Three-dimensional winding imaging denoising method and system based on physical constraint optimization algorithm

Abstract

The application provides a three-dimensional winding imaging denoising method and system based on a physical constraint optimization algorithm, and belongs to the field of ultrasonic imaging denoising, and comprises the following steps: collecting ultrasonic echo signals of a transformer winding and performing pretreatment to generate a joint coordinate matrix; reorganizing the joint coordinate matrix into a three-dimensional winding data parameter matrix, constructing a physical constraint term based on the amplitude-energy coupling relationship of ultrasonic waves in the winding medium, and constructing a fitness function for evaluating the quality of coordinate data in combination with the physical constraint term; initializing iteration parameters and solution space boundary parameters of the physical constraint optimization algorithm, taking the three-dimensional winding data parameter matrix as initial data, performing multiple rounds of iteration optimization until the convergence condition is met; and generating a denoised three-dimensional winding image by using the optimized coordinate parameters, through mutual checking of coordinate systems and three-dimensional point cloud reconstruction. The application effectively suppresses the interference of specific frequency band electromagnetic pulse noise, and improves the precision and clarity of three-dimensional winding imaging.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of ultrasonic imaging denoising, and particularly relates to a three-dimensional winding imaging denoising method and system based on a physical constraint optimization algorithm. BACKGROUND

[0002] As an important method for non-invasive visualization detection of internal structures of an object using high-frequency sound waves (usually in the range of 1-20 MHz), the principle of ultrasonic imaging technology is based on the reflection, scattering and attenuation characteristics of ultrasonic waves in a medium, and an image is generated by receiving and processing echo signal data. The key links of this technology include: high-frequency sound waves are generated by a transducer under the excitation of an electrical signal; the signal quality is optimized by a signal processing system (covering modules such as beamforming, filtering, amplification and digitization); and finally the processing results are presented by a display system. Thanks to its advantages such as non-invasive and real-time, ultrasonic imaging is not only widely used in medical diagnosis (such as cardiovascular and superficial organ examination), but also plays an important role in the field of industrial non-destructive testing, such as composite material layer analysis and fault detection of internal structures of power equipment (such as transformer windings).

[0003] Currently, ultrasonic imaging technology is rapidly developing in the direction of intelligentization, standardization and multi-modal fusion. The introduction of artificial intelligence has promoted the realization of intelligent scanning assistance, adaptive image optimization (such as deep learning denoising and contrast enhancement). At the same time, the differences brought by different equipment manufacturers and application scenarios also promote the increasing attention to standardization construction to ensure the safety, reliability and comparability of detection results. It is particularly worth noting that the deep fusion with other imaging modalities (such as MRI / CT) and frontier technologies (such as AI) is promoting the evolution of ultrasonic from a single morphological examination tool to a comprehensive platform integrating diagnosis and monitoring. This multi-modal fusion aims to break through the limitations of traditional ultrasound in resolution, penetration depth and functional imaging, and with the development of 5G telemedicine, new materials and other technologies, it brings new opportunities to the industrial application field including intelligent detection of power equipment.

[0004] However, in the three-dimensional ultrasonic imaging detection of power transformer windings, due to the influence of the complex electromagnetic environment and air humidity inside the transformer station, the ultrasonic signals received by the transducer are prone to noise interference, and some frequency signals cannot be removed by the filtering device in the signal processing stage, thereby affecting the three-dimensional winding imaging effect and making it difficult to accurately detect deformation faults. At the same time, some new denoising algorithms are needed to improve the accuracy of three-dimensional winding imaging technology in ultrasonic detection devices for transformer station environments.

[0005] The prior art document (CN119273565A) discloses an image denoising neural network architecture search method based on an improved multi-objective marine predator algorithm, which can optimize the denoising model, but needs to be independently trained for each candidate grid, and repeated iterations result in high computational cost, which is not suitable for practical application. SUMMARY

[0006] To solve the problems in the prior art, the application provides a three-dimensional winding imaging denoising method and system based on a physical constraint optimization algorithm.

[0007] The application adopts the following technical solutions.

[0008] The first aspect of the application provides a three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm, including the following steps:

[0009] Collecting ultrasonic echo signals of a transformer winding and performing preprocessing to generate a joint coordinate matrix containing polar coordinate system and Cartesian coordinate system parameters;

[0010] Reorganizing the joint coordinate matrix into a three-dimensional winding data parameter matrix in a time series data structure, combining the three-dimensional winding data parameter matrix, constructing a physical constraint term based on the amplitude and energy coupling relationship of ultrasonic waves in the winding medium, and combining the physical constraint term to construct a fitness function for evaluating the quality of coordinate data;

[0011] Initializing the iteration parameters and solution space boundary parameters of the physical constraint optimization algorithm, taking the three-dimensional winding data parameter matrix as the initial data, and performing multiple rounds of iteration optimization, each round of iteration including: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, one-dimensional optimization of the remaining coordinate parameters, and updating the coordinate parameter position based on the fitness function until the convergence condition is met;

[0012] Using the optimized coordinate parameters, generating a denoised three-dimensional winding image through coordinate system mutual verification and three-dimensional point cloud reconstruction.

[0013] Optionally, the collecting ultrasonic echo signals of a transformer winding and performing preprocessing includes:

[0014] Collecting original echo signals through an ultrasonic transducer array arranged on the surface of the transformer, and the transducer array is arranged according to the spatial distribution of the winding structure;

[0015] The original echo signals are sequentially subjected to band-pass filtering, beam synthesis and quadrature demodulation processing, wherein the band-pass filtering processing suppresses environmental electromagnetic noise and mechanical noise outside the operating frequency band of the ultrasonic transducer, the beam synthesis processing enhances the echo signal energy from the winding reflection surface, and the quadrature demodulation processing converts the high-frequency echo signal into a low-frequency baseband signal containing amplitude and phase information.

[0016] Optionally, the generating a joint coordinate matrix comprising polar coordinate system parameters and Cartesian coordinate system parameters comprises:

[0017] The baseband signal obtained after preprocessing is subjected to scan transformation processing to map and generate coordinate parameters in a polar coordinate system, including a distance parameter, an azimuth angle parameter, and an elevation angle parameter, wherein the distance parameter represents the straight-line distance from the winding detection point to the ultrasonic wave emission source, the azimuth angle parameter represents the azimuth angle of the winding detection point in the horizontal direction, and the elevation angle parameter represents the pitch angle of the winding detection point in the vertical direction.

[0018] The polar coordinate parameters are converted into Cartesian coordinate parameters through a coordinate conversion relationship, and the Cartesian coordinate parameters represent the three-dimensional spatial position of the winding detection point.

[0019] The polar coordinate system coordinate parameters and the Cartesian coordinate system coordinate parameters corresponding to the same time sampling point are associated and combined to generate a joint coordinate matrix.

[0020] Optionally, the constructing a physical constraint based on the amplitude and energy coupling relationship of the ultrasonic wave in the winding medium comprises:

[0021] According to the propagation characteristics of the ultrasonic wave in the winding medium of the oil-immersed transformer, an amplitude-coordinate attenuation model of the spatial coordinate point and the reflected sound wave amplitude is established, wherein the reflected sound wave amplitude attenuates with the increase of the propagation distance.

[0022] Based on the reflected sound wave amplitude, the sound wave energy flux corresponding to the current spatial coordinate parameter is calculated in combination with the physical characteristics of the transformer medium density and the sound speed.

[0023] The rated amplitude and the rated energy of the ultrasonic transducer factory calibration are taken as reference benchmarks.

[0024] Based on the amplitude-coordinate attenuation model and the current spatial coordinate parameter, the reflected sound wave amplitude is calculated inversely, and the reflected sound wave amplitude and the sound wave energy flux corresponding to the current spatial coordinate parameter are compared with the corresponding rated reference benchmarks to form amplitude residuals and energy residuals.

[0025] Based on the amplitude residuals and the energy residuals, a physical constraint term for constraining the coordinate parameter optimization process is constructed.

[0026] Optionally, the constructing a fitness function for evaluating the quality of the coordinate data in combination with the physical constraint comprises:

[0027] The spatial coordinate parameter values of the three-dimensional winding data parameter matrix at a plurality of consecutive acquisition time points are obtained.

[0028] For each spatial coordinate parameter, the normalized change amount between adjacent acquisition time points is calculated.

[0029] The normalized variation of each spatial coordinate parameter is multiplied by a pre-configured weight coefficient, and then superimposed and summed to generate a weighted variation evaluation term;

[0030] The weighted variation evaluation term is combined with a physical constraint term to generate a fitness function.

[0031] Optionally, the random selection of multiple coordinate parameters for multi-dimensional joint optimization includes:

[0032] In each iteration, a plurality of coordinate parameters are randomly selected from the three-dimensional winding data parameter matrix to construct a multi-dimensional optimization subspace;

[0033] In the multi-dimensional optimization subspace, the current optimal parameter position is searched based on the fitness function;

[0034] The coordinate difference between the current position and the optimal parameter position in the corresponding parameter dimension is calculated;

[0035] The coordinate difference is scaled according to a pre-set movement step factor to generate a parameter position adjustment amount;

[0036] The parameter position adjustment amount is added to the current position to obtain an updated parameter value;

[0037] The fitness value corresponding to the updated parameter value is calculated based on the fitness function. If it is better than the current value, the update is accepted, otherwise the current value is retained;

[0038] The updated parameter value is subjected to solution space boundary judgment, and the parameter value exceeding the solution space boundary is corrected, and the solution space boundary is defined by the physical size range of the transformer winding structure.

[0039] Optionally, the one-dimensional optimization of the remaining coordinate parameters includes:

[0040] For each remaining coordinate parameter that is not selected to participate in multi-dimensional joint optimization, the following operations are respectively performed in its single parameter dimension:

[0041] In the current iteration, two different historical parameter values are randomly selected as references;

[0042] The difference vector between the two reference values is calculated, and a random disturbance factor is added to generate a one-dimensional search direction vector;

[0043] The current value of the parameter is adjusted along the direction vector to obtain an updated parameter value;

[0044] The fitness value corresponding to the updated parameter value is calculated based on the fitness function. If it is better than the current value, the update is accepted, otherwise the current value is retained;

[0045] The updated parameter value is subjected to solution space boundary judgment, and the parameter value exceeding the solution space boundary is corrected.

[0046] Optionally, the convergence condition is determined after the multi-dimensional optimization and one-dimensional optimization of all parameters in each iteration are completed, including:

[0047] When a round of iteration is completed, the fitness function value of all updated parameters is calculated;

[0048] If the updated fitness function value reaches a maximum value, and the winding deformation sensitive parameter is not greater than a preset sensitive threshold, it is determined that the convergence is reached, and the iteration is terminated;

[0049] Otherwise, the next round of iteration is continued until the maximum number of iterations is reached.

[0050] Optionally, the three-dimensional winding image after denoising is generated by mutual verification of the coordinate system and three-dimensional point cloud reconstruction, including:

[0051] The optimized polar coordinate system coordinate parameters and Cartesian coordinate system coordinate parameters are mutually converted according to the spatial geometric conversion relationship, the consistency between the converted coordinate parameters and the corresponding optimized coordinate parameters under the target coordinate system is judged by calculating the spatial position deviation between the converted coordinate parameters and the corresponding optimized coordinate parameters under the target coordinate system, and the abnormal data points are identified and removed;

[0052] The weighted average values of the verified polar coordinate system coordinate parameters and Cartesian coordinate system coordinate parameters are calculated respectively to obtain the final spatial coordinate sequence after denoising;

[0053] Based on the final spatial coordinate sequence, a three-dimensional spatial structure model of the transformer winding is constructed by using a three-dimensional point cloud reconstruction algorithm;

[0054] The three-dimensional spatial structure model is subjected to surface rendering and visualization processing to generate and output a three-dimensional winding imaging image after denoising.

[0055] The present application provides a three-dimensional winding imaging denoising system based on a physical constraint optimization algorithm in the second aspect, based on the three-dimensional winding imaging denoising method based on the physical constraint optimization algorithm in the first aspect of the present application, the system comprises:

[0056] The signal acquisition and preprocessing module, the data reorganization and modeling module, the optimization algorithm execution module and the three-dimensional imaging module;

[0057] The signal acquisition and preprocessing module is used to execute the acquisition of the ultrasonic echo signal of the transformer winding and the preprocessing to generate a joint coordinate matrix containing polar coordinate system and Cartesian coordinate system parameters;

[0058] The data reorganization and modeling module is configured to reorganize the joint coordinate matrix into a three-dimensional winding data parameter matrix of time series data structure, and based on the three-dimensional winding data parameter matrix, construct a physical constraint term based on an amplitude and energy coupling relationship of ultrasonic waves in winding medium, and construct a fitness function for evaluating coordinate data quality in combination with the physical constraint term;

[0059] The optimization algorithm execution module is configured to execute iteration parameters and solution space boundary parameters of the initialized physical constraint optimization algorithm, take the three-dimensional winding data parameter matrix as initial data, and perform multiple rounds of iteration optimization, each round of iteration including: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, performing one-dimensional optimization on the remaining coordinate parameters, and updating the coordinate parameter positions based on the fitness function until a convergence condition is met.

[0060] The three-dimensional imaging module is configured to generate a denoised three-dimensional winding image through coordinate system mutual verification and three-dimensional point cloud reconstruction using the optimized coordinate parameters.

[0061] The three-dimensional winding imaging denoising method based on the physical constraint optimization algorithm provided by the present application has the following advantages compared with the prior art:

[0062] The three-dimensional winding imaging denoising method based on the physical constraint optimization algorithm provided by the present application adopts a multi-dimensional one-dimensional hybrid search optimization strategy based on energy and amplitude constraints, and is suitable for the optimization and denoising process of three-dimensional winding coordinate multi-dimensional data. Moreover, the optimization dimension can be changed according to the array dimension size during the optimization process, and compared with the traditional ultrasonic three-dimensional imaging optimization and denoising algorithm, the global optimal solution can be found more efficiently and the optimization efficiency can be adjusted flexibly.

[0063] The three-dimensional winding imaging denoising method based on the physical constraint optimization algorithm provided by the present application proposes a method of multi-dimensional one-dimensional joint parameter optimization for three-dimensional winding coordinate data according to the coupling relationship between amplitude energy and winding coordinates in noisy ultrasonic signals. For some denoising problems that cannot be solved by electromagnetic interference in substations, the multi-dimensional optimization method based on physical constraints can efficiently solve non-linear multi-modal problems and effectively remove noise interference that cannot be removed by band-pass filtering in three-dimensional winding imaging. BRIEF DESCRIPTION OF DRAWINGS

[0064] Figure 1 is a denoising optimization flowchart of the physical constraint optimization algorithm provided according to an embodiment of the present application;

[0065] Figure 2 is a three-dimensional winding deformation trend visualization display interface schematic diagram provided according to an embodiment of the present application;

[0066] Figure 3Fig. 1 is a schematic diagram of a three-dimensional winding deformation extreme point visual display interface provided according to an embodiment of the present application;

[0067] Figure 4 Fig. 4 is a schematic diagram of a transformer winding generation report interface provided according to an embodiment of the present application;

[0068] Figure 5 Fig. 5 is a flow chart of a method provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. The described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those skilled in the art without creative efforts based on the spirit of the present application shall fall within the protection scope of the present application.

[0070] In the three-dimensional ultrasonic imaging detection of a power transformer winding, the high-intensity pulse noise of a specific frequency band introduced by the complex electromagnetic environment of a transformer substation makes the pulse amplitude energy of a certain frequency band larger, resulting in the potential error jump of the ultrasonic signal receiving circuit and further resulting in the mispositioning noise data contained in the three-dimensional winding coordinate data. Moreover, the noise spectrum and the effective ultrasonic signal are highly overlapped, making it difficult to effectively eliminate the noise by using the traditional band-pass filtering method. The interference on the three-dimensional spatial coordinate parameter data is particularly significant: in the Cartesian coordinate system , the electromagnetic pulse noise with the same frequency as the ultrasonic wave will interfere with the amplitude and phase parameters of the echo signal, induce abnormal jump and deviation of the positioning point coordinates, and cause the distortion of the spatial position and geometric shape of the winding structure; in the polar coordinate system (based on the radial distance, azimuth angle and pitch angle of the transducer), the radial distance measurement is directly interfered (represented as abnormal radial fluctuation) and the azimuth angle and pitch angle deviation is introduced, which destroys the spatial relationship representation of the target point relative to the sound source. The coordinate data disturbance caused by the electromagnetic pulse noise causes the final reconstructed three-dimensional winding image to be blurred, distorted and have a decreased spatial resolution, which seriously hinders the accurate identification and positioning of the subtle deformation and other key defects of the winding. Therefore, it is urgent to develop a new signal processing scheme capable of efficiently suppressing such specific frequency band pulse noise, so as to improve the coordinate accuracy and state detection reliability of the three-dimensional ultrasonic imaging.

[0071] Under this background, as Figure 5 shown, the present application provides a three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm in Embodiment 1, which includes the following steps:

[0072] Step 1, collect the ultrasonic echo signals of the transformer winding and perform preprocessing to generate a joint coordinate matrix containing the polar coordinate system and Cartesian coordinate system parameters.

[0073] Preferably, the step 1 comprises:

[0074] Step 1.1, collecting the original echo signal by the ultrasonic transducer array arranged on the surface of the transformer, the transducer array is arranged according to the spatial distribution of the winding structure;

[0075] Further preferably, the step 1.1 comprises:

[0076] The ultrasonic transducer array device transmits a sequence of high-frequency pulse waves through the excitation circuit, part of the signal is reflected along the original path after the ultrasonic wave contacts the winding surface, and the ultrasonic echo signal received by the transducer makes the piezoelectric ceramic sheet vibrate and converts mechanical energy into electrical energy signal through its piezoelectric effect.

[0077] Specifically, since different types of winding deformation will affect the reflection of ultrasonic signals, the type of transformer winding fault can be judged by comparing the size of the reflected echo energy.

[0078] Step 1.2, preprocessing the original electrical signal to generate a preprocessed baseband signal, the preprocessing includes bandpass filtering, beam synthesis and quadrature demodulation processing;

[0079] Further preferably, the step 1.2 comprises:

[0080] Bandpass filtering is used to filter out noise outside the working frequency band of the transducer, but its suppression ability to in-band noise is limited, and such noise needs to be filtered out by the optimization algorithm, which is one of the main goals of the present application. Then, beam synthesis is performed, which adjusts the time delay of each array element by coordinating multiple signal processing algorithms to realize the coherent superposition of sound wave fronts, thereby enhancing the effective signal. Further, through quadrature demodulation processing, the high-frequency ultrasonic signal received by the transducer is converted into a low-frequency baseband signal, and the amplitude and phase information thereof are extracted, providing a basis for subsequent image reconstruction.

[0081] Step 1.3, mapping the preprocessed baseband signal to a joint coordinate matrix containing polar coordinate system and Cartesian coordinate system parameters through scan conversion, and generating structured data for subsequent iterative optimization after dynamic compression, frame correlation and gray scale mapping optimization.

[0082] Further preferably, the step 1.3 comprises:

[0083] The back-end image processing is a key step for converting the echo signal into a high-quality visual image, which includes dynamic compression, frame correlation, scan conversion and gray scale mapping;

[0084] Dynamic compression is the process of compressing the wide dynamic range of an ultrasonic echo signal to a narrow dynamic range that the display can show, in order to preserve details and avoid signal saturation or loss.

[0085] Frame correlation utilizes temporal continuity to suppress random noise and improve image smoothness;

[0086] The scanning transformation converts ultrasonic data in polar coordinates into Cartesian coordinates to adapt to the display pixel array. Specifically, it first generates polar coordinate parameters through scanning transformation, and then converts the polar coordinate parameters into Cartesian coordinate parameters through spatial coordinate transformation relationships. The winding dataset of this invention contains two sets of data in polar coordinates and Cartesian coordinates.

[0087] Gray-scale mapping is a technique that improves the visual quality of an image by changing the gray level of pixels. Gray-scale mapping can use mapping functions such as linear, logarithmic, and step quantization to compress data and preserve details in dark areas.

[0088] The signal demodulation used in this invention is quadrature demodulation, which utilizes the orthogonality of sine and cosine signals to decompose the high-frequency signal into two orthogonal components, corresponding to the real and imaginary parts of the signal, respectively.

[0089] The combined polar and Cartesian coordinate matrix generated after scanning transformation of the preprocessed baseband signal. .

[0090] For example, based on this joint coordinate matrix, the present invention can adopt a five-dimensional + one-dimensional joint optimization structure. The five-dimensional space optimization involves randomly selecting five parameters from the six parameters of the joint dataset of polar and Cartesian coordinate systems for simultaneous optimization. Due to the strict correlation between the two coordinate systems, the five-dimensional optimization simultaneously retains the values ​​in the polar coordinate system. The physical properties of sound wave propagation and the Cartesian coordinate system It provides three-dimensional geometric representation capabilities and reduces the amount of optimization data to avoid parameter redundancy, but the remaining parameter still participates in the constraint calculation. For example, a five-dimensional joint parameter space. With individual parameters pass Strict correlation, and through geometric transformation equations:

[0091]

[0092] Rigid constraints are established to form a hypersurface manifold for the five-dimensional parameters. The algorithm expressions for one-dimensional optimization and five-dimensional optimization are consistent, but the computational cost is greatly reduced. It can accurately mark contaminated data points and form closed-loop constraints with the five-dimensional parameter space transformation, thereby disrupting the inevitable continuity of coordinate jumps caused by electromagnetic pulse noise and significantly improving the noise detection rate.

[0093] The application is based on the coupling relationship between the amplitude and energy of the ultrasonic signal, and the amplitude size in the ultrasonic signal is constrained according to the energy relationship of the signal wave emitted and recovered by the ultrasonic wave to avoid noise interference in the signal, and then a fitness function is constructed to constrain the multi-dimensional one-dimensional optimization process.

[0094] Step 2, reorganize the joint coordinate matrix into a three-dimensional winding data parameter matrix of time series data structure, combine the three-dimensional winding data parameter matrix, construct a physical constraint based on the coupling relationship between the amplitude and energy of the ultrasonic wave in the winding medium, and construct a fitness function for evaluating the quality of coordinate data in combination with the physical constraint.

[0095] Step 2.1, reorganize the joint coordinate matrix into a time series data structure to generate a three-dimensional winding data parameter matrix;

[0096] Further preferably, the step 2.1 comprises:

[0097] The joint coordinate matrix of step 1 is reorganized into a time series data structure, the row index corresponds to the acquisition time, and the column vector contains six-dimensional coordinate parameters;

[0098] The three-dimensional winding data parameter matrix is pointed by the time series, and consists of six parameters in the Cartesian coordinate system and the polar coordinate system , the row index corresponds to the acquisition time, and the column vector contains six-dimensional coordinate parameters, represents the straight-line distance between the winding coordinate point and the coordinate origin; according to the distance attenuation characteristic of the ultrasonic wave, in order to control the attenuation rate of the ultrasonic wave to be less than 5%, it is necessary to control ≤1.5m, and represents the angle between the straight line of the winding coordinate point and the coordinate origin in the polar coordinate system and axis and axis, since the position of the transducer in space is located on the front surface of the transformer, then the change limit is , the change limit is ;

[0099] Each row of the matrix represents the coordinate data of the acquisition point in each time period.

[0100] Step 2.2, construct a physical constraint based on the coupling relationship between the amplitude and energy of the ultrasonic wave in the winding medium, and construct a fitness function for evaluating the quality of coordinate data.

[0101] Further preferably, the step 2.2 comprises:

[0102] The construction of the fitness function is based on the coupling relationship between amplitude and energy to establish a weighted evaluation function, and the quantitative data fluctuation is calculated by normalizing the three-dimensional winding coordinate change. Specifically:

[0103] 1) According to the propagation characteristics of ultrasonic waves in the medium of oil-immersed transformer windings, an amplitude-coordinate attenuation model is established between the spatial coordinate point and the reflected sound wave amplitude, where the reflected sound wave amplitude attenuates with the increase of propagation distance:

[0104] According to the three-dimensional winding coordinate data, the amplitude-coordinate attenuation model is established by ultrasonic point diffusion propagation attenuation mechanism. Assuming that the target point emits or scatters amplitude attenuates with distance, the amplitude of the ultrasonic wave related to the coordinate point can be obtained:

[0105]

[0106] wherein, represents the reflected point position The reflected sound pressure amplitude unit is the unit of pressure Because the essence of sound pressure amplitude is sound pressure; is the direction factor about the angle direction in polar coordinates, which is a dimensionless coefficient, only describing the ratio of energy distribution in different directions; is the propagation path length, which refers to the distance between the reference emission point and the reflection point , with the unit of m; is the attenuation index, which is in the free field case , and is the nominal equivalent scattering intensity or source intensity, which is set to 101 Because after dividing the overall result must return to the unit of sound pressure .

[0107] 2) Based on the reflected sound wave amplitude, combined with the physical properties of transformer medium density and sound velocity, the ultrasonic wave energy flux corresponding to the current spatial coordinate parameter is calculated, and the expression is:

[0108]

[0109] wherein, is the medium density of the transformer oil passed by the ultrasonic wave, is the sound velocity, both of which can be corrected by oil temperature, is the focused peak value, with the unit of .

[0110] 3) The rated amplitude and rated energy of the ultrasonic transducer are used as reference benchmarks, and the amplitude and energy parameters of the single ultrasonic pulse wave of the transducer can be obtained according to the transducer factory setting parameters, and the rated values of the amplitude and energy parameters are and , so that the rated values of the amplitude and energy are compared with the inversion values of the amplitude and energy derived according to the coordinates in each iteration, the amplitude residual and the energy residual are formed, and the physical constraint term for constraining the coordinate parameter optimization process is constructed based on the amplitude residual and the energy residual.

[0111] Further preferably, the constructing the fitness function comprises:

[0112] Obtaining the spatial coordinate parameter values of the three-dimensional winding data parameter matrix at a plurality of continuous acquisition time points;

[0113] For each spatial coordinate parameter, the normalized change amount between adjacent acquisition time points is calculated respectively;

[0114] The normalized change amount of each spatial coordinate parameter is multiplied by a preconfigured weight coefficient and then superimposed and summed to generate a weighted change evaluation term;

[0115] The weighted change evaluation term is combined with the physical constraint term to generate a fitness function.

[0116] Specifically, the coupling relationship between the energy amplitude and the three-dimensional coordinate point is established, and the optimal solution of the three-dimensional winding coordinate error denoising is maintained by the physical constraint method:

[0117]

[0118] wherein all coordinate parameters are normalized to dimensionless values, represent the fitness function value after the iteration, starting from 0; is an array of population positions containing six parameter information, which is updated after each iteration to calculate and values, is any integer in [0, 5], and the index n=0-5 corresponds to the six parameters in order: n=0→ , n=1→ , n=2→ , n=3→ , n=4→ , n=5→vertical coordinate , for example, ; represent the proportion of each parameter in the optimal fitness function value, and The subscript corresponds to the optimization of the winding position coordinate parameters in this invention due to the measurement process. The parameter value is compared with the angle measurement value. and It will be more accurate, so strengthen polar coordinates. The proportion, weakening and The proportion, for example, The value is [1.5, 0.75, 0.75, 1, 1, 1].

[0119] Step 3: Initialize the iterative parameters and solution space boundary parameters of the physical constraint optimization algorithm. Using the three-dimensional winding data parameter matrix as the initial data, perform multiple rounds of iterative optimization. Each round of iteration includes: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, performing one-dimensional optimization on the remaining coordinate parameters, and updating the coordinate parameter positions based on the fitness function until the convergence condition is met.

[0120] Preferably, in step 3, initializing the iterative parameters and solution space boundary parameters of the physical constraint optimization algorithm includes:

[0121] Physical constraint optimization algorithm parameter configuration: Set the maximum number of iterations, the step size factor, and the solution space boundary parameters, including:

[0122] Initialize optimization algorithm parameters: Set the maximum number of iterations to balance convergence speed and solution accuracy by setting the maximum iteration parameter;

[0123] Set the movement step factor The search rate is determined by controlling the size of the global exploration step size. Setting a reasonable step size factor helps to converge to the global optimum quickly.

[0124] Set boundary parameters for multidimensional joint optimization and , , are the lower and upper bounds of the optimization variables, respectively, used to define the legal range of the solution space.

[0125] Preferably, such as Figure 1 As shown, using the three-dimensional winding data parameter matrix as initial data, the process is executed through multiple rounds of iteration: five coordinate parameters are randomly selected for multi-dimensional joint optimization, the remaining coordinate parameter is optimized in one dimension, and the coordinate parameters are updated based on the fitness function until the termination condition is met.

[0126] More preferably, the mathematical representation of the initial three-dimensional winding data parameter matrix is ​​shown as follows: The six parameters Normalization is performed to reduce computational cost and accelerate iteration, making the sequence more conducive to optimization iteration. The normalization expression is:

[0127]

[0128] wherein, is the maximum distance from the transducer original coordinate point to the transformer winding, since is in the range of [- , ] it needs to be processed by absolute value so that the range of each normalized parameter must be in [0, 1], is the normalized parameter value of the polar radius in the polar coordinate system, is the normalized parameter value of the azimuth angle in the polar coordinate system, is the normalized parameter value of the elevation angle in the polar coordinate system, is the normalized parameter value of the in the Cartesian coordinate system, is the normalized parameter value of the in the Cartesian coordinate system, is the normalized parameter value of the in the Cartesian coordinate system.

[0129] Further preferably, in each round of complete iteration, the algorithm performs the following steps:

[0130] Multi-dimensional joint optimization: randomly select 3 to 5 coordinate parameters from the current parameter set, and perform joint optimization in the multi-dimensional subspace formed by these parameters to update the values of these parameters.

[0131] One-dimensional optimization: for the remaining parameters that have not been selected (i.e. 6 minus the number of selected parameters), sequentially perform one-dimensional optimization in their respective single dimensions to update their parameter values.

[0132] Solution space boundary judgment: perform solution space boundary judgment on the updated parameter values, and correct the parameter values that exceed the solution space boundary, which is defined by the physical size range of the transformer winding structure.

[0133] Fitness evaluation and convergence judgment: after completing the multi-dimensional optimization and one-dimensional optimization of the current round (i.e. all 6 parameters obtain new values), recalculate the fitness function value based on the updated parameter set, and detect whether the convergence condition is met:

[0134] If the fitness function value reaches the maximum value, and the deformation sensitive parameter value is not greater than the preset sensitive threshold, it is determined that the convergence has been reached, and the iteration is terminated in advance;

[0135] Otherwise, continue the next iteration until the preset maximum number of iterations is reached.

[0136] Further preferably, the mathematical model of the multi-dimensional joint optimization is:

[0137]

[0138]

[0139] wherein, represents the data adjustment made based on the current time three-dimensional winding data parameter matrix parameter value, represents the three-dimensional winding data parameter matrix current time parameter information, represents the moving step factor, represents the current best data parameter dimension, represents 3-5 dimensions randomly selected from the parameters, represents a random rational number randomly extracted in the interval, represents different random optimization modes, wherein the upper right superscript represents the time after the second iteration, since the data amount and algorithm flow of each iteration are consistent, so the time of each iteration is set to be fixed , .

[0140] The mathematical model of the one-dimensional optimization is:

[0141]

[0142]

[0143] wherein, is the information storage of the three-dimensional winding data parameter matrix, and are two random numbers between (-1, 1) respectively, as random disturbance factors, is the time required to reach the maximum number of iterations, and are the parameter information of the two column vector arrays randomly selected in dimension respectively;

[0144] wherein, is a control parameter, which changes with the number of iterations, ranging from 0 to / 2, ensuring that each angle direction can be updated. If the optimization problem is not greater than 3, i.e. only single parameter needs to be optimized, optimization according to the one-dimensional optimization mathematical model can reduce data operation amount and approach the optimal fitness value more quickly. The mutual conversion of multi-dimensional and one-dimensional optimization modes is conducive to improving the optimization rate of three-dimensional data coordinates.

[0145] Preferably, the multi-dimensional and one-dimensional joint optimization strategy randomly selects two of the parameter comparison values as reference, and the mathematical update expression is as follows:

[0146]

[0147]

[0148] in, For the optimal value and The difference between the corrected values ​​in each iteration.

[0149] Preferably, the innovative behavior of the algorithm lies in the fact that the necessary deletion of certain data solves the problem of large coordinate data errors caused by abrupt changes in amplitude at a certain frequency band of the signal. If the error of a certain set of data is severe, unnecessary data can be deleted, and a new set of superior data can be slowly reconstructed based on other data sets. The mathematical model for updating the matrix parameters is as follows:

[0150]

[0151] In the formula This refers to the size of the matrix data.

[0152] Preferably, the condition for regenerating the data is that the winding coordinate matrix parameter values ​​cannot exceed the specified boundary conditions, that is: and In accordance with the regulations Optimize internally; otherwise, discard the data and regenerate it. and respectively correspond to the boundary parameters of multidimensional joint optimization and .

[0153] After each iteration of optimizing the population position, it is necessary to... The parameters are inversely normalized to obtain optimized parameter values, and then the first... Fitness function value after the second iteration The formula for inverse normalization of the winding position coordinate parameters is shown below:

[0154]

[0155] More preferably, the convergence condition includes:

[0156] Optimize iterative comparison of two adjacent time steps The parameter values ​​are normalized, and the fitness function value is calculated for each iteration until the maximum value of the fitness function is found and the deformation sensitivity parameter of the winding is not greater than the preset sensitivity threshold. Then the algorithm optimization iteration stops. The maximum value of the fitness function represents the minimum effect of noise on the three-dimensional winding imaging.

[0157] Specifically, the deformation sensitivity parameter of the winding is calculated as follows:

[0158]

[0159] In the present embodiment, the optimization process of the present application adopts a five-dimensional and one-dimensional combined search strategy. Specifically, in each round of iteration, five parameters are randomly selected in the normalized for multi-dimensional combined optimization; after the multi-dimensional optimization is completed, one-dimensional optimization is performed on the remaining one parameter, and the one-dimensional optimization result is mutually verified using the optimized five parameters. After each iteration optimization, the fitness function value is calculated based on the coupling relationship between the amplitude and the energy, until the optimal solution is searched.

[0160] Further, according to the parameter coupling of the parameter matrix, the multi-dimensional search and one-dimensional search mode in the optimization problem are combined, and if the dimension of the optimization problem is not greater than 3, the one-dimensional exploration mode will be enabled to update the position. The present application designs two update strategies to adapt to the optimization characteristics of the data parameter matrix. Based on the differentiated parameters between arrays, a bidirectional multi-dimensional search strategy is used to update the parameter information.

[0161] Step 4, using the optimized coordinate parameters, generating a denoised three-dimensional winding image through coordinate system mutual verification and three-dimensional point cloud reconstruction.

[0162] Preferably, the step 4 comprises: importing the optimized data, generating a winding three-dimensional image, wherein:

[0163] The data optimized based on the physical constraint optimization algorithm is imported into a three-dimensional winding imaging software system, and according to the relationship between the polar coordinate system data and the coordinates in the Cartesian coordinate system:

[0164]

[0165] A mutual verification mechanism is formed to judge the reliability of the data.

[0166] Generating a three-dimensional winding image: using the three-dimensional winding imaging GUI interface that has been built to convert the data in the polar coordinate system into data in the Cartesian coordinate system, and taking the average of the original data, a clear, accurate and noise-free three-dimensional winding structure model is constructed using a three-dimensional point cloud reconstruction algorithm. Based on the generated three-dimensional model, a high-precision three-dimensional winding imaging image is rendered and output. The image should be able to intuitively show the overall shape, spatial arrangement, details of key parts and smooth surface after denoising of the winding. At the same time, the built-in software is called to analyze the deformation trend of the winding and the winding deformation extreme point.

[0167] Further preferably, the step 4 further comprises generating a fault report:

[0168] The reconstructed three-dimensional model is automatically analyzed to detect potential winding deformation and generate a structured winding fault diagnosis report automatically according to the analysis result.

[0169] Specifically, the report analysis includes the detected fault type and the position coordinates of the fault point in the three-dimensional space, and the overall state of the winding is evaluated according to the severity or quantitative index of the winding.

[0170] It should be noted that the physical constraint optimization algorithm of the present application includes multi-dimensional and one-dimensional joint optimization, which optimizes the fitness function of the coordinate parameters through a multi-dimensional and one-dimensional two-stage search strategy to specifically suppress the pulse noise generated by the electromagnetic interference of the substation in a specific frequency band. First, the noise perception fitness function is used to jointly evaluate the data quality in the frequency domain and space, forcing the algorithm to preferentially reduce the interference frequency band energy and the conversion error of the two-coordinate system (polar coordinate / Cartesian coordinate); the data reconstruction formula reconstructs the deteriorated data points, and the coordinate parameters are reconstructed according to the number of iterations and the data size to eliminate the electromagnetic noise interference in the ultrasonic signal at a specific frequency.

[0171] The present application provides a three-dimensional winding imaging denoising system based on a physical constraint optimization algorithm in embodiment 2, based on the three-dimensional winding imaging denoising method based on the physical constraint optimization algorithm in embodiment 1, the system comprises:

[0172] The signal acquisition and preprocessing module, the data reorganization and modeling module, the optimization algorithm execution module and the three-dimensional imaging module;

[0173] The signal acquisition and preprocessing module is used to execute the acquisition of the ultrasonic echo signal of the transformer winding and perform preprocessing to generate a joint coordinate matrix containing polar coordinate system and Cartesian coordinate system parameters;

[0174] The data reorganization and modeling module is used to execute the reorganization of the joint coordinate matrix into a three-dimensional winding data parameter matrix of time series data structure, and based on the three-dimensional winding data parameter matrix, a physical constraint term is constructed based on the amplitude and energy coupling relationship of the ultrasonic wave in the winding medium, and a fitness function for evaluating the coordinate data quality is constructed combined with the physical constraint term;

[0175] The optimization algorithm execution module is used to initialize the iteration parameters and the solution space boundary parameters of the physical constraint optimization algorithm, and takes the three-dimensional winding data parameter matrix as the initial data to perform multiple rounds of iteration optimization, each round of iteration including: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, one-dimensional optimization is performed on the remaining coordinate parameters, and the coordinate parameter position is updated based on the fitness function until the convergence condition is met;

[0176] The three-dimensional imaging module is used for performing the three-dimensional winding imaging denoising by using the optimized coordinate parameters, mutual checking by a coordinate system and three-dimensional point cloud reconstruction to generate a three-dimensional winding image after denoising.

[0177] In order to more clearly introduce the outstanding substantial features of the present application and the significant progress brought to the prior art, an application example of the present application is introduced below in Embodiment 3.

[0178] The application example specifically includes the following: Figure 2 、 Figure 3 and Figure 4 Embodiments of the present application are described in detail below with reference to the accompanying drawings, and the application example specifically includes the following:

[0179] The three-dimensional winding imaging denoising scheme based on the physical constraint optimization algorithm of the present application first analyzes the specific mathematical model of the physical constraint optimization algorithm, which can flexibly handle multi-dimensional data optimization problems based on amplitude-energy coupling constraints, and is very suitable for the optimization of three-dimensional data coordinate parameters. The working principles of each part in the ultrasonic echo signal processing are analyzed, the working process of the three-dimensional winding imaging software system is analyzed, and how to apply the physical constraint optimization algorithm to imaging denoising is analyzed. Finally, the iteration process of the physical constraint optimization algorithm applied to three-dimensional winding imaging denoising is established, software simulation is performed according to the steps of the physical constraint optimization algorithm, and an optimized three-dimensional winding image is generated to verify whether the algorithm can improve the three-dimensional winding imaging effect.

[0180] The physical constraint optimization algorithm designed by the present application is a multi-dimensional one-dimensional joint optimization transformer winding ultrasonic detection imaging denoising algorithm based on physical constraints. The algorithm realizes the adaptive suppression of electromagnetic interference and environmental noise in the ultrasonic signal by introducing an amplitude-energy joint constraint model in a multi-dimensional space and combining a multi-dimensional joint optimization algorithm. Specifically, the algorithm first establishes a reference model of amplitude and energy using the factory calibration parameters of the transducer, and compares the inverted amplitude and inverted energy with the rated reference value in the iterative optimization process to form a physical consistency residual. Secondly, the spatial orientation and propagation path characteristics obtained by the multi-dimensional one-dimensional joint optimization estimation are combined to embed the residual term into a multi-objective optimization function, realizing the joint optimization of minimum reconstruction error and optimal continuity of three-dimensional coordinate field distribution. Through the above process, the physical constraint optimization algorithm designed by the present application not only significantly improves the signal-to-noise ratio of transformer winding ultrasonic detection imaging, but also enhances the physical interpretability and convergence stability of the inversion result, providing reliable support for high-precision three-dimensional winding coordinate imaging reconstruction.

[0181] In this embodiment, the three-dimensional winding imaging denoising process of the physical constraint optimization algorithm includes the following steps:

[0182] Initialize a three-dimensional winding data parameter matrix. The winding matrix is composed of six parameters in the time sequence as the pointer, the Cartesian coordinate system and the polar coordinate system Each row of the matrix represents the coordinate data of each time acquisition point.

[0183] In the global integration stage, the mutual influence of the six parameters is optimized according to the formula in the three-dimensional winding imaging optimization problem, and the fitness function is constructed according to the coupling relationship between the amplitude and the energy to seek the optimal solution of the three-dimensional winding coordinate data.

[0184] The global optimization stage includes multi-dimensional one-dimensional joint optimization and data reconstruction. The multi-dimensional one-dimensional joint optimization adopts a bidirectional search strategy, calculates the optimization difference value between the row coordinates of the time sequence and the row coordinates of the previous time according to the formula, and calculates the joint optimization data update according to the optimization difference value.

[0185] By embedding the physical constraint optimization algorithm designed by the application into the three-dimensional winding imaging software system through MATLAB software, the application simulates the ultrasonic signal that cannot filter out noise by adding an adaptive noise signal. The application designs a control experiment to analyze the signal filtering effect by observing the three-dimensional winding imaging graph after optimization from the signal-to-noise ratio improvement (SNR improvement), reconstruction error (MSE), signal fidelity, and observing the three-dimensional winding imaging graph after optimization.

[0186] Table 1 Comparison test results of key performance indicators of different denoising algorithms

[0187]

[0188] From the data in Table 1 and the imaging effect in the drawing, it can be seen that the interference of the noise of the optimized three-dimensional winding imaging is effectively improved.

[0189] The three-dimensional winding imaging denoising technology based on the physical constraint optimization algorithm of the application adopts a multi-dimensional one-dimensional combined search method of three-dimensional winding imaging denoising based on the physical constraint optimization algorithm to optimize the three-dimensional ultrasonic winding data. The algorithm has higher calculation efficiency and is conducive to improving the noise interference in the three-dimensional winding imaging, and helps to realize accurate and rapid positioning of the transformer winding fault. It has important significance for improving the stability of power equipment and preventing the instability of power system work.

[0190] The above implementation cases are only preferred embodiments of the application and do not limit the application in any form. Although the application has been disclosed as above with the preferred embodiments, it is not intended to limit the application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the application should fall within the scope of protection of the technical solution of the application.

Claims

1. A three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm, characterized in that, Includes the following steps: The ultrasonic echo signal of the transformer winding is collected and preprocessed to generate a joint coordinate matrix containing parameters of polar coordinate system and Cartesian coordinate system; The joint coordinate matrix is ​​reorganized into a three-dimensional winding data parameter matrix of time-series data structure. Based on the amplitude and energy coupling relationship of ultrasonic waves in the winding medium, a physical constraint term is constructed using the physical constraint term to construct a fitness function for evaluating the quality of coordinate data. The iterative parameters and solution space boundary parameters of the physical constraint optimization algorithm are initialized. Using the three-dimensional winding data parameter matrix as initial data, multiple rounds of iterative optimization are performed. Each round of iteration includes: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, performing one-dimensional optimization on the remaining coordinate parameters, and updating the coordinate parameter positions based on the fitness function until the convergence condition is met. Using the optimized coordinate parameters, a denoised 3D winding image is generated through coordinate system cross-verification and 3D point cloud reconstruction.

2. The three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The process of acquiring and preprocessing the ultrasonic echo signals from the transformer windings includes: The original echo signal is acquired by an ultrasonic transducer array arranged on the surface of the transformer, the transducer array being arranged according to the spatial distribution of the winding structure; The original echo signal is sequentially subjected to bandpass filtering, beamforming, and quadrature demodulation. Bandpass filtering suppresses environmental electromagnetic noise and mechanical noise outside the operating frequency band of the ultrasonic transducer. Beamforming enhances the echo signal energy from the winding reflector. Quadrature demodulation converts the high-frequency echo signal into a low-frequency baseband signal containing amplitude and phase information.

3. The three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 2, characterized in that: The generation of the joint coordinate matrix containing parameters of the polar coordinate system and the Cartesian coordinate system includes: The preprocessed baseband signal is scanned and transformed to generate coordinate parameters in the polar coordinate system, including distance parameter, azimuth parameter and elevation parameter. The distance parameter represents the straight-line distance between the winding detection point and the ultrasonic emission source, the azimuth parameter represents the azimuth angle of the winding detection point in the horizontal direction, and the elevation parameter represents the pitch angle of the winding detection point in the vertical direction. The polar coordinate parameters are converted to Cartesian coordinate parameters through coordinate transformation relationships. The Cartesian coordinate parameters represent the three-dimensional spatial position of the winding detection point. By associating and combining the polar coordinate system coordinate parameters and the Cartesian coordinate system coordinate parameters corresponding to the sampling points at the same time, a joint coordinate matrix is ​​generated.

4. The three-dimensional winding imaging and denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The physical constraints constructed based on the amplitude and energy coupling relationship of ultrasonic waves in the winding medium include: Based on the propagation characteristics of ultrasound in the winding medium of oil-immersed transformers, an amplitude-coordinate attenuation model of spatial coordinate points and reflected sound wave amplitude is established, in which the reflected sound wave amplitude attenuates as the propagation distance increases; Based on the reflected sound wave amplitude, and combined with the physical characteristics of transformer dielectric density and sound velocity, the sound wave energy flux corresponding to the current spatial coordinate parameters is calculated. Use the rated amplitude and rated energy of the ultrasonic transducer as the reference standard. The amplitude of the reflected sound wave is calculated based on the amplitude-coordinate attenuation model and the current spatial coordinate parameters. The amplitude of the reflected sound wave and the sound wave energy flux corresponding to the current spatial coordinate parameters are compared with the corresponding rated reference to form the amplitude residual and energy residual. Physical constraints are constructed based on the magnitude residual and energy residual to constrain the coordinate parameter optimization process.

5. A three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The fitness function constructed in conjunction with the physical constraints for evaluating the quality of coordinate data includes: Obtain the spatial coordinate parameter values ​​of the three-dimensional winding data parameter matrix at multiple consecutive acquisition time points; For each spatial coordinate parameter, calculate its normalized change between adjacent acquisition time points; The normalized changes of each spatial coordinate parameter are multiplied by the pre-configured weighting coefficients and then summed to generate a weighted change evaluation item. The weighted change evaluation term is combined with the physical constraint term to generate the fitness function.

6. The three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The random selection of multiple coordinate parameters for multidimensional joint optimization includes: In each iteration, multiple coordinate parameters are randomly selected from the three-dimensional winding data parameter matrix to construct a multi-dimensional optimization subspace; Within the multidimensional optimization subspace, the current optimal parameter position is searched based on the fitness function; Calculate the coordinate difference between the current position and the optimal parameter position in the corresponding parameter dimension; The coordinate difference is scaled according to a preset movement step factor to generate a parameter position adjustment amount; The parameter position adjustment amount is added to the current position to obtain the updated parameter value; Calculate the fitness value corresponding to the updated parameter value based on the fitness function. If the fitness value is better than the current value, accept the update; otherwise, retain the current value. The updated parameter values ​​are used to determine the solution space boundary, and parameter values ​​that exceed the solution space boundary are corrected. The solution space boundary is defined by the physical size range of the transformer winding structure.

7. The three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The one-dimensional optimization of the remaining coordinate parameters includes: For each remaining coordinate parameter that was not selected to participate in the multidimensional joint optimization, perform the following operations on its single parameter dimension: In the current iteration, two different historical parameter values ​​are randomly selected as references; Calculate the difference vector between the two different historical parameter values, and superimpose a random perturbation factor to generate a one-dimensional search direction vector; Adjust the current value of the parameter along the direction vector to obtain the updated parameter value; Calculate the fitness value corresponding to the updated parameter value based on the fitness function. If the fitness value is better than the current value, accept the update; otherwise, retain the current value. The updated parameter values ​​are used to determine the solution space boundaries, and parameter values ​​that exceed the solution space boundaries are corrected.

8. A three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The convergence condition is determined after all parameters have undergone multidimensional and one-dimensional optimization in each iteration, including: At the end of one iteration, calculate the fitness function value after updating all parameters; If the updated fitness function value reaches its maximum value and the winding deformation sensitivity parameter is not greater than the preset sensitivity threshold, then convergence is determined and the iteration is terminated. Otherwise, continue to the next iteration until the maximum number of iterations is reached.

9. A three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm according to claim 1, characterized in that: The process of generating a denoised 3D winding image through coordinate system cross-verification and 3D point cloud reconstruction includes: The optimized polar coordinate system coordinate parameters and Cartesian coordinate system coordinate parameters are converted to each other according to the spatial geometric transformation relationship. By calculating the spatial position deviation between the converted coordinate parameters and the corresponding optimized coordinate parameters in the target coordinate system, the consistency between the two is judged, and abnormal data points are identified and removed. The weighted average of the verified polar coordinate system coordinate parameters and Cartesian coordinate system coordinate parameters is calculated to obtain the final denoised spatial coordinate sequence. Based on the final spatial coordinate sequence, a three-dimensional point cloud reconstruction algorithm is used to construct a three-dimensional spatial structure model of the transformer winding; The three-dimensional spatial structure model is subjected to surface rendering and visualization processing to generate and output a denoised three-dimensional winding image.

10. A three-dimensional winding imaging denoising system based on a physical constraint optimization algorithm, comprising the three-dimensional winding imaging denoising method based on a physical constraint optimization algorithm as described in any one of claims 1-9, characterized in that, The system includes: The module includes a signal acquisition and preprocessing module, a data reconstruction and modeling module, an optimization algorithm execution module, and a 3D imaging module. The signal acquisition and preprocessing module is used to acquire the ultrasonic echo signal of the transformer winding and perform preprocessing to generate a joint coordinate matrix containing parameters of polar coordinate system and Cartesian coordinate system. The data recombination and modeling module is used to reconstruct the joint coordinate matrix into a three-dimensional winding data parameter matrix with a time series data structure, and to construct physical constraint terms based on the amplitude and energy coupling relationship of ultrasonic waves in the winding medium in combination with the three-dimensional winding data parameter matrix, and to construct a fitness function for evaluating the quality of coordinate data in combination with the physical constraint terms. The optimization algorithm execution module is used to execute the iterative parameters and solution space boundary parameters of the initial physical constraint optimization algorithm. Using the three-dimensional winding data parameter matrix as the initial data, it performs multiple rounds of iterative optimization. Each round of iteration includes: randomly selecting multiple coordinate parameters for multi-dimensional joint optimization, performing one-dimensional optimization on the remaining coordinate parameters, and updating the coordinate parameter positions based on the fitness function until the convergence condition is met. The three-dimensional imaging module is used to generate a denoised three-dimensional winding image by using optimized coordinate parameters, coordinate system cross-verification, and three-dimensional point cloud reconstruction.

Images