Water conservancy project online monitoring and abnormity diagnosis system based on edge calculation

Abstract

The invention relates to the technical field of water conservancy project structure health monitoring and intelligent sensing, in particular to a water conservancy project online monitoring and abnormity diagnosis system based on edge calculation, which comprises a data acquisition module used for acquiring current water level data, current temperature data and actually measured structure response data; the reference reconstruction module is used for generating an ideal state vector; the fault simulation module is used for generating a theoretical damaged simulation state vector; the topology judgment module is used for calculating a first difference value between the actual measurement structure response data and the ideal state vector, calculating a second difference value between the theoretical damaged simulation state vector and the ideal state vector, and calculating the topology similarity based on the first difference value and the second difference value; the risk assessment module is used for generating a structure safety risk index in combination with the fault strength scalar and the topological similarity; according to the method, the problems of data transmission lag and low online calculation efficiency in a traditional centralized processing mode are effectively solved, and the real-time requirement of an engineering site is met.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring and intelligent sensing technology for water conservancy projects, specifically to an online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing. Background Technology

[0002] In the field of safety monitoring of water conservancy projects, key structures such as dams and gates are in a complex environment with continuous changes in water level and temperature. By deploying sensor arrays such as global navigation satellite systems or vertical coordinate instruments, displacement vectors and environmental monitoring data that reflect the actual physical deformation state of the structure can be collected in real time. Existing monitoring and diagnostic solutions generally rely on centralized offline data processing or forward simulation based on high-fidelity finite element models. This approach demands extremely high computational resources and has a long response cycle. Since the deformation response of hydraulic structures is the result of the coupling effect of environmental loads and structural stiffness, and the measured data is often mixed with sensor drift and background noise, traditional methods based on single threshold judgment or simple difference calculation are difficult to effectively isolate environmental effects and cannot achieve low-latency accurate calculations at the edge. This limitation makes it difficult for the system to accurately identify minor structural damage from aliased signals, and it is prone to false alarms or missed alarms due to environmental interference, failing to meet the needs of real-time quantitative assessment and graded early warning of structural health status. Therefore, how to achieve rapid decoupling and benchmark reconstruction of environmental loads at edge computing nodes, and eliminate noise interference by analyzing the topological differences of residual vectors, thereby improving the real-time performance and accuracy of structural anomaly diagnosis, is a technical problem that urgently needs to be solved. Summary of the Invention

[0003] The purpose of this invention is to provide an online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing. This system enables rapid decoupling and baseline reconstruction of environmental loads at the edge, effectively eliminates environmental interference by analyzing the topological differences in residual vectors, and improves the real-time performance and accuracy of structural anomaly diagnosis. Specifically, the technical solution of this invention is as follows: An online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing includes: The data acquisition module is used to acquire current water level data, current temperature data, and measured structural response data; The baseline reconstruction module is used to convert the current water level data and current temperature data into a total environmental load vector, and to process the total environmental load vector using a preset reduced-order compliance matrix to generate an ideal state vector. The fault simulation module is used to call preset fault topology operators and preset fault intensity scalars, and combine them with the total environmental load vector to generate a theoretically damaged simulation state vector. The topology decision module is used to calculate the first difference between the measured structural response data and the ideal state vector, calculate the second difference between the theoretical damaged simulated state vector and the ideal state vector, and calculate the topological similarity based on the first and second differences. The risk assessment module is used to generate a structural safety risk index by combining a fault intensity scalar and topological similarity.

[0004] Preferably, the current water level data and current temperature data are converted into a total environmental load vector, including: Call the preset water level correction factor and the preset temperature correction factor; The current water level data is converted into a water pressure load, and the water pressure load is weighted using a water level correction coefficient to generate a weighted water pressure load. The current temperature data is converted into a thermal stress load, and the thermal stress load is weighted using a temperature correction factor to generate a weighted thermal stress load. The weighted water pressure load and the weighted thermal stress load are superimposed to generate the total environmental load vector.

[0005] Preferably, the total environmental load vector is processed using a preset reduced-order compliance matrix to generate an ideal state vector, including: Obtain the reduced-order compliance matrix derived from the high-fidelity finite element model after modal truncation; Calculate the product of the reduced compliance matrix and the total environmental load vector; The product is defined as an ideal state vector.

[0006] Preferably, the theoretically damaged simulation state vector is generated by combining the total environmental load vector, including: Call the identity matrix and the reduced-order compliance matrix; Calculate the product of the fault strength scalar and the fault topology operator; Calculate the difference between the identity matrix and the product, and invert the difference to generate the compliance magnification matrix; Calculate the product of the reduced compliance matrix, the compliance amplification matrix, and the total environmental load vector to generate the theoretically damaged simulation state vector.

[0007] Preferably, the calculation of topological similarity based on the first difference and the second difference includes: Define the first difference as the actual residual vector; The second difference is defined as the theoretical residual vector; Based on the cosine similarity algorithm, the degree of parallelism between the actual residual vector and the theoretical residual vector in the direction of the vector space is calculated; The degree of parallelism is defined as topological similarity.

[0008] Preferably, based on the cosine similarity algorithm, the parallelism between the actual residual vector and the theoretical residual vector in the vector space direction is calculated, including: Introduce a pre-defined numerical stability term; Calculate the dot product of the actual residual vector and the theoretical residual vector; Calculate the magnitude of the actual residual vector and the first sum of the numerical stability terms; Calculate the magnitude of the theoretical residual vector and the second sum of the numerical stability terms; The degree of parallelism is obtained by dividing the dot product by the product of the first and second sums.

[0009] Preferably, a structural safety risk index is generated by combining a fault intensity scalar and topological similarity, including: Invoke the preset design limit strength, preset confidence threshold, and preset decision sensitivity constant; Calculate the ratio of the failure strength scalar to the design ultimate strength to generate the failure degree component; Calculate the difference between the topological similarity and the confidence threshold; Based on the difference and the decision sensitivity constant, the confirmation confidence component is calculated using the Sigmoid function. The structural safety risk index is generated by weighting and summing the damage degree component and the confirmation confidence component.

[0010] Preferably, the methods for obtaining the water level correction factor and the temperature correction factor include: Collect historical health monitoring data of the structure; Based on the principle of least squares, an optimization objective is constructed for the error function between historical health monitoring data and theoretical monitoring values; By minimizing the error function, the water level correction coefficient and the temperature correction coefficient are calculated.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. By introducing a reduced-order compliance matrix derived from a high-fidelity model, this system avoids the high computational cost of directly running a large finite element model on the monitoring terminal, significantly reducing computational latency. This design enables the system to be deployed at edge nodes close to the hydraulic structure itself, achieving low-latency local benchmark reconstruction and anomaly diagnosis. It effectively solves the problems of data transmission lag and low online computation efficiency in the traditional centralized processing mode, meeting the real-time requirements of the engineering site. 2. This system establishes an environmental load decoupling mechanism based on unified physical dimensions, which can transform multi-source heterogeneous water level and temperature data into a total environmental load vector and reconstruct the ideal state vector under the current environment. This mechanism can accurately separate the elastic deformation component caused solely by water level fluctuations and temperature changes from the measured data and construct a dynamic theoretical zero point, thereby overcoming the shortcomings of traditional methods in effectively distinguishing between environmental effects and structural damage and improving the accuracy of diagnostic benchmarks. 3. This system innovatively adopts a topological decision method based on cosine similarity. By calculating the parallelism between the actual residual vector and the theoretical residual vector in the vector space, it quantitatively evaluates the geometric fit of the deformation morphology. It focuses on analyzing the spatial distribution characteristics of the displacement field rather than the absolute amplitude of a single node, effectively avoiding the interference of sensor zero-point drift and environmental background noise on the diagnostic results, and significantly improving the robustness of the system in identifying minor damage in complex signal environments. 4. This system constructs a dual risk assessment model that integrates the degree of physical damage and the confidence level of topological confirmation. It uses nonlinear functions to suppress the signals of suspected faults with low similarity and enhance the confidence level of damage patterns with high similarity. This mechanism generates an intuitive structural safety risk index, which greatly reduces false warnings caused by accidental errors while ensuring that no serious structural hidden dangers are missed. It provides a quantitative and reliable scientific basis for the operation and maintenance decision-making of water conservancy projects. Attached Figure Description

[0012] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0014] Example 1: Please see Figure 1 An online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing, comprising: The data acquisition module is used to acquire current water level data, current temperature data, and measured structural response data; The baseline reconstruction module is used to convert the current water level data and current temperature data into a total environmental load vector, and to process the total environmental load vector using a preset reduced-order compliance matrix to generate an ideal state vector. The fault simulation module is used to call preset fault topology operators and preset fault intensity scalars, and combine them with the total environmental load vector to generate a theoretically damaged simulation state vector. The topology decision module is used to calculate the first difference between the measured structural response data and the ideal state vector, calculate the second difference between the theoretical damaged simulated state vector and the ideal state vector, and calculate the topological similarity based on the first and second differences. The risk assessment module is used to generate a structural safety risk index by combining a fault intensity scalar and topological similarity.

[0015] This embodiment discloses an online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing. The system is deployed at an edge computing node close to the water conservancy structure and achieves low-latency structural health diagnosis through localized processing. The system specifically includes a data acquisition module, a baseline reconstruction module, a fault simulation module, a topology decision module, and a risk assessment module. The data acquisition module is used to perceive the external environment and internal status of water conservancy projects in real time; this module synchronously acquires current water level data through sensor arrays deployed at key parts of the dam or gate. Current temperature data and measured structural response data Among them, the measured structural response data Specifically, it refers to the displacement vectors of key structural nodes acquired through the Global Navigation Satellite System (GNSS) or a vertical coordinate instrument, which can reflect the actual physical deformation state of the structure at the current moment; The baseline reconstruction module is used to construct the theoretical zero point of the structure under the current environment; this module will collect the current water level data. and current temperature data Transformed into a total environmental load vector with unified physical dimensions And using a preset reduced-order compliance matrix The total environmental load vector is linearly mapped to generate the ideal state vector. The ideal state vector It characterizes the displacement response that should theoretically be exhibited under the current combination of water level and temperature, assuming the structure is in a completely undamaged and healthy state; This embodiment employs an analytical-synthetic inverse logic; the fault simulation module does not perform only one forward deduction, but rather iterates within a preset solution space based on an optimized search strategy, such as particle swarm optimization or grid search; the fault simulation module is used to deduce possible fault modes in the digital space; this module calls the candidate fault topology operators under the current iteration step. and preset fault intensity scale Combined with the aforementioned generated total environmental load vector The theoretically damaged simulation state vector is generated through forward mechanics deduction. This vector represents the deformation pattern that the assumed structure should exhibit under the current environmental load when it suffers a specific degree of damage at a specific location. Based on this, the topology decision module performs the core difference analysis; this module calculates the measured structural response data. With the ideal state vector The first difference between them is defined as the real residual vector. Simultaneously calculate the theoretically damaged simulation state vector. With the ideal state vector The second difference between them is defined as the theoretical residual vector. This module is further based on real-world residual vectors. and theoretical residual vector Calculate topological similarity To quantitatively assess the degree of agreement between the measured anomaly patterns and the hypothetical fault patterns in spatial geometry; The risk assessment module is used to output the final decision indicators; this module combines a fault intensity scalar. and the calculated topological similarity A structural safety risk index is generated through a weighted fusion algorithm. This index is used to intuitively reflect the current safety status of water conservancy projects and serves as the basis for triggering graded early warnings.

[0016] Example 2: The current water level and temperature data are converted into a total environmental load vector, including: Call the preset water level correction factor and the preset temperature correction factor; The current water level data is converted into a water pressure load, and the water pressure load is weighted using a water level correction coefficient to generate a weighted water pressure load. The current temperature data is converted into a thermal stress load, and the thermal stress load is weighted using a temperature correction factor to generate a weighted thermal stress load. The weighted water pressure load and the weighted thermal stress load are superimposed to generate the total environmental load vector; The methods for obtaining water level correction factors and temperature correction factors include: Collect historical health monitoring data of the structure; Based on the principle of least squares, an optimization objective is constructed for the error function between historical health monitoring data and theoretical monitoring values; By minimizing the error function, the water level correction coefficient and the temperature correction coefficient are calculated.

[0017] This embodiment further details the specific process by which the baseline reconstruction module transforms environmental data into a total environmental load vector, as well as the calibration method for the relevant correction coefficients; During system operation, the baseline reconstruction module calls the preset water level correction coefficient. and preset temperature correction factor Specifically, the module will display the current water level data. Input the preset water pressure function Converted to standard water pressure load, this function It is a unit load vector lookup table or interpolation function generated offline based on the shape function of the discrete mesh on the outer surface of the structure. During online runtime, it only needs to be called via linear interpolation, without requiring real-time integration at the edge; water pressure function. Based on hydrostatic pressure formula The pressure values ​​at each mesh node on the wetted surface of the structure are calculated, and the distributed pressure is transformed into an equivalent nodal force vector acting on the structural nodes through finite element shape function integration; and the water level correction coefficient is used. The water pressure load is weighted to generate a weighted water pressure load. At the same time, the current temperature data Input the preset thermal stress function Converted into standard thermal stress load, and using a temperature correction factor. The thermal stress load is weighted to generate a weighted thermal stress load. Finally, the module performs vector superposition of the weighted water pressure load and the weighted thermal stress load to generate the total environmental load vector. This process enables the transformation of environmental scalars with different physical dimensions into a unified force vector at mechanical nodes. To ensure the accuracy of the above conversion process, a water level correction coefficient is used. and temperature correction factor The parameters need to be obtained through a rigorous calibration process; the specific calibration method is as follows: collect historical health monitoring datasets of the structure during its historical health operation cycle; to distinguish real-time variables, the historical water level sample data used for calibration is denoted as... Historical temperature sample data is recorded as The corresponding historical measured displacement sample data is denoted as Based on the principle of least squares, an optimization objective is constructed for the error function between historical health monitoring data and theoretical monitoring values. The optimization objective The aim is to find an optimal set of and This makes the theoretical displacement calculated based on this coefficient consistent with historical measured displacement sample data. The sum of squared Euclidean distances between them is minimized; this is achieved by applying the error function. By minimizing the solution, the water level correction coefficient that best fits the structural characteristics of the specific hydraulic engineering project is calculated. and temperature correction factor .

[0018] Example 3: The total environmental load vector is processed using a preset reduced-order compliance matrix to generate an ideal state vector, including: Obtain the reduced-order compliance matrix derived from the high-fidelity finite element model after modal truncation; Calculate the product of the reduced compliance matrix and the total environmental load vector; The product is defined as an ideal state vector.

[0019] This embodiment elaborates on the technical details of generating ideal state vectors using a reduced-order compliance matrix; In this embodiment, the reduced-order compliance matrix The data originates from a high-fidelity finite element model of the water conservancy project. Specifically, it is a low-dimensional matrix representing the force-displacement mapping relationship of the structure in an undamaged state, extracted by modal truncation or Guyan reduction techniques from the high-fidelity model. To address the dimensionality matching problem between the physical space and the reduced-order space, the benchmark reconstruction module first calls a preset load projection matrix. This matrix is ​​the transpose of the model condensation transformation matrix, with dimensions of . ,in To preserve the degrees of freedom, The total number of degrees of freedom for the entire model; degrees of freedom retained. The selection principle is based on the node degrees of freedom corresponding to the actual sensor deployment location, to ensure a direct mapping between the physical space measured data and the reduced-order space simulation data in the topology; to adapt to the limited storage space and computing power resources of the edge computing gateway, this embodiment retains the number of degrees of freedom. The order of magnitude is controlled to be between 10^1 and 10^2, for example, 50-200 key node degrees of freedom, which is much smaller than the total number of degrees of freedom of the entire model. ; Module calculates the load projection matrix With the total environmental load vector The product of these terms generates the generalized load vector. Calculate the reduced-order compliance matrix With generalized load vector The matrix product; the result of this product operation is directly defined as the ideal state vector. ,Right now Through this step, the system can quickly calculate the theoretical reference displacement under the current environment based on the physical model.

[0020] Example 4: By combining the total environmental load vector, a theoretically damaged simulation state vector is generated, including: Call the identity matrix and the reduced-order compliance matrix; Calculate the product of the fault strength scalar and the fault topology operator; Calculate the difference between the identity matrix and the product, and invert the difference to generate the compliance magnification matrix; Calculate the product of the reduced compliance matrix, the compliance amplification matrix, and the total environmental load vector to generate the theoretically damaged simulation state vector.

[0021] This embodiment details the logic of the fault simulation module in generating the theoretically damaged simulation state vector; This process is based on the principles of continuum damage mechanics and aims to simulate the mechanical behavior of a structure after its stiffness degrades; the module calls the identity matrix. and the preset reduced-order compliance matrix Here, the reduced-order compliance matrix Also derived from the finite element model, it characterizes the elastic properties of the structure when it is undamaged; the module calculates the fault strength scalar. With fault topology operators product Among them, the fault topology operator It is a sparse matrix where only the elements corresponding to the fault locations are non-zero, specifically constructed as follows: This is a diagonal matrix or block diagonal matrix with the same dimension as the reduced-order compliance matrix. The elements on the diagonal corresponding to the degrees of freedom of the assumed fault node are 1, and all other elements are 0. This allows matrix operations to filter out only the stiffness influence term at the fault location; the fault strength scalar... The range of values ​​is ; Module calculates the identity matrix The difference from the above product Then, the matrix inversion operation is performed on the difference to generate the compliance amplification matrix. Thanks to the aforementioned modal truncation and order reduction processing, the dimension of the matrix to be inverted here is relatively low, i.e., m×m. Edge computing nodes can directly complete the calculation in milliseconds using LU decomposition or block inversion algorithms, meeting the real-time requirements of online monitoring. This compliance amplification matrix physically represents the compliance amplification effect caused by damage. The module calculates the reduced-order compliance matrix. Flexibility Amplification Matrix and generalized load vector That is, as defined in Example 3 The product of these elements generates the theoretically damaged simulation state vector. Its mathematical expression is: This process precisely reconstructs the deformation field of the structure under specific damage modes in digital space.

[0022] Example 5: Topological similarity is calculated based on the first difference and the second difference, including: Define the first difference as the actual residual vector; The second difference is defined as the theoretical residual vector; Based on the cosine similarity algorithm, the degree of parallelism between the actual residual vector and the theoretical residual vector in the direction of the vector space is calculated; The degree of parallelism is defined as topological similarity; Based on the cosine similarity algorithm, the parallelism between the actual residual vector and the theoretical residual vector in the vector space direction is calculated, including: Introduce a pre-defined numerical stability term; Calculate the dot product of the actual residual vector and the theoretical residual vector; Calculate the magnitude of the actual residual vector and the first sum of the numerical stability terms; Calculate the magnitude of the theoretical residual vector and the second sum of the numerical stability terms; The degree of parallelism is obtained by dividing the dot product by the product of the first and second sums.

[0023] This embodiment describes in detail the specific algorithm for calculating topological similarity in the topology decision module; Measured structural response data Subtract the ideal state vector The first difference obtained is defined as the actual residual vector. ; to transform the theoretically damaged simulation state vector Subtract the ideal state vector The second difference obtained is defined as the theoretical residual vector. ; Calculate the real residual vector based on the cosine similarity algorithm. With the theoretical residual vector Parallelism in the direction of the vector space; to avoid division by zero error when the system is in an ideal healthy state, i.e., the magnitude of the residual vector approaches zero, this embodiment introduces a preset numerical stability term in the calculation. This numerical stability term For a very small positive number, for example, taking the value of ; The specific calculation steps include: calculating the actual residual vector. With the theoretical residual vector dot product ; Calculate the magnitude of the actual residual vector and the first sum of the numerical stability terms. ; Calculate the magnitude of the theoretical residual vector and the second sum of the numerical stability term. The degree of parallelism, or topological similarity, is obtained by dividing the dot product by the product of the first and second sums. Its calculation formula is: This method effectively eliminates the interference of sensor drift and environmental background noise on diagnostic results by comparing the direction of the vector rather than its magnitude.

[0024] It should be noted that the above embodiments 4 and 5 describe the calculation process for a single hypothetical failure mode. In actual operation, the system executes an automated iterative optimization strategy: the system has a pre-set library of fault topology operators containing multiple typical damage modes, such as abnormal uplift pressure in the dam foundation and instability of the gate arm. The fault simulation module works in conjunction with the topology decision module to traverse every topology operator in the library. And within the interval [0,1), it traverses the fault intensity scalar with a preset step size, such as 0.01. For each group ( , Combining these elements, the process is repeated to generate theoretically damaged simulation state vectors. And calculate topological similarity The system compares the topological similarity generated by all iterations and selects the group with the highest similarity value. , As the diagnostic conclusion at the current moment, the dominant damage mode that contributes the most to the overall structural flexibility is identified, and the parameters corresponding to the optimal solution are passed to the risk assessment module.

[0025] Example 6: By combining fault intensity scalar and topological similarity, a structural safety risk index is generated, including: Invoke the preset design limit strength, preset confidence threshold, and preset decision sensitivity constant; Calculate the ratio of the failure strength scalar to the design ultimate strength to generate the failure degree component; Calculate the difference between the topological similarity and the confidence threshold; Based on the difference and the decision sensitivity constant, the confirmation confidence component is calculated using the Sigmoid function. The structural safety risk index is generated by weighting and summing the damage degree component and the confirmation confidence component.

[0026] This embodiment details the specific logic behind the risk assessment module's generation of the structural safety risk index; This module calls the preset design limit strength. Preset confidence threshold and the preset decision sensitivity constant Among them, the design ultimate strength The maximum permissible damage ratio is derived from dam design specifications, for example, a value of 0.8; confidence threshold. Derived from the 99th percentile value of the statistical analysis of the system background noise similarity distribution; decision sensitivity constant. This is a constant used to adjust the slope of the Sigmoid function, for example, a value of 15; before proceeding with risk calculation, the system first iterates through the topological similarity results generated by the fault simulation module in all iteration steps, locking in the topological similarity values. Extract the fault intensity scalar from the largest set of iterative data. and topological similarity This serves as the input data for this module; During the calculation process, the module calculates the aforementioned extracted fault intensity scalar. With design ultimate strength The ratio of the two values ​​generates a component representing the degree of damage. This component linearly reflects the physical severity of the damage; simultaneously, topological similarity is calculated. With confidence threshold The difference, and based on this difference and the decision sensitivity constant The confidence component is calculated using the Sigmoid function, specifically in the form of... This component non-linearly reflects the reliability of the diagnostic results. When the similarity is below the threshold, this component approaches 0, suppressing false alarms; when the similarity is above the threshold, this component rapidly approaches 1. The module calls the preset physical damage weight coefficient. For example, a value of 0.6 and a preset confidence weight coefficient. For example, the value is 0.4, and it satisfies... Using formulas The structural safety risk index is generated by weighting and summing the damage severity component and the confidence level component. Through this dual verification mechanism, the system can significantly reduce false alarms caused by environmental noise while ensuring that no serious faults are missed.

[0027] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing, characterized in that, include: The data acquisition module is used to acquire current water level data, current temperature data, and measured structural response data; The baseline reconstruction module is used to convert the current water level data and current temperature data into a total environmental load vector, and to process the total environmental load vector using a preset reduced-order compliance matrix to generate an ideal state vector. The fault simulation module is used to call preset fault topology operators and preset fault intensity scalars, and combine them with the total environmental load vector to generate a theoretically damaged simulation state vector. The topology decision module is used to calculate the first difference between the measured structural response data and the ideal state vector, calculate the second difference between the theoretical damaged simulated state vector and the ideal state vector, and calculate the topological similarity based on the first and second differences. The risk assessment module is used to generate a structural safety risk index by combining a fault intensity scalar and topological similarity.

2. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 1, characterized in that, The current water level and temperature data are converted into a total environmental load vector, including: Call the preset water level correction factor and the preset temperature correction factor; The current water level data is converted into a water pressure load, and the water pressure load is weighted using a water level correction coefficient to generate a weighted water pressure load. The current temperature data is converted into a thermal stress load, and the thermal stress load is weighted using a temperature correction factor to generate a weighted thermal stress load. The weighted water pressure load and the weighted thermal stress load are superimposed to generate the total environmental load vector.

3. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 1, characterized in that, The total environmental load vector is processed using a preset reduced-order compliance matrix to generate an ideal state vector, including: Obtain the reduced-order compliance matrix derived from the high-fidelity finite element model after modal truncation; Calculate the product of the reduced compliance matrix and the total environmental load vector; The product is defined as an ideal state vector.

4. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 1, characterized in that, By combining the total environmental load vector, a theoretically damaged simulation state vector is generated, including: Call the identity matrix and the reduced-order compliance matrix; Calculate the product of the fault strength scalar and the fault topology operator; Calculate the difference between the identity matrix and the product, and invert the difference to generate the compliance magnification matrix; Calculate the product of the reduced compliance matrix, the compliance amplification matrix, and the total environmental load vector to generate the theoretically damaged simulation state vector.

5. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 1, characterized in that, Topological similarity is calculated based on the first difference and the second difference, including: Define the first difference as the actual residual vector; The second difference is defined as the theoretical residual vector; Based on the cosine similarity algorithm, the degree of parallelism between the actual residual vector and the theoretical residual vector in the direction of the vector space is calculated; The degree of parallelism is defined as topological similarity.

6. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 5, characterized in that, Based on the cosine similarity algorithm, the parallelism between the actual residual vector and the theoretical residual vector in the vector space direction is calculated, including: Introduce a pre-defined numerical stability term; Calculate the dot product of the actual residual vector and the theoretical residual vector; Calculate the magnitude of the actual residual vector and the first sum of the numerical stability terms; Calculate the magnitude of the theoretical residual vector and the second sum of the numerical stability terms; The degree of parallelism is obtained by dividing the dot product by the product of the first and second sums.

7. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 1, characterized in that, By combining fault intensity scalar and topological similarity, a structural safety risk index is generated, including: Invoke the preset design limit strength, preset confidence threshold, and preset decision sensitivity constant; Calculate the ratio of the failure strength scalar to the design ultimate strength to generate the failure degree component; Calculate the difference between the topological similarity and the confidence threshold; Based on the difference and the decision sensitivity constant, the confirmation confidence component is calculated using the Sigmoid function. The structural safety risk index is generated by weighting and summing the damage degree component and the confirmation confidence component.

8. The online monitoring and anomaly diagnosis system for water conservancy projects based on edge computing according to claim 2, characterized in that, The methods for obtaining water level correction factors and temperature correction factors include: Collect historical health monitoring data of the structure; Based on the principle of least squares, an optimization objective is constructed for the error function between historical health monitoring data and theoretical monitoring values; By minimizing the error function, the water level correction coefficient and the temperature correction coefficient are calculated.

Images