Electric transmission line steel pipe pole bearing platform evaluation method based on prefabricated pipe pile

Abstract

The invention discloses a power transmission line steel pipe pole bearing platform assessment method based on a prefabricated pipe pile, and relates to the technical field of power transmission line steel pipe pole bearing platform assessment, and the method comprises the following steps: exciting an elastic wave of a bearing platform-pipe pile interface through a sound wave emission device, and receiving a sound wave signal through an ultrasonic array sensor at the top of the pipe pile; carrying out feature separation on the sound wave signal, and evaluating the sound velocity relative deviation according to a separation result; on the basis of an evaluation result, constructing an optimal sound velocity gradient interval by adopting a dynamic optimization algorithm comprising a Gaussian mixture model, a density peak detection algorithm and NSGA-II (Non-dominated Sorting Genetic Algorithm-II) optimization; identifying a gradient mutation region in the interval, and expanding a defect detection range by taking the region as a reference; and constructing a defect characteristic spectrum based on the expanded defect detection range, carrying out void defect positioning according to the defect characteristic spectrum, and outputting a grading evaluation result, thereby solving the problems that the defect detection range is limited and the accuracy of bearing platform evaluation is reduced due to the fact that a sound velocity gradient interval is not dynamically optimized when sound velocity is used for evaluation in a traditional method.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line steel pipe pole cap evaluation, and more particularly to a transmission line steel pipe pole cap evaluation method based on prefabricated pipe piles. Background Art

[0002] With the continuous development of infrastructure construction, particularly in industries like power and transportation, steel tubular poles and their associated cap structures are widely used in key projects such as transmission lines, communication towers, and bridges. Due to long-term external environmental pressures, climate change, and potential quality issues during construction, the safety and stability of these structures often face varying degrees of risk. The connection area between precast tubular piles and steel tubular pole caps is particularly important, as it is a critical load-bearing component, and its health directly affects the stability and reliability of the entire structure.

[0003] In recent years, with the development of acoustic imaging and ultrasonic testing technologies, structural health monitoring methods based on acoustic signals have gained widespread application in engineering practice. The propagation characteristics of acoustic signals can effectively reflect the internal defects of materials. By exciting elastic waves within a structure and analyzing their reflected and scattered waves, defects such as cracks and cavities can be accurately detected. However, existing acoustic detection methods still have some shortcomings, mainly in terms of acoustic signal processing, defect location accuracy, and optimization of sound velocity inversion models.

[0004] For example, patent publication CN115586258A discloses a method for assessing internal corrosion of substation grounding flat steel using spiral-guided ultrasound. By measuring the propagation velocity of ultrasound within the pantograph slide at varying degrees of aging, a relationship is established between the ultrasonic propagation velocity and the pantograph slide's aging. This method provides a scientific and rapid assessment method for pantograph slide aging, resolving the inaccuracies associated with current manual visual assessments. This method provides valuable theoretical guidance for assessing pantograph slide aging.

[0005] The above disclosed technical solutions have at least the following technical problems: Traditional assessment methods fail to accurately assess the relative deviation of sound velocity, resulting in limited detection accuracy. In particular, in the case of complex structures or uneven materials, potential defect locations cannot be accurately captured. In addition, there is no dynamic optimization of the sound velocity gradient interval, resulting in a limited range and accuracy of defect detection, which in turn reduces the accuracy of transmission line steel pipe pole cap assessment.

[0006] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for evaluating steel pipe pole caps for transmission lines based on prefabricated pipe piles, which expands the defect detection range by constructing an optimal sound velocity gradient interval to solve the problem that traditional methods do not dynamically optimize the sound velocity gradient interval when evaluating by sound velocity, resulting in a limited range of defect detection and a reduction in the accuracy of cap evaluation.

[0008] To achieve the above object, the present invention provides the following technical solutions: An evaluation method for steel pipe caps of transmission lines based on prefabricated pipe piles comprises the following steps: exciting elastic waves at the cap-pile interface by an acoustic wave transmitting device, and receiving the acoustic wave signals by an ultrasonic array sensor on the top of the pipe pile; performing feature separation on the acoustic wave signals, and evaluating the relative deviation of sound velocity based on the separation results; constructing an optimal sound velocity gradient interval based on the evaluation results using a dynamic optimization algorithm including a Gaussian mixture model, a density peak detection algorithm, and NSGA-II optimization; identifying a gradient mutation region in the interval, and expanding the defect detection range based on the region; constructing a defect feature map based on the expanded defect detection range, locating void defects based on the defect feature map, and outputting a graded evaluation result.

[0009] In a preferred embodiment, the elastic waves at the interface between the foundation and the pipe pile are excited by the acoustic wave emitting device, and the acoustic wave signals are received by the ultrasonic array sensor on the top of the pipe pile. Specifically, a multi-source annular acoustic wave emitting array is arranged on the surface of the foundation to excite pulse signals of different frequencies according to a preset time sequence; the ultrasonic array sensor group on the top of the pipe pile is synchronously started to collect the time domain sequence of the reflected wave and the scattered wave; the time domain sequence is divided into time windows, and the acoustic wave signals are extracted based on a preset wave velocity inversion model.

[0010] In a preferred embodiment, the feature separation of the acoustic wave signal and the evaluation of the relative deviation of the sound speed are specifically as follows: a spatial domain filtering algorithm is used to reduce the noise of the acoustic wave signal to obtain a reduced noise signal; the reduced noise signal is subjected to empirical mode decomposition to obtain a principal component; the time-frequency eigenvector of the principal component is extracted and the offset between the time-frequency eigenvector and a preset dispersion curve is calculated; based on the offset, a cross-correlation algorithm is used to calculate a relative deviation matrix of the sound speed; and the relative deviation of the sound speed is evaluated according to the relative deviation matrix of the sound speed to obtain a relative deviation evaluation matrix of the sound speed.

[0011] In a preferred embodiment, based on the evaluation results, a dynamic optimization algorithm including a Gaussian mixture model, a density peak detection algorithm and NSGA-II optimization is used to construct the optimal sound speed gradient interval, specifically: the sound speed relative deviation evaluation matrix is ​​input into the Gaussian mixture model, and a density peak detection algorithm is used to identify a number of gradient communities, and a first sound speed gradient topology map is constructed based on the spatial distribution of the gradient community and the gradient change direction; a first constraint condition is constructed based on the first sound speed gradient topology map, including a physical range constraint of the sound speed gradient, a confidence interval constraint within each Gaussian distribution community, and a maximum change rate constraint between adjacent sampling points; based on the first constraint condition, a Monte Carlo method is used to perform random sampling within a preset probability density distribution to obtain a number of first sound speed gradient samples; and a preset distributed community algorithm is used to perform NSGA-II optimization on the first sound speed gradient samples to obtain the optimal sound speed gradient interval.

[0012] In a preferred embodiment, the preset distributed community algorithm is used to perform NSGA-II optimization on several first sound velocity gradient samples to obtain the optimal sound velocity gradient interval, specifically: several first sound velocity gradient samples are taken as individuals and grouped according to preset community labels to obtain several first populations; the NSGA-II algorithm is used to perform non-dominated sorting on each first population to obtain several Pareto front solution sets; several Pareto front solution sets are migrated across communities to obtain a first solution set distribution; the first population is split and merged based on the first solution set distribution to obtain a reconstructed community structure; and NSGA-II optimization is re-performed based on the reconstructed community structure to obtain the optimal sound velocity gradient interval.

[0013] In a preferred embodiment, the cross-community migration of several Pareto frontier solution sets to obtain a first solution set distribution is specifically as follows: calculating the crowding distance of several Pareto frontier solution sets, and screening several Pareto frontier solution sets according to the crowding distance to obtain several boundary solutions; calculating the gradient similarity matrix of several boundary solutions, and generating community migration instructions according to the gradient similarity matrix; and cross-community migration of several boundary solutions according to the community migration instructions to obtain the first solution set distribution.

[0014] In a preferred embodiment, the first population is divided and merged based on the first solution set distribution to obtain a reconstructed community structure, specifically: the number of migrations of the first solution set distribution is counted, and a merge judgment is made based on the number of migrations; if a merge reconstruction is performed, the main gradient direction vector of each community in the first solution set distribution is extracted; and the communities are divided and merged based on the main gradient direction vector to obtain a reconstructed community structure.

[0015] In a preferred embodiment, a gradient mutation area is identified in the interval, and the defect detection range is expanded based on the area, specifically: the pipe pile is constructed into a three-dimensional spatial grid according to axial stratification and radial sectorization; within the optimal sound velocity gradient interval, the absolute value of the gradient change rate of adjacent spatial grids is calculated; based on the absolute value of the gradient change rate, the gradient mutation of the grid is identified to generate an initial defect seed point set; with the preset seed point as the center, spatial expansion is performed along the axial and radial directions of the pipe pile to obtain an expanded area; and the expanded area is merged with the initial defect seed point set to obtain the defect detection range.

[0016] In a preferred embodiment, a defect feature map is constructed based on the expanded defect detection range, and void defects are located according to the defect feature map and a graded evaluation result is output, specifically: the defect detection range is divided into a three-dimensional voxel grid according to a preset resolution; multi-band acoustic wave energy attenuation imaging is performed on each voxel to generate an initial defect energy map; a defect feature tensor is constructed based on the initial defect energy map; a three-dimensional convolutional neural network is used to extract void area boundary features in the defect feature tensor; based on the void area boundary features, defects are divided into three levels: slight, medium and severe, and a defect graded evaluation result including position, size and level is output.

[0017] The technical effects and advantages of the evaluation method for steel pipe pole caps for power transmission lines based on prefabricated pipe piles of the present invention are as follows: 1. The present invention uses an acoustic wave transmitter to excite elastic waves at the interface between the foundation and the pile. An ultrasonic array sensor at the top of the pile receives reflected and scattered wave signals. By separating the characteristics of these acoustic wave signals and removing noise, the relative deviation of sound velocity can be accurately assessed, providing high-quality signal data for subsequent defect detection. The key technology in this process is the extraction of the time-frequency feature vectors of the acoustic wave signals. The relative deviation matrix of sound velocity is calculated using a cross-correlation algorithm. Through precise evaluation, an accurate assessment of the relative deviation of sound velocity is obtained, laying the foundation for subsequent defect location. Based on the evaluation results, a dynamic optimization algorithm is used to construct the optimal sound velocity gradient interval. By combining a Gaussian mixture model with a density peak detection algorithm, regions of sudden change in the sound velocity gradient can be effectively identified, thereby precisely defining the optimal sound velocity gradient interval. This optimal gradient interval not only optimizes the sound wave propagation path but also ensures higher accuracy and sensitivity in detecting structural defects within this interval. By optimizing the sound velocity gradient, external noise interference can be reduced, the accuracy of defect location can be improved, and more reliable assessment results can be provided.

[0018] 2. By constructing an optimal sound velocity gradient range, the present invention further expands the scope of defect detection. By identifying and expanding regions of gradient mutation, more potential defect areas can be covered, especially in complex structural environments, enabling more comprehensive detection of possible structural defects such as cracks and voids. By establishing a defect feature map within this expanded region and extracting defect features using a three-dimensional convolutional neural network, the present invention achieves efficient defect location and graded assessment, significantly improving detection accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the flow of the method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to the present invention; Figure 2 This is the sound velocity relative deviation matrix thermal map of the present invention; Figure 3 This is the first sound velocity gradient topology diagram of the present invention; Figure 4 This is the three-dimensional defect energy spectrum diagram of the present invention. DETAILED DESCRIPTION

[0020] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] Example 1, Figure 1 The present invention provides a method for evaluating steel pipe pole caps for transmission lines based on prefabricated pipe piles, comprising the following steps: S1, the elastic wave at the interface between the cap and the pile is excited by the acoustic wave transmitting device, and the acoustic wave signal is received by the ultrasonic array sensor on the top of the pile; In this example, the acoustic wave transmitter excites elastic waves at the interface between the cap and the pile, and the ultrasonic array sensor on the top of the pile receives the acoustic wave signal. Specifically: A multi-source annular acoustic wave emission array is arranged on the surface of the platform to stimulate pulse signals of different frequencies according to a preset time sequence; Synchronously start the ultrasonic array sensor group on the top of the pile to collect the time domain sequence of reflected waves and scattered waves; The time domain sequence is divided into time windows, and the acoustic wave signal is extracted based on the preset wave velocity inversion model.

[0022] It should be noted that acoustic signals refer to the wave signals generated by elastic waves excited by an acoustic wave transmitter, propagating through the interface between the pile cap and the pile, and then undergoing reflection and scattering. These acoustic signals contain key information about the internal properties of the medium (e.g., the pile cap and pile cap structure), such as material density, defect location, and the structural integrity of the interface. Receiving these acoustic signals through an ultrasonic array sensor system captures characteristics such as wave propagation time and frequency variations, which can be used to analyze and assess the health of the pile cap. A velocity inversion model mathematically models the propagation characteristics of acoustic signals (such as propagation time and frequency) to infer velocity variations. This model, based on a specific inversion algorithm and combining time-domain data from actual acoustic signals, performs velocity inversion calculations. By comparing the measured velocity with the preset value in the theoretical model, the relative deviation in the acoustic velocity of the medium can be determined, reflecting the physical condition of the pile cap and pile cap, and identifying potential defect areas. This method utilizes velocity variations to identify potential defects such as voids and cracks in the structure, providing an effective basis for defect location and assessment.

[0023] Furthermore, the acoustic signal acquisition process involves deploying a multi-source annular acoustic wave transmitting array on the cap surface, stimulating pulse signals of different frequencies according to a preset time sequence to generate elastic waves, and simultaneously activating an ultrasonic array sensor group on the top of the pile to receive the time domain sequence of reflected and scattered waves. The acquired time domain signals are then subjected to time window segmentation processing, and effective acoustic wave signal characteristics are extracted based on a pre-established wave velocity inversion model. This achieves high-precision excitation and signal capture of elastic waves at the cap-pile interface, providing a data foundation for subsequent sound velocity analysis. This solution effectively improves the spatiotemporal resolution and signal-to-noise ratio of the interface acoustic wave signal through the time-sequential excitation and synchronous reception design of the multi-source annular array, combined with time window segmentation and wave velocity inversion models.

[0024] S2, feature separation of the acoustic signal and evaluation of the relative deviation of the sound velocity based on the separation results; In this example, the acoustic signal is feature separated, and the relative deviation of the sound velocity is evaluated based on the separation results. Specifically: The spatial domain filtering algorithm is used to reduce the noise of the sound wave signal to obtain a noise-reduced signal; Perform empirical mode decomposition on the denoised signal to obtain the principal components; Extracting the time-frequency eigenvector of the principal component and calculating the offset between the time-frequency eigenvector and the preset dispersion curve; Based on the offset, the cross-correlation algorithm is used to calculate the relative deviation matrix of the speed of sound; The sound speed relative deviation is evaluated according to the sound speed relative deviation matrix to obtain a sound speed relative deviation evaluation matrix.

[0025] In this example, the specific calculation formula of the spatial domain filtering algorithm is as follows:

[0026] in, is the noise reduction signal, is the weight coefficient of the i-th ultrasonic array sensor, is the original time domain signal received by the i-th ultrasonic array sensor, is the number of ultrasound array sensors.

[0027] In this example, the calculation formula of the sound speed relative deviation matrix is ​​as follows:

[0028] in, is the relative deviation of the sound velocity between the i-th ultrasonic array sensor and the j-th ultrasonic array sensor, is the time-frequency feature vector, is the offset between the time-frequency eigenvector corresponding to the i-th ultrasonic array sensor and the preset dispersion curve, is the offset between the time-frequency eigenvector corresponding to the jth ultrasonic array sensor and the preset dispersion curve, is the number of principal components.

[0029] It should be noted that spatial filtering is a technique used to remove noise, particularly high-frequency noise components from acoustic signals. During this process, acoustic signals are first received by a sensor array and converted into digital signals. Spatial filtering is then processed using a spatial filtering algorithm. Spatial filtering analyzes the spatial distribution of acoustic signals and uses specific mathematical models (such as high-pass, low-pass, or band-pass filtering) to suppress noise. By filtering out noise, the original reflected and scattered wave components are retained, resulting in clearer signals and facilitating subsequent feature extraction and analysis.

[0030] Furthermore, dispersion curves describe the relationship between frequency and wave velocity as waves propagate through a medium. In sound wave propagation, waves of different frequencies may propagate at different speeds. By showing the relationship between frequency and wave velocity, dispersion curves reveal the elastic properties, density, and structural characteristics of the material. By studying dispersion curves, we can understand the distribution of the medium and identify structural defects. For example, significant changes in wave velocity may indicate defects such as cracks or voids in the medium.

[0031] Finally, after calculating the sound velocity relative deviation matrix, the next step is to evaluate these deviations. The evaluation process involves analyzing the relative deviation values ​​of the sound velocity and scoring each area according to a preset standard or model. For example, if the deviation values ​​in certain areas are large, it may indicate that there are defects (such as cracks or voids) in that area. By comprehensively evaluating all the deviation values, a sound velocity relative deviation evaluation matrix is ​​generated, which is used to represent the health status of each inspection area, such as Figure 2 As shown, the heatmap matrix visually reflects the deviation distribution between the signals received by different sensors. Abnormal areas (red / dark blue) may indicate abnormal material density or connection defects. This evaluation matrix can help engineers identify possible abnormal areas in piles or caps, providing data support for subsequent defect location and assessment.

[0032] S3, based on the evaluation results, a dynamic optimization algorithm including Gaussian mixture model, density peak detection algorithm and NSGA-II optimization is used to construct the optimal sound velocity gradient interval; In this example, based on the evaluation results, a dynamic optimization algorithm including a Gaussian mixture model, a density peak detection algorithm, and NSGA-II optimization is used to construct the optimal sound velocity gradient interval, specifically: The sound velocity relative deviation evaluation matrix is ​​input into the Gaussian mixture model, and a density peak detection algorithm is used to identify several gradient communities. The first sound velocity gradient topology map is constructed based on the spatial distribution of the gradient communities and the gradient change direction. Constructing a first constraint condition based on the first sound velocity gradient topology graph, including a physical range constraint of the sound velocity gradient, a confidence interval constraint within each Gaussian distribution community, and a maximum change rate constraint between adjacent sampling points; Based on the first constraint, a Monte Carlo method is used to perform random sampling within a preset probability density distribution to obtain a number of first sound velocity gradient samples; The preset distributed community algorithm is used to perform NSGA-II optimization on several first sound velocity gradient samples to obtain the optimal sound velocity gradient interval.

[0033] In this example, the specific calculation formula of the Gaussian mixture model is as follows:

[0034] in, is the probability density of the observed data point x, is the weight of the t-th Gaussian distribution, is the mean of the t-th Gaussian distribution, is the covariance matrix of the t-th Gaussian distribution, is the probability density function of the t-th Gaussian distribution, is the number of Gaussian distributions, that is, the number of gradient communities that need to be identified.

[0035] It should be noted that the density peak detection algorithm is a clustering method based on the local density and distance of data points. In this step, the algorithm is used to analyze the density distribution of different regions in the sound speed relative deviation matrix. By calculating the local density of each data point (i.e., the density of the surrounding area) and the distance to the point farthest from that point, the density peak of each data point is determined. Based on these density peaks, the density peak detection algorithm can effectively identify multiple gradient communities (i.e., regions that exhibit significant differences in sound speed gradient changes). This method can construct a first sound speed gradient topology map, showing the structure of the sound speed gradient distribution and the relationship between different gradient regions.

[0036] In addition, the first constraint refers to the restrictions imposed on the range and regularity of the sound velocity gradient when constructing the optimal sound velocity gradient interval. Specifically, the constraints may include limiting the maximum value, minimum value, and rate of change of the sound velocity gradient, or requiring the gradient to remain consistent within certain specific areas. The mathematical form of the first constraint can be expressed as the following joint constraint:

[0037] in, is the sound velocity gradient at position x, and are the minimum and maximum values ​​of the sound velocity gradient (determined by the physical properties of the sound propagation medium, such as , ), and are the mean and covariance matrix of the t-th Gaussian distribution (obtained by clustering the Gaussian mixture model), k is the confidence coefficient (usually k=2, corresponding to a 95% confidence interval), is the maximum allowed rate of change (gradient difference threshold within unit distance), For adjacent sampling points , The spatial distance between two points.

[0038] These constraints ensure that the optimization process does not deviate from the actual physical phenomena and avoid constructing unreasonable sound speed gradient intervals. For example, if the sound speed changes very drastically in a certain area, it may be necessary to limit the rate of change of the gradient. and the gradient range within the community ( ), to ensure that the variation trend of this interval complies with the actual physical constraints.

[0039] In this example, a preset distributed community algorithm is used to perform NSGA-II optimization on several first sound velocity gradient samples to obtain the optimal sound velocity gradient interval, specifically: A number of first sound velocity gradient samples are taken as individuals and grouped according to preset community labels to obtain a number of first populations; The NSGA-II algorithm is used to perform non-dominated sorting on each first population and obtain several Pareto frontier solution sets; Perform cross-community migration on several Pareto frontier solution sets to obtain the first solution set distribution; Based on the distribution of the first solution set, the first population is divided and merged to obtain the reconstructed community structure; Based on the reconstructed community structure, NSGA-II optimization was performed again to obtain the optimal sound speed gradient range.

[0040] For example, suppose that during the evaluation of the steel pipe pole cap of a certain transmission line, the elastic wave on the cap-pile interface is first excited by the acoustic wave transmitting device, and the acoustic wave signal from the cap to the pile area is received by the ultrasonic array sensor. After processing, the sound velocity relative deviation matrix of the area is obtained, and it is divided into several gradient communities by the Gaussian mixture model (GMM). These communities are identified by the density peak detection algorithm, and the first sound velocity gradient topology map is constructed based on the spatial distribution of the gradient community and the gradient change direction. Specifically: each gradient community is a topological node, and the node attributes include the community center coordinates, the average gradient value and the gradient main direction vector. If the two community spatial areas meet the absolute value of the average gradient value difference less than the preset threshold, and the gradient main direction vector similarity is greater than the preset threshold, then an edge is established between the two gradient communities, such as Figure 3 As shown, different colors represent multiple sound velocity gradient communities automatically divided by the Gaussian mixture model, showing the spatial distribution pattern of gradient changes, which helps to identify potential heterogeneous structural areas.

[0041] Next, based on the structure of these gradient communities, the preset first constraint was applied to define the range of the sound velocity gradient. For example, suppose the sound velocity gradient intervals are set to [1800, 2200] m / s, [1800, 2200] m / s, and [1800, 2200] m / s, and the maximum rate of gradient change within these intervals must not exceed 100 m / s / km. Based on these constraints, a Monte Carlo method was used to randomly sample within this probability distribution, generating several sound velocity gradient samples ranging from 1800 m / s to 2200 m / s.

[0042] Subsequently, the NSGA-II algorithm was used to perform a multi-objective optimization on these sound velocity gradient samples, with the goal of minimizing the volatility of gradient changes and enhancing the uniformity of the gradient distribution. Specifically, the NSGA-II algorithm performed a non-dominated sorting on each sound velocity gradient sample, dividing the samples into multiple populations. Multiple rounds of crossover, mutation, and selection were performed based on the Pareto front solution set, ultimately resulting in a set of optimal sound velocity gradient intervals. These optimal solutions represent the equilibrium state of the sound velocity gradient distribution while satisfying constraints such as maximum gradient change rate and minimum volatility.

[0043] Ultimately, after optimization, the optimal sound velocity gradient intervals of [1850, 2150] m / s, [1850, 2150] m / s, and [1850, 2150] m / s were obtained. This interval represents the most stable and representative region of sound velocity variation within the steel pipe pile cap area of ​​this transmission line. This optimal interval allows engineers to accurately identify potential defects, such as cracks or voids, at the cap-pile interface, providing crucial decision support for further structural reinforcement or repair.

[0044] In this example, several Pareto frontier solution sets are migrated across communities to obtain the first solution set distribution, specifically: Calculate the crowding distance for several Pareto frontier solution sets, and screen several Pareto frontier solution sets according to the crowding distance to obtain several boundary solutions; Calculate the gradient similarity matrix of several boundary solutions and generate community migration instructions based on the gradient similarity matrix; According to the community migration instruction, several boundary solutions are migrated across communities to obtain the first solution set distribution.

[0045] It should be noted that, first, the crowding distance of each sample in the Pareto front solution set is set, which is an indicator to measure the density of sample solutions in the target space. Assume that the calculated crowding distance is as follows:

[0046] The larger the crowding distance, the sparser the solution is in the target space. Based on the crowding distance, several boundary solutions are screened out. These solutions are located at the "edge" of the solution set, that is, the farthest away from other solutions. For example, suppose 10 boundary solutions are screened out from the solution set, specifically: , by calculating the gradient similarity between these boundary solutions and generating community migration instructions. Gradient similarity is defined as the similarity measure between boundary solutions. The specific calculation formula is:

[0047] in, is the gradient similarity between the mth boundary solution and the nth boundary solution, is the mth boundary solution, is the nth boundary solution, and are the maximum and minimum values ​​of the boundary solution, respectively.

[0048] It should be noted that based on the similarity matrix, community migration instructions are generated to indicate the migration path of the boundary solutions in the solution space. These migration instructions enable solutions with lower similarity (i.e., relatively distant solutions) to exchange information, promoting comprehensive exploration of the solution space.

[0049] Then, according to the generated migration instructions, perform cross-community migration operations. This process includes the following steps: Cross-community migration: Boundary solutions are migrated across communities according to the migration instructions. That is, boundary solutions with lower similarity will be exchanged to different communities for further optimization and merging.

[0050] Update the solution set: After the migration operation, a new solution set is obtained. The distribution of these solutions is more even, and the solution space is more fully explored.

[0051] In this example, the first population is split and merged based on the first solution set distribution to obtain the reconstructed community structure, specifically: Count the number of migrations of the first solution set distribution, and make merging decisions based on the number of migrations; If merging and reconstruction are performed, the main gradient direction vector of each community in the first solution set distribution is extracted; The communities are divided and merged based on the main direction vector of the gradient to obtain the reconstructed community structure.

[0052] It's important to note that we first analyze the distribution of the first set of solutions obtained through cross-community migration optimization, counting the number of migrations each solution has experienced within the solution set. This migration count reflects how each solution has evolved within the solution space. If a solution migrates multiple times across multiple optimization iterations, this indicates that its position in the solution space is unstable and may require further adjustment or reassignment to a different community. This migration count helps determine the stability and importance of solutions, and based on this information, we can decide whether to merge communities.

[0053] Next, if the clustered areas of solutions with a large number of migrations have similar gradient change trends (i.e., their main gradient directions are similar), the communities can be split or merged by extracting the "main gradient direction vectors" (i.e., the main direction of gradient change) of these solutions. The main gradient direction vector is a mathematical description that represents the dominant direction of gradient change in the solution space and can be obtained by calculating the gradient vector of the solution (e.g., based on the gradient change rate or other optimization indicators). If the main gradient directions of multiple solutions are similar, the communities containing these solutions can be merged so that these solutions are optimized in the same community; conversely, if the main gradient directions differ significantly, these solutions can be split to form a new community structure.

[0054] Through the splitting and merging process, a new, reconstructed community structure is ultimately obtained, making the distribution of solutions more reasonable and better able to meet the optimization objectives. This reconstruction process helps to increase the diversity of the solution set and avoid excessive concentration in a single local area, thereby making the optimization of the entire sound velocity gradient range more comprehensive and balanced.

[0055] S4, identifying a gradient mutation region in the interval, and expanding the defect detection range based on the region; In this example, a gradient mutation region is identified within the interval, and the defect detection range is expanded based on the region, specifically: The piles are divided into axial layers and radial sectors to construct a three-dimensional space grid; In the optimal sound velocity gradient range, the absolute value of the gradient change rate of adjacent spatial grids is calculated; Based on the absolute value of the gradient change rate, the grid is subjected to gradient mutation identification to generate an initial defect seed point set; Taking the preset seed point as the center, spatial expansion is performed along the axial and radial directions of the pile to obtain the expansion area; The extended area is combined with the initial defect seed point set to obtain the defect detection range.

[0056] It should be noted that the pile structure was first divided into layers every 0.5 meters in the axial direction and sectors every 30 degrees in the radial direction to construct a three-dimensional spatial grid system, forming a total of 240 independent grid cells with 20 axial layers and 12 radial sectors. Within the optimal sound velocity gradient range (assuming it is 0.05-0.12 km / s), the absolute value of the gradient change rate of each grid cell relative to the adjacent cells was calculated. When the value exceeded 0.08 km / s / m, it was determined to be a gradient mutation area, thereby identifying 15 initial defect seed points. With each seed point as the center, spatial expansion was performed by extending the range of 1 meter in the axial direction and 60 degrees in the radial direction. Finally, the expanded area was merged with the initial seed point set to form a defect detection range containing 42 grid cells. This range covers the possible void defect areas on the pile surface, providing an accurate spatial positioning basis for the subsequent construction of defect feature maps.

[0057] S5, construct a defect feature map based on the expanded defect detection range, locate the void defect according to the defect feature map and output the graded evaluation result.

[0058] In this example, a defect feature map is constructed based on the expanded defect detection range. The void defect is located based on the defect feature map and a graded evaluation result is output. Specifically, Divide the defect detection range into a three-dimensional voxel grid according to a preset resolution; Multi-band acoustic wave energy attenuation imaging is performed on each voxel to generate an initial defect energy map, such as Figure 4 As shown, the energy intensity distribution of the defect space area is displayed. The red area represents the location with greater attenuation, which may contain structural defects such as voids and cracks. Based on the initial defect energy map, a defect feature tensor is constructed; A 3D convolutional neural network will be used to extract the boundary features of the void area in the defect feature tensor; Based on the boundary characteristics of the void area, defects are divided into three levels: mild, moderate, and severe, and the defect grading assessment results including location, size, and grade are output.

[0059] It should be noted that the expanded defect detection range is first divided into a three-dimensional voxel grid of 5cm×5cm×5cm, and each voxel unit performs acoustic wave energy attenuation imaging in three frequency bands of 10kHz, 20kHz and 30kHz. The initial defect energy spectrum is generated by calculating the signal attenuation coefficient of each frequency band; based on the spectrum, a defect feature tensor containing three dimensions of acoustic wave attenuation rate, energy distribution gradient, and dispersion characteristics is constructed, and a pre-trained three-dimensional convolutional neural network (containing 5 convolutional layers and 3 fully connected layers) is used to extract defect boundary features. After Softmax classification, the network output obtains the precise boundary coordinates of the void area and the defect grade score (divided into mild, moderate and severe levels), and finally outputs a graded assessment report containing the defect location, size and severity.

[0060] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0061] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0062] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0063] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0064] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0065] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Evaluation method for steel pipe pole caps for transmission lines based on prefabricated pipe piles, characterized by: The following steps are involved: The acoustic wave transmitter excites elastic waves at the interface between the cap and the pile, and the ultrasonic array sensor on the top of the pile receives the acoustic wave signal. Separate the characteristics of the acoustic wave signal and evaluate the relative deviation of the sound velocity based on the separation results; Based on the evaluation results, a dynamic optimization algorithm including Gaussian mixture model, density peak detection algorithm and NSGA-II optimization is used to construct the optimal sound velocity gradient interval; Identifying a gradient mutation region in the interval and expanding the defect detection range based on the region; A defect feature map is constructed based on the expanded defect detection range, and the void defects are located according to the defect feature map and the graded evaluation results are output.

2. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 1, characterized in that: The acoustic wave emitting device excites elastic waves at the interface between the cap and the pile, and the ultrasonic array sensor on the top of the pile receives the acoustic wave signal, specifically: A multi-source annular acoustic wave emission array is arranged on the surface of the platform to stimulate pulse signals of different frequencies according to a preset time sequence; Synchronously start the ultrasonic array sensor group on the top of the pile to collect the time domain sequence of reflected waves and scattered waves; The time domain sequence is divided into time windows, and the acoustic wave signal is extracted based on the preset wave velocity inversion model.

3. The evaluation method for steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 2, characterized in that: The feature separation of the acoustic wave signal and the evaluation of the relative deviation of the sound velocity according to the separation result are specifically as follows: The spatial domain filtering algorithm is used to reduce the noise of the sound wave signal to obtain a noise-reduced signal; Perform empirical mode decomposition on the denoised signal to obtain the principal components; Extracting the time-frequency eigenvector of the principal component and calculating the offset between the time-frequency eigenvector and the preset dispersion curve; Based on the offset, the cross-correlation algorithm is used to calculate the relative deviation matrix of the speed of sound; The sound speed relative deviation is evaluated according to the sound speed relative deviation matrix to obtain a sound speed relative deviation evaluation matrix.

4. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 3, characterized in that: Based on the evaluation results, a dynamic optimization algorithm including a Gaussian mixture model, a density peak detection algorithm, and NSGA-II optimization is used to construct the optimal sound velocity gradient interval, specifically: The sound velocity relative deviation evaluation matrix is ​​input into the Gaussian mixture model, and a density peak detection algorithm is used to identify several gradient communities. The first sound velocity gradient topology map is constructed based on the spatial distribution of the gradient communities and the gradient change direction. Constructing a first constraint condition based on the first sound velocity gradient topology graph, including a physical range constraint of the sound velocity gradient, a confidence interval constraint within each Gaussian distribution community, and a maximum change rate constraint between adjacent sampling points; Based on the first constraint, a Monte Carlo method is used to perform random sampling within a preset probability density distribution to obtain a number of first sound velocity gradient samples; The preset distributed community algorithm is used to perform NSGA-II optimization on several first sound velocity gradient samples to obtain the optimal sound velocity gradient interval.

5. The evaluation method for steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 4, characterized in that: The preset distributed community algorithm is used to perform NSGA-II optimization on a number of first sound velocity gradient samples to obtain the optimal sound velocity gradient interval, specifically: A number of first sound velocity gradient samples are taken as individuals and grouped according to preset community labels to obtain a number of first populations; The NSGA-II algorithm is used to perform non-dominated sorting on each first population and obtain several Pareto frontier solution sets; Perform cross-community migration on several Pareto frontier solution sets to obtain the first solution set distribution; Based on the distribution of the first solution set, the first population is divided and merged to obtain the reconstructed community structure; Based on the reconstructed community structure, NSGA-II optimization was performed again to obtain the optimal sound speed gradient range.

6. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 5, characterized in that: The cross-community migration of several Pareto frontier solution sets is performed to obtain the first solution set distribution, which is specifically: Calculate the crowding distance for several Pareto frontier solution sets, and screen several Pareto frontier solution sets according to the crowding distance to obtain several boundary solutions; Calculate the gradient similarity matrix of several boundary solutions and generate community migration instructions based on the gradient similarity matrix; According to the community migration instruction, several boundary solutions are migrated across communities to obtain the first solution set distribution.

7. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 6, characterized in that: The first population is divided and merged based on the first solution set distribution to obtain a reconstructed community structure, specifically: Count the number of migrations of the first solution set distribution, and make merging decisions based on the number of migrations; If merging and reconstruction are performed, the main gradient direction vector of each community in the first solution set distribution is extracted; The communities are divided and merged based on the main direction vector of the gradient to obtain the reconstructed community structure.

8. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 7, characterized in that: Identify the gradient mutation area in the interval and expand the defect detection range based on the area, specifically: The piles are divided into axial layers and radial sectors to construct a three-dimensional space grid; In the optimal sound velocity gradient range, the absolute value of the gradient change rate of adjacent spatial grids is calculated; Based on the absolute value of the gradient change rate, the grid is subjected to gradient mutation identification to generate an initial defect seed point set; Taking the preset seed point as the center, the space is expanded along the axial and radial directions of the pile to obtain the expanded area. The extended area is combined with the initial defect seed point set to obtain the defect detection range.

9. The method for evaluating steel pipe pole caps for power transmission lines based on prefabricated pipe piles according to claim 8, characterized in that: The defect feature map is constructed based on the expanded defect detection range, and the void defect is located according to the defect feature map and a graded evaluation result is output, specifically: Divide the defect detection range into a three-dimensional voxel grid according to a preset resolution; Perform multi-band acoustic wave energy attenuation imaging on each voxel to generate an initial defect energy map; Based on the initial defect energy map, the defect feature tensor is constructed by integrating the acoustic wave attenuation rate, energy distribution gradient and dispersion characteristics; A 3D convolutional neural network will be used to extract the boundary features of the void area in the defect feature tensor; Based on the boundary characteristics of the void area, defects are divided into three levels: mild, moderate, and severe, and the defect grading assessment results including location, size, and grade are output.

Images