Dam monitoring analysis method and system based on superconducting magnetic quantum sensor

Abstract

The application discloses a dam monitoring analysis method and system based on a superconducting magnetic quantum sensor, relates to the technical field of dam safety monitoring and intelligent analysis, and comprises the following steps: constructing a FOC-DFT signal processing combined with a DPD cost function to realize high-resolution positioning of an abnormal source, enhance weak signal extraction and noise resistance, realizing independent extraction of stress, seepage and crack characteristics by combining spatial clustering with PICA blind source decomposition, improving the resolution capability for multi-physical coupling effects, realizing inversion calculation of stress, seepage and magnetic permeability parameters by combining a multi-physical field coupling model with sparse Bayesian inversion and compressed sensing, improving the accuracy of abnormal mechanism analysis, and realizing trend prediction and risk assessment of dam hidden dangers by combining LSTM time series modeling with a comprehensive abnormal scoring mechanism, and improving the foresight and intelligent level of early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dam safety monitoring and intelligent analysis, in particular to a dam monitoring and analysis method and system based on superconducting magnetic quantum sensors. BACKGROUND

[0002] As an important infrastructure of water conservancy projects, the safe and stable operation of the dam is directly related to the flood control and waterlogging prevention capacity and the economic and social development of the surrounding area. With the development of monitoring technology, the evaluation of the health status of the dam has gradually changed from traditional manual inspection to automation and intelligence. In recent years, with the development of the Internet of Things, geophysical exploration and artificial intelligence technology, the academic and engineering communities have gradually explored methods based on electromagnetic field monitoring, geological survey data fusion and three-dimensional modeling to improve the accuracy of dam hazard identification and early warning.

[0003] For complex dam safety problems, the prior art has the following deficiencies: in signal processing, traditional frequency domain analysis methods cannot effectively distinguish magnetic field disturbances caused by different physical mechanisms such as stress, seepage and cracks, resulting in insufficient accuracy of abnormal source positioning; in the aspects of prediction and early warning, most methods lack time series modeling capability and cannot make trend inferences on potential risks of the dam. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a dam monitoring and analysis method and system based on superconducting magnetic quantum sensors to solve the problem that traditional frequency domain analysis methods cannot effectively distinguish magnetic field disturbances caused by different physical mechanisms such as stress, seepage and cracks, resulting in insufficient accuracy of abnormal source positioning, and most methods lack time series modeling capability and cannot make trend inferences on potential risks of the dam.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides a dam monitoring and analysis method based on superconducting magnetic quantum sensors, which comprises,

[0008] A three-dimensional model of the dam is constructed, superconducting magnetic quantum sensors are assigned to each grid, superconducting magnetic quantum sensors are used to collect dam magnetic field signal data, FOC-DFT processing is performed, a DPD cost function is constructed, and the position of the abnormal source is determined;

[0009] Based on the weighted spatial magnetic field data of the abnormal source position, the grid is divided into sub-grids using the k-means algorithm, the frequency components are extracted using the fast Fourier transform, and the characteristic components of stress, seepage and cracks are decomposed by the PICA algorithm;

[0010] respectively, for the magnetic field-stress coupling model, the stress characteristics of the sub-grid points are solved using linear regression method, for the magnetic field-conductivity response model, the percolation characteristics of the sub-grid points are calculated using sparse Bayesian learning inversion, and for the optimization objective function of the magnetic field-conductivity response, the magnetic permeability characteristics of the sub-grid points are calculated using compressed sensing inversion;

[0011] The extracted features are generated into a tensor, an LSTM network model is constructed to predict a trend tensor, and a comprehensive anomaly score is calculated;

[0012] A visual interface is constructed to display a three-dimensional model of the dam and abnormal points, and dam monitoring data generated by analysis is stored.

[0013] As a preferred scheme of the dam monitoring and analysis method based on the superconducting magnetic quantum sensor, wherein: the dam magnetic field signal data is collected, FOC-DFT processing is performed, a DPD cost function is constructed, and the abnormal source position is determined, including:

[0014] Geological survey data of the dam are collected, including soil conductivity, permeability coefficient and stress distribution;

[0015] Based on the design drawings of the dam, a three-dimensional model of the dam is constructed using three-dimensional software, and the three-dimensional model is discretized into an initial grid, and each grid point records coordinates and corresponding geological survey data;

[0016] The geological survey data of each grid is normalized, and a comprehensive risk factor of each grid point is calculated based on weighted linear combination;

[0017] Based on historical regression statistics, a high-risk threshold is set, and grid points greater than the high-risk threshold are selected from the comprehensive risk factor to generate a high-risk point set;

[0018] The high-risk point set is orthogonally projected onto the two-dimensional dam foundation plane of the dam to generate two-dimensional projection coordinates, and Delaunay triangulation is performed to generate an initial triangle set;

[0019] The cross product method is used to calculate the triangle in the initial triangle set that covers the most high-risk points, and based on the coverage range of the maximum triangle, the experience method is used to perform local grid encryption in the maximum triangle and its adjacent area;

[0020] Based on the boundary of the maximum triangle, the incremental Delaunay triangulation algorithm is used to generate a two-dimensional non-uniform grid, and the three-dimensional coordinates are restored by back-projection to the three-dimensional space;

[0021] Based on the three-dimensional space grid point vertex, the superconducting magnetic quantum sensor is distributed, and if the node exceeds the budget of the number of sensors, the vertex with a high comprehensive risk factor is preferentially selected;

[0022] The magnetic field signal collected by the superconducting magnetic quantum sensor is subjected to environmental parameter correction, and the corrected magnetic field signal is subjected to FOC-DFT processing;

[0023] The FOC-DFT processing includes constructing a fourth-order cumulant FOC matrix, performing singular value decomposition on the fourth-order cumulant FOC matrix, separating out a signal subspace and a noise subspace, and extracting a magnetic field feature corresponding to the signal subspace, constructing a normalized DFT matrix, and fusing the signal subspace vector into a DPD cost function;

[0024] By minimizing the DPD cost function, the position of the abnormal source is determined, and a denoised magnetic field feature vector is generated.

[0025] As a preferred scheme of the dam monitoring and analysis method based on the superconducting magnetic quantum sensor, wherein: the weighted spatial magnetic field data is decomposed into stress, seepage and crack characteristic components by the PICA algorithm, including:

[0026] The denoised magnetic field feature data is mapped into a three-dimensional space grid, and spatial weighted magnetic field data is generated based on the magnetic field feature vector;

[0027] Based on the magnetic field feature vector, the spatial coordinates of the abnormal source are located using the direction of arrival (DOA), the distance matrix of the grid points and the abnormal source is calculated, and the k-means algorithm is used for clustering. Based on the clustering result, a sub-grid set is generated, the frequency components of the spatial weighted magnetic field data on each sub-grid are extracted using the fast Fourier transform, and the PICA algorithm is used to decompose the stress, seepage and crack characteristic components;

[0028] The PICA algorithm decomposes the stress, seepage and crack characteristic components, including centering and whitening processing for each frequency component, constructing an initial mixing matrix for each sub-grid using the azimuth and elevation, adding frequency prior weight to each column of the initial mixing matrix, defining a PICA optimization objective function, maximizing the non-Gaussianity and adding spatial and frequency prior regularization;

[0029] The PICA optimization objective function is solved using fixed point iteration optimization, and an independent component matrix of each sub-grid is obtained, and the characteristic components are reconstructed.

[0030] As a preferred scheme of the dam monitoring and analysis method based on the superconducting magnetic quantum sensor, wherein: the construction of the magnetic field-stress coupling model, the magnetic field-conductivity response model and the crack magnetic field model includes:

[0031] The stress magnetic field component, the seepage magnetic field component and the fracture magnetic field component are time-averaged respectively to generate a static stress magnetic field, a static magnetic field and a static fracture magnetic field, and a magnetic field-stress coupling model, a magnetic field-conductivity response model and a fracture magnetic field model are constructed respectively;

[0032] For the magnetic field-stress coupling model, a linear regression method is directly used to solve the stress characteristics of the sub-grid points, and a central finite difference method is used to calculate the stress gradient of the stress characteristics in space three-dimensionally;

[0033] For the magnetic field-conductivity response model, a sparse Bayesian learning inversion is used to calculate the seepage characteristics of the sub-grid points;

[0034] The sparse Bayesian learning inversion is used to calculate the seepage value of the sub-grid points, including discretizing the magnetic field-conductivity response model and constructing an optimization objective function of the magnetic field-conductivity response;

[0035] An expectation maximization algorithm is used to solve the optimization objective function of the magnetic field-conductivity response to obtain the conductivity characteristics of each sub-grid, and a Delaunay triangulation is used to generate a seepage path;

[0036] The optimization objective function of the magnetic field-conductivity response is used to calculate the magnetic permeability characteristics of the sub-grid points by using a compressed sensing inversion;

[0037] The compressed sensing inversion is used to calculate the magnetic permeability characteristics of the sub-grid points, including discretizing the fracture magnetic field model to construct an optimization objective function of the fracture magnetic field;

[0038] A basis pursuit denoising algorithm is used to solve the optimization objective function of the fracture magnetic field to obtain the magnetic permeability characteristics of each sub-grid point.

[0039] As a preferred scheme of the dam monitoring analysis method based on the superconducting magnetic quantum sensor, the method comprises the following steps:

[0040] The stress characteristics, the seepage value and the magnetic permeability of each sub-grid point are normalized and merged according to time to generate a sequence tensor of time series;

[0041] An LSTM network model is constructed, historical sequence tensors are collected for training, the trained LSTM network model is used to predict the sequence tensor, and a prediction trend tensor is obtained;

[0042] For each sub-grid point and time step, a comprehensive anomaly score is calculated, a historical regression estimate is used to set an anomaly score threshold, and the sub-grid points greater than the anomaly score threshold are marked.

[0043] As a preferred scheme of the dam monitoring analysis method based on the superconducting magnetic quantum sensor, wherein: the constructed visual interface displays the three-dimensional model of the dam and the abnormal points, including:

[0044] The visual interface is constructed using the Three.js library to display the three-dimensional model of the dam, and based on the marked sub-grid points, the three-dimensional model of the dam is marked at the corresponding position;

[0045] Add mouse hovering function, when the mouse hovers over the sub-grid point, display the abnormal score and coordinates;

[0046] Support perspective adjustment, users can rotate, zoom in or out the perspective by dragging the mouse, or click on the abnormal point to jump to the corresponding position.

[0047] As a preferred scheme of the dam monitoring analysis method based on the superconducting magnetic quantum sensor, wherein: the dam monitoring data generated by the analysis is stored, including:

[0048] The dam monitoring data generated by the analysis is stored in the database, and security access measures are set, the database stores the data in the cloud backup, and regularly checks the integrity of the stored data and the backup data, and generates an integrity detection record after the detection is completed. Synchronously stored in the database.

[0049] Secondly, the application provides a dam monitoring analysis system based on a superconducting magnetic quantum sensor, comprising,

[0050] The collection and processing module is used to construct a three-dimensional model of the dam, assign a superconducting magnetic quantum sensor to each grid, collect dam magnetic field signal data using the superconducting magnetic quantum sensor, perform FOC-DFT processing, construct a DPD cost function, and determine the abnormal source position;

[0051] The extraction and decomposition module is used to weight the spatial magnetic field data based on the abnormal source position, use the k-means algorithm to divide the grid into sub-grids, use the fast Fourier transform to extract the frequency component, and use the PICA algorithm to decompose the stress, seepage and crack characteristic components;

[0052] The construction and calculation module is used to construct a magnetic field-stress coupling model, a magnetic field-conductivity response model and a crack magnetic field model respectively, for the magnetic field-stress coupling model, the stress characteristics of the sub-grid points are solved using linear regression method, for the magnetic field-conductivity response model, the seepage characteristics of the sub-grid points are calculated using sparse Bayesian learning inversion, and the magnetic permeability characteristics of the sub-grid points are calculated using compressed sensing inversion for the optimization objective function of the magnetic field-conductivity response;

[0053] The prediction module is used to generate a tensor of the extracted features, construct an LSTM network model to predict a trend tensor, and calculate a comprehensive abnormal score;

[0054] A visualization storage module is configured to build a visualization interface to display a three-dimensional model of the dam and abnormal points, and to store dam monitoring data generated by analysis.

[0055] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the dam monitoring analysis method based on a superconducting magnetic quantum sensor according to the first aspect of the present application.

[0056] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements any step of the dam monitoring analysis method based on a superconducting magnetic quantum sensor according to the first aspect of the present application.

[0057] The present application has the following beneficial effects: the present application realizes high-resolution positioning of abnormal sources by FOC-DFT signal processing combined with DPD cost function construction, enhances weak signal extraction and noise resistance, realizes independent extraction of stress, seepage and crack characteristics by spatial clustering combined with PICA blind source decomposition, improves the resolution capability of multi-physical coupling effect, realizes inversion calculation of stress, seepage and magnetic permeability parameters by multi-physical field coupling model combined with sparse Bayesian inversion and compressed sensing, improves the accuracy of abnormal mechanism analysis, realizes trend prediction and risk assessment of dam hidden dangers by LSTM time series modeling combined with comprehensive abnormal scoring mechanism, and improves the foresight and intelligent level of early warning. BRIEF DESCRIPTION OF DRAWINGS

[0058] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0059] Fig. 1 The flowchart of the dam monitoring analysis method based on a superconducting magnetic quantum sensor in embodiment 1.

[0060] Fig. 2 The schematic diagram of the dam monitoring analysis system based on a superconducting magnetic quantum sensor in embodiment 1.

[0061] Fig. 3 The signal processing flowchart in embodiment 1.

[0062] Fig. 4 The feature decomposition and inversion flowchart in embodiment 1. DETAILED DESCRIPTION

[0063] In order to make the above objectives, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0064] In the following description, a lot of specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0065] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or selected from other embodiments.

[0066] Embodiment 1, reference Figs. 1 to 4 For the first embodiment of the present application, the embodiment provides a dam monitoring analysis method based on superconducting magnetic quantum sensor, comprising the following steps:

[0067] S1, constructing a three-dimensional model of the dam, assigning a superconducting magnetic quantum sensor to each grid, collecting dam magnetic field signal data using the superconducting magnetic quantum sensor, performing FOC-DFT processing, constructing a DPD cost function, and determining the location of the abnormal source;

[0068] Preferably, collect the geological survey data of the dam, including soil conductivity, permeability coefficient and stress distribution;

[0069] Based on the design drawings of the dam, use three-dimensional software (such as CAD) to construct a three-dimensional model of the dam, and discretize the three-dimensional model into initial grids, and record the coordinates and corresponding geological survey data of each grid point;

[0070] The geological survey data of each grid is normalized, and the comprehensive risk factor of each grid point is calculated based on weighted linear combination, formula:

[0071] ,

[0072] Wherein, is the comprehensive risk factor of the lth monitoring point, , and are the conductivity, permeability coefficient and stress distribution of the lth monitoring point, respectively;

[0073] Based on historical regression statistics, set a high risk threshold, select grid points greater than the high risk threshold from the comprehensive risk factor, and generate a high risk point set;

[0074] Orthogonal projection of the high-risk point set to the two-dimensional dam foundation plane of the dam generates two-dimensional projection coordinates, and Delaunay triangulation is performed to generate an initial triangle set;

[0075] The cross product method is used to calculate the triangle in the initial triangle set that covers the most high-risk points, and based on the coverage range of the maximum triangle, the experience method is used to perform local grid encryption in the maximum triangle and its adjacent area;

[0076] Based on the boundary of the maximum triangle, a two-dimensional non-uniform grid is generated using the incremental Delaunay triangulation algorithm, and is projected back to the three-dimensional space to restore the three-dimensional coordinates;

[0077] Based on the vertices of the three-dimensional space grid points, superconducting magnetic quantum sensors are assigned, and if the number of nodes exceeds the budget of the sensor number, the vertices with high comprehensive risk factors are preferentially selected;

[0078] The magnetic field signals collected by the superconducting magnetic quantum sensor are subjected to environmental parameter correction, and the corrected magnetic field signals are subjected to FOC-DFT processing;

[0079] The FOC-DFT processing includes constructing a fourth-order cumulant FOC matrix, formula:

[0080] ,

[0081] Wherein, is the fourth-order cumulant FOC matrix, is the mathematical expectation calculation symbol, , , and are different four sampling times, and are the corrected magnetic field signal feature data and the complex conjugate of the corrected magnetic field signal feature data, respectively;

[0082] Singular value decomposition is performed on the fourth-order cumulant FOC matrix to separate the signal subspace and the noise subspace (the right singular vector represents the signal subspace), and the magnetic field features corresponding to the signal subspace are extracted, formula:

[0083] ,

[0084] Wherein, is the denoised magnetic field feature data at time t, is the denoised magnetic field signal subspace vector matrix, and H is the conjugate transpose symbol;

[0085] A normalized DFT matrix is constructed, formula:

[0086] ,

[0087] wherein, is a normalized DFT matrix with size , is the number of superconducting magnetic quantum sensors, is the imaginary unit, and are the indices of the rows and columns of the DFT matrix, respectively, representing the sensor number and the frequency domain component;

[0088] Based on the normalized DFT matrix, the signal subspace vector is incorporated into the DPD cost function, formula:

[0089] ,

[0090] wherein, is the DPD cost function, representing the likelihood of the anomaly source in the spatial angle , and are the azimuth and elevation angles inside the dam, respectively;

[0091] By minimizing the DPD cost function, the anomaly source position is determined, and the denoised magnetic field feature vector is generated, formula:

[0092] ,

[0093] wherein, is the magnetic field feature vector at time t, and are the azimuth and elevation angles at time t, respectively.

[0094] By orthogonally projecting the high-risk point set to the two-dimensional dam foundation plane and performing Delaunay triangulation, the precise reconstruction of the geometric relationship of the high-risk area is realized. By using the cross product method to find the triangle that covers the most high-risk points, local grid encryption in the area where the risk is most concentrated is realized. Through the incremental Delaunay algorithm, a non-uniform grid is generated and back projected to the three-dimensional space, realizing a three-dimensional monitoring pattern that takes into account the whole and the local. By performing environmental correction on the magnetic field signals collected by the sensors and using FOC-DFT to construct a high-order cumulant matrix, the suppression of non-Gaussian noise and the highlighting of abnormal signals are realized. By using the normalized DFT matrix, the signal subspace information is introduced into the DPD cost function, realizing the modeling of the likelihood of the anomaly source in the angle space.

[0095] S2, based on the anomaly source position weighted spatial magnetic field data, using the k-means algorithm to divide the grid into sub-grids, using the fast Fourier transform to extract the frequency components, and using the PICA algorithm to decompose the stress, seepage and crack feature components;

[0096] Preferably, the de-noised magnetic field feature data is mapped into a three-dimensional spatial grid, and the spatially weighted magnetic field data is generated based on the magnetic field feature vector, formula:

[0097] ,

[0098] wherein, is the spatially weighted distribution magnetic field data, is the three-dimensional coordinate and time of the grid point, respectively, is a linear interpolation function, and a triangular interpolation method is used here, is the grid point at the coordinate , is the spatial weight at , and are calculated using NumPy, and are the corresponding theoretical azimuth and elevation angles, which are calculated by grid coordinate back-projection,

[0099] Based on the magnetic field feature vector, the spatial coordinates of the abnormal source are located using the direction of arrival (DOA), the distance matrix of the grid point and the abnormal source is calculated, and the k-means algorithm is used for clustering. Based on the clustering result, a sub-grid set is generated, the frequency component of the spatially weighted magnetic field data on each sub-grid is extracted using the fast Fourier transform, and the PICA algorithm is used for decomposition into stress, seepage and fracture feature components;

[0100] The PICA algorithm decomposes into stress, seepage and fracture feature components, including centering and whitening processing for each frequency component, using the direction angle and elevation angle to construct an initial mixing matrix for each sub-grid, and adding a frequency prior weight to each column of the initial mixing matrix, formula:

[0101] ,

[0102] wherein, is the adjusted mixing matrix, is the sub-grid point index, is the component index, is the sub-grid index, is the spatial weight of the th sub-grid, is the total number of sub-grids, is the The frequency prior weight of each component is set based on expert experience, and the weights of the stress component, the seepage component and the fracture component are set respectively, wherein the frequency components are divided, such as low frequency <1Hz for the stress component, medium frequency 1-10Hz for the seepage component, and high frequency >10Hz for the fracture component;

[0103] The PICA optimization objective function is defined, the non-Gaussianity is maximized, and spatial and frequency prior regularization is added, and the formula is:

[0104]

[0105] wherein, PICA is an optimization objective function, is an independent component matrix of a sub-grid k, is a non-Gaussianity expectation, is a transpose of a mixing matrix of the kth sub-grid, is a jth component of the mixing matrix of the kth sub-grid, used for measuring the independence of the component, is a non-Gaussianity measurement function, is a regularization parameter of data fitting error, is a transpose symbol, is whitened data of the sub-grid k, is a spatial-frequency prior regularization parameter, set based on an experimental method;

[0106] The fixed point iteration optimization is used to solve the PICA optimization objective function, to obtain the independent component matrix of each sub-grid, and to be reconstructed into a characteristic component.

[0107] By mapping the denoised magnetic field characteristic data to a three-dimensional space grid, spatially weighted magnetic field data is generated, the spatial resolution and signal-to-noise ratio of the magnetic field data are improved, the problem of low positioning accuracy caused by noise interference in traditional magnetic field data processing is solved, through the DOA algorithm and k-means clustering, high-precision positioning and spatial segmentation of the abnormal source are realized, through the centralization and whitening processing of the frequency components of each sub-grid, an initial mixing matrix is constructed, and an adjustment based on the frequency prior weight is added to each column, so that the decomposition result is more in line with the geophysical law, and the physical interpretation of the stress, seepage and fracture components is enhanced.

[0108] S3, a magnetic field-stress coupling model, a magnetic field-conductivity response model and a fracture magnetic field model are respectively constructed, for the magnetic field-stress coupling model, a linear regression method is used to solve the stress characteristics of the sub-grid points, for the magnetic field-conductivity response model, a sparse Bayesian learning inversion is used to calculate the seepage characteristics of the sub-grid points, and for the optimization objective function of the magnetic field-conductivity response, a compressed sensing inversion is used to calculate the magnetic permeability characteristics of the sub-grid points;

[0109] ​Preferably, the stress magnetic field component, the seepage magnetic field component and the fracture magnetic field component are time-averaged respectively to generate static stress magnetic field, static magnetic field and static fracture magnetic field, and a magnetic field-stress coupling model, a magnetic field-conductivity response model and a fracture magnetic field model are constructed respectively, formula:

[0110] ,

[0111] ,

[0112] ,

[0113] wherein, , and are the average stress magnetic field, the average seepage magnetic field and the average fracture magnetic field at the sub-grid point coordinates , is a magnetic-force coupling coefficient, which is measured by dam material permeability experiment, is a stress feature at the sub-grid point coordinates , is a vacuum permeability (constant), is a conductivity feature of the sub-grid point , is a current density of the sub-grid point r, which is obtained based on a geological model, is a sub-grid unit volume, is a discrete grid set of a monitoring area, is a permeability feature of the sub-grid point r, is a stress gradient of the sub-grid point r, is a seepage path Boolean value of the sub-grid point r;

[0114] For the magnetic field-stress coupling model, a linear regression method is directly used to solve the stress feature of the sub-grid point, and a central finite difference method is used to calculate the stress gradient of the stress feature in space three-dimension;

[0115] For the magnetic field-conductivity response model, a sparse Bayesian learning inversion is used to calculate the seepage feature of the sub-grid point;

[0116] The sparse Bayesian learning inversion used to calculate the seepage value of the sub-grid point includes discretizing the magnetic field-conductivity response model, and constructing an optimization objective function of the magnetic field-conductivity response, formula:

[0117] ,

[0118] wherein, is the optimization objective function of the magnetic field-conductivity response, is a data fitting error, is a forward operator matrix of current density, which is set based on current density and sub-grid geometry, is a sparse prior, is an index of traversing the midpoint of the sub-grid, is a regularization parameter of the sparse prior;

[0119] solving the optimization objective function of the magnetic field-conductivity response using an expectation maximization algorithm to obtain the conductivity feature of each sub-grid, using Delaunay triangulation to generate a percolation path, formula:

[0120] ,

[0121] wherein, is a percolation path distribution (three-dimensional Boolean number), indicates that the path exists, otherwise indicates that the path does not exist, is a conductivity feature threshold value set based on geological experiments;

[0122] The optimization objective function of the magnetic field-conductivity response is calculated using a compressed sensing inversion to calculate the magnetic permeability feature of the sub-grid point;

[0123] The compressed sensing inversion calculates the magnetic permeability feature of the sub-grid point, including discretizing the fracture magnetic field model, and constructing an optimization objective function of the fracture magnetic field, formula:

[0124] ,

[0125] wherein, is an optimization objective function of the fracture magnetic field, is a data fitting error of the discretized fracture magnetic field model, is a forward operator matrix of magnetic permeability, is a regularization parameter of the discretized fracture magnetic field;

[0126] The base pursuit denoising algorithm is used to solve the optimization objective function of the fracture magnetic field to obtain the magnetic permeability feature of each sub-grid point.

[0127] By modeling different physical quantities respectively, the decoupling and quantitative characterization of the three key elements inside the dam or geological body are realized, the mapping relationship between a large amount of measurement data and the theoretical model can be quickly established through the linear regression method, the high computational burden caused by complex numerical solution is avoided, the inversion accuracy is improved while maintaining the sparsity of the model through sparse Bayesian learning, overfitting and redundant solutions are effectively avoided, and the fracture magnetic field features can be accurately restored under the condition of insufficient sampling through compressed sensing inversion, which significantly reduces the dependence on the amount of observation data.

[0128] S4, generate the extracted feature tensor, construct the LSTM network model to predict the trend tensor, and calculate the comprehensive anomaly score;

[0129] Preferably, the stress feature, seepage value and magnetic permeability of each sub-grid point are normalized and merged according to time to generate a time sequence sequence tensor;

[0130] An LSTM network model is constructed, historical sequence tensors are collected for training, and the trained LSTM network model is used to predict the sequence tensor to obtain a predicted trend tensor;

[0131] For each sub-grid point and time step, the comprehensive anomaly score is calculated, and the formula is:

[0132] ,

[0133] Wherein, is the sub-grid point The anomaly score in the future time , , and are weighted weights, and the importance proportions of the three tensors are set based on expert experience;

[0134] Based on historical regression estimation, set an anomaly score threshold, and mark the sub-grid points greater than the anomaly score threshold.

[0135] By constructing a long short-term memory (LSTM) network model, using historical sequence tensors for training, and using the trained model to predict the input sequence tensor, a predicted trend tensor is generated, realizing high-precision prediction of the future trend of a complex system with multiple parameters (stress, seepage, magnetic permeability). Through weighted summation, multiple parameters (stress, seepage, magnetic permeability) are integrated into a single anomaly score index, quantifying the abnormality of the sub-grid point in the future time step.

[0136] S5, build a visual interface to display the dam three-dimensional model and abnormal points, and store the dam monitoring data generated by the analysis;

[0137] Preferably, the Three.js library is used to build a visual interface to display the dam three-dimensional model, and based on the marked sub-grid points, the corresponding positions of the dam three-dimensional model are marked;

[0138] Add a mouse hover function, which displays the anomaly score and coordinates when the mouse hovers over the sub-grid point;

[0139] Support perspective adjustment, users can rotate, zoom in or out of the perspective by dragging the mouse, or jump to the corresponding position by clicking the abnormal point.

[0140] Through the Three.js library, the system can generate a high-precision three-dimensional model of the dam and intuitively present the spatial distribution of the abnormal points. The positions of the abnormal points are distinguished from the normal area through markers, realizing accurate positioning and visual expression of the abnormal points in the three-dimensional space.

[0141] Further, the generated dam monitoring data is stored in the database, and security access measures are set. The database stores the data for cloud backup, and regularly checks the integrity of the stored data and the backup data. After the check is completed, an integrity check record is generated and stored in the database.

[0142] By storing the monitoring data in the database and setting security access measures, the system realizes reliable storage and protection of the data, significantly improving the integrity and traceability of the data.

[0143] The embodiment also provides a dam monitoring and analysis system based on a superconducting magnetic quantum sensor, which comprises:

[0144] The collection and processing module is configured to construct a three-dimensional model of the dam, assign a superconducting magnetic quantum sensor to each grid, collect dam magnetic field signal data using the superconducting magnetic quantum sensor, perform FOC-DFT processing, construct a DPD cost function, and determine the position of the abnormal source.

[0145] The extraction and decomposition module is configured to weight the spatial magnetic field data based on the position of the abnormal source, use a k-means algorithm to divide the grid into subgrids, use a fast Fourier transform to extract frequency components, and use a PICA algorithm to decompose stress, seepage, and crack feature components.

[0146] The construction and calculation module is configured to construct a magnetic field-stress coupling model, a magnetic field-conductivity response model, and a crack magnetic field model, respectively. For the magnetic field-stress coupling model, the stress characteristics of the subgrid points are solved using a linear regression method. For the magnetic field-conductivity response model, the seepage characteristics of the subgrid points are calculated using sparse Bayesian learning inversion. For the optimization objective function of the magnetic field-conductivity response, the magnetic permeability characteristics of the subgrid points are calculated using compressed sensing inversion.

[0147] The prediction module is configured to generate a tensor from the extracted features, construct an LSTM network model to predict a trend tensor, and calculate a comprehensive abnormal score.

[0148] The visualization storage module is configured to construct a visualization interface to display the three-dimensional model of the dam and the abnormal points, and store the generated dam monitoring data.

[0149] The embodiment also provides a computer device suitable for the dam monitoring and analysis method based on a superconducting magnetic quantum sensor, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the dam monitoring and analysis method based on the superconducting magnetic quantum sensor as proposed in the above embodiment.

[0150] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved through WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0151] The embodiment also provides a storage medium having a computer program stored thereon. The program is executed by a processor to realize the dam monitoring and analysis method based on the superconducting magnetic quantum sensor as proposed in the above embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0152] To sum up, the application realizes high-resolution positioning of abnormal sources by FOC-DFT signal processing combined with DPD cost function construction, enhances weak signal extraction and noise resistance, realizes independent extraction of stress, seepage and crack characteristics by spatial clustering combined with PICA blind source decomposition, improves the resolution capability of multi-physical coupling effect, realizes inversion calculation of stress, seepage and magnetic permeability and the like parameters by multi-physical field coupling model combined with sparse Bayesian inversion and compressed sensing, improves the accuracy of abnormal mechanism analysis, realizes trend prediction and risk assessment of dam hidden dangers by LSTM time series modeling combined with comprehensive abnormal scoring mechanism, and improves the foresight and intelligent level of early warning.

[0153] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application and not to limit the application, although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the application, which should be covered in the scope of the claims of the application.

Claims

1. A dam monitoring analysis method based on superconducting magnetic quantum sensors, characterized in that: The application relates to a dam monitoring method based on magnetic field and machine learning. The application comprises the following steps: building a three-dimensional model of a dam, assigning superconducting quantum magnetic sensors to each grid, collecting dam magnetic field signal data using the superconducting quantum magnetic sensors, performing FOC-DFT processing, building a DPD cost function, and determining the position of an abnormal source; based on the weighted spatial magnetic field data of the abnormal source position, using the k-means algorithm to divide the grid into a sub-grid, using the fast Fourier transform to extract the frequency component, and using the PICA algorithm to decompose the stress, seepage and crack characteristic components; respectively building a magnetic field-stress coupling model, a magnetic field-conductivity response model and a crack magnetic field model, for the magnetic field-stress coupling model, using a linear regression method to solve the stress characteristics of the sub-grid points, for the magnetic field-conductivity response model, using sparse Bayesian learning inversion to calculate the seepage characteristics of the sub-grid points, and for the optimization objective function of the magnetic field-conductivity response, using compressed sensing inversion to calculate the magnetic permeability characteristics of the sub-grid points; generating a tensor by extracting the features, building an LSTM network model to predict a trend tensor, and calculating a comprehensive anomaly score; 2. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 1, wherein: building a visual interface to display the three-dimensional model of the dam and the abnormal points, and storing the dam monitoring data generated by the analysis. The application comprises the following steps: collecting dam geological survey data, including soil conductivity, permeability coefficient and stress distribution; based on the design drawings of the dam, using three-dimensional software to build a three-dimensional model of the dam, and discretizing the three-dimensional model into initial grids, each grid point recording coordinates and corresponding geological survey data; normalizing the geological survey data of each grid, and based on weighted linear combination, calculating the comprehensive risk factor of each grid point; based on historical regression statistics, setting a high-risk threshold, and selecting grid points greater than the high-risk threshold from the comprehensive risk factor to generate a high-risk point set; orthogonal projecting the high-risk point set to the two-dimensional dam foundation plane of the dam to generate two-dimensional projection coordinates, and performing Delaunay triangulation to generate an initial triangle set; using the cross product method to calculate the triangle in the initial triangle set that covers the most high-risk points, and based on the coverage range of the maximum triangle, using the empirical method to perform local grid encryption in the maximum triangle and its adjacent area; based on the boundary of the maximum triangle, using the incremental Delaunay triangulation algorithm to generate a two-dimensional non-uniform grid, and back-projecting it back to the three-dimensional space to restore the three-dimensional coordinates; based on the vertices of the three-dimensional space grid points, assigning superconducting quantum magnetic sensors, and if the number of nodes exceeds the budget of the sensors, preferentially selecting the vertices with high comprehensive risk factors; correcting the magnetic field signals collected by the superconducting quantum magnetic sensors according to the environmental parameters, and performing FOC-DFT processing on the corrected magnetic field signals; the FOC-DFT processing comprises the following steps: building a fourth-order cumulant FOC matrix, performing singular value decomposition on the fourth-order cumulant FOC matrix to separate a signal subspace and a noise subspace, extracting the magnetic field characteristics corresponding to the signal subspace, building a normalized DFT matrix, and integrating the signal subspace vector into a DPD cost function. By minimizing the DPD cost function, the abnormal source position is determined, and the denoised magnetic field feature vector is generated.

3. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 2, wherein: The weighted spatial magnetic field data is decomposed into stress, seepage and fracture feature components by the PICA algorithm, including: The denoised magnetic field feature data is mapped into a three-dimensional spatial grid, and the spatial weighted magnetic field data is generated based on the magnetic field feature vector; Based on the magnetic field feature vector, the spatial coordinates of the abnormal source are located using the direction of arrival (DOA), the distance matrix between the grid points and the abnormal source is calculated, and the k-means algorithm is used for clustering. Based on the clustering result, a sub-grid set is generated. The frequency components of the spatial weighted magnetic field data on each sub-grid are extracted using the fast Fourier transform, and the PICA algorithm is used to decompose them into stress, seepage and fracture feature components. The PICA algorithm decomposes into stress, seepage and fracture feature components, including centering and whitening processing for each frequency component, constructing an initial mixing matrix for each sub-grid using the azimuth and elevation, adding frequency prior weight to each column of the initial mixing matrix, defining the PICA optimization objective function, maximizing the non-Gaussianity and adding spatial and frequency prior regularization. The PICA optimization objective function is solved using fixed-point iteration optimization to obtain the independent component matrix of each sub-grid, and is reconstructed into feature components.

4. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 3, wherein: The construction of the magnetic field-stress coupling model, the magnetic field-conductivity response model and the fracture magnetic field model, and the solving of the model, include: The time average is performed on the stress magnetic field component, the seepage magnetic field component and the fracture magnetic field component respectively to generate the static stress magnetic field, the static magnetic field and the static fracture magnetic field, and the magnetic field-stress coupling model, the magnetic field-conductivity response model and the fracture magnetic field model are constructed respectively; For the magnetic field-stress coupling model, the stress feature of the sub-grid point is directly solved using the linear regression method, and the stress gradient of the stress feature in the spatial three-dimensional space is calculated by the central finite difference method; For the magnetic field-conductivity response model, the seepage feature of the sub-grid point is calculated using sparse Bayesian learning inversion; The sparse Bayesian learning inversion is used to calculate the seepage value of the sub-grid point, including discretizing the magnetic field-conductivity response model and constructing the optimization objective function of the magnetic field-conductivity response; The optimization objective function of the magnetic field-conductivity response is solved using the expectation maximization algorithm to obtain the conductivity feature of each sub-grid, and the seepage path is generated using Delaunay triangulation; The optimization objective function of the magnetic field-conductivity response is calculated using the compressed sensing inversion to calculate the magnetic permeability feature of the sub-grid point; The compressed sensing inversion calculates the magnetic permeability feature of the sub-grid point, including discretizing the fracture magnetic field model and constructing the optimization objective function of the fracture magnetic field; The optimization objective function of the fracture magnetic field is solved using the basis pursuit denoising algorithm to obtain the magnetic permeability feature of each sub-grid point.

5. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 4, wherein: The construction of the LSTM network model to predict the trend tensor and calculate the comprehensive anomaly score, includes: The stress feature, seepage value and magnetic permeability of each sub-grid point are normalized and merged according to time to generate a sequence tensor of time series; The LSTM network model is constructed, historical sequence tensors are collected for training, the trained LSTM network model is used to predict the sequence tensors, and a prediction trend tensor is obtained; For each sub-grid point and time step, a comprehensive anomaly score is calculated, an anomaly score threshold is set based on historical regression estimation, and sub-grid points greater than the anomaly score threshold are marked.

6. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 5, wherein: The constructed visualization interface displays a three-dimensional dam model and abnormal points, including: The visualization interface is constructed using the Three.js library to display the three-dimensional dam model, and based on the marked sub-grid points, the three-dimensional dam model is marked at the corresponding position; A mouse hovering function is added, which displays the anomaly score and coordinates when the mouse hovers over the sub-grid point; Support for adjusting the viewing angle, users can rotate and zoom the viewing angle by dragging the mouse, or click on the abnormal point to jump to the corresponding position.

7. The superconducting magnetic quantum sensor based dam monitoring analysis method of claim 6, wherein: The dam monitoring data generated by the analysis is stored, including: The dam monitoring data generated by the analysis is stored in the database, and security access measures are set. The database stores the data in the cloud backup, and periodically checks the integrity of the stored data and the backup data. After the detection is completed, the integrity detection record is stored in the database.

8. A dam monitoring analysis system based on superconducting magnetic quantum sensors, based on the dam monitoring analysis method based on superconducting magnetic quantum sensors according to any one of claims 1 to 7, characterized in that: including, The collection and processing module is used to construct a three-dimensional model of the dam, assign superconducting quantum sensor to each grid, collect dam magnetic field signal data using superconducting quantum sensor, perform FOC-DFT processing, construct DPD cost function, and determine the abnormal source position; The extraction and decomposition module is used to weight the spatial magnetic field data based on the abnormal source position, use the k-means algorithm to divide the grid into sub-grid, use the fast Fourier transform to extract the frequency component, and use the PICA algorithm to decompose the stress, seepage and crack characteristic components; The construction and calculation module is used to construct the magnetic field-stress coupling model, the magnetic field-conductivity response model and the crack magnetic field model respectively. For the magnetic field-stress coupling model, the stress characteristics of the sub-grid point are solved using linear regression method. For the magnetic field-conductivity response model, the seepage characteristics of the sub-grid point are calculated using sparse Bayesian learning inversion. For the optimization objective function of the magnetic field-conductivity response, the magnetic permeability characteristics of the sub-grid point are calculated using compressed sensing inversion. The prediction module is used to generate a tensor of extracted features, construct an LSTM network model to predict a trend tensor, and calculate a comprehensive anomaly score. The visualization storage module is used to construct a visualization interface to display a three-dimensional dam model and abnormal points, and store dam monitoring data generated by analysis. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the dam monitoring analysis method based on superconducting quantum sensor in any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the dam monitoring analysis method based on superconducting quantum sensor in any one of claims 1-7.

Images