pending verification" risk label to "low risk", providing a structural safety basis for subsequent graded early warning, ensuring that dam operation and maintenance personnel are clear about the current structural status, and avoiding over-control or risk omission.

[0043] In this embodiment, the integrated early warning module 500 includes, for example: Figure 6 As shown, there are multi-dimensional evaluation units, five-level early warning judgment units, and hierarchical reporting units.

[0044] Among them, the multi-dimensional evaluation unit is used to integrate the anomaly level, the verification results and the central mapping status, and evaluate the overall performance of the dam in combination with the digital mapping status; the five-level early warning judgment unit is used to determine the normal, level IV, level III, level II and level I early warning status according to the environmental risk and dam safety early warning standards; the graded reporting unit triggers corresponding measures such as encrypted observation, on-site inspection and emergency response according to the early warning information reporting rules.

[0045] It can be understood that the multi-dimensional evaluation unit in the embodiments of the present application integrates the anomaly level, review result, and central mapping status, and combines the digital mapping status to evaluate the overall behavior of the dam. It can break through the limitations of single indicators, comprehensively judge the dam status from multiple dimensions such as the degree of anomaly, structural safety, and digital model feedback, avoid overall misjudgment caused by local data anomalies, and improve the systematicness of safety evaluation; the five-level early warning determination unit determines the normal, level-IV, level-III, level-II, and level-I early warning status according to environmental risks and dam safety early warning standards, and can achieve hierarchical control according to the severity of risks, avoid insufficient response or resource waste caused by ambiguous early warning levels, and ensure that the early warning matches the actual risks; the hierarchical reporting unit triggers corresponding measures such as encrypted observation, on-site inspection, and emergency response according to the early warning information reporting rules, and can accurately associate the early warning level with the disposal means. For example, level-IV early warning starts encrypted observation, and level-I early warning triggers an emergency response, shortening the risk disposal chain and improving the pertinence and timeliness of dam safety control.

[0046] For example, in the safety control of a certain concrete gravity dam, the multi-dimensional evaluation unit integrates three types of core information: anomaly level (2 seepage pressure measurement points at the dam foundation are determined to be slightly abnormal, and 1 displacement measurement point is determined to be generally abnormal), structural review result (the calculated value of the dam body stress is 11.2 MPa, meeting the C18030 concrete compressive standard of 11.32 MPa; the slope stability safety factor is 3.4, meeting the allowable value of the basic combination in the specification of 3.5), and central digital mapping status (the model shows that the deviation between the dam body deformation and the measured value is 0.2 mm, and the fitting degree reaches 99%). Combining the overall stress distribution of the dam body by digital mapping, the overall behavior of the dam is comprehensively evaluated as "there are slight risks locally, and the overall safety is controllable", avoiding misjudging the overall safety status only based on a single abnormal measurement point. The five-level early warning determination unit analyzes according to the early warning standard: in terms of environmental risks, the upstream water level is 1090 m (lower than the design flood level of 1100 m), and the single-day rainfall is 45 mm (lower than the rainstorm threshold of 50 mm), so there is no environmental risk; in terms of dam safety, combining the multi-dimensional evaluation results, the proportion of slightly abnormal measurement points is 5%, and the proportion of generally abnormal measurement points is 2.5%, which does not reach the high-level early warning threshold. Finally, it is determined as level-IV early warning (blue early warning), neither underestimating local risks nor over-upgrading the early warning level. The hierarchical reporting unit triggers corresponding measures according to the rules: pushes the level-IV early warning result to the management personnel of the production technology department and the power station person in charge, synchronously triggers encrypted observation (the collection frequency of seepage pressure measurement points is increased from 1 time / hour to 1 time / 10 minutes, and the displacement measurement points are increased from 1 time / 2 hours to 1 time / 30 minutes), and arranges maintenance personnel to conduct on-site inspections within 24 hours (focusing on verifying cracks and water seepage conditions around the abnormal measurement points). By accurately associating the early warning level with the disposal measures, while ensuring the safety of the dam, it avoids resource waste and improves the risk control efficiency.

[0047] This application proposes an online dam safety monitoring system based on multi-source information fusion. This system constructs a real-time mapping system between the physical dam and its digital model through a digital twin central module. It combines a data assimilation algorithm with multi-source real-time data from environmental quantities, measurements, and inspections to correct the model. This, along with a data acquisition module's two-layer invalid identification and valid dataset construction, effectively addresses issues such as scattered data storage and invalid data interference, improving the reliability and correlation analysis capabilities of monitoring data in complex operating environments. Furthermore, relying on the monitoring identification module's collaborative identification using multiple methods such as the distribution pattern method and the improved envelope domain method, and combining an attention mechanism to strengthen the extraction of key measurement point features, this system improves performance in scenarios with weak measurement patterns and local structural anomalies. The anomaly identification accuracy is improved. Simultaneously, the structural verification module integrates the finite difference method, limit equilibrium method, and PINN algorithm for calculation, combined with real-time state mapping using a digital twin model, to perform spatiotemporal correlation analysis on parameters such as dam stress and stability safety factor, providing multi-dimensional data support for dam safety assessment. The comprehensive early warning module optimizes early warning thresholds through the GA algorithm, and, in conjunction with a five-level hierarchical reporting and response mechanism, can respond in real time to anomaly levels and structural verification results, triggering corresponding response measures. This effectively avoids the escalation of risks due to delayed early warnings, controls the misjudgment and omission rates in dam safety monitoring, improves the timeliness and accuracy of dam operation safety management, reduces manual monitoring costs, and ensures the stability of watershed flood control and water resource allocation. Therefore, it solves the problems of data fragmentation, single anomaly identification, and lack of early warning levels in existing technologies.

[0048] The following will illustrate a specific embodiment of an online dam safety monitoring system based on multi-source information fusion, such as... Figure 7 As shown, it includes: In the online safety monitoring practice of a concrete double-curvature arch dam (maximum dam height 210m, crest elevation 1135m, crest arch width 12m, base arch width 63m), the digital twin central module first constructs a real-time mapping system between the physical dam and the digital model. Based on the arch dam construction drawings, the dam foundation geological survey report (including the distribution of fresh granite and the characteristics of rock mass fissure development), and material mechanical parameters, a 1:1 digital twin model is built, accurately importing the concrete mechanical parameters (C18036 concrete compressive strength 12.14MPa, elastic modulus 3.5×10⁻⁶) for different regions of the dam body. 4 MPa; C18030 concrete compressive strength 11.32 MPa, elastic modulus 3.2 × 10⁻⁶ MPa. 4The model incorporates parameters such as MPa, deformation modulus of the dam foundation rock mass (50 GPa for fresh granite, 25 GPa for weathered granite), dam body joint spacing (15 m / segment), and arch crown beam profile curve parameters (radius of curvature 810 m, central angle 85°) to ensure that the model's geometric and mechanical properties are consistent with the physical dam height. During the flood season, the module uses a data assimilation algorithm to fuse multi-source real-time data to correct the model. It receives daily data on upstream water level (fluctuating range 1080-1120m, maximum daily increase 5m), dam displacement (among 31 horizontal displacement measuring points, the radial displacement of the PL14-1 measuring point on the arch crown changes by an average of 0.2mm daily), dam foundation seepage pressure (among 76 vertical displacement measuring points, the uplift pressure at the TC4-8 measuring point remains stable at 0.35MPa, fluctuating ±0.02MPa), and ambient temperature (average daily 25-32℃). Addressing the 1.5mm deviation between the radial displacement calculated by the initial model at the PL14-1 measuring point and the measured value, the elastic modulus of the C18036 concrete in the upper part of the dam body is adjusted to 3.42×10⁻⁶. 4 MPa, reducing the deviation to within 0.3mm. Simultaneously, as a data hub, the module pushes standardized benchmark values ​​to downstream modules daily, including the radial displacement benchmark value of 116.15mm at the PL14-1 measuring point at a water level of 1120m, the uplift pressure benchmark value of 0.3MPa at the TC4-8 measuring point, and the allowable value of 3.5 for the dam abutment stability safety factor specified in the standard. When the structural verification module reports that the calculated value of the dam abutment stability safety factor of 3.3 deviates from the standard value by 0.2, the friction angle within the dam foundation rock mass is adjusted from 45° to 44.2°, resulting in a safety factor of 3.52, achieving iterative optimization of the model.

[0049] The data acquisition module synchronously collects environmental data, measured data, and inspection data, comprehensively covering key information on dam operation. Environmental data collection is conducted as required: upstream water level is collected once every 10 minutes, with a data accuracy of ±0.01m; daily rainfall is collected once per hour, with a single measurement error ≤2%, and a maximum daily monitoring value of 95mm during the flood season; inflow is classified into early warning thresholds according to standards such as 100-year and 500-year return periods, with a real-time monitoring accuracy of ±5m³ / s. Measured data covers 28 horizontal displacement measuring points (e.g., the measured radial displacement of PL14-3 on a certain day was 115.1mm), 72 vertical displacement measuring points (e.g., the vertical displacement of TC3-6 was stable at -18.5mm), and 38 seepage pressure measuring points (e.g., the measured seepage pressure of TC4-8 was 0.36MPa). All measuring instruments operate stably, and the validity of the data stored is ≥98%. Inspection data was recorded according to the established process, including minor cracks on the dam surface (2 locations with a length ≤3cm and a width ≤0.1mm), slight seepage on the right bank abutment (flow rate 0.015L / min, clear water), and slight erosion of the concrete of the dam crest wave wall (area approximately 0.5㎡). The recording format was standardized as "Inspection Time-Location-Defect Type-Description-Photo Number". After data collection, the data was transmitted to the digital twin center for standardized processing, unifying displacement data to millimeters, pressure data to megapascals, and timestamps to "Year-Month-Day Hour:Minute:Second". An average of 12,000 data entries were processed daily, eliminating format differences between different devices. Subsequently, invalid data was filtered using a two-layer identification model: The first layer used an error judgment method to remove one horizontal displacement data point exceeding the instrument's range (154mm, range 150mm) and one data point that violated the logic of "increased seepage pressure with rising water level" (when the upstream water level rose by 2m, the measured value of a certain seepage pressure measuring point decreased by 0.03MPa); The second layer used a mathematical model evaluation method, identifying two abnormal data points with displacement residuals exceeding 3σ (residual values ​​3.2σ and 3.5σ) through a statistical regression model (multiple correlation coefficient 0.88), and identifying one low-probability deviation in seepage pressure data point (probability of occurrence 0.008) using a low-probability method (significance level 0.01). For the five suspected invalid data points, a recall was initiated (abnormal displacement measuring points were resampled 3 times, and seepage pressure measuring points were resampled 2 times) and manual verification was performed (confirming one case as a sensor malfunction and four cases as temporary electromagnetic interference). Finally, a valid dataset of 11,987 data points was constructed and transmitted back to the digital twin central processing unit.

[0050] The monitoring and identification module uses a valid dataset and real-time simulated benchmark values ​​pushed by the digital twin central control to identify risks and defects and determine anomaly levels. For the horizontal displacement monitoring point group of dam section 9#-18#, the correlation between monitoring points was calculated using the distribution law method. It was found that the correlation between monitoring point PL14-3 and the adjacent monitoring points PL14-2 and PL14-4 decreased from the historical average of 0.91 to 0.65, which is lower than the normal threshold of 0.8, indicating that there is a local anomaly at this monitoring point. For the dam foundation seepage pressure data, the improved envelope domain method was used to analyze the historical data of the past 3 years (after removing the maximum and minimum 10% of the data, the remaining data has a maximum value of 0.32MPa and a minimum value of 0.28MPa). The calculated envelope domain range was 0.27-0.33MPa (including a monitoring error of 0.02MPa). It was found that the measured seepage pressure of TC4-8 was 0.36MPa, which exceeded this range, and was identified as a seepage pressure anomaly. According to the four-level anomaly judgment rules, the radial displacement of measuring point PL14-3 differs from the benchmark value by 2.6 mm, and the change over three consecutive days has decreased from 0.5 mm to 0.2 mm (trend convergence). Furthermore, the measured value of 115.1 mm ≤ the control value of 133 mm, thus it is judged as a slight anomaly. The seepage pressure at measuring point TC4-8 exceeds the envelope range by 0.03 MPa, showing an upward trend over four consecutive days (daily increase of 0.01 MPa, trend divergence). The measured value of 0.36 MPa ≤ the control value of 0.4 MPa, thus it is judged as a general anomaly. For the remaining measuring points (such as PL14-1 and TC3-6), the difference between the measured values ​​and the benchmark values ​​is within 0.4 mm or 0.02 MPa, with no trend change, thus it is judged as normal. There are no seriously abnormal measuring points. The module synchronously records the anomaly identification criteria (correlation calculation results, envelope range, trend change curve) and pushes the anomaly results to the digital twin central control and comprehensive early warning module in real time.

[0051] The structural verification module uses a digital twin that is updated in real time by a digital twin centralization system, and outputs verification results by combining multiple calculation methods. It calls upon the centrally optimized mesh model (the mesh size in the key area of ​​the dam abutment arch seat is adjusted from 4m×4m to 1.8m×1.8m, refining the rock mass fissures and dam body joints), and uses the finite difference method to calculate the dam body stress. The maximum compressive stress at the 1135m elevation of the arch crown beam is found to be 11.8MPa, compared to the allowable compressive strength of 12.14MPa for the basic composite of C18036 concrete, a deviation of 0.34MPa. For the stress concentration area at the junction of dam section #14 and the dam foundation, the PINN algorithm is used for correction, reducing the compressive stress to 11.6MPa, with the deviation reduced to within 0.54MPa. The stability safety factor of the dam abutment was calculated using the limit equilibrium method. The initial value of 3.42 was lower than the allowable value of 3.5 for the basic combination (deviation 0.08). Combined with the slight seepage information of the right bank dam abutment fed back by the monitoring and identification module, the shear strength parameters of the dam foundation rock mass were checked. It was found that the internal friction angle of 45° deviated from the actual geological conditions. After adjusting it to 44.2°, the safety factor was recalculated and corrected to 3.52, which meets the specifications. At the same time, the seepage pressure reduction coefficient of the dam foundation was calculated. Three seepage pressure measuring points were selected in dam section 14, and the average value was 0.65 (within the normal range of 0.5-0.8), indicating that the seepage prevention effect of the dam foundation is good. The module feeds back the verification results (dam stress qualified, dam abutment stability met the standard, seepage pressure reduction coefficient normal) to the digital twin center. Based on the results, the center updates the original "pending verification" risk label to "low risk" and simultaneously pushes the verification report to the comprehensive early warning module.

[0052] The integrated early warning module combines anomaly levels, verification results, and central mapping status to conduct early warning and response. It integrates the anomaly levels from the monitoring and identification module (1 minor anomaly, 1 general anomaly, anomaly measurement points accounting for 2.8%), the pass / fail results from the structural verification module, and the model mapping status of the digital twin central system (model-measured deformation fit 98.8%, calculation deviation ≤0.3mm after parameter correction). A progressive reasoning mechanism is used to assess the overall dam behavior, determining that there are controllable risks in a localized area (anomalies limited to a single measurement point, with no spreading trend), and that the overall dam is safe and stable. The early warning threshold is optimized using the GA algorithm: the radial displacement threshold for measurement point PL14-3 is slightly adjusted from 133mm to 132mm to improve early warning sensitivity; the upper limit of the seepage pressure envelope for measurement point TC4-8 is optimized from 0.33MPa to 0.32MPa, allowing for earlier detection of anomaly trends. Combined with environmental risks (upstream water level 1110m < design flood level 1118m, daily rainfall 48mm < rainstorm threshold 50mm), the warning is classified as Level IV (blue). According to the tiered reporting rules, Level IV early warning information is pushed to relevant power plant management personnel and power plant managers in the Production Technology Department. The information includes the location of the abnormal monitoring point, the type of abnormality, the verification results, the warning level, and recommended measures. At the same time, the response mechanism is triggered: intensified monitoring is started (the data collection frequency of PL14-3 monitoring point is increased from once / hour to once / 10 minutes, and the frequency of TC4-8 monitoring point is increased from once / hour to once / 8 minutes). Maintenance personnel are arranged to conduct on-site verification within 20 hours (focusing on checking the cracks in the dam body around PL14-3 monitoring point and the seepage in the dam foundation gallery corresponding to TC4-8 monitoring point). Daily feedback of verification results and intensified monitoring data is required to form a closed-loop management of "early warning-response-feedback".

[0053] In summary, the embodiments of this application utilize a physical-digital real-time mapping system and data assimilation algorithm constructed through a digital twin central module. This system can dynamically adjust model parameters to match the actual operating conditions of the concrete double-curvature arch dam (such as water level fluctuations and temperature changes during the flood season). Combined with a benchmark value push and iterative optimization mechanism, it provides high-precision data support for each module, ensuring the accuracy of monitoring and analysis. Relying on the multi-source data acquisition, standardized processing, and two-layer invalid identification model of the data acquisition module, it can comprehensively cover environmental quantities, measurements, and inspection data, efficiently eliminating abnormal data. Combined with suspected data recall and manual verification, it constructs a high-efficiency dataset, laying a reliable data foundation for subsequent monitoring and identification. Furthermore, by leveraging the distribution pattern method and improved envelope domain of the monitoring and identification module… The system employs a four-level anomaly judgment rule to accurately identify risks and defects such as dam displacement and seepage pressure, enabling scientific definition of anomaly levels and avoiding misjudgments or omissions. Through the finite difference method, PINN algorithm, and limit equilibrium method in the structural verification module, key indicators such as dam stress and abutment stability safety factor can be accurately calculated. The results are then used to update risk labels, providing a quantitative assessment basis for dam structural safety. Simultaneously, the comprehensive early warning module integrates multi-dimensional information and optimizes early warning thresholds and the five-level reporting mechanism through the GA algorithm. This allows for rapid triggering of response measures such as encrypted observation and on-site verification, effectively shortening anomaly handling time, improving the timeliness and pertinence of dam safety management, reducing safety risks, and providing strong support for the long-term stable operation of the dam.

[0054] Secondly, referring to the accompanying drawings, an online monitoring method for dam safety based on multi-source information fusion is described according to an embodiment of this application.

[0055] like Figure 8 As shown, this online monitoring method for dam safety based on multi-source information fusion includes the following steps: In step S101, environmental quantity data, measurement data, and inspection data are acquired, and a digital twin model of the physical dam is constructed.

[0056] It is understood that the embodiments of this application can comprehensively cover key information on dam operation by synchronously acquiring environmental quantities, measurements, and inspection data, thus avoiding monitoring blind spots; at the same time, the construction of a digital twin model of the physical dam can accurately reproduce the structural characteristics and operating status of the dam, providing a concrete platform for subsequent analysis. The combination of the two provides a real working condition basis for the model and reserves an analysis path for data processing, effectively laying the foundation for monitoring.

[0057] In step S102, based on environmental quantity data, measured data, and inspection data, invalid data is identified through a two-layer identification model and adaptive verification. A valid dataset is constructed by combining suspected data recall and manual verification, and the twin model is initially calibrated.

[0058] Among them, the two-layer identification model refers to the model that initially filters invalid data through error judgment, and then further identifies abnormal data using statistical regression model, improved envelope domain method, low probability method and singular spectrum method, so as to realize the dual verification of dam monitoring data and improve the validity of data.

[0059] It should be noted that the formula for the statistical regression model is:

[0060]

[0061]

[0062]

[0063] in, The first in the dam monitoring project One observation value; The number of independent variables participating in the regression model; These are the regression coefficients; For the first The i-th observation of each independent variable; The power of the independent variable; This is the random error term; The total number of observation data sets; To minimize; This is a predicted value; For residuals; This is a vector of regression coefficients; A matrix consisting of the observed values ​​of the independent variable; for The transpose of the matrix; A vector composed of the observed values ​​of the dam monitoring effect; For a specific residual; The remaining standard deviation; The first effect of the dam monitoring effect obtained from the regression model One predicted value.

[0064] The improved envelope domain method formula is as follows:

[0065]

[0066] in, The upper limit threshold for the monitored indicators; This refers to the lower threshold of the monitoring indicator; The maximum value of the displacement data after initial screening; This represents the minimum value of the displacement data after initial screening. The annual rate of change; The degree of data dispersion; To monitor error values; for A monitoring data point from a specific year; This is the average value.

[0067] The low-probability method is:

[0068]

[0069]

[0070]

[0071] in, For a sample space; For a certain monitoring point; The number of samples; The mean of the sample; The standard deviation is the sample standard deviation. This represents the probability of an anomaly. It is the probability density function; The derivative of the integral variable.

[0072] The formula for singular spectrum analysis is:

[0073]

[0074]

[0075] in, The trajectory matrix; Matrix elements; To measure the length of the time span of the data; For the embedding dimension; These are measured values; For the first The contribution rate of each reconstructed component to the overall variation of dam monitoring data; Trajectory matrix The One eigenvalue; For fractional statistics; This refers to the residuals or biases in the dam monitoring data; The median; It is the interquartile range.

[0076] It is understood that the embodiments of this application use a two-layer identification model to first eliminate obviously invalid data that exceeds the range or violates physical logic through error judgment, and then use mathematical models (such as statistical regression and low probability methods) to identify hidden abnormal data. The dual verification can significantly reduce the false judgment rate of a single identification method, effectively filtering invalid information and avoiding the omission of potential abnormal data, thus laying the foundation for the subsequent construction of high-quality and effective datasets and ensuring the accuracy of dam safety monitoring and analysis.

[0077] In step S103, based on the effective dataset and combined with the real-time simulated benchmark value pushed by the digital twin center, risk defects are monitored through the distribution law method and the improved envelope domain method; the difference between the measured value and the benchmark value is calculated, and the anomaly level is determined according to the four-level rule. At the same time, the real-time model of the digital twin center is called, and the finite difference method and PINN algorithm are combined to calculate the dam stress and stability safety factor, generate the structural review results, and update the risk label.

[0078] Among them, the PINN algorithm is an algorithm that embeds physical laws (such as mechanical equations and conservation laws) into the neural network training process to achieve accurate calculation and correction of physical quantities such as stress and displacement of engineering structures such as dams. The formula is as follows:

[0079]

[0080]

[0081] in, For the predicted physical quantities of the dam; Spatial coordinates; It is a time variable; It is a neural network model; For neural network parameters; This is the total loss function; For data fitting loss; For physical constraint weights; Loss due to physical constraints; for Normal stress in the direction; for Normal stress in the direction; Shear stress; It is a partial derivative.

[0082] It is understood that the embodiments of this application embed the physical laws (such as mechanical equilibrium equations) related to the dam structure into the neural network training through the PINN algorithm. This can accurately correct the deviations in local stress concentration areas in traditional calculations such as the finite difference method. This not only improves the calculation accuracy of key indicators such as dam stress and displacement, but also eliminates the need to rely on a large number of measured samples. This provides efficient and accurate calculation support for dam structure verification and ensures the reliability of structural safety assessment.

[0083] In step S104, based on the anomaly level and review results, combined with the real-time mapping status assessment of the digital central model, the early warning threshold is optimized through the GA algorithm to provide real-time early warning, and the reporting level is specified according to the five-level rule to trigger the response mechanism.

[0084] The GA algorithm refers to an intelligent algorithm that uses iterative optimization to find the optimal solution and is used for adaptive adjustment of parameters such as dam early warning thresholds. The formula is:

[0085]

[0086]

[0087] in, The fitness value for the combination of warning thresholds; False alarm rate; This refers to the false negative rate. It is the minimum value; For the first The selection probability of an individual with a threshold value; for The fitness value of each threshold individual; For newly generated threshold individuals; For the first One parent threshold individual to be crossed; For the first One parent threshold individual to be crossed; The threshold individual before mutation; This represents the mutation probability.

[0088] It is understood that the embodiments of this application, through the GA algorithm, can adaptively optimize the early warning threshold based on the dam's historical monitoring data and safety standards. This avoids the subjectivity and lag of manually setting thresholds, and allows the thresholds to dynamically match changes in the dam's operating conditions, improving the accuracy of early warning judgments. This provides a scientific basis for triggering subsequent graded early warnings and response measures, and reduces the problem of excessive or insufficient early warnings.

[0089] According to the embodiments of this application, a method for online monitoring of dam safety based on multi-source information fusion is proposed. This method constructs a real-time mapping system between the physical dam and the digital model through a digital twin central module. It combines a data assimilation algorithm with multi-source real-time data from environmental quantities, measurements, and inspections to correct the model. This is complemented by a data acquisition module that performs two-layer invalid identification and constructs a valid dataset. This effectively addresses issues such as scattered data storage and invalid data interference, improving the reliability and correlation analysis capabilities of monitoring data under complex operating environments. Furthermore, the method relies on the collaborative identification of multiple methods, including the distribution pattern method and the improved envelope domain method, within the monitoring identification module. It also incorporates an attention mechanism to enhance the extraction of key measurement point features, improving the performance of abnormal monitoring in scenarios with weak measurement patterns and local structural anomalies. To improve identification accuracy, the structural verification module integrates the finite difference method, limit equilibrium method, and PINN algorithm for calculation, combined with real-time state mapping using a digital twin model. This enables spatiotemporal correlation analysis of parameters such as dam stress and stability safety factor, providing multi-dimensional data support for dam safety assessment. The comprehensive early warning module optimizes early warning thresholds through the GA algorithm, and, in conjunction with a five-level reporting and response mechanism, can respond in real-time to anomaly levels and structural verification results, triggering corresponding measures. This effectively avoids the escalation of risks due to delayed early warnings, controls the misjudgment and omission rates in dam safety monitoring, improves the timeliness and accuracy of dam operation safety management, reduces manual monitoring costs, and ensures the stability of watershed flood control and water resource allocation. Therefore, it solves the problems of data fragmentation, single anomaly identification, and lack of early warning levels in existing technologies.

[0090] The following will illustrate a method for online monitoring of dam safety based on multi-source information fusion through a specific embodiment, such as... Figure 9 As shown, it includes: In the online safety monitoring of a concrete gravity dam (maximum dam height 150m, dam crest length 420m, dam crest elevation 580m), the first step was to simultaneously conduct multi-source data acquisition and digital twin model construction. For environmental data, upstream water level (monitoring range 520-565m, maximum daily fluctuation 3m) and downstream water level (monitoring range 510-515m) were collected every 15 minutes using water level sensors. Rainfall was collected every hour using a tipping bucket rain gauge (maximum daily measured value during the flood season 78mm). Inflow was recorded in real time using a flow monitoring station (monitoring range 200-1800 m³ / s). For measurement data, 22 horizontal displacement measurement points on the dam crest (e.g., the daily average horizontal displacement of measurement point #1 on the dam crest) were obtained using a GNSS displacement monitoring system. 0.12mm), 18 vertical displacement measuring points on the dam foundation (e.g., the vertical displacement of measuring point 5 on the dam foundation is stable at -12.3mm), 30 uplift pressure measuring points on the dam foundation (e.g., the daily average uplift pressure of measuring point 12 on the dam foundation is 0.28MPa) were collected using a piezometer, and 15 stress measuring points on the dam body (e.g., the daily average compressive stress of measuring point 7 on the dam body is 3.2MPa) were recorded using a stress sensor; the inspection data were recorded daily on-site by the operation and maintenance personnel, including 2 microcracks ≤2cm in length on the dam surface and 1 slight seepage point on the left bank dam foundation with a flow rate of 0.01L / min. Simultaneously, based on the gravity dam construction drawings, geological survey report (dam foundation is moderately weathered sandstone, deformation modulus 35GPa), and material parameters (dam body C18030 concrete compressive strength 11.32MPa, elastic modulus 3.2×10⁻⁶), data were collected. 4 Based on MPa), a 1:1 physical dam digital twin model was constructed to accurately reproduce the structural features of the dam body, such as joints (20m / segment) and the layout of the dam foundation anti-seepage curtain, and the initial model construction was completed.

[0091] After data collection, invalid data identification and initial calibration of the twin model were carried out based on multi-source data. First, invalid data was screened using a two-layer identification model: the first layer used error judgment to remove one horizontal displacement data point exceeding the instrument's range (instrument maximum range 100mm, measured value 105mm) and one data point violating the physical logic of "increased uplift pressure with rising water level" (when the upstream water level rose 1.5m, the uplift pressure at measuring point 15 on the dam foundation actually decreased by 0.02MPa); the second layer used a mathematical model for evaluation, calculating the displacement data residuals using a statistical regression model (multiple correlation coefficient 0.86), identifying two abnormal data points with residuals exceeding 3σ (residual values ​​3.1σ and 3.3σ), and then using a low-probability method (significance level 0.01) to identify one low-probability stress deviation data point (probability of occurrence 0.007). For these four suspected invalid data points, a suspected data recall was initiated (abnormal displacement measuring points were re-collected three times, and uplift pressure and stress measuring points were each re-collected twice). Maintenance personnel were then assigned to conduct on-site verification (confirming one data point as a sensor malfunction and three as temporary electromagnetic interference). Ultimately, a valid dataset of 9862 data points was constructed. Subsequently, the valid data was substituted into the digital twin model, comparing the calculated horizontal displacement of the dam crest (e.g., the calculated value of 0.11 mm at measuring point #1) with the measured value (0.12 mm), and the uplift pressure of the dam foundation (the calculated value of 0.27 MPa at measuring point #12) with the measured value (0.28 MPa). The elastic modulus of the dam concrete in the model was adjusted (from 3.2 × 10⁻⁶). 4 MPa is corrected to 3.15 × 10 4 MPa), permeability coefficient of dam foundation rock mass (from 1×10⁻ 6 cm / s corrected to 1.2 × 10⁻ 6 (cm / s) to complete the initial calibration of the twin model, reducing the deviation between the calculated and measured values ​​to within 0.02 mm (displacement) and 0.01 MPa (pressure). Based on the calibrated valid dataset, combined with the real-time simulation benchmark values ​​pushed by the digital twin center (such as the horizontal displacement benchmark value of 0.11mm at the No. 1 measuring point on the dam crest, the uplift pressure benchmark value of 0.27MPa at the No. 12 measuring point on the dam foundation, and the stress benchmark value of 3.1MPa in the dam body), risk and defect monitoring and structural verification are carried out. Regarding risk and defect monitoring, the distribution law method was used to calculate the correlation between the horizontal displacement measuring points of dam sections 8-15. It was found that the correlation between measuring point 10 on the dam crest and the adjacent measuring points 9 and 11 dropped from the historical average of 0.89 to 0.67 (below the normal threshold of 0.8), indicating a local anomaly at this measuring point. For the uplift pressure data of the dam foundation, the improved envelope domain method was used to analyze the historical data of the past two years (after removing the maximum and minimum values ​​by 10%, the maximum value is 0.29 MPa and the minimum value is 0.25 MPa). The envelope domain range was calculated to be 0.24-0.30 MPa (including a monitoring error of 0.01 MPa). It was found that the measured uplift pressure of measuring point 18 on the dam foundation was 0.32 MPa, which was outside the range and identified as an anomaly in seepage pressure. The differences between the measured values ​​and the benchmark values ​​were calculated: the displacement difference at measuring point #10 on the dam crest was 0.23 mm, and the uplift pressure difference at measuring point #18 on the dam foundation was 0.05 MPa. According to the fourth-level rule, the displacement difference at measuring point #10 on the dam crest was small and the trend was converging, indicating a slight anomaly; the uplift pressure difference at measuring point #18 on the dam foundation was large and continuously increasing, indicating a general anomaly. Simultaneously, the real-time model of the digital twin was invoked, and the finite difference method was used to calculate the dam body stress (the maximum compressive stress of the dam body was 4.8 MPa, lower than the allowable compressive strength of C18030 concrete of 11.32 MPa). The PINN algorithm was used to correct the stress concentration area at the junction of the dam foundation and the dam body (the corrected stress decreased from 5.2 MPa to 5.0 MPa). Then, the limit equilibrium method was used to calculate the dam foundation's anti-sliding stability safety factor (3.6 under the basic combination, higher than the allowable value of 3.0 in the specification). The structural verification results were generated, and the original "to be assessed" risk label was updated to "local medium risk, overall safe."

[0092] Finally, comprehensive evaluation and early warning response are carried out by combining the anomaly level, the results of structural review, and the real-time mapping status of the digital twin central model (the fitting degree of the model and the measured deformation is 98.5%, and the stress calculation deviation ≤ 0.2 MPa). First, the early warning thresholds are optimized through the GA algorithm: the early warning threshold for the horizontal displacement at the dam crest is fine-tuned from 0.3 mm to 0.28 mm, and the early warning threshold for the uplift pressure at the dam foundation is optimized from 0.31 MPa to 0.30 MPa to improve the early warning sensitivity. Then, in accordance with the five-level early warning rules, combined with the environmental risks (the upstream water level 550 m < the design flood level 560 m, the daily rainfall 45 mm < the rainstorm threshold 50 mm, no environmental risks), it is comprehensively determined as a level-IV (blue) early warning. According to the five-level reporting rules, the early warning information is pushed to the management personnel of the production technology department and the power station responsible person, including the positions of the abnormal measurement points (10# at the dam crest, 18# at the dam foundation), the types of abnormalities (local displacement abnormality, uplift pressure exceeding the limit), the review results (qualified dam body stress, up-to-standard dam foundation stability), and the early warning level. At the same time, the response mechanism is triggered: encrypted observation is started (the acquisition frequency of the 10# measurement point at the dam crest is increased from 1 time per hour to 1 time per 10 minutes, and the 18# measurement point at the dam foundation is increased from 1 time per hour to 1 time per 8 minutes), on-site verification is arranged within 18 hours for the maintenance personnel (focusing on checking the cracks in the dam body around the 10# measurement point at the dam crest and the integrity of the anti-seepage curtain corresponding to the 18# measurement point at the dam foundation), and it is required to feedback the verification results and encrypted observation data daily to form a closed-loop management of "evaluation - early warning - response" to ensure that the dam safety risks are timely controllable.

[0093] In summary, in the embodiment of the present application, by synchronously collecting multi-source data and constructing a digital twin model, the operation information of the gravity dam is comprehensively covered, and the structural characteristics are accurately reproduced, laying a reliable foundation for monitoring; through double-layer identification and manual review to eliminate invalid data, the model is calibrated (the deviation ≤ 0.02 mm / 0.01 MPa), improving the reliability of the data and the model; relying on professional methods to identify risk defects and determine the anomaly level, combined with algorithms to quantify the dam body stress (4.8 MPa < 11.32 MPa) and the stability coefficient (3.6 > 3.0), providing a scientific evaluation basis; optimizing the early warning threshold through the GA algorithm and triggering a level-IV early warning response (verification within 18 hours), the abnormal disposal efficiency is increased by 40%, forming a closed-loop management, effectively ensuring the stable operation of the dam, and having both adaptability and control accuracy.

[0094] Figure 10 It is a schematic structural diagram of the electronic device provided by the embodiment of the present application. The electronic device may include: A memory 1001, a processor 1002, and a computer program stored on the memory 1001 and executable on the processor 1002.

[0095] When the processor 1002 executes the program, it implements a dam safety online monitoring method based on multi-source information fusion provided in the above embodiment.

[0096] Furthermore, electronic devices also include: Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0097] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0098] The memory 1001 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0099] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0100] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0101] The processor 1002 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0102] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for online monitoring of dam safety based on multi-source information fusion.

[0103] Furthermore, this application also provides a computer program product, including a computer program or instructions, which, when executed, implement the above-mentioned method for online monitoring of dam safety based on multi-source information fusion.

[0104] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0105] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0106] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0107] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0108] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium. When executed, the program includes one or a combination of the steps of the method embodiments.

[0109] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A dam safety online monitoring system based on multi-source information fusion, characterized in that, include: The digital twin central module includes a data acquisition module, a monitoring and identification module, a structural verification module, and a comprehensive early warning module; among them, The digital twin central module is used to construct a real-time mapping system between the physical dam and the digital model. It integrates multi-source real-time data to correct the model through a data assimilation algorithm, and serves as a hub to push data and benchmark values, receive feedback, and perform iterative optimization. The data acquisition module is used to collect environmental quantity data, measurement data and inspection data, and transmit them to the digital twin central storage standardization. It identifies invalid data through a two-layer identification model and adaptive review, and constructs an effective dataset by combining suspected data recall and manual review and then sends it back. The monitoring and identification module monitors risk defects based on effective datasets and central real-time simulation benchmarks, using distribution pattern method and improved envelope domain method, and determines the anomaly level according to four-level rules. The structural verification module is based on the central real-time digital twin, and combines the finite difference method and PINN algorithm to establish a structural model, and outputs the verification results to the central system to update the risk label. The integrated early warning module is used to combine the anomaly level, review results and central mapping status assessment, optimize the early warning threshold through the GA algorithm, provide real-time early warning, clarify the reporting level according to the five-level rule, and trigger the response mechanism.

2. The dam safety online monitoring system based on multi-source information fusion according to claim 1, characterized in that, The digital twin central module includes a physical digital mapping unit, a data assimilation and correction unit, a data hub unit, and an iterative optimization unit. The physical digital mapping unit is used to construct a digital model of the monitored object, matching the geometric and mechanical characteristics of the physical dam. The data assimilation and correction unit is used to fuse multi-source real-time data to correct model parameters. The data hub unit is used to receive standardized data, push real-time simulation benchmark values, and receive module feedback. The iterative optimization unit is used to update the model by combining structural verification deviations and early warning feedback.

3. The dam safety online monitoring system based on multi-source information fusion according to claim 1, characterized in that, The data acquisition module includes a multi-source data acquisition unit, a data standardization unit, a two-layer invalid identification unit, and a valid dataset construction unit. The multi-source data acquisition unit acquires environmental quantity data, measured data, and inspection data. The data standardization unit converts the data into a unified format and stores it. The two-layer invalid identification unit identifies invalid data through error judgment, statistical regression, and low-probability methods. The valid dataset construction unit combines suspected data recall with manual review to filter valid data.

4. The dam safety online monitoring system based on multi-source information fusion according to claim 1, characterized in that, The monitoring and identification module includes a benchmark comparison unit, a multi-method anomaly identification unit, a four-level anomaly determination unit, and a result push unit. The benchmark comparison unit is used to calculate the difference between the measured value and the benchmark value pushed by the central system. The multi-method anomaly identification unit is used to identify risks and defects through the distribution law method and the improved envelope domain method. The four-level anomaly determination unit is used to determine the status as normal, minor anomaly, general anomaly, and serious anomaly. The result push unit is used to push the anomaly result and trigger the initial alarm.

5. The dam safety online monitoring system based on multi-source information fusion according to claim 1, characterized in that, The structural verification module includes a model calling unit, a multi-method structural calculation unit, a calculation result verification unit, and a result feedback unit. The model calling unit is used to acquire the central digital model and optimize the mesh. The multi-method structural calculation unit is used to calculate the dam stress and stability safety factor using the finite difference method, limit equilibrium method, arch-beam load distribution method, and PINN algorithm. The calculation result verification unit is used to compare the calculated values ​​with the measured values ​​and adjust the parameters. The result feedback unit is used to provide feedback on the verification results and update the risk labels.

6. The dam safety online monitoring system based on multi-source information fusion according to claim 1, characterized in that, The comprehensive early warning module includes a multi-dimensional evaluation unit, a five-level early warning determination unit, and a graded reporting unit. The multi-dimensional evaluation unit integrates anomaly levels, verification results, and central mapping status to evaluate the overall performance of the dam in conjunction with the digital mapping status. The five-level early warning determination unit determines the early warning status as normal, Level IV, Level III, Level II, and Level I based on environmental risk and dam safety early warning standards. The graded reporting unit triggers corresponding measures such as encrypted observation, on-site inspection, and emergency response according to the early warning information reporting rules.

7. A method for online monitoring of dam safety based on multi-source information fusion, applicable to any one of claims 1-6, characterized in that, include: Acquire environmental data, measurement data, and inspection data; simultaneously, construct a digital twin model of the physical dam. Based on the environmental data, measured data, and inspection data, invalid data is identified through a two-layer identification model and adaptive verification. A valid dataset is constructed by combining suspected data recall and manual verification, and the twin model is initially calibrated. Based on the effective dataset and combined with the real-time simulated benchmark values ​​pushed by the digital twin center, risk defects are monitored through the distribution law method and the improved envelope domain method; the difference between the measured value and the benchmark value is calculated, and the anomaly level is determined according to the four-level rule. At the same time, the real-time model of the digital twin center is called, and the finite difference method and PINN algorithm are combined to calculate the dam stress and stability safety factor, generate structural review results and update risk labels. Based on the aforementioned anomaly level and review results, combined with the real-time mapping status assessment of the digital central model, the early warning threshold is optimized through the GA algorithm to provide real-time early warning, and the reporting level is clearly defined according to the five-level rule to trigger the response mechanism.

8. An electronic device, characterized in that, include: The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the online monitoring method for dam safety based on multi-source information fusion as described in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they implement the online monitoring method for dam safety based on multi-source information fusion as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed, they implement the online monitoring method for dam safety based on multi-source information fusion as described in claim 7.

Images