Parameter inversion method, device and equipment for main reinforcement of concrete pole and medium

Abstract

This invention discloses a method, apparatus, equipment, and medium for parameter inversion of the main reinforcement of concrete poles. It includes: acquiring and parsing raw data to obtain a pure static magnetic feature sequence; calculating the spatial coordinate offset values ​​of the pure static magnetic feature sequence on the first and second coordinate axes respectively, fusing the spatial coordinate offset values ​​of the first and second coordinate axes to obtain a topology navigation signal curve; acquiring multiple local first numerical points on the topology navigation signal curve, and taking the point with the smallest value among adjacent first numerical points as the second numerical point; obtaining multiple feature slices; constructing a multidimensional feature mapping vector; and inputting the multidimensional feature mapping vector into a pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole. By processing the acquired signal, this invention effectively solves the problems of conventional deep learning models lacking physical prior constraints and facing high difficulty in parameter inversion.

Description

Technical Field

[0001] This invention relates to the technical field of dimensional detection using artificial intelligence methods, and particularly to a method, apparatus, equipment, and medium for parameter inversion of the main reinforcement of concrete poles. Background Technology

[0002] Ordinary reinforced concrete poles are crucial infrastructure for power and communication networks, containing multiple main concrete reinforcement bars arranged longitudinally along their circumference. The diameter, protective layer thickness (the thickness of the cement concrete layer encasing the reinforcement bars, i.e., the lift-off distance), and distribution of these main reinforcement bars directly determine the pole's structural strength and service life. In practical construction scenarios, individually inspecting the diameter and location of each individual main reinforcement bar is impractical. Therefore, non-destructive testing (NDT) is typically performed on pre-determined poles already in service. Currently, NDT primarily employs electromagnetic induction or simple magnetic positioning techniques to scan the reinforcement bars. These methods can only preliminarily determine the location and embedment depth (lift-off distance) of the reinforcement bars, and their accuracy is highly susceptible to interference from the scanning environment.

[0003] With the rapid development of artificial intelligence, intelligent algorithms are gradually being applied to the field of non-destructive testing. However, at present, there are still many technical challenges in applying them to the circumferential scanning inspection of utility poles.

[0004] First, traditional thresholding methods struggle to extract effective signals characterizing diameter. Due to the curved surface of the pole, probe tilting and lifting abruptly occur during circumferential scanning, leading to severe overlap of disturbance signals from the densely packed main reinforcement bars of the concrete pole. Traditional fixed threshold or spatial window methods cannot achieve high-fidelity separation of the main reinforcement signals from a single concrete pole, causing characteristic parameters reflecting diameter changes (such as amplitude and energy) to be obscured by severe signal interweaving, thus failing to support accurate diameter calculation.

[0005] Secondly, fluctuations in scanning speed cause the diameter inversion model to fail. Most existing technologies rely on sampling at equal time intervals, but in engineering settings, it's difficult to maintain a constant scanning speed. This speed fluctuation causes nonlinear stretching or compression of the magnetic field signal along the time axis, resulting in severe drift of the frequency domain features extracted from the sampling points. Since diameter features are closely related to spectral morphology, this dimensional distortion directly leads to a precipitous drop in the diameter prediction accuracy of the inversion model under real-world conditions such as non-uniform scanning in engineering settings.

[0006] Furthermore, directly applying conventional deep learning approaches to pole circumferential scanning and detection—for example, some solutions attempt to introduce purely data-driven models—lacks a physical prior mechanism. Additionally, the highly similar effects of rebar diameter and lift-off depth on magnetic field disturbances make it difficult for the model to understand the physical distribution of magnetic dipoles, leading to confusion and misjudgments. Diameter and depth are also difficult to decouple. Moreover, the combination of diameter, burial depth, and adhesion conditions increases exponentially, requiring massive amounts of labeled data for training. Clearly, collecting large amounts of high-precision data in real-world engineering scenarios is impractical.

[0007] In summary, existing technical solutions primarily focus on the qualitative identification and spatial geometric positioning of the main reinforcement bars of concrete poles. However, they lack effective quantitative characterization and high-precision decoupling and inversion methods for the core physical parameter reflecting the structural strength of the pole—the diameter. Therefore, there is an urgent need for a technical solution that can integrate static and magnetic disturbance physical constraints for adaptive signal stripping and possess scanning speed interference compensation capabilities, thereby achieving synchronous and high-precision inversion of the diameter and lift-off parameters of the main reinforcement bars of concrete poles. Summary of the Invention

[0008] The purpose of this invention is to provide a method, apparatus, equipment, and medium for parameter inversion of the main reinforcement of concrete poles, in order to solve the problems of lack of physical prior constraints and high difficulty in parameter inversion in the conventional deep learning model pure data-driven mode of non-destructive testing, and to effectively improve the inversion detection accuracy of the main reinforcement parameters of concrete poles.

[0009] To achieve the above objectives, the technical solution adopted by this invention is: a parameter inversion method for the main reinforcement of concrete poles, comprising the following steps: S1: Collect raw data and parse the raw data to obtain a pure static magnetic feature sequence; wherein, the raw data is multi-channel static magnetic detection data collected by using a static magnetic differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions; S2: Calculate the spatial coordinate offset values ​​of the pure magnetic static feature sequence on the first and second coordinate axes respectively, and fuse the spatial coordinate offset values ​​of the first and second coordinate axes to obtain the topology navigation signal curve; wherein, a coordinate system is established with the center of the horizontal section at one end of the pole as the origin, the tangent along the circumference of the section is taken as the direction of the first coordinate axis, and the axial direction extending from one end of the pole to the other end is taken as the direction of the second coordinate axis; the second coordinate axis is perpendicular to the horizontal section; S3: Based on a preset protrusion threshold, obtain multiple local first value points on the topology navigation signal curve, and take the point with the smallest value among adjacent first value points as the second value point; cut the topology navigation signal curve based on the second value point to obtain multiple feature slices; S4: Extract multi-dimensional signal features from each feature slice and construct a multi-dimensional feature mapping vector; S5: Input the multidimensional feature mapping vector into the pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole; wherein, the pre-trained nonlinear regression model is trained based on the known feature mapping vector of the main reinforcement of the concrete pole.

[0010] This invention incorporates prior physical knowledge into the collected raw data. Specifically: it uses dual probes to acquire differential signals, eliminating environmental interference; it calculates and fuses spatial offset values ​​on different coordinate axes, effectively compensating for the slight probe posture deviation and single-axis signal attenuation caused by the curved surface of the pole; it uses a protrusion threshold mechanism to filter local numerical points, with each first numerical point corresponding to a single rebar, and uses a second numerical point to segment the feature slices corresponding to each rebar, effectively removing interference signals from overlapping main reinforcement bars of the concrete pole and achieving complete signal separation for each rebar; and it constructs a multi-dimensional feature mapping vector to complete the conversion from the time domain to the spatial domain, decoupling the feature parameters from the scanning speed and avoiding the sharp drop in diameter prediction accuracy caused by non-uniform scanning on site. These processes provide reliable data for subsequent inversion calculations, improving the accuracy of inversion detection.

[0011] Further, in S1, the radial direction from the center of the cross-section to the circumference is taken as the direction of the third coordinate axis, and the first coordinate axis and the third coordinate axis are orthogonal in the horizontal cross-section; the magnetic static differential detection device includes a first probe and a second probe arranged symmetrically, the first probe and the second probe have opposite polarities of sensitivity vectors on the first coordinate axis, the same polarity of sensitivity vectors on the second coordinate axis, and opposite polarities of sensitivity vectors on the third coordinate axis; the analysis of the original data specifically includes: processing the original data of the first coordinate axis and the second coordinate axis collected by the first probe and the second probe, extracting differential disturbance components in a directional manner, and obtaining a pure magnetic static disturbance feature sequence.

[0012] By utilizing the symmetrical arrangement of the first and second probes with opposite polarities of sensitivity vectors in the tangential and radial directions, and based on the physical response characteristics, the common-mode interference of the external environmental geomagnetic field widely distributed in the field is accurately filtered out. This allows for the precise extraction of the local distortion components (i.e., differential-mode disturbance components) of the tangential and axial static magnetic field caused by the main reinforcement of the concrete pole, ultimately outputting a pure static magnetic disturbance feature sequence with a high signal-to-noise ratio.

[0013] Furthermore, S2 specifically includes the following steps: S2.1: Calculate the statistical median of the pure static magnetic characteristic sequence on the first and second coordinate axes respectively; S2.2: Calculate the spatial coordinate offset between the magnetic field value of each equidistant physical space grid point within the first and second coordinate axes and the statistical median of the corresponding coordinate axis; S2.3: The spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis are superimposed by a scalar to obtain the topology navigation signal; S2.4: Based on the topology navigation signal, construct the topology navigation signal curve.

[0014] By employing a dual-axis signal calculation method, specifically using the statistical median as the reference value for the magnetic field along each axis, the calculation error caused by the overall baseline deviation is eliminated. Based on this, a spatial offset value is calculated, which can intuitively reflect the degree of distortion of the local magnetic field relative to the reference. This effectively compensates for the single-axis signal reception angle deviation, loss of static magnetic features, and signal attenuation caused by the curvature of the pole itself when the probe scans along the pole's circumferential direction. It fully preserves the spatial distribution characteristics of the magnetic field, providing a reliable topology navigation basis for subsequent signal slicing and feature extraction.

[0015] Furthermore, in S3, the magnitude of the spuriousness threshold is dynamically adjusted based on the number of local first numerical points of the topology navigation signal curve, specifically including: When the number of local first numerical points exceeds the preset upper limit threshold, the protrusion threshold is increased; When the number of local first numerical points is detected to be lower than the preset lower limit threshold, the protrusion threshold is reduced.

[0016] By dynamically adjusting the protrusion threshold, it can adapt to the differentiated magnetic field signals generated by the main reinforcement bars of concrete poles in different areas and with different arrangement densities, accurately locate the local first value point, improve the accuracy of signal cutting and feature slicing, and ensure the signal separation effect of a single main reinforcement bar.

[0017] Furthermore, in S3, when the amplitude of the first numerical point exceeds the first preset threshold and the amplitude of the second numerical point exceeds the second preset threshold, it is determined that multiple steel bars are stuck together, and an adhesion state warning is triggered. By marking abnormal states in advance, normal single-reinforcement signals and multi-reinforcement overlapping signals are distinguished, avoiding interference from adhesion signals with subsequent feature extraction and parameter inversion, thus ensuring the accuracy of the detection results.

[0018] Furthermore, S3 also includes: mapping the topology navigation signal curve to an equidistant physical space grid based on the physical displacement parameters of the first probe and the second probe, to obtain the mapped topology navigation signal curve; Based on a preset protrusion threshold, multiple local first value points of the mapped topology navigation signal curve are obtained, and the point with the smallest value among adjacent first value points is taken as the second value point. Based on the second numerical point, the mapped topology navigation signal curve is cut to obtain multiple feature slices.

[0019] By mapping the non-uniform topological navigation signal curve in the time domain to an equidistant physical space grid through interpolation and resampling, and by binding the signal amplitude to the actual scanning position of the probe one by one, the time-domain nonlinear distortion caused by uneven manual scanning speed and start-stop disturbances is eliminated, thereby obtaining a steady-state static magnetic disturbance distribution field that varies with spatial position.

[0020] Furthermore, S4 also includes: performing circumferential angle positioning of the main reinforcement of the concrete pole through circumferential scanning to obtain positioning results, extracting the positioning results and multi-dimensional signal features of each feature slice, constructing a multi-dimensional feature mapping vector, collecting circumferential angle data through the angle encoder of the static magnetic differential detection device, completing the circumferential positioning of the main reinforcement of the concrete pole, so that the signal features match the actual position of the reinforcement, and accurately marking the detection points.

[0021] Based on the same concept, the present invention also provides a parameter inversion device for the main reinforcement of concrete poles, which is used to implement the above method, the device comprising: The data acquisition and analysis module is used to acquire raw data and analyze the raw data to obtain a pure static magnetic feature sequence; wherein, the raw data is multi-channel static magnetic detection data acquired by using a static magnetic differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions; The magnetic field signal enhancement module is used to calculate the spatial coordinate offset values ​​of the pure static magnetic feature sequence on the first coordinate axis and the second coordinate axis respectively, and fuse the spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis to obtain the topology navigation signal curve; wherein, a coordinate system is established with the center of the horizontal section at one end of the pole as the origin, the tangent along the circumference of the section is taken as the direction of the first coordinate axis, and the axial direction extending from one end of the pole to the other end is taken as the direction of the second coordinate axis; the second coordinate axis is perpendicular to the horizontal section; The feature slice acquisition module is used to acquire multiple local first value points on the topology navigation signal curve based on a preset protrusion threshold, and take the point with the smallest value among adjacent first value points as the second value point; and cut the topology navigation signal curve based on the second value point to obtain multiple feature slices. The multidimensional feature mapping vector construction module is used to extract multidimensional signal features from each feature slice and construct multidimensional feature mapping vectors. The parameter prediction module is used to input the multidimensional feature mapping vector into a pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole; wherein, the pre-trained nonlinear regression model is trained based on the known feature mapping vector of the main reinforcement of the concrete pole.

[0022] Based on the same concept, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0023] Based on the same concept, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described above.

[0024] Compared with existing technologies, the beneficial effects of this invention are as follows: By employing active permanent magnet excitation combined with a forward and reverse differential dual-probe array to acquire dual-axis data, a magnetic field physical prior mechanism is introduced, and differential operations are used to eliminate common-mode interference of the geomagnetic field caused by the environment. This allows for the directional extraction of local static magnetic field differential-mode disturbance signals caused by the main reinforcement of concrete poles, solving the problem of interference signals caused by the scanning environment mixing with the detected main reinforcement signals, making it impossible to identify the true main reinforcement signals. Furthermore, by using the median to calculate the offset value and completing dual-axis coordinate fusion, signal loss and interference caused by pole attitude are eliminated, solving the problems of signal attenuation and zero-crossing point loss in traditional single-axis methods. Finally, by employing a local numerical point segmentation method, the signals of the main reinforcement of multiple concrete poles are effectively distinguished, providing accurate and reliable data support for subsequent inversion. Attached Figure Description

[0025] To more clearly illustrate the technical method of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only one embodiment of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart illustrating the parameter inversion method for the main reinforcement of concrete poles in an embodiment of the present invention. Figure 2 This is a schematic diagram of the hardware layout and sensor spatial arrangement of active magnetostatic detection using permanent magnets in an embodiment of the present invention. Figure 3 This is a three-dimensional spatial structure diagram of the pole and the arrangement of the main reinforcement bars of the concrete pole inside it, as shown in an embodiment of the present invention. Figure 4 This is a two-dimensional top view of the pole and the arrangement of the main reinforcement bars of the concrete pole inside in an embodiment of the present invention; Figure 5 This is a diagram of the original magnetic static disturbance signal of the first channel acquired by the magnetic static differential detection device in this embodiment of the invention; Figure 6 This is a diagram of the original magnetic static disturbance signal of the second channel acquired by the magnetic static differential detection device in this embodiment of the invention; Figure 7 This is an adaptive segmentation effect diagram of the one-dimensional topology navigation signal curve in an embodiment of the present invention; Figure 8 This is a scatter plot showing the evaluation of the diameter parameter prediction error by the integrated tree regression model in this embodiment of the invention. Figure 9 This is a scatter plot showing the evaluation of the prediction error of the extraction parameter by the integrated tree regression model in this embodiment of the invention.

[0027] exist Figure 2 middle 1. First probe; 2. Second probe; 3. Main reinforcement of the concrete pole to be tested; 4. N pole; 5. Permanent magnet; 6. S pole. Detailed Implementation

[0028] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other. For ease of description, the terms "upper," "lower," "left," and "right" used below only indicate that they correspond to the upper, lower, left, and right directions in the accompanying drawings and do not limit the structure.

[0029] The technical methods of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] The technical methods of this application will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0031] The core control and data transmission unit of the hardware device used in the embodiments of the present invention is built using an STM32 microcontroller combined with a W5500 Ethernet chip to support the operation and data interaction between modules.

[0032] Figure 1 A flowchart illustrating the parameter inversion method for the main reinforcement of concrete poles provided in an embodiment of the present invention is shown, as follows: Figure 1 As shown, the method includes the following steps: S1: Collect raw data and parse the raw data to obtain a pure static magnetic feature sequence; wherein, the raw data is multi-channel static magnetic detection data collected by using a static magnetic differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions.

[0033] Figure 2The schematic diagram of the detection method according to an embodiment of the present invention is shown. The detection hardware adopts a forward and reverse orthogonal differential dual-probe array structure. Two sets of magnetic probes, namely the first probe 1 and the second probe 2, are symmetrically arranged to the side (or directly below) the permanent magnet 5. The two ends of the permanent magnet 5 correspond to the N pole 4 and the S pole 6, respectively, and are used to actively excite the main reinforcement 3 of the concrete pole to be tested below. The sensing probe unit adopts a miniaturized arc-shaped adaptive design, which can stably fit the curved surface of the pole with a diameter between 150mm and 550mm. The nominal lift-off gap between the probe and the surface of the pole under test is maintained within a preset effective lift-off range, which is preferably about 10mm.

[0034] Figure 3 and Figure 4 The diagram shows a three-dimensional spatial structure diagram and a two-dimensional top view diagram of the main reinforcement arrangement of the pole and the concrete pole inside. The main reinforcement to be measured is arranged longitudinally inside the pole.

[0035] The magnetic static differential detection device includes a first probe 1 and a second probe 2 arranged symmetrically. The sensitivity vectors of the first probe 1 and the second probe 2 have opposite polarities on the first coordinate axis (tangential, i.e., the X-axis), the same polarity on the second coordinate axis (axial, i.e., the Z-axis), and opposite polarities on the third coordinate axis (radial, i.e., the Y-axis). A coordinate system is established as follows: the origin is the center of the horizontal section at one end of the pole; the tangential direction along the circumference of the section is taken as the first coordinate axis direction; the axial direction extending from one end of the pole to the other is taken as the second coordinate axis direction; the second coordinate axis is perpendicular to the horizontal section; and the radial direction from the center of the section to the circumference is taken as the third coordinate axis direction. The first and third coordinate axes are orthogonal within the horizontal section. During actual scanning, the acquired multi-channel raw magnetic static data contains significant external environmental geomagnetic background interference signals and system noise that dynamically fluctuate with the scanning location. For poles of different physical sizes, the sampling frequency of the acquisition module is set to an adaptive adjustable range (preferably including 100Hz, 200Hz, 400Hz or 1000Hz) to adapt to the consistency of spatial sampling density of poles of different diameters under constant scanning angular velocity.

[0036] Figure 5 and Figure 6 The diagrams show the original magnetic field disturbance signals of the first and second channels acquired by the first and second probes of the magnetic field differential detection device in this embodiment of the invention. Each channel contains data along three coordinate axes. The X and Y axes of the two channels have opposite signal directions, while the Z axis has the same signal direction.

[0037] The analysis of the raw data specifically includes: processing the raw data of the first and second coordinate axes collected by the first and second probes, extracting differential disturbance components in a directional manner, and obtaining a pure magnetic static disturbance feature sequence. Based on the differential signal formed by the forward and reverse spatial topology arrays, differential suppression processing is performed using its physical response characteristics, i.e., the polarity mapping relationship based on spatial symmetry distribution, to accurately filter out the wide-area distributed common-mode interference of the external environmental geomagnetic field, and to extract the local distortion component of the static magnetic field (i.e., differential-mode disturbance component) caused by the main reinforcement of the concrete pole, finally outputting a pure magnetic static disturbance feature sequence with a high signal-to-noise ratio.

[0038] To address the severe geomagnetic fluctuation interference in open spaces, this invention employs active permanent magnet excitation combined with a forward and reverse differential dual-probe array. Through physical-level differential computation, it accurately filters out widely distributed geomagnetic common-mode interference and directionally extracts local static magnetic field differential-mode disturbance signals caused by the main reinforcement of concrete poles. This significantly improves the signal-to-noise ratio and adaptability to the field environment from a hardware perspective.

[0039] S2: Because the curved surface of the pole causes slight deviations in probe attitude and attenuation of single-axis signals, a physical dual-axis orthogonal fusion strategy is adopted: within a time window, the spatial coordinate offset values ​​of the pure magnetic static feature sequence on the first and second coordinate axes are calculated respectively, and the spatial coordinate offset values ​​of the first and second coordinate axes are fused to obtain the topology navigation signal curve. Specifically, the following steps are included: S2.1: Calculate the global statistical median of the pure static magnetic characteristic sequence on the first and second coordinate axes respectively, and use it as the steady-state field-free background benchmark; S2.2: Calculate the spatial coordinate offset between the magnetic field values ​​of each equidistant physical space grid point within the first and second coordinate axes and the statistical median of the corresponding coordinate axis, i.e., calculate the absolute difference of each sampling point relative to the background reference. S2.3: The spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis are superimposed by time-corresponding scalar values ​​and fused to generate a topology navigation signal that is resistant to attitude disturbances; S2.4: Based on the topology navigation signal, a topology navigation signal curve is constructed so that when the probe loses or attenuates the single-axis static magnetic feature due to the curvature of the pole during circumferential scanning, energy compensation is performed through the deviation of the other coordinate axis, effectively overcoming the detection blind zone caused by the single-axis zero crossing point.

[0040] To address the challenge of single-axis signal attenuation or loss of zero-crossing points when scanning along the curved surface of a utility pole, this invention extracts the median of the steady-state signal as a physical background benchmark based on a norm minimization criterion, and scalarly superimposes the spatial absolute baseline deviations of the two axes. This strategy fully utilizes the energy complementarity of the static magnetic field in the orthogonal dimensions of space to generate a one-dimensional topological navigation signal curve resistant to attitude interference, overcoming the problems of baseline drift and detection blind zones with extremely low computational cost.

[0041] S3: Based on a preset protrusion threshold, obtain multiple local first numerical points (i.e., spatial maxima) representing the physical center of the main reinforcement of the concrete pole in the topology navigation signal curve, and take the point with the smallest value among adjacent first numerical points as the second numerical point; cut the topology navigation signal curve based on the second numerical point to obtain multiple feature slices. Adaptive segmentation of the signal is achieved using the energy minima as the boundary, stripping away the overlapping and intertwined static magnetic disturbance signals to obtain independent feature slices.

[0042] In one embodiment, based on the physical displacement parameters of the first and second probes, the topology navigation signal curve is mapped to an equidistant physical space grid to obtain the mapped topology navigation signal curve. By combining the probe physical displacement parameters (such as encoder pulse data), a spatial resampling algorithm is used to map the navigation signal acquired in the time domain to the physical space domain. By binding the signal amplitude to the actual scanning position of the probe one by one, the time-domain nonlinear distortion caused by uneven manual scanning speed and start-stop disturbances can be eliminated, thereby obtaining a steady-state static magnetic disturbance distribution field that varies with spatial position. Based on a preset protrusion threshold, multiple local first value points of the mapped topology navigation signal curve are obtained, and the point with the smallest value among adjacent first value points is taken as the second value point. Based on the second numerical point, the mapped topology navigation signal curve is cut to obtain multiple feature slices.

[0043] In one embodiment, when the amplitude of the first numerical point exceeds the first preset threshold and the amplitude of the second numerical point exceeds the second preset threshold, it is determined that multiple steel bars are stuck together, and an adhesion status warning is triggered.

[0044] When multiple reinforcing bars are tightly bonded, the equivalent magnetization cross-sectional area increases, and the resulting magnetic static disturbance signal no longer exhibits a clear, independent bimodal or multimodal pattern, but rather a wider, more diffuse, and integrated single-peak pattern. This is reflected in the subsequent mapping vector, where the time-domain maxima (peak characteristic) increase accordingly, and the trough spacing (span characteristic) obtained based on the segmentation also widens simultaneously. By jointly evaluating the extreme values ​​and span characteristics in the multidimensional feature vector, reliable data support is provided for effectively distinguishing between independent main reinforcing bars and multiple bonded states in concrete poles.

[0045] In one embodiment, the protrusion threshold is dynamically adjusted based on the number of local first numerical points in the topology navigation signal curve. For example, if the number of detected first numerical points is greater than a preset upper threshold, it is determined that a large number of noise pseudo-peaks are misidentified as maxima, and the protrusion threshold is raised to filter invalid clutter and reduce redundant first numerical points. If the number of detected first numerical points is less than a preset lower threshold, it is determined that the real rebar peaks are missed by the threshold, and the protrusion threshold is lowered to detect buried effective maxima, adapting to signal strength changes caused by different lift depths and different diameter rebars.

[0046] Figure 7 This diagram illustrates the adaptive segmentation effect of the one-dimensional topology navigation signal curve in an embodiment of the present invention. The horizontal axis represents the actual physical location scanned by the probe, and the vertical axis represents the magnetic field combined with the absolute baseline deviation, i.e., the one-dimensional topology navigation signal curve generated in step S2. For the signal overlap caused by the dense arrangement of the main reinforcement bars on concrete poles, a protrusion threshold mechanism based on a dynamic scaling factor (preferably 0.15 in this embodiment) is introduced. The system accurately retrieves local maxima points (such as...) in the spatial domain navigation signal that satisfy this threshold in the... Figure 7 The peaks of the star-shaped markers, each maximum point uniquely corresponds to the physical center of the main reinforcement (i.e., steel bar) of a concrete pole.

[0047] Between adjacent wave crests (i.e., the centers of the main reinforcement bars of adjacent concrete poles), the system dynamically retrieves energy minimum points as natural physical cutting boundaries (e.g., Figure 7 The algorithm can adaptively adjust the cutting width according to the real-time extracted rebar spacing, and adaptively divide the continuous circumferential magnetic disturbance signal into feature slices that characterize the main reinforcement of the independent concrete pole. This achieves high-fidelity stripping of continuously overlapping disturbance signals and provides independent feature slices for subsequent S4.

[0048] This invention improves upon traditional isochronous sampling logic by innovatively introducing physical space mapping technology to reconstruct and map the original data onto an equidistant physical space grid. This improvement ensures that the extracted spatial features remain constant at the physical scale regardless of scanning speed fluctuations or start / stop operations, completely resolving the signal dimensional distortion and spectral drift problems caused by non-uniform scanning.

[0049] S4: Extract multi-dimensional signal features from each feature slice and construct a multi-dimensional feature mapping vector. This transforms the detected physical signal into structured data, enabling subsequent recognition and computation.

[0050] For the independent feature slices obtained by S3 segmentation, multidimensional feature extraction is performed on the pure micromagnetic feature sequences corresponding to the first and second channels respectively. Let the discrete signal sequence after removing local minima within the slice be X={x1,x2,...,x...}N}; where x n Let N represent the amplitude of the static magnetic disturbance at the nth discrete sampling point, where n = 1, 2, ..., N; N is the sequence length of the feature slice after spatial domain mapping (i.e., the number of spatial quantization points), and its value is determined by the physical induction width of the magnetic field distortion caused by the main reinforcement of the concrete pole on the pole surface and the spatial axis sampling resolution of the system. By switching the calculation reference N from the time domain to the spatial domain, the feature parameters and the scanning speed are essentially decoupled.

[0051] The following section details the construction process of the multidimensional feature mapping vector. The construction of the 11-dimensional temporal spatial features specifically includes: extracting the mean and minimum values ​​of the original signal sequence as the original signal representation (2 dimensions); to eliminate the background noise caused by the environmental magnetic field, zero drift correction is performed on the sequence, and the relative mean, minimum, maximum, and peak-to-peak values ​​of the corrected sequence are calculated as the zero drift correction representation (4 dimensions); further, the standard deviation, median, 5th percentile, and 95th percentile of the sequence are calculated to obtain the statistical dispersion features (4 dimensions); simultaneously, the root mean square value Xrms of the sequence is calculated as the energy concentration feature (1 dimension), and its calculation formula is as follows: ; The aforementioned 11-dimensional time-domain features can reflect the overall strength and local distortion gradient of the static magnetic disturbance field, and have a strong nonlinear correlation with the cross-sectional area (diameter) of the reinforcing bar and the sensor sensing distance (lift-off).

[0052] By performing a Discrete Fast Fourier Transform (FFT) on the signal, a 4-dimensional frequency domain spatial feature is constructed to characterize the spectral evolution of the magnetic field signal caused by differences in rebar specifications or burial depth. Specifically, this includes: extracting the energy proportions of the low-frequency band (first 20% of frequency points) and the high-frequency band (last 80% of frequency points) to construct energy structure features; and extracting spectral complexity features, specifically including calculating spectral entropy and defining the total energy. ,in, Let represent the power spectral density value corresponding to the k-th frequency point, and ε be a local constant. The normalized probability distribution is calculated by... Extracting power spectrum information entropy Then, the normalized dominant frequency position, i.e. the position index of the frequency point corresponding to the maximum power spectral density, is extracted in the spectrum to obtain the frequency center feature.

[0053] The aforementioned four-dimensional frequency domain features can effectively capture the "thickness" of a signal's waveform. Typically, as the diameter of the reinforcing bar increases or the lift-off depth decreases, the dominant frequency of the signal shifts to lower frequencies, and the spectral entropy exhibits a regular change.

[0054] In one embodiment, the circumferential angle of the main reinforcement of the concrete pole is located by circumferential scanning to obtain the positioning result. The positioning result and the multi-dimensional signal features of each feature slice are extracted to construct a multi-dimensional feature mapping vector to eliminate the difference between encoder resolution and unit dimensions. Specifically, the absolute angular position θ of the main reinforcement of the concrete pole on the circumference of the pole is calculated by the following formula: ; Where L is the absolute physical displacement parameter of the probe in the current circumferential scan; R is the radius of the concrete base of the pole; k is the proportional conversion coefficient related to the resolution of the positioning sensor and the transmission mechanism. The system can dynamically calibrate the value of k according to the encoder accuracy of the actual scanning hardware.

[0055] This invention constructs a feature mapping vector containing 15-dimensional high-order spatial domain and spatial frequency domain statistical features to deeply explore the nonlinear physical laws between signal amplitude, energy distribution and the equivalent magnetization volume of the main reinforcement of concrete poles.

[0056] S5: Input the multidimensional feature mapping vector into the pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole; wherein, the pre-trained nonlinear regression model is trained based on the known feature mapping vector of the main reinforcement of the concrete pole.

[0057] The constructed 15-dimensional feature mapping vector is input into a pre-trained nonlinear regression model, such as an ensemble tree regression model, preferably a random forest regressor, which outputs the predicted diameter and predicted lift-off in parallel. To ensure the accuracy of the nonlinear mapping, the ensemble tree regression model is trained during the training phase using the 15-dimensional feature mapping vector of the calibrated samples as input features, and the actual physical diameter and actual lift-off depth of the corresponding main reinforcement bars of the concrete pole as supervision labels, through supervised learning training by minimizing the mean squared error (MSE) criterion.

[0058] During the training phase, this embodiment of the invention constructs a wide-range calibration dataset: the diameter of the measured rebar covers 5mm to 24mm, and the lift-off depth covers 10mm to 25mm. For each "diameter-lift-off" specification combination, only 20 to 25 lightweight samples need to be collected. Figure 8 and Figure 9 The model evaluation results show that, through this gray box mapping mechanism, this method can effectively control the diameter and lift-off prediction error (MAE) of most measured samples within 1.0 mm while significantly reducing the amount of data collected.

[0059] By fusing the features of two sets of probes, the model's sensitivity to minute changes in the diameter of the main reinforcement bars of concrete poles was significantly enhanced. Combined with an integrated tree regression mapping manifold, the "data hunger" problem of pure deep learning models was solved, achieving high-precision inversion with a small sample size. The diameter prediction error was effectively controlled within 1.0 mm, meeting the quantitative requirements of power systems for assessing the service life of poles.

[0060] The method of this invention overcomes the technical shortcomings of existing technologies, such as the susceptibility of uniaxial static magnetic signals to interference from curved surfaces, the difficulty in separating overlapping signals of main reinforcement bars in dense concrete poles, signal dimensional distortion caused by scanning speed fluctuations, and the susceptibility of conventional data-driven models to disturbances and their reliance on massive amounts of data. It effectively eliminates geomagnetic interference and single-dimensional blind zones, breaks through the limitations of fixed spatial windows, and achieves high-fidelity adaptive separation and high-precision parameter inversion of the concealed main reinforcement bars inside the pole.

[0061] Example 2 Based on the same concept, embodiments of the present invention also provide a parameter inversion device for the main reinforcement of concrete poles, the device comprising: The data acquisition and analysis module is used to acquire and analyze raw data to obtain a pure magnetic static feature sequence. The raw data is multi-channel magnetic static detection data acquired by using a magnetic static differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions. The module acquires the multi-channel raw magnetic static data of the pole circumferential scan, analyzes the differential signal, and eliminates external environmental geomagnetic background interference. The magnetic field signal enhancement module is used to calculate the spatial coordinate offset values ​​of the pure static magnetic feature sequence on the first coordinate axis and the second coordinate axis respectively, and fuse the spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis to obtain the topology navigation signal curve; wherein, a coordinate system is established with the center of the horizontal section at one end of the pole as the origin, the tangent along the circumference of the section is taken as the direction of the first coordinate axis, and the axial direction extending from one end of the pole to the other end is taken as the direction of the second coordinate axis; the second coordinate axis is perpendicular to the horizontal section; The feature slice acquisition module is used to acquire multiple local first value points on the topology navigation signal curve based on a preset protrusion threshold, and take the point with the smallest value among adjacent first value points as the second value point; cut the topology navigation signal curve based on the second value point to obtain multiple feature slices; interpolate and resample the one-dimensional topology navigation signal curve by combining displacement parameters to construct an equidistant physical space grid, and search for energy minimum points in adjacent maximum value intervals in the grid as physical cutting boundaries, so as to adaptively divide the continuous magnetic static disturbance signal into independent feature slices; The multidimensional feature mapping vector construction module is used to extract multidimensional signal features of each feature slice in equidistant physical space and spatial frequency domain, and construct multidimensional feature mapping vectors. The parameter prediction module is used to input the multidimensional feature mapping vector into a pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters of the main reinforcement of the concrete pole. The pre-trained nonlinear regression model is trained based on the known feature mapping vectors of the main reinforcement of the concrete pole. The feature mapping vectors are then input into an ensemble tree regression mapping manifold to output the corresponding predicted diameter parameters and predicted lift-off parameters of the main reinforcement of the concrete pole, and the circumferential angle position of the main reinforcement is calculated in conjunction with the radius of curvature.

[0062] Example 3 Based on the same concept, embodiments of the present invention also provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in Embodiment 1 above.

[0063] Example 4 Based on the same concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described in Embodiment 1 above.

[0064] In the embodiments disclosed in this application, a computer storage medium may be a tangible medium that may contain or store programs for use by or in conjunction with an instruction execution system, apparatus, or device. The computer storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of computer storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0065] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0066] The above embodiments should be understood as being used only to illustrate the present invention more clearly, and not to limit the scope of the present invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art fall within the scope defined by the appended claims.

Claims

1. A parameter inversion method for a concrete pole reinforcement, characterized in that, Includes the following steps: S1: Collect raw data and parse the raw data to obtain a pure static magnetic feature sequence; wherein, the raw data is multi-channel static magnetic detection data collected by using a static magnetic differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions; S2: Calculate the spatial coordinate offset values ​​of the pure magnetic static feature sequence on the first and second coordinate axes respectively, and fuse the spatial coordinate offset values ​​of the first and second coordinate axes to obtain the topology navigation signal curve; wherein, a coordinate system is established with the center of the horizontal section at one end of the pole as the origin, the tangent along the circumference of the section is taken as the direction of the first coordinate axis, and the axial direction extending from one end of the pole to the other end is taken as the direction of the second coordinate axis; the second coordinate axis is perpendicular to the horizontal section; S3: Based on a preset protrusion threshold, obtain multiple local first value points on the topology navigation signal curve, and take the point with the smallest value among adjacent first value points as the second value point; cut the topology navigation signal curve based on the second value point to obtain multiple feature slices; S4: Extract multi-dimensional signal features from each feature slice and construct a multi-dimensional feature mapping vector; S5: Input the multidimensional feature mapping vector into the pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole; wherein, the pre-trained nonlinear regression model is trained based on the known feature mapping vector of the main reinforcement of the concrete pole.

2. The method of claim 1, wherein, In S1, the radial direction from the center of the cross-section to the circumference is taken as the direction of the third coordinate axis. The first coordinate axis and the third coordinate axis are orthogonal in the horizontal cross-section. The magnetic static differential detection device includes a first probe and a second probe arranged symmetrically. The first probe and the second probe have opposite polarities in their sensitivity vectors on the first coordinate axis, the same polarity in their sensitivity vectors on the second coordinate axis, and opposite polarities in their sensitivity vectors on the third coordinate axis. The analysis of the original data specifically includes: processing the original data of the first coordinate axis and the second coordinate axis collected by the first probe and the second probe, extracting the differential disturbance component in a directional manner, and obtaining a pure magnetic static disturbance feature sequence.

3. The method of claim 1, wherein, S2 specifically includes the following steps: S2.1: Calculate the statistical median of the pure static magnetic characteristic sequence on the first and second coordinate axes respectively; S2.2: Calculate the spatial coordinate offset between the magnetic field value of each equidistant physical space grid point within the first and second coordinate axes and the statistical median of the corresponding coordinate axis; S2.3: The spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis are superimposed by a scalar to obtain the topology navigation signal; S2.4: Based on the topology navigation signal, construct the topology navigation signal curve.

4. The method of claim 1, wherein, In S3, the spuriousness threshold is dynamically adjusted based on the number of local first numerical points on the topology navigation signal curve, specifically including: When the number of local first numerical points exceeds the preset upper limit threshold, the protrusion threshold is increased; When the number of local first numerical points is detected to be lower than the preset lower limit threshold, the protrusion threshold is reduced.

5. The method of claim 1, wherein, In S3, when the amplitude of the first numerical point exceeds the first preset threshold and the amplitude of the second numerical point exceeds the second preset threshold, it is determined that multiple steel bars are stuck together, and a sticking state warning is triggered.

6. The method of claim 2, wherein, S3 also includes: mapping the topology navigation signal curve to an equidistant physical space grid based on the physical displacement parameters of the first probe and the second probe, to obtain the mapped topology navigation signal curve; Based on a preset protrusion threshold, multiple local first value points of the mapped topology navigation signal curve are obtained, and the point with the smallest value among adjacent first value points is taken as the second value point. Based on the second numerical point, the mapped topology navigation signal curve is cut to obtain multiple feature slices.

7. The method of claim 1, wherein, S4 also includes: performing circumferential angle positioning of the main reinforcement of the concrete pole by circumferential scanning, obtaining positioning results, extracting the positioning results and multi-dimensional signal features of each feature slice, and constructing a multi-dimensional feature mapping vector.

8. A parameter inversion device for the main reinforcement of a concrete pole, used to implement the method according to any one of claims 1 to 7, characterized in that, The device includes: The data acquisition and analysis module is used to acquire raw data and analyze the raw data to obtain a pure static magnetic feature sequence; wherein, the raw data is multi-channel static magnetic detection data acquired by using a static magnetic differential detection device to perform a circumferential scan of the pole under permanent magnet excitation conditions; The magnetic field signal enhancement module is used to calculate the spatial coordinate offset values ​​of the pure static magnetic feature sequence on the first coordinate axis and the second coordinate axis respectively, and fuse the spatial coordinate offset values ​​of the first coordinate axis and the second coordinate axis to obtain the topology navigation signal curve; wherein, a coordinate system is established with the center of the horizontal section at one end of the pole as the origin, the tangent along the circumference of the section is taken as the direction of the first coordinate axis, and the axial direction extending from one end of the pole to the other end is taken as the direction of the second coordinate axis; the second coordinate axis is perpendicular to the horizontal section; The feature slice acquisition module is used to acquire multiple local first value points on the topology navigation signal curve based on a preset protrusion threshold, and take the point with the smallest value among adjacent first value points as the second value point; and cut the topology navigation signal curve based on the second value point to obtain multiple feature slices. The multidimensional feature mapping vector construction module is used to extract multidimensional signal features from each feature slice and construct multidimensional feature mapping vectors. The parameter prediction module is used to input the multidimensional feature mapping vector into a pre-trained nonlinear regression model for multivariate inversion calculation to obtain the predicted size parameters and predicted lift-off distance parameters corresponding to the main reinforcement of the concrete pole; wherein, the pre-trained nonlinear regression model is trained based on the known feature mapping vector of the main reinforcement of the concrete pole. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 7.

Images