A Defect Three-Dimensional Tomographic Imaging Device and Method Based on Triaxial Orthogonal Coil Magnetic Vector Modulation
The three-dimensional tomographic imaging device for defects using triaxial orthogonal coil magnetic vector modulation solves the problems of directional blind spots and noise interference introduced by mechanical motion in the three-dimensional imaging of defects in electromagnetic detection, and achieves efficient and accurate three-dimensional imaging of defects.
Patent Information
- Application Number
- CN202611297218.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-25
- Publication Date
- 2026-11-17
AI Technical Summary
Existing electromagnetic detection technologies have directional blind spots in three-dimensional defect imaging, cannot analyze the three-dimensional shape of defects, and mechanical motion introduces positioning errors and noise interference, which cannot meet the needs of online rapid detection.
The three-dimensional tomographic imaging device for defects using triaxial orthogonal coil magnetic vector modulation achieves omnidirectional electromagnetic imaging and magnetic gradient tensor matrix reconstruction without mechanical movement by using a triaxial orthogonal coil array and a high-density triaxial magnetic sensor array, combined with a current control system and a signal processing system. This eliminates blind spots and is immune to probe attitude drift interference.
It achieves high-fidelity three-dimensional tomographic imaging of internal defects in components, eliminates blind spots caused by unidirectional excitation, improves detection efficiency and imaging fidelity, and can accurately locate defect boundaries and inversion depth.
Smart Images

Figure CN123112221A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electromagnetic nondestructive testing technology, specifically to a three-dimensional tomographic imaging device and method for defects based on triaxial orthogonal coil magnetic vector modulation. Background Technology
[0002] In critical equipment fields such as aerospace, rail transportation, pressure vessels, nuclear power, and marine engineering, internal defects in metal components, such as cracks, porosity, inclusions, and corrosion pits, are major causes of structural failure. Accurate detection, location, and quantitative evaluation of these defects are crucial for ensuring the safe operation of equipment. Electromagnetic nondestructive testing (EMT) is widely used in the inspection of magnetically and electrically conductive components due to its advantages of being non-contact, having a fast response, requiring no coupling agent, and being sensitive to surface and near-surface defects. However, current electromagnetic testing technologies still have significant limitations in three-dimensional imaging of defects.
[0003] Traditional eddy current and magnetic flux leakage detection methods often employ fixed-direction unidirectional or bidirectional excitation magnetic fields. When the defect extension direction tends to be parallel to the excitation magnetic field or the direction of induced eddy currents, the defect's disturbance to the magnetic field weakens drastically, resulting in a weak detection signal or even being drowned out by noise, forming a directional blind zone. This makes it difficult to effectively illuminate the entire boundary of internal defects with arbitrary spatial orientation and complex morphology. To alleviate the directional sensitivity problem, some solutions achieve multi-angle scanning by rotating probes or rotating coils. However, the introduced mechanical motion mechanism brings positioning errors, vibration artifacts, and motion wear, and the detection speed is limited by the mechanical motion speed, making it unsuitable for online rapid detection requirements.
[0004] Most existing methods only provide two-dimensional projection information of defects on the scanning surface. They cannot resolve the absolute depth of the defect from the detection surface, nor can they reconstruct the three-dimensional morphology of the defect inside the component like medical CT. At the same time, background magnetic field fluctuations caused by probe attitude drift and lift-off fluctuations can seriously interfere with defect signal extraction, resulting in reduced imaging fidelity. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a defect three-dimensional tomographic imaging device and method based on triaxial orthogonal coil magnetic vector modulation, so as to realize high-fidelity three-dimensional tomographic imaging of defects arbitrarily distributed in the internal space of components, overcome the problem of defect omission blind zone caused by unidirectional excitation in traditional electromagnetic detection, and simultaneously infer the lateral position, absolute depth and three-dimensional topological morphology of defects under the premise of single-sided bonding and no mechanical movement, thus mathematically eliminating the background magnetic field interference caused by probe attitude drift.
[0006] To achieve the above objectives, the first aspect of the present invention provides a three-dimensional tomographic imaging device for defects based on triaxial orthogonal coil magnetic vector modulation, comprising a triaxial orthogonal coil array, a high-density triaxial magnetic sensor array, a current control system, and a signal processing system.
[0007] The triaxial orthogonal coil array consists of three groups of six electromagnetic coils arranged concentrically along the X, Y, and Z axes in a spatial rectangular coordinate system. The coils of each axis are concentrically arranged around the geometric center to form a static magnetic field cavity.
[0008] A high-density triaxial magnetic sensor array is rigidly fixed on a non-magnetic substrate, with one side of the array closely attached to the surface of the component under test and placed in a static magnetic field cavity. The array consists of a matrix of multiple full-field triaxial magnetic sensor chips, each chip having rigidly integrated mutually orthogonal x, y, z triaxial sensing units for static synchronous capture of spatial three-dimensional distorted magnetic field signals caused by defects.
[0009] The current control system is independently electrically connected to each group of coils in the triaxial orthogonal coil array. It is used to independently adjust the amplitude ratio and polarity of the excitation current in the triaxial coils, so as to control the magnetic vector synthesized by each coil to perform full-circumferential spherical uniform grid multi-point spatial angle switching in space, and perform all-round electromagnetic vision of the tested component without any mechanical movement.
[0010] The signal processing system is connected to a high-density triaxial magnetic sensor array to perform multi-channel digital phase-locked demodulation of distorted magnetic field signals and convert them into digital magnetic field matrices. Combined with the spatial geometric spacing of the triaxial magnetic sensor chips, a symmetric full-component three-dimensional magnetic gradient tensor matrix is calculated using spatial finite difference. The principal eigenvalue with the largest absolute value in the matrix is extracted to lock the defect boundary. The defect depth is then analyzed based on the spatial attenuation law of the tensor field strength. Finally, the defect boundary data and depth data obtained from all spatial angles pointing downwards are fused in a three-dimensional voxel grid to reconstruct the spatial three-dimensional topological morphology of the defect.
[0011] The second aspect of the present invention provides a defect three-dimensional tomographic imaging method based on the above-mentioned device, including a spherical mesh multi-point magnetic excitation step S1, a distortion signal acquisition and demodulation step S2, a magnetic gradient tensor calculation and principal eigenvalue extraction step S3, a defect depth analysis step S4, and a three-dimensional voxel mesh multi-source inverse projection fusion tomography step S5.
[0012] Compared with the prior art, the embodiments of the present invention bring the following beneficial effects: First, omnidirectional electromagnetic illumination eliminates blind spots in detection. This invention achieves this by independently adjusting the amplitude ratio and polarity of the excitation current of a triaxial orthogonal coil, allowing the synthesized magnetic vector to switch at multiple angles in space according to a uniform spherical grid, applying multiple excitation magnetic fields in different directions to the same defect. Regardless of the defect orientation, there are always several excitation directions that can effectively "illuminate" the defect, fundamentally eliminating blind spots caused by unidirectional excitation.
[0013] Second, multi-angle excitation is achieved without mechanical movement. The direction switching of the synthesized magnetic vector is achieved entirely through electronically controlled modulation of the current ratio and polarity, without requiring any mechanical rotation or movement of the coil or probe. This avoids mechanical positioning errors, motion inertia delays, and mechanical vibration artifacts, thus improving detection efficiency and system reliability.
[0014] Third, attitude drift immunity. This invention uses the principal eigenvalue with the largest absolute value of the three-dimensional magnetic gradient tensor matrix as the defect boundary criterion. This principal eigenvalue is a spatial rotation invariant of the magnetic gradient tensor and is independent of the attitude of the observation coordinate system. Therefore, it can mathematically cancel the background magnetic field amplitude interference caused by probe attitude deflection and lift-off fluctuations, and achieve isotropic robust locking of the defect boundary.
[0015] Fourth, high-fidelity 3D reconstruction in the near-field with single-sided contact. The sensor array operates in the near-field region with a high signal-to-noise ratio, with one side closely attached to the component surface. Combined with tensor reconstruction based on the prior knowledge of the non-dispersion and non-rotational physical properties of magnetic fields according to Maxwell's equations, a complete and symmetrical magnetic gradient tensor can be reconstructed non-destructively using only single-sided near-field measurements. Then, the depth is analyzed by combining the spatial attenuation law of the tensor field strength, and fused in a 3D voxel mesh through multi-angle inverse projection of the spherical surface. Finally, the spatial 3D topological morphology of the defect is reproduced with high fidelity, like a medical CT tomographic image. Simulation verification shows that the lateral positioning error of the internal spherical defect can be controlled to the sub-millimeter level, and the depth inversion error is close to zero, verifying the high fidelity and feasibility of the solution.
[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.
[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description
[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0019] Figure 1 This is a schematic diagram of the defect three-dimensional tomographic imaging device and the defect / defect-free reference differential fusion processing link of the present invention; Figure 2This is a schematic diagram of the spatial orthogonal concentric nested structure of the triaxial orthogonal coil array of the present invention; Figure 3 This is a schematic diagram of the two-dimensional rectangular grid arrangement of the high-density triaxial magnetic sensor array of the present invention; Figure 4 This is a schematic diagram illustrating the principle of the synthesized magnetic vector of the present invention performing 72 discrete spherical orientation switching, wherein... N =72=12×6; Figure 5 This is a flowchart illustrating the three-dimensional tomographic imaging method for defects of the present invention; Figure 6 This is a schematic diagram of the topological morphology of the three-dimensional point cloud in the defect space reconstructed by the present invention.
[0020] In the figure: 1. Triaxial orthogonal coil array; 11. X-axis coil pair; 12. Y-axis coil pair; 13. Z-axis coil pair; 2. High-density triaxial magnetic sensor array; 21. Triaxial magnetic sensor chip; 22. Non-magnetic substrate; 3. Current control system; 4. Signal processing system; 41. Multi-channel digital phase-locked amplifier module; 5. Component under test; 6. Defect; 7. External industrial actuator; 100. Static magnetic field cavity. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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.
[0022] Currently, electromagnetic detection technology employs unidirectional or mechanically rotating excitation magnetic fields, resulting in inherent directional blind spots. Furthermore, mechanical motion introduces positioning errors, vibration artifacts, and speed bottlenecks. Additionally, it can only provide two-dimensional projection information of defects, failing to resolve depth or reconstruct three-dimensional morphology. Moreover, it lacks mathematical immunity to engineering interferences such as probe attitude drift and lift-off fluctuations, leading to severely insufficient image fidelity. Therefore, this invention provides a three-dimensional tomographic imaging device and method for defects based on triaxial orthogonal coil magnetic vector modulation. This method eliminates blind spots and immunizes against probe attitude interference without mechanical motion through omnidirectional modulation of the electrically controlled magnetic vector of triaxial orthogonal coils. Combined with unilateral near-field magnetic gradient tensor measurement and multi-angle inverse projection fusion, it achieves high-fidelity three-dimensional tomographic imaging of defects.
[0023] To facilitate understanding of this embodiment, a detailed description of the defect three-dimensional tomographic imaging device and method based on triaxial orthogonal coil magnetic vector modulation disclosed in this embodiment of the invention will be provided first.
[0024] The device provided in this application embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.
[0025] like Figure 1 As shown, the present invention provides a three-dimensional tomographic imaging device for defects based on triaxial orthogonal coil magnetic vector modulation, comprising a triaxial orthogonal coil array, a high-density triaxial magnetic sensor array, a current control system, and a signal processing system. The current control system is independently electrically connected to each group of coils in the triaxial orthogonal coil array; the signal processing system is communicatively connected to the high-density triaxial magnetic sensor array. These components cooperate to achieve three-dimensional tomographic imaging of defects. The specific structure and function of each component will be described in detail below.
[0026] like Figure 2 As shown, the three sets of coils in the triaxial orthogonal coil array are: a pair of coils placed opposite each other along the X-axis, a pair of coils placed opposite each other along the Y-axis, and a pair of coils placed opposite each other along the Z-axis. Each of the three sets of opposite coils generates an excitation magnetic field in its respective axial direction. Based on the fundamental principle of electromagnetic induction, the magnetic dipole moments of any two sets of coils are orthogonal to each other. Under ideal symmetry, the net magnetic flux through the winding loop of one coil approaches zero, thus reducing the spatial electromagnetic mutual inductance coefficient between any two sets of coils to near zero. To further reduce the coupling components introduced by actual winding and assembly errors, in this embodiment, the winding centers of the three sets of coils are strictly coincident with the geometric center of the device. Simultaneously, due to the axially symmetrical distribution of the windings about the geometric center, stray magnetic flux components introduced by winding or assembly errors cancel each other out between the two opposite coils, thereby reducing the magnetic field components in non-target directions. Therefore, the X, Y, and Z-axis currents can be independently adjusted, reducing the inter-axis coupling effect and providing an electrical decoupling basis for the subsequent synthesis of magnetic vectors in arbitrary directions.
[0027] like Figure 2 As shown, in this embodiment, the triaxial orthogonal coil array, from the innermost layer to the outermost layer, consists of: an inner layer of X-axis opposed coils, a middle layer of Y-axis opposed coils, and an outer layer of Z-axis opposed coils. These three coils are arranged coaxially nested around the geometric center of the device. In this example, each group of coils uses a pair of coaxially placed Helmholtz coils. This structure can obtain a highly uniform magnetic field region at the midpoint of the two coils, facilitating overlap with the central cavity where the sensor array is located.
[0028] Because the winding radius of the inner X-axis coil is smaller than that of the middle Y-axis coil, and the winding radius of the middle Y-axis coil is smaller than that of the outer Z-axis coil, according to the Helmholtz coil central magnetic field relationship, when the number of turns and current are constant, the central magnetic field strength decreases as the coil radius increases. If all three sets of coils use the same number of turns and wire diameter design, when a current of equal amplitude is applied, the magnetic field strength contributed by the outer Z-axis coil at the central cavity will be significantly weaker than that of the inner X-axis coil. This leads to anisotropic deviations in the magnetic field conversion coefficients along different axes within the central cavity region, thus affecting the directional accuracy of the subsequently synthesized magnetic vector.
[0029] To compensate for the magnetic field response deviation caused by the difference in spatial nesting radius of the coils in different axes, this embodiment adopts an impedance matching design with unequal number of turns and unequal wire diameter. Specifically, the inner X-axis coil with a smaller winding radius uses relatively fewer turns and is configured with appropriate wire specifications according to the operating current requirements; the middle Y-axis coil with a moderate winding radius uses a moderate number of turns and wire specifications; and the outer Z-axis coil with the largest winding radius uses relatively more turns and is configured with a larger wire diameter to improve the coil's allowable current carrying capacity and the ampere-turns generated per unit current, thereby compensating for the attenuation of the central magnetic field caused by the increase in coil radius. By differentiating and calibrating the number of turns, wire specifications, and excitation parameters of the three sets of coils, the magnetic field response per unit current generated by the three sets of coils in the central cavity region is kept consistent, achieving consistency in the three-axis magnetic field conversion coefficient, thus ensuring stable amplitude accuracy of the synthesized magnetic vector during switching in different spatial directions.
[0030] like Figure 3 As shown, the high-density triaxial magnetic sensor array in this embodiment consists of multiple full-field triaxial magnetic sensor chips rigidly fixed on a non-magnetic substrate. Specifically, this embodiment employs a tunnel magnetoresistive (TMR) sensor chip with triaxial magnetic field detection capability. This chip integrates mutually orthogonal x, y, and z-axis magnetic sensing units, which can statically and synchronously output magnetic field component signals in three orthogonal directions, providing a complete three-dimensional magnetic field raw data foundation for subsequent magnetic gradient tensor matrix calculations.
[0031] The high-density triaxial magnetic sensor array is arranged in a conformal, two-dimensional rectangular grid topology, tightly attached to the surface of the component, with one side attached to the surface of the component under test. Each triaxial magnetic sensor chip is uniformly arranged along the row and column directions, with both row spacing Δx and column spacing Δy set to 2 mm (satisfying the quantitative range requirement of 1 mm ≤ Δx, Δy ≤ 5 mm in the claims). This value takes into account two factors: firstly, a smaller array spacing can increase spatial sampling density, which is more conducive to capturing details of local magnetic field distortion caused by defects and improving the spatial resolution of defect boundary localization; secondly, too small a spacing will lead to a significant increase in the number of sensor chips and signal acquisition channels required for the same detection area, increasing the array manufacturing cost and the hardware complexity of the signal processing system. In this embodiment, considering the above trade-off between resolution and cost, 2 mm is selected as the preferred value for row and column spacing, satisfying the quantitative range described in the claims while also taking into account engineering feasibility, thereby meeting the requirement for high spatial resolution magnetic field sampling in the three-dimensional tomographic imaging of defects.
[0032] In this embodiment, the excitation current output by the current control system is an alternating current that is a simple harmonic quadrature phase-modulated current, that is, a sinusoidal alternating excitation current with strictly synchronized frequency and phase is introduced into each group of coils of the three-axis orthogonal coil array. By adjusting the amplitude ratio and relative polarity of the current components of each axis, a spatial magnetic vector in a preset direction is synthesized.
[0033] The operating frequency f of the excitation current is pre-calibrated and selected based on the electromagnetic properties of the material of the component under test. This embodiment uses a conductive metal component as an example. Considering the influence of the skin effect on detection sensitivity: if the frequency is too high, eddy currents concentrate on the surface of the component, resulting in a shallow skin depth, making it difficult to effectively respond to deeply embedded defects within the component; if the frequency is too low, although the skin depth increases, the signal-to-noise ratio and demodulation efficiency of the magnetic field distortion signal decrease accordingly. This embodiment considers the above factors and selects an operating frequency f within the range of 1kHz to 10kHz to balance skin depth and detection sensitivity.
[0034] The total number of discrete switching directions N of the multi-point spatial angle pointing switching of the circumferential spherical uniform grid must meet the quantitative range requirement of 32≤N≤128. The value of N directly affects the completeness and detection efficiency of multi-angle electromagnetic illumination: the larger the value of N, the denser the coverage of the spherical excitation direction, which is more conducive to eliminating unidirectional excitation blind spots and improving the completeness of defect detection, but at the same time, it will also increase the number of excitation and demodulation, and prolong the detection cycle; if the value of N is too small, detection blind spots may remain in the sparse areas of the spherical grid, weakening the core technical effect of eliminating unidirectional excitation blind spots of this invention. This embodiment comprehensively balances detection completeness and detection efficiency, and preferably takes N=72, which satisfies the quantitative range of the claims while taking into account engineering feasibility and detection cycle.
[0035] In this embodiment, the signal processing system integrates a multi-channel digital phase-locked amplifier module. This module shares the same clock reference source with the current control system. The current control system generates an excitation reference signal based on the clock reference source and provides the reference signal to the digital phase-locked amplifier module, thereby ensuring that the demodulation reference signal of the digital phase-locked amplifier module and the actual excitation current maintain a stable synchronous relationship in frequency and phase, avoiding demodulation errors caused by frequency drift or phase drift between two independent clock sources.
[0036] Based on this strict synchronization, the digital lock-in amplifier module uses orthogonal demodulation to perform product-integral operations on the original voltage time-domain signals output by each triaxial magnetic sensor chip in the high-density triaxial magnetic sensor array at each spatial angle, with sinusoidal and cosine reference signals of the same frequency, respectively, to extract the in-phase and quadrature components of the signals. Since the disturbance signals caused by environmental stray noise and mechanical vibration are usually out of frequency with the excitation frequency and have random phases, their contributions tend to cancel each other out after product integration within a preset integration time window. The effective magnetic field distortion signal components that are strictly in sync with the excitation frequency and cocorrelated are coherently accumulated and retained, thereby effectively filtering out the disturbance signals caused by environmental stray noise and mechanical vibration, and improving the signal-to-noise ratio of the magnetic field distortion signal.
[0037] In this embodiment, after acquiring the digital magnetic field matrix pointing at each spatial angle, the signal processing system, in conjunction with the spatial geometric spacing Δx and Δy of the triaxial magnetic sensor chip, calculates the full-component three-dimensional magnetic gradient tensor matrix G using spatial finite difference. Its mathematical expression is:
[0038] Each element in the matrix represents the first-order rate of change of the three magnetic field components Bx, By, and Bz along the three spatial directions X, Y, and Z, respectively, and is used to describe the variation characteristics of the local magnetic field distortion caused by the defect in the spatial region.
[0039] The sum of the diagonal elements of the matrix (i.e., the trace of the matrix) is constrained to zero in this embodiment, i.e., it satisfies:
[0040] This constraint is not an artificially imposed mathematical condition, but rather stems from the fundamental physical laws of magnetic fields themselves. Since the high-density triaxial magnetic sensor array is located in the air region outside the measured component, where there is no magnetic monopole source, according to Gauss's law of magnetic fields in Maxwell's equations, i.e. Expanding this, we arrive at the constraint that the trace of the aforementioned matrix is zero. Therefore, the zero-trace constraint of the magnetic gradient tensor matrix is a physical constraint determined by the actual propagation laws of the magnetic field, independent of defect morphology, spatial location, and excitation direction, and remains valid under different detection angles and spatial sampling positions. In the subsequent 3D reconstruction process, the signal processing system uses this non-dispersion constraint of the magnetic field as a physical prior condition, jointly optimizing the solution with the spatial finite-difference calculation results. This compensates for the uncertain gradient components caused by the limited spatial sampling direction during unilateral magnetic field measurement, thereby improving the stability of the 3D magnetic gradient tensor calculation and providing a reliable data foundation for defect boundary localization and depth inversion.
[0041] Performing eigenvalue decomposition on the magnetic gradient tensor matrix G, since G is a symmetric magnetic gradient tensor matrix, yields three real eigenvalues. Extract the one with the largest absolute value as the principal feature value. ,Right now:
[0042] Since the magnetic gradient tensor matrix G is a symmetric real matrix, its eigenvalue decomposition results have a clear physical meaning: matrix G essentially describes the first-order rate of change distribution of the magnetic field in a local region of space. When the overall attitude of the sensor array undergoes a slight tilt or rotation, although the measurement coordinate system changes accordingly and the values of each component of the magnetic gradient tensor in the new coordinate system will change accordingly, the eigenvalues of the magnetic gradient tensor have coordinate rotation invariance and do not change with the selection of the measurement coordinate system, i.e., the principal eigenvalues. It possesses rotational invariance. This mathematical property makes... It can effectively reduce the interference caused by the measurement coordinate system deviation introduced by factors such as changes in coordinate direction and local undulations on the surface of components, and remains sensitive only to the intensity and spatial distribution of magnetic field distortion itself.
[0043] Based on the above characteristics, this embodiment iterates through the principal eigenvalues at each node within the measurement plane of the high-density triaxial magnetic sensor array. ,get Spatial distribution across the entire measurement plane. When a region within the measurement plane is adjacent to a node... A significant abrupt change indicates a discontinuous transition in the local rate of change of the magnetic field in that region. The location of this abrupt change is determined to correspond to the geometric edge of the defect on the currently measured surface. By setting a threshold for determining the abrupt change, [the following is considered:] By extracting the set of continuous nodes in the spatial distribution whose gradient exceeds the threshold, the surface boundary profile of the defect under the current excitation angle can be locked.
[0044] Based on locking the boundary of the defect surface, this embodiment further calculates the first-order tensor Frobenius norm of the magnetic gradient tensor matrix G. This is used to characterize the overall strength of the magnetic field gradient at the current measuring point. Combined with the three real eigenvalues already obtained in claim 8... , The formula for calculation is:
[0045] Since the intensity of magnetic field distortion gradually decreases with increasing vertical depth from the defect, The spatial distribution along the vertical Z-axis also exhibits a corresponding monotonic decay law. To characterize this decay law, this embodiment further calculates... First-order spatial rate of change along the vertical Z-axis It should be noted that, because the high-density triaxial magnetic sensor array 4 is attached to the surface of the component being measured on only one side, measurement points cannot be directly set up along the Z-axis. It cannot be obtained through direct measurement, but needs to be obtained indirectly by combining the partial derivatives of the vertical component of the magnetic field derived analytically in claim 7 and the chain rule of partial derivatives. The specific calculation method is explained in step S4 of the subsequent method embodiment.
[0046] and algebraic ratio Essentially, this reflects the characteristic scale of magnetic field distortion intensity attenuation: the deeper the defect is buried, the longer the spatial attenuation distance that the magnetic field distortion undergoes before reaching the measurement surface. Its absolute level decreases accordingly, while its rate of change with depth... The ratio of the two also decreases accordingly. The electromagnetic properties of the component increase monotonically with increasing defect depth. Based on this monotonic relationship, this embodiment pre-calibrates the electromagnetic properties of the tested component material using a standard defect test block with a known burial depth. Ratio and actual physical depth The correspondence between them is used to obtain the shape correction coefficient. In actual testing, the solution at the current measuring point is... Ratio multiplied by shape correction factor This allows us to inversely solve for the absolute physical depth of the defect relative to the currently measured surface. Its expression is:
[0047] The shape correction coefficient in the above formula It is used to compensate for the differences in magnetic field attenuation characteristics of components with different materials and geometries. It must be pre-calibrated for the specific material being tested before it can be used.
[0048] In an optional embodiment of the present invention, such as Figure 1As shown, the device also includes an external industrial actuator. After completing the multi-source fusion reconstruction of the three-dimensional voxel mesh and reconstructing the spatial three-dimensional topological shape of the defect, the signal processing system further converts the reconstruction result into real-time control commands and sends them to the external industrial actuator via a communication interface. The control commands include at least the following information: the two-dimensional planar coordinates of the defect in the coordinate system of the measured component surface, and the absolute physical depth information of the defect relative to the measurement surface. And the action type identifier to be executed.
[0049] Based on the received control commands, the external industrial actuator drives its carrying machining tool to the physical coordinates corresponding to the defect, and performs one of the following two optional processing methods according to the action type identifier: Firstly, physical coloring marking method, which involves applying visual markings to the surface of the component corresponding to the defect using marking devices such as inkjet or paint dotting, for subsequent manual re-inspection or archiving and traceability. This method is suitable for detection scenarios where the defect only needs to be located and marked and no physical treatment is required. Secondly, the laser stress-relieving beveling method involves using a laser processing device to cut at the corresponding defect location according to preset beveling geometry parameters, thereby reducing the concentration of residual stress around the defect area and inhibiting further crack propagation. This method is suitable for inspection scenarios where confirmed crack-like defects require immediate engineering treatment.
[0050] By automatically generating the aforementioned control commands and linking them with external industrial actuators, this invention enables integrated closed-loop operations from defect detection and 3D positioning to on-site processing, reducing the need for manual secondary positioning and improving the automation level of defect detection and maintenance of complex components.
[0051] The specific embodiments of the device of the present invention have been described in detail above with reference to the accompanying drawings. The following description, in conjunction with... Figure 5 The specific implementation of the defect three-dimensional tomographic imaging method based on the above-mentioned device is further described below. The method is applied to the aforementioned device and includes five steps: S1 multi-point magnetic excitation of spherical mesh, S2 distortion signal acquisition and demodulation, S3 magnetic gradient tensor calculation and principal eigenvalue extraction, S4 defect depth analysis, and S5 three-dimensional voxel mesh multi-source inverse projection fusion tomography. Each step is executed sequentially, and after S2 to S4 are completed, it is determined whether all N spatial angles have been traversed: if not, the process returns to step S1 to switch to the next excitation angle and continue execution; if the process has been completed, step S5 is executed to complete the reconstruction of the three-dimensional morphology of the defect.
[0052] S1. Multi-point magnetic excitation steps for spherical mesh The current control system 5 independently adjusts the amplitude ratio and polarity of the excitation current in the triaxial orthogonal coil array 1, and controls the synthetic magnetic vector to perform multi-point spatial angle switching of the full circumferential spherical uniform grid in N directions in space, so as to perform multi-angle electromagnetic illumination of the inside of the tested component without any mechanical movement.
[0053] Specifically, the multi-point scanning direction of the spherical grid is obtained by discretizing the spatial spherical coordinates, and the current control system 5 inputs AC excitation currents of the same frequency and phase into the X-axis, Y-axis, and Z-axis opposing coils, respectively. The equations for the amplitudes of the currents on each axis are as follows:
[0054] In the formula, The reference amplitude of the basic excitation current The operating frequency of alternating current. The angle between the current spatial synthesized magnetic vector and the Z-axis. Let X be the angle between the projection of the spatially synthesized magnetic vector onto the XY plane and the X-axis, and the number of discrete directions be preferably... N =72; dynamically changed via electronic control and The step combination drives the synthesized magnetic vector to spontaneously complete the switching of the magnetic polarization direction of the isotropic sphere.
[0055] S2. Distortion Signal Acquisition and Demodulation Steps A high-density triaxial magnetic sensor array 4, which is attached to the surface of the component under test on one side, is controlled to synchronously capture the spatial three-dimensional distorted magnetic field signal caused by the defect. The multi-channel digital lock-in amplifier module is used to perform orthogonal demodulation and amplitude conversion to obtain the high-purity digital magnetic field absolute amplitude vector of each measuring node under the current excitation angle.
[0056] Specifically, under the nth spatial angle direction, the triaxial magnetic sensor chip 41 at each grid node (i,j) synchronously reads out the original voltage time-domain signal vectors in three orthogonal directions in space:
[0057] The multi-channel digital phase-locked amplifier module uses the excitation source frequency f and initial phase of the current control system 5 as reference signals to generate a sinusoidal reference signal RI(t)=cos(2πft) and a cosine reference signal RQ(t)=sin(2πft). It performs product integration demodulation on the input time-domain signal of each channel to obtain the in-phase component. and orthogonal components The formula for its calculation is:
[0058] In the formula, T is the integration period. Since environmental stray noise and mechanical vibration interference are usually not at the same frequency as the excitation frequency, their contributions tend to cancel each other out after the above product integration. That is, through coherent filtering, environmental stray noise and mechanical vibration interference that are not at the same frequency as the excitation frequency are drained, and only the effective signal components that are strictly in sync with the excitation frequency and have the same correlation are retained.
[0059] Furthermore, the demodulated in-phase and quadrature components are multiplied by the pre-calibrated sensor sensitivity coefficients for each channel. Converted to the current spherical excitation angle Next, node The absolute amplitude vector of the static spatial magnetic field at that location:
[0060] The absolute amplitude vector of the magnetic field This forms the basis for the original digital magnetic field data required for the subsequent step S3 to calculate the magnetic gradient tensor matrix.
[0061] S3. Steps for Calculating the Magnetic Gradient Tensor and Extracting Principal Eigenvalues After completing step S2, the signal processing system obtains the three-dimensional magnetic field response matrix corresponding to each node of the high-density triaxial magnetic sensor array under each spatial excitation direction n. The digital magnetic field matrix is used as the input data for calculating the magnetic gradient tensor. Combined with the spatial geometric spacing of the triaxial magnetic sensor chip 41, and through spatial finite difference, and with the physical prior constraints of Maxwell's equations regarding the non-divergent and non-rotational magnetic field in the passive, current-free region of air, a complete three-dimensional magnetic gradient tensor matrix is calculated. Furthermore, eigenvalue decomposition is performed on the magnetic gradient tensor matrix, and the principal eigenvalue with the largest absolute value among the three real eigenvalues is extracted to obtain the defect boundary response characteristics under the current excitation direction.
[0062] Specifically, by combining the row spacing Δx and column spacing Δy of the two-dimensional rectangular grid, the spatial first-order partial derivatives of the magnetic field quantity along the horizontal X-axis and Y-axis are obtained using the central finite difference formula:
[0063] Then, the magnetic field divergence in the air-free region is introduced to be zero. ·B=0 and curl is zero The physical prior hard constraint ×B=0 analytically extrapolates the vertical axis auto-derivative and cross partial derivative, which cannot be directly measured:
[0064] Because the high-density triaxial magnetic sensor array 4 in this embodiment is arranged on the surface of the component under test using a single-sided bonding method, the sensor can only directly acquire the magnetic field component information near the detection surface, and cannot directly obtain part of the magnetic field gradient component in the vertical direction through physical measurement. Therefore, this embodiment utilizes the non-dispersion and non-rotational physical constraint relationship of the magnetic field satisfied by the air region to perform analytical compensation for the gradient component that cannot be directly measured, thereby realizing the reconstruction of the complete three-dimensional magnetic gradient tensor without moving the sensor array or performing double-sided detection.
[0065] In this way, a complete symmetric 3×3 magnetic gradient tensor matrix can be reconstructed without loss in the near field on one side of the surface. :
[0066] As can be seen from the above formula, the matrix is symmetric about the main diagonal, and the sum of the diagonal elements is always zero, which is completely consistent with the physical prior constraints described in claim 7 of the aforementioned device embodiment.
[0067] In obtaining the matrix Then, eigenvalue decomposition is performed on it to obtain three real eigenvalues. Extract the principal eigenvalue with the largest absolute value. Its mathematical definition equation is as follows:
[0068] In the formula, This is a sign function used to preserve the original sign information of the principal eigenvalues. Because the matrix... The eigenvalues are mathematical invariants under coordinate system rotation transformations, utilizing... This full-space rotational invariance can completely cancel out the background magnetic field amplitude noise caused by irregular probe orientation deflection. This is achieved by traversing each node (i,j) within the measurement plane of the high-density triaxial magnetic sensor array. The spatial distribution of the defect on the entire measurement plane is obtained; by extracting the abrupt maximum points in the spatial distribution, the stable positioning of the defect geometric contour boundary under the current nth excitation angle can be achieved.
[0069] S4. Defect Depth Analysis Steps Based on the spatial vertical attenuation law of tensor field strength, the first-order tensor Frobenius norm of the three-dimensional magnetic gradient tensor matrix and its spatial first-order rate of change along the vertical direction are analytically calculated. By obtaining the dimensionless algebraic ratio of the two, the absolute physical depth of the defect from the surface of the current measuring point is solved inversely.
[0070] Specifically, first solve the matrix First-order tensor Frobenius norm :
[0071] Secondly, the norm is calculated using the following chain of partial derivative product equations. First-order spatial rate of change along the vertical Z-axis Since the high-density triaxial magnetic sensor array adopts a single-sided arrangement, it cannot directly obtain the vertical magnetic field gradient change in the defect direction. Therefore, this embodiment uses the continuous spatial variation relationship of the magnetic field to measure the tensor field strength norm. The rate of change along the vertical direction is solved indirectly. Specifically, the rate of change is... Treating it as a function of the spatial magnetic field state variable, according to the chain rule, its rate of change along the Z direction is expanded into a combination of the rates of change of each magnetic field component. Combined with the magnetic field gradient constraint relationship obtained in step S3, we obtain:
[0072] Finally, based on the algebraic ratio between the two, the absolute physical depth of the defect from the current sensor measurement plane is solved. :
[0073] In the formula, This is a shape correction coefficient used to compensate for the influence of different material conductivity, permeability, and defect geometry on the spatial attenuation of the magnetic field. This parameter is obtained through pre-calibration using a standard defect test block: under the same excitation frequency, excitation current, and sensor array arrangement as the actual test, a standard test block with a defect of known depth is tested, the F1 / F2 ratio at the corresponding location is calculated, and the corresponding value is deduced from the known defect depth. Value. After calibration using multiple defect samples of different depths, a shape correction coefficient corresponding to the current tested material is established.
[0074] S5. Steps for 3D Voxel Mesh Multi-Source Backprojection Fusion Tomography The defect geometric edge boundary data and absolute physical depth data obtained from all N spatial angles are combined and fused using multi-source spherical multi-angle inverse projection probabilistic overlap in a 3D voxel mesh constructed in computer virtual space. This process filters out artifact noise and activates solid voxels, reconstructing the spatial 3D point cloud topology of the defect. The defect geometric edge boundary data originates from the principal eigenvalue abrupt change regions extracted under different spatial excitation directions in step S3, and the absolute physical depth data originates from the depth matrix calculated in step S4. The signal processing system 8 spatially registers the aforementioned two-dimensional spatial boundary information with the depth information and uses this as input data for three-dimensional voxel fusion. Specifically, this includes the following three sub-steps: A. Establish a virtual grid In the computer virtual space, a defect three-dimensional voxel space matrix representing a discrete three-dimensional spatial mesh is initialized and constructed:
[0075] The three coordinate axes of the voxel space matrix correspond to the X direction, Y direction and Z direction perpendicular to the detection surface in the coordinate system of the surface of the component under test, respectively. The range of the X and Y directions is determined by the coverage area of the magnetic sensor array, and the range of the Z direction is determined according to the preset maximum detection depth.
[0076] B. Spherical multi-angle inverse projection accumulation Iterate through all n=1 to N spherical excitation directions, and transfer the calculated trajectories of the two-dimensional surface boundary nodes of the defect from each direction to the absolute physical depth corresponding to the voxel space matrix along the inverse projection matrix corresponding to the current excitation angle. The following probability density function mapping accumulation is performed on the slice layer:
[0077] In the formula, H is the step gating function activated when the principal feature value is greater than the preset edge threshold γ, and σ is the depth mapping standard deviation coefficient.
[0078] C. Entity Activation and Point Cloud Output Set the final entity activation threshold Γ if and only if the cumulative score of the voxel points satisfies When the voxel point is determined to be a defect entity, its three-dimensional coordinates in the spatial grid are extracted, artifact noise from a single illumination is removed, and the spatial three-dimensional point cloud topology of the defect is output with high fidelity.
[0079] Through the above-mentioned loop processing of S1 to S5, that is, after completing S2 to S4, it is determined whether N directions have been traversed. If not, return to S1 to switch to the next excitation angle. If the traversal has been completed, execute S5. Finally, the complete spatial three-dimensional shape of the defect is reconstructed. The three-dimensional point cloud topology includes at least the defect spatial coordinates, boundary contour and corresponding depth information.
[0080] Taking an aerospace aluminum alloy flat plate component as an example, a crack-like defect model is set inside the component, with its direction parallel to the surface of the component. The defect model is about 20 mm long, about 0.1 mm wide, and about 3 mm buried.
[0081] In this embodiment, the device parameters are set as follows: the row spacing Δx and column spacing Δy of the high-density triaxial magnetic sensor array are both 2 mm; the total number of discrete excitation directions N is 72; the excitation current operating frequency f is 5 kHz (within the preferred range of 1 kHz to 10 kHz); and the preset edge threshold γ, depth mapping standard deviation coefficient σ, entity activation threshold Γ, and shape correction coefficient are also specified. The parameters are determined based on pre-established standard defect calibration data.
[0082] The detection is performed according to steps S1 to S5 as described in the method embodiment. The expected implementation process of each step is as follows: S1 (Spherical mesh multi-point magnetic excitation): The current control system sequentially drives a three-axis orthogonal coil array according to 72 discrete spatial angles to synthesize magnetic vectors in the corresponding directions, and performs full-circumferential electromagnetic illumination on the aluminum alloy component to obtain excitation response data in multiple spatial directions.
[0083] S2 (Distortion Signal Acquisition and Demodulation): A high-density triaxial magnetic sensor array synchronously captures distorted magnetic field signals at various excitation angles. After demodulation and amplitude conversion by a multi-channel digital phase-locked amplifier module, the absolute amplitude vector of the digital magnetic field at each measurement point node at the corresponding excitation angle is obtained.
[0084] S3 (Magnetic Gradient Tensor Calculation and Principal Eigenvalue Extraction): By combining finite-difference spatial operations with Δx and Δy, and with prior physical constraints of non-dispersion and non-rotational magnetic fields, a complete symmetric three-dimensional magnetic gradient tensor matrix is reconstructed, and then the principal eigenvalues are extracted. Since the crack-like defect model in this embodiment is parallel to the component surface, the spatial abrupt change in the principal eigenvalues is expected to be significantly reduced at a few excitation angles where the synthesized magnetic vector direction is parallel to the crack direction; while at most other excitation angles, a significant spatial abrupt change in the principal eigenvalues can be observed at the corresponding crack position, thereby locking the surface boundary profile of the crack at the current excitation angle.
[0085] S4 (Defect In-Depth Analysis): For the magnetic gradient tensor matrix obtained under each excitation angle, calculate its Frobenius norm. and the first-order rate of change along the vertical direction Combined with pre-calibrated shape correction coefficients The depth information of the defect relative to the detection surface is estimated. Under the excitation angle with a high defect response, the solved depth value should be concentrated around the preset depth of 3 mm.
[0086] S5 (3D Voxel Mesh Multi-Source Inverse Projection Fusion Tomography): The boundary and depth data calculated from all 72 excitation angles are accumulated using a probability density function mapping in a 3D voxel mesh. By fusing multi-angle data, the influence of misjudgment caused by weak response or artifacts at individual excitation angles (especially those where the magnetic vector direction is nearly parallel to the crack direction) can be reduced. The final reconstructed 3D point cloud topology is expected to clearly reproduce the spatial morphology that is basically consistent with the pre-set crack in length, direction, and depth.
[0087] It should be noted that the detection process and expected results described in this embodiment are derived from the working principle of the technical solution of the present invention and the mathematical relationship of each step, and are used to illustrate the feasibility of the technical solution of the present invention. The specific detection effect may vary due to factors such as the material of the tested component, the type of defect, and the on-site detection environment.
[0088] In the triaxial orthogonal coil array of the present invention, each group of coils, in addition to the Helmholtz coils used in the aforementioned device embodiments, can also be replaced with a pair of saddle-shaped coils. Saddle-shaped coils have better surface adaptation capabilities than Helmholtz coils; therefore, when the tested component is a non-planar geometry such as a pipe or curved shell, saddle-shaped coils are preferred to improve the spatial adaptation capability between the coil array and the component surface. Conversely, when the tested component is a flat plate, Helmholtz coils are preferred to obtain higher magnetic field uniformity in the central cavity region.
[0089] External industrial actuators can choose to perform either physical color marking or laser stress-relieving beveling. For inspection scenarios where only defect location needs to be identified for subsequent manual review or archiving, physical color marking is preferred. For scenarios where defects are identified as cracks and require further engineering treatment to reduce stress concentration in the defect area, laser stress-relieving beveling is preferred. These two methods can also be combined within the same external industrial actuator according to on-site inspection needs. The signal processing system automatically selects the appropriate action type and issues control commands based on the defect type determination.
[0090] In summary, this invention, through a triaxial orthogonal coil array combined with a current control system, achieves full-circumferential multi-angle electromagnetic illumination without any mechanical movement, reducing the risk of blind zones caused by the mismatch between the defect direction and the excitation direction in traditional unidirectional excitation methods. By extracting the principal eigenvalues of the magnetic gradient tensor matrix for defect boundary determination, and utilizing the invariance of this eigenvalue under coordinate system rotation transformations, it helps reduce the interference of probe attitude drift and background magnetic field fluctuations on the detection results, thus improving the stability and consistency of the detection results. Through analytical calculation of the spatial attenuation law of the magnetic gradient tensor field strength, the inverse solution of the absolute physical depth of the defect is realized. Furthermore, the boundary data and depth data obtained from multiple excitation angles are fused using multi-source probability in a three-dimensional voxel grid, achieving a high-fidelity three-dimensional imaging effect similar to medical CT tomography, simultaneously acquiring the planar position, depth, and spatial three-dimensional morphology information of the defect. The above technical effects correspond and reinforce each other with the beneficial effects described in the "Summary of the Invention" section of this invention, verifying the feasibility and effectiveness of the technical solution of this invention.
[0091] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than that shown in the embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result.
[0092] In the description of the embodiments of this application, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments of this application. In the embodiments of this application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of this application, as well as the features of different embodiments or examples.
[0093] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of embodiments of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0094] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present application should be included within the scope of protection of the present application.
Claims
1. A defect three-dimensional tomographic imaging device based on triaxial orthogonal coil magnetic vector modulation, characterized in that, include: The triaxial orthogonal coil array consists of three groups of six electromagnetic coils arranged orthogonally and concentrically along the X-axis, Y-axis and Z-axis in a spatial rectangular coordinate system. The coils of each axis are concentrically arranged around the geometric center to form a static magnetic field cavity. A high-density triaxial magnetic sensor array is rigidly fixed on a non-magnetic substrate, and is attached to the surface of the component under test on one side and placed in the static magnetic field cavity. The high-density triaxial magnetic sensor array is composed of a dot matrix of multiple full-field triaxial magnetic sensor chips. Each of the triaxial magnetic sensor chips has rigidly integrated mutually orthogonal x, y, z triaxial sensing units inside, which are used to statically and synchronously capture the spatial three-dimensional distorted magnetic field signal caused by defects. The current control system is independently electrically connected to each group of coils in the triaxial orthogonal coil array; the current control system is used to independently adjust the amplitude ratio and polarity of the excitation current in the triaxial coils, so as to control the magnetic vector synthesized by each coil to perform full-circumferential spherical uniform grid multi-point spatial angle pointing switching in space, so as to perform all-round electromagnetic vision of the tested component without any mechanical movement. The signal processing system is communicatively connected to the high-density triaxial magnetic sensor array. The signal processing system is used to perform multi-channel digital phase-locked demodulation on the distorted magnetic field signal and convert it into a digital magnetic field matrix. Combined with the spatial geometric spacing of the triaxial magnetic sensor chip, a symmetric full-component three-dimensional magnetic gradient tensor matrix is calculated through spatial finite difference. The principal eigenvalue with the largest absolute value in the matrix is extracted to lock the defect boundary. The defect depth is analyzed according to the spatial attenuation law of the tensor field strength. Finally, the defect boundary data and depth data obtained by pointing all spatial angles downward are fused in a three-dimensional voxel grid to reconstruct the spatial three-dimensional topological morphology of the defect.
2. The defect three-dimensional tomographic imaging device based on triaxial orthogonal coil magnetic vector modulation according to claim 1, characterized in that, The three sets of coils in the triaxial orthogonal coil array are a pair of coils placed opposite each other along the X-axis, a pair of coils placed opposite each other along the Y-axis, and a pair of coils placed opposite each other along the Z-axis. The three sets of coils are arranged in a spatially complementary orthogonal configuration on the geometric winding, so that the spatial electromagnetic mutual inductance coefficient between the coils of each axis approaches zero.
3. The defect three-dimensional tomographic imaging device based on triaxial orthogonal coil magnetic vector modulation according to claim 1, characterized in that: The full-field triaxial magnetic sensor chip is a triaxial tunnel magnetoresistive sensor chip; the high-density triaxial magnetic sensor array adopts a conformal two-dimensional rectangular grid topology arrangement that is closely attached to the surface of the component, and the row spacing of the grid is... Column spacing The quantitative coverage range is 1mm≤ , ≤5mm.
4. The defect three-dimensional tomographic imaging device based on triaxial orthogonal coil magnetic vector modulation according to claim 1, characterized in that: The excitation current output by the current control system is an alternating current supplied with a simple harmonic quadrature phase-modulated current; the multi-point spatial angle pointing switch of the circumferential spherical uniform grid has a total number of discrete switching directions. N The quantitative range satisfies 32≤ N ≤128.
5. A three-dimensional tomographic imaging device for defects based on triaxial orthogonal coil magnetic vector modulation according to claim 1, characterized in that: The mathematical expression of the symmetric full-component three-dimensional magnetic gradient tensor matrix G calculated by the signal processing system using spatial finite difference is as follows: The diagonal elements of the matrix are based on the physical prior of non-dispersion magnetic fields in the air-source-free region, and rigidly satisfy the trace eigenvalue constraint: 。 6. A method for three-dimensional tomographic imaging of defects based on triaxial orthogonal coil magnetic vector modulation, characterized in that, The method of the three-dimensional tomographic imaging device for defects based on triaxial orthogonal coil magnetic vector modulation as described in any one of claims 1 to 5 includes the following steps: S1. Spherical grid multi-point magnetic excitation step: The amplitude ratio and polarity of the excitation current in the triaxial orthogonal coil array are independently adjusted by the current control system to control the synthesized magnetic vector to perform multi-point spatial angle pointing switching of the spherical uniform grid in N directions in space, and to perform multi-angle electromagnetic illumination on the interior of the tested component; S2. Distortion signal acquisition and demodulation step: The high-density triaxial magnetic sensor array is controlled to synchronously acquire the spatial three-dimensional distorted magnetic field signal caused by the defect, and the multi-channel digital lock-in amplifier module is used to perform orthogonal demodulation and amplitude conversion to obtain the digital magnetic field absolute amplitude vector of each measuring node under the current excitation angle; S3. S4. Magnetic Gradient Tensor Calculation and Principal Eigenvalue Extraction Steps: Combining the spatial geometric spacing of the triaxial magnetic sensor chip, and using the physical prior hard constraint of non-dispersion and non-rotational magnetic field through spatial finite difference distribution, the symmetric full-component three-dimensional magnetic gradient tensor matrix is inversely solved; the matrix is decomposed into eigenvalues, and the principal eigenvalue with the largest absolute value among the three real eigenvalues is extracted. The spatial distribution of the principal eigenvalue is used to lock the geometric edge boundary of the defect; S5. Defect Depth Analysis Steps: Based on the spatial vertical attenuation law of the tensor field strength, the first-order tensor Frobenius norm of the three-dimensional magnetic gradient tensor matrix and its spatial first-order rate of change along the vertical direction are analytically calculated. The absolute physical depth of the defect from the current measuring point surface is inversely solved by obtaining the dimensionless algebraic ratio of the two; S6. Three-dimensional Voxel Mesh Multi-Source Inverse Projection Fusion Tomography Steps: All The geometric edge boundary data of the defect obtained from the spatial angle downwards and the absolute physical depth data are combined with the three-dimensional voxel mesh to construct a multi-source spherical multi-angle inverse projection accumulation and fusion, filter out artifact noise and activate solid voxels, and reconstruct the spatial three-dimensional point cloud topology of the defect.
7. A method for three-dimensional tomographic imaging of defects based on triaxial orthogonal coil magnetic vector modulation according to claim 6, characterized in that, In step S1, the amplitude ratio and polarity of the excitation current in the independently adjustable triaxial coil are specifically modulated using the following time-domain digital modulation formula: when the synthesized magnetic vector needs to point to the nth spatial discrete angle in the spherical grid... At that time, the current control system inputs AC excitation currents of the same frequency and phase into the opposing coils of the X-axis, Y-axis, and Z-axis, respectively. The equations for the amplitudes of the currents on each axis are as follows: In the formula, The reference amplitude of the basic excitation current The operating frequency of alternating current. The angle between the current spatial synthesized magnetic vector and the Z-axis. Let be the angle between the projection of the spatially synthesized magnetic vector onto the XY plane and the X-axis, and let the discrete direction number be quantified as follows. N =72; Dynamic changes via electronic control and The step combination drives the synthesized magnetic vector to spontaneously complete the switching of the magnetic polarization direction of the isotropic sphere.
8. A method for three-dimensional tomographic imaging of defects based on triaxial orthogonal coil magnetic vector modulation according to claim 6, characterized in that, In step S3, the symmetric full-component three-dimensional magnetic gradient tensor matrix The specific solution process is as follows: First, combine the row spacing of the two-dimensional rectangular grid. Column spacing The spatial first-order partial derivatives of the magnetic field quantity along the horizontal X-axis and Y-axis are obtained using the central finite difference formula: Then, the magnetic field divergence in the air-free region is introduced to be zero. And the curl is zero Based on the physical prior hard constraints, the vertical axis autoderivative and cross partial derivatives, which cannot be directly measured, are analytically extrapolated: In this way, a complete symmetric 3×3 magnetic gradient tensor matrix can be reconstructed without loss in the near field on one side of the surface. : 。 9. A method for three-dimensional tomographic imaging of defects based on triaxial orthogonal coil magnetic vector modulation according to claim 8, characterized in that, In step S3, the mathematical definition of extracting the principal eigenvalue with the largest absolute value is as follows: for the matrix Performing eigenvalue decomposition yields three real eigenvalues. Extract the principal eigenvalue with the largest absolute value. : in For functions with positive and negative signs, use The mathematical invariant properties of the full-space rotation completely cancel out the background magnetic field amplitude noise caused by the irregular attitude deflection of the probe, and by traversing each node The spatial mutation maximum point rigidly locks the geometric contour boundary of the defect under the current excitation.
10. A method for three-dimensional tomographic imaging of defects based on triaxial orthogonal coil magnetic vector modulation according to claim 9, characterized in that, In step S5, the process of multi-source inverse projection fusion in the three-dimensional voxel mesh specifically includes the following steps: A. Establishing a virtual mesh: Initialize and construct a defective three-dimensional voxel space matrix V(x,y,z)=0 representing a discrete three-dimensional space mesh in the computer virtual space, where x∈[1,M],y∈[1,N],z∈[1,K]; B. Spherical multi-angle inverse projection accumulation: Traverse all n=1 to N For each spherical excitation direction, the calculated trajectories of the two-dimensional surface boundary nodes of the defect in each direction are projected along the inverse projection matrix corresponding to the current excitation angle to the absolute physical depth corresponding to the voxel space matrix. The following probability density function mapping accumulation is performed on the slice layer: In the formula, This is a step gate function activated when the principal eigenvalue is greater than a preset edge threshold γ. C. Entity Activation and Point Cloud Output: Set the final entity activation threshold Γ, if and only if the cumulative score of the voxel points satisfies the following condition. When the voxel point is determined to be a defect entity, its three-dimensional coordinates in the spatial grid are extracted, artifact noise from a single illumination is removed, and the spatial three-dimensional point cloud topology of the defect is output.
