Multi-frequency composite electromagnetic excitation based magnetic ring multi-dimensional nondestructive testing system and method

Abstract

The application relates to the technical field of magnetic ring detection, and discloses a multi-dimensional nondestructive detection system and method for a magnetic ring based on multi-frequency composite electromagnetic excitation, which solves the problems of information acquisition, composite excitation and comprehensive diagnosis in the existing magnetic ring detection. The method comprises the following steps: applying multi-frequency composite excitation, collecting and preprocessing multi-dimensional responses; extracting features, fusing and inverting through deep learning, and acquiring magnetic properties, material components and defect information. The system comprises the following modules: composite excitation, multi-modal sensing, data collection and preprocessing, data fusion and inversion, and a control module. Through the above technical scheme, the application realizes comprehensive and efficient nondestructive detection of multi-dimensional electromagnetic information of a magnetic ring, overcomes the existing limitations, and significantly improves the detection accuracy and comprehensive diagnosis capability.

Description

Technical Field

[0001] This invention relates to the field of magnetic ring testing technology, and in particular to a multi-dimensional non-destructive testing system and method for magnetic rings based on multi-frequency composite electromagnetic excitation. Background Technology

[0002] Magnetic rings, as indispensable key functional components in modern industrial production, are widely used in high-tech fields such as motors, sensors, encoders, medical equipment, and aerospace due to their unique magnetic and electrical properties. The uniformity of their internal magnetic properties, the consistency of their material composition, and the presence of potential internal defects directly determine the performance stability, accuracy, reliability, and lifespan of the final product. As application scenarios place increasingly stringent demands on performance indicators—for example, the requirements for high-temperature resistance and high-speed performance in new energy vehicle drive motors, and the requirements for angular resolution and long-term stability in precision encoders—the quality control standards for magnetic rings are constantly rising, posing higher-dimensional and deeper challenges to testing technologies.

[0003] In the existing magnetic ring detection technology system, various schemes have been developed for evaluating specific aspects. For example, a magnetic ring encoder and a method for detecting the absolute angle of the magnetic ring encoder disclosed in patent document CN117405150B achieve high-precision calibration of the magnetic ring rotation angle through a clever combination design of multiple sets of Hall elements and multiple pairs of pole magnets. Specifically, its design principle focuses on utilizing the sensitivity of Hall elements to the spatial distribution of magnetic fields. By accurately measuring the magnetic field signal generated by the preset magnetic pole structure on the surface of the magnetic ring, the precise rotation angle of the magnetic ring relative to the reference position can be calculated. This technology provides a practical and effective solution for industrial applications that require high-precision angle detection on large-diameter shaft parts, and has played an important role in improving the positioning and control accuracy of rotating machinery for a certain period of time. However, the core focus of this technical solution is on the macroscopic magnetic field distribution of the magnetic ring and its application in angle measurement. Its detection method is essentially limited to the acquisition and analysis of magnetic field signals on or near the surface of the magnetic ring. This Hall effect-based magnetic field induction mode primarily responds to the static or low-frequency magnetic properties of the magnetic ring. It lacks the necessary detection capabilities for the deep material composition, microstructural defects, or electromagnetic response characteristics under high-frequency operating conditions within the magnetic ring. Therefore, it cannot achieve multi-dimensional, in-depth, non-destructive testing of the uniformity of magnetic properties, material composition consistency, and potential internal defects within the magnetic ring.

[0004] Based on this, a search also revealed a patent document with publication number CN113588996B, which discloses a magnetic ring clamp and testing device. This patent mainly aims to solve the signal interference problem that may be caused by the curling of external wires in the traditional testing process by optimizing the mechanical structure and automated process of the testing device, and significantly improves the automated testing efficiency of the overall parameters of the magnetic ring. Its design focuses on providing a stable and repeatable magnetic ring positioning and contact mechanism to ensure that the conventional electrical parameters of the magnetic ring can be obtained quickly and accurately in batch testing. Through mechanical automation and standardized measurement processes, this technology has made significant progress in improving testing efficiency and reducing human intervention errors, and is particularly suitable for scenarios of rapid screening and quality control of the basic electrical parameters of magnetic rings. However, the testing content of this technical solution still mainly focuses on the overall electrical parameters of the magnetic ring, such as resistance, inductance, or average magnetic flux, and fails to delve into the more complex electromagnetic physical characteristics inside the magnetic ring. In terms of its working principle, this solution mostly relies on traditional single-frequency or DC excitation methods for parameter measurement. The inherent limitation of single-frequency excitation lies in its inability to fully reveal the electromagnetic response characteristics of materials over a wide frequency range. This makes it often inadequate for identifying subtle internal defects such as microcracks, voids, and non-magnetic inclusions, or for determining whether there are local deviations in material composition. These defects and deviations typically exhibit unique spectral characteristics under the influence of electromagnetic fields at different frequencies, and detection using a single frequency will lose this crucial information.

[0005] However, as modern industry places increasingly stringent demands on the performance of magnetic rings, and as product integration and the complexity of working environments continue to rise, the inherent limitations of the aforementioned single-dimensional or traditionally based testing technologies are becoming increasingly apparent when facing new challenges. The actual performance of a magnetic ring is influenced by a combination of multiple internal factors, including local non-uniformity of magnetic permeability, spatial fluctuations in electrical conductivity, even subtle differences in microscopic crystal structure and elemental composition, as well as various potential geometric or material defects. Existing testing methods, whether focusing on macroscopic magnetic field measurements based on angle calibration or on overall electrical parameter evaluation based on single-frequency excitation, can only provide a partial view of the magnetic ring's performance. The fundamental contradiction lies in the fact that, in order to improve testing efficiency or achieve high accuracy in a specific dimension, traditional methods often sacrifice the ability to comprehensively acquire multi-physics information within the magnetic ring. For example, while high-precision angle detection systems (such as CN117405150B) can accurately sense magnetic pole distribution, the static or low-frequency magnetic fields they employ cannot effectively excite or detect high-frequency electromagnetic effects such as eddy currents and dielectric losses within the magnetic ring, thus missing crucial information closely related to material conductivity, high-frequency magnetic loss, and deep defects. Traditional single-frequency excitation detection (such as CN113588996B), although capable of automatically measuring overall parameters, essentially reflects the complex electromagnetic properties within the magnetic ring at a specific single frequency. It cannot effectively distinguish and accurately locate defects of different depths and properties, nor can it reveal the dynamic electromagnetic behavior of the material at different frequencies. This inherent limitation in principle prevents existing technologies from comprehensively, efficiently, and non-destructively acquiring the integrated electromagnetic physics information within the magnetic ring in a single system and process when faced with comprehensive testing requirements that simultaneously assess magnetic property uniformity, material composition consistency, and potential defects. In short, existing technologies suffer from irreconcilable contradictions regarding breadth, depth, accuracy, and efficiency. This fragmented information acquisition and limited diagnostic capabilities not only lead to insufficient or inefficient comprehensive judgment of product quality issues, but may also cause some key but hidden defects to go undetected in a timely manner, thus posing a potential threat to the reliability and safety of downstream products.

[0006] Therefore, how to construct a non-destructive testing system and method based on the principle of multi-frequency composite electromagnetic excitation, capable of collaboratively acquiring, deeply integrating, and performing integrated inversion analysis of multi-dimensional electromagnetic physical information inside the magnetic ring, thereby overcoming the significant shortcomings of existing technologies in single-dimensional detection, low-frequency / single-frequency excitation limitations, and comprehensive diagnostic capabilities, has become a key technical challenge that urgently needs to be solved in the field of magnetic ring testing technology. Summary of the Invention

[0007] To address the significant shortcomings of existing magnetic ring inspection systems in multi-dimensional information acquisition, composite electromagnetic excitation design, and comprehensive diagnostic capabilities, this invention provides a multi-dimensional non-destructive testing system and method for magnetic rings based on multi-frequency composite electromagnetic excitation. This invention aims to achieve simultaneous and efficient detection of the uniformity of internal magnetic properties, material composition consistency, and potential defects in magnetic rings through an innovative composite excitation module, collaborative sensor array, and multi-modal data fusion and inversion algorithm, thereby meeting the demands of modern industry for high-quality magnetic ring inspection.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation, comprising:

[0009] The step of applying composite electromagnetic excitation is used to apply multi-frequency composite electromagnetic excitation to the magnetic ring to be tested through the composite electromagnetic excitation module. The multi-frequency composite electromagnetic excitation includes low-frequency magnetic field excitation, medium- and high-frequency eddy current excitation and broadband dielectric excitation.

[0010] The step of capturing response signals is used to capture in real time the multidimensional response signals of the magnetic ring under test under the multi-frequency composite electromagnetic excitation through a multimodal collaborative sensor array module. The multidimensional response signals include magnetic field changes, eddy current signals, impedance spectra, and dielectric responses.

[0011] The data preprocessing step is used to perform high-precision acquisition, time synchronization, filtering and noise reduction, and data normalization on the multidimensional response signal through a high-precision data acquisition and preprocessing module.

[0012] The data fusion and inversion step is used to extract the multimodal features of the magnetic ring based on the preprocessed multidimensional response signal through the multimodal data fusion and inversion module, and to fuse and invert the multimodal features through a deep learning network to output the magnetic property uniformity, material composition consistency and potential defect information of the magnetic ring.

[0013] The system control steps are used to coordinate the collaborative operation of the composite electromagnetic excitation module, the multimodal collaborative sensor array module, the high-precision data acquisition and preprocessing module, and the multimodal data fusion and inversion module through the system control and human-machine interaction module, and to provide a human-machine interaction interface.

[0014] To further realize the present invention, the following technical solutions may be preferred:

[0015] Preferably, the step of applying the composite electromagnetic excitation includes:

[0016] A low-frequency magnetic field excitation is applied to the magnetic ring to be tested by a low-frequency magnetic field excitation unit;

[0017] A medium-to-high frequency eddy current excitation is applied to the magnetic ring under test by a medium-to-high frequency eddy current excitation unit;

[0018] A broadband dielectric excitation is applied to the magnetic ring to be tested by a broadband dielectric excitation unit;

[0019] The low-frequency magnetic field excitation unit consists of a pair of Helmholtz coils; the medium- and high-frequency eddy current excitation unit consists of a flat spiral coil; and the broadband dielectric excitation unit consists of a pair of parallel plate capacitors.

[0020] Preferably, the step of capturing the response signal includes:

[0021] The uniformity signal of magnetic field distribution under low-frequency magnetic field excitation is captured by a three-axis fluxgate sensor array;

[0022] The high-sensitivity Hall sensor array captures the secondary magnetic field change signal induced by medium- and high-frequency eddy currents.

[0023] The complex impedance spectrum signal under broadband dielectric excitation was captured by a broadband dielectric response sensor array.

[0024] The triaxial fluxgate sensor array comprises multiple sensors evenly distributed along the circumference of the magnetic ring, each capable of measuring the magnetic field strength in three orthogonal directions; the high-sensitivity Hall sensor array comprises multiple sensors evenly distributed along the radial and axial directions of the magnetic ring, each Hall sensor used to measure local magnetic field strength changes; and the broadband dielectric response sensor array consists of multiple high-frequency probes.

[0025] Preferably, the data preprocessing step includes:

[0026] The multimodal collaborative sensor array module captures data from each sensor synchronously using a high-resolution analog-to-digital converter.

[0027] Hardware-level timestamp synchronization of data from various sensors is achieved through a general-purpose timer chip and a high-precision clock source.

[0028] The acquired signal is noise-suppressed using a digital low-pass filter, a band-pass filter, and an adaptive Wiener filter.

[0029] Sensor data with different dimensions and amplitudes are subjected to minimum-maximum normalization or Z-score normalization.

[0030] Preferably, the data fusion and inversion step includes:

[0031] Extract the magnetic field intensity gradient, harmonic content, and hysteresis loop parameters from the magnetic field change signal;

[0032] The real part, imaginary part, phase angle, and amplitude attenuation curve of the eddy current impedance are extracted from the eddy current signal using wavelet transform and Fourier transform.

[0033] The complex dielectric constant, dielectric loss tangent, and relaxation time distribution are extracted from the dielectric response signal.

[0034] Preferably, the data fusion and inversion step further includes:

[0035] The extracted magnetic field feature vector, eddy current feature matrix, and dielectric spectrum feature vector are input into the deep learning network for fusion. The deep learning network is a deep convolutional recurrent network based on an attention mechanism.

[0036] The deep convolutional recurrent network structure includes multiple convolutional layers, each followed by batch normalization and ReLU activation functions, then multiple bidirectional long short-term memory network layers, and finally a multi-head attention layer.

[0037] Preferably, the data fusion and inversion step further includes:

[0038] The output layer of the deep convolutional recurrent network is connected to a multilayer perceptron to invert the two-dimensional distribution of magnetic domains, residual magnetic flux density, and coercivity inside the magnetic ring.

[0039] Invert the three-dimensional distribution of electrical conductivity, magnetic permeability and dielectric constant of the material;

[0040] Identify the location, size, and shape of defects such as cracks, pores, and inclusions.

[0041] Preferably, the method further includes:

[0042] The results display and diagnostic report generation module displays the distribution of magnetic properties, material parameters, and defect locations inside the magnetic ring in a 3D visualization interface, and generates a customized report that includes detection parameters, raw data curves, inversion results, defect details, and quality assessment level.

[0043] A multi-dimensional non-destructive testing system for magnetic rings based on multi-frequency composite electromagnetic excitation, applicable to the above-mentioned methods, includes:

[0044] A composite electromagnetic excitation module is used to apply multi-frequency composite electromagnetic excitation to the magnetic ring to be tested. The multi-frequency composite electromagnetic excitation includes low-frequency magnetic field excitation, medium- and high-frequency eddy current excitation and broadband dielectric excitation.

[0045] A multimodal collaborative sensor array module is used to capture in real time the multidimensional response signal of the magnetic ring under test under the multi-frequency composite electromagnetic excitation. The multidimensional response signal includes magnetic field change, eddy current signal, impedance spectrum and dielectric response.

[0046] A high-precision data acquisition and preprocessing module is used to perform high-precision acquisition, time synchronization, filtering and noise reduction, and data normalization on the multidimensional response signal.

[0047] The multimodal data fusion and inversion module is used to extract multimodal features of the magnetic ring based on the preprocessed multidimensional response signal, and to fuse and invert the multimodal features through a deep learning network to output information on the magnetic property uniformity, material composition consistency and potential defects of the magnetic ring.

[0048] The system control and human-machine interaction module is used to coordinate and control the collaborative work of the composite electromagnetic excitation module, the multimodal collaborative sensor array module, the high-precision data acquisition and preprocessing module, and the multimodal data fusion and inversion module, and to provide a human-machine interaction interface.

[0049] Preferably, the composite electromagnetic excitation module includes:

[0050] A low-frequency magnetic field excitation unit is used to apply a low-frequency magnetic field excitation to the magnetic ring to be tested. The low-frequency magnetic field excitation unit is composed of a pair of Helmholtz coils.

[0051] A medium-high frequency eddy current excitation unit is used to apply medium-high frequency eddy current excitation to the magnetic ring to be tested. The medium-high frequency eddy current excitation unit is composed of a flat spiral coil.

[0052] A broadband dielectric excitation unit is used to apply broadband dielectric excitation to the magnetic ring to be tested. The broadband dielectric excitation unit consists of a pair of parallel plate capacitors.

[0053] The beneficial effects of this invention are:

[0054] This invention provides a multi-dimensional non-destructive testing system and method for magnetic rings based on multi-frequency composite electromagnetic excitation. By combining low-frequency magnetic field excitation, mid-to-high-frequency eddy current excitation, and broadband dielectric excitation, it can fully excite electromagnetic responses of different scales and physical properties within the magnetic ring, overcoming the limitations of existing single-frequency or single-mode excitation methods. The multi-modal collaborative sensor array can simultaneously capture various heterogeneous data such as magnetic field changes, eddy current signals, impedance spectra, and dielectric responses, providing rich raw data support for multi-dimensional information within the magnetic ring. The high-precision data acquisition and preprocessing module ensures data accuracy and temporal synchronization, laying the foundation for subsequent processing. The multi-modal data fusion and inversion module, particularly employing a deep convolutional recurrent network based on an attention mechanism, achieves deep fusion and intelligent analysis of multi-source heterogeneous data, accurately inverting the magnetic property uniformity and material composition consistency of the magnetic ring, and precisely locating internal defects. This system-level solution overcomes the problem that existing technologies can only assess magnetic property uniformity, material composition, or internal defects individually, achieving comprehensive, efficient, and non-destructive acquisition of electromagnetic physical information within the magnetic ring. This invention significantly improves testing efficiency and provides more accurate and comprehensive quality judgment criteria through an integrated system and process, thereby reducing production costs and improving the reliability of the final product. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the overall technical architecture of the detection system of the present invention;

[0056] Figure 2 This is a schematic diagram of the framework of the composite electromagnetic excitation module of the present invention;

[0057] Figure 3 This is a schematic diagram of the framework of the multimodal collaborative sensor array module of the present invention;

[0058] Figure 4 This is a schematic diagram of the deep learning network structure of the present invention;

[0059] Figure 5 This is a schematic diagram of the framework of the result display and diagnostic report generation module of the present invention;

[0060] Figure 6 This is a logic flowchart of the detection method of the present invention;

[0061] Figure 7 This is a schematic diagram of the multi-level interaction relationships and data flow between the various functional modules of the present invention. Detailed Implementation

[0062] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0063] The technical solutions of the embodiments 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, 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.

[0064] Example 1

[0065] The multi-dimensional non-destructive testing system for magnetic rings based on multi-frequency composite electromagnetic excitation proposed in this embodiment refers to... Figures 1 to 5 Its structure includes: a composite electromagnetic excitation module, a multimodal collaborative sensor array module, a high-precision data acquisition and preprocessing module, a multimodal data fusion and inversion module, as well as a control and analysis unit and a result output and display module.

[0066] The composite electromagnetic excitation module is responsible for generating and applying a multi-frequency composite electromagnetic field to the magnetic ring under test. This module internally includes a multi-frequency signal source, a power amplifier, an impedance matching unit, and an excitation coil array. The multi-frequency signal source can generate excitation signals with a wide frequency range and controllable waveforms, including but not limited to sine waves, square waves, and swept-frequency signals. These signals can range in frequency from Hertz to megahertz to enable the detection of different depths and electromagnetic response characteristics of the magnetic ring. The power amplifier amplifies the weak excitation signal output from the signal source, providing sufficient drive current to generate the required alternating magnetic field strength in the excitation coil array. The impedance matching unit ensures maximum power transfer between the power amplifier and the excitation coil array, reducing signal reflection loss and improving excitation efficiency. The excitation coil array consists of multiple precision-wound coils, which can be arranged according to the geometry of the magnetic ring under test and the detection requirements, such as a ring-shaped, planar array, or local excitation configuration. By precisely controlling the excitation phase and amplitude of different coils, a composite electromagnetic field with specific spatial distribution and temporal variation characteristics can be formed, thereby exciting various electromagnetic responses within the magnetic ring.

[0067] The multimodal collaborative sensor array module is used to sense and acquire various physical response signals of the magnetic ring under test in real time under combined electromagnetic excitation. This module consists of an induction coil, a Hall sensor, a magnetoresistive sensor, and a temperature sensor. The induction coil captures the induced electromotive force generated by changes in the magnetic field inside and around the magnetic ring through the principle of electromagnetic induction. Its output signal reflects the changes in the conductivity, permeability, and eddy current distribution caused by defects in the magnetic ring. The Hall sensor and magnetoresistive sensor are used to measure the static magnetic field distribution on the surface of the magnetic ring with high precision and the dynamic magnetic field changes under excitation. These data can reveal the uniformity of magnetic characteristic parameters such as remanence and coercivity of the magnetic ring, as well as local demagnetization or magnetization anomalies. The temperature sensor is used to monitor the temperature rise of the magnetic ring during the excitation process, paying particular attention to local hot spots, because abnormal temperature rises are often related to concentrated energy loss caused by internal defects or material inhomogeneities. These sensors are arranged in an array to achieve comprehensive coverage of the magnetic ring area and high spatial resolution acquisition. The arrangement strategy takes into account the geometric characteristics of the magnetic ring and the expected defect types.

[0068] The high-precision data acquisition and preprocessing module is responsible for converting the analog signals acquired by the multimodal collaborative sensor array module into digital signals and performing preliminary quality improvement processing. This module includes a high-precision analog-to-digital converter (ADC), an anti-aliasing filter, a signal amplifier, and a clock synchronization unit. The high-precision ADC digitizes the sensor signals with a high sampling rate and high bit depth, ensuring the original accuracy of the data. The anti-aliasing filter removes high-frequency noise exceeding half the sampling frequency before ADC conversion, preventing aliasing. The signal amplifier appropriately amplifies the signal according to the output characteristics of different sensors, ensuring it falls within the optimal input range of the ADC. The clock synchronization unit ensures accurate timestamps on data from different sensors using a unified clock reference, providing a time alignment basis for subsequent multimodal data fusion. Furthermore, this module performs preliminary digital filtering to suppress power frequency interference, random noise, and other contaminants.

[0069] The multimodal data fusion and inversion module is the core intelligent processing unit of this system, responsible for in-depth analysis and comprehensive evaluation of preprocessed multimodal data. This module employs a high-performance processor and large-capacity memory, and runs a specially developed multimodal data fusion network and electromagnetic inversion algorithm. The multimodal data fusion network can extract features, align, and deeply fuse heterogeneous data from different sensor modes, establishing complex correlations between various modal information. The electromagnetic inversion algorithm, based on the fused features and combined with the physical model of the magnetic ring, uses iterative calculations or machine learning models to invert the three-dimensional distribution of electromagnetic physical parameters such as conductivity, permeability, and geometric defects within the magnetic ring. This module aims to extract hidden information from multi-source data that is difficult to obtain using traditional single-detection methods, improving the accuracy of defect identification and the quantitative analysis capability of material properties.

[0070] The Control and Analysis Unit (CMU), as the central control hub of the system, is responsible for coordinating and managing the collaborative work of all the aforementioned modules. This unit includes a main control processor, system memory, and various interfaces for configuring excitation parameters, controlling the data acquisition process, initiating data processing and inversion tasks, and performing preliminary analysis of the detection results. It is also responsible for diagnosing the system's own operating status and processing user input.

[0071] The results output and display module presents the final test results, obtained from the multimodal data fusion and inversion module, to the user in an intuitive and easy-to-understand format. This module includes a high-resolution display, a report generator, and a data storage interface. The display can show the distribution of electromagnetic properties, defect locations, sizes, and types, as well as material homogeneity assessment results, within the magnetic ring in various formats, such as 3D visualization, pseudo-color images, graphs, and tables. The report generator can automatically generate detailed test reports based on preset templates, including all key data, analysis conclusions, and diagnostic recommendations. The data storage interface is used to permanently store raw data, intermediate processing results, and the final report, facilitating traceability and further analysis.

[0072] Example 2

[0073] The multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation proposed in this embodiment is a detailed description of the specific operation process and data processing logic of the above system. (Refer to...) Figure 6 and Figure 7 The steps include: clamping and initializing the magnetic ring to be tested, generating and applying composite electromagnetic excitation signals, acquiring multimodal collaborative response signals, high-precision preprocessing of acquired data, multimodal feature extraction and fusion, inversion of the internal electromagnetic properties of the magnetic ring and defect identification, and analysis and output of multidimensional detection results.

[0074] S1, clamping and initialization of the magnetic ring under test.

[0075] This step is the initial stage of the testing process, ensuring the stability and controllability of the magnetic ring under test during the testing process, and providing standardized starting conditions for subsequent electromagnetic excitation and signal acquisition. Specific operations include: precisely placing the magnetic ring under test in the system's preset fixture, ensuring that the geometric center of the magnetic ring is highly aligned with the central axis of the excitation coil array and sensor array. This process utilizes a high-precision mechanical positioning device, with a repeatability accuracy down to the micrometer level. System initialization includes functional self-tests and parameter configurations for the composite electromagnetic excitation module, the multimodal collaborative sensor array module, and the high-precision data acquisition and preprocessing module. The functional self-test verifies the connectivity and operating status of each sensor, such as detecting zero-point drift in the Hall sensor and whether the induction coil has an open or short circuit. Parameter configuration, based on the material type, geometric dimensions of the magnetic ring under test, and the expected target characteristics to be detected, sets the frequency range, amplitude, and phase of the excitation signal, as well as the sampling frequency and gain of the sensors. For example, ferrite magnetic rings, with their high permeability, may require lower frequency excitation to achieve deeper penetration; while sintered NdFeB magnetic rings, with their high coercivity, may require specific pulse excitation modes. Meanwhile, environmental calibration is also a crucial component of the initialization process. By acquiring background noise and environmental field data under conditions without the test ring or a standard ring, a baseline is established for background cancellation in subsequent data processing. This eliminates the influence of external factors such as the geomagnetic field and power grid interference on the measurement results, improving the signal-to-noise ratio and accuracy of the detection.

[0076] S2, Generation and application of composite electromagnetic excitation signal.

[0077] This step is the core initiation stage of multi-dimensional magnetic ring detection. By applying a carefully designed electromagnetic excitation to the magnetic ring under test, a rich electromagnetic response is induced within it. This step is further broken down into sub-steps:

[0078] S201, multi-frequency excitation signal generation.

[0079] The system's internal multi-frequency signal source generates a series of electromagnetic excitation signals with specific frequencies and waveforms based on preset parameters. These signals are not single frequencies, but rather optimized composite frequency spectra, such as simultaneously containing low-frequency sine waves, mid-frequency swept signals, and high-frequency pulse signals. Low-frequency signals, such as those from 50 Hz to 5 kHz, have strong penetrating power and are mainly used to detect deeper macroscopic structural defects and uniformity within the magnetic ring, being sensitive to changes in the overall permeability and conductivity of the magnetic ring. Mid-frequency signals, such as those from 5 kHz to 50 kHz, provide high-resolution detection of surface and subsurface defects in the magnetic ring, being sensitive to the material's dielectric constant and magnetic loss characteristics. High-frequency signals, such as those from 50 kHz to 5 MHz, are mainly focused on detecting surface or near-surface micro-defects and material properties sensitive to the skin effect. In terms of waveform selection, sine waves are used for steady-state response analysis, swept signals are used to obtain broadband response characteristics, and pulse signals are used for transient response analysis to capture the nonlinear characteristics of the magnetic ring material. The amplitude and phase accuracy of the generated signal are guaranteed by a high-precision digital-to-analog converter and a phase-locked loop circuit, ensuring the stability and repeatability of the excitation field. The signal amplitude range is, for example, from 0.1 volt to 10 volts, and the phase accuracy is, for example, less than 0.1 degree.

[0080] S202, excitation signal applied.

[0081] The generated multi-frequency composite electromagnetic excitation signal is amplified by a power amplifier and then drives an excitation coil array through an impedance matching unit. The excitation coil array converts the electrical signal into a spatially distributed alternating magnetic field according to a preset excitation mode, such as alternating magnetic field excitation or rotating magnetic field excitation. In alternating magnetic field excitation mode, the excitation coils can be arranged radially, axially, or circumferentially along the magnetic ring, generating the main magnetic field components in these directions respectively, thereby probing the anisotropic characteristics of the magnetic ring in different directions. For example, one set of coils can be radially excited, and another set can be circumferentially excited; they can work alternately or synchronously. When working alternately, the responses in different directions can be analyzed independently; when working synchronously, the coupled responses under multi-directional excitation can be studied. The magnetic field strength is, for example, between 0.01 Tesla and 0.1 Tesla, and the magnetic field gradient can reach 0.005 Tesla per millimeter. The uniformity and stability of the excitation field are jointly determined by the coil geometry, winding process, and the stability of the driving current. Through this multi-frequency composite excitation, a variety of electromagnetic responses are generated inside the magnetic ring, including eddy currents, magnetic domain wall motion, and dielectric polarization. These responses carry rich information about the magnetic ring material composition, microstructure, and defects.

[0082] S3, multimodal collaborative response signal acquisition.

[0083] While applying composite electromagnetic excitation, the multimodal collaborative sensor array module acquires various physical responses of the magnetic ring in real time. This step is further refined into sub-steps:

[0084] S301, Induced voltage signal acquisition.

[0085] An array of induction coils uses the principle of electromagnetic induction to capture the induced electromotive force generated by changes in the eddy current field inside a magnetic ring. When an alternating magnetic field passes through the magnetic ring, eddy currents are induced within it. The magnitude, distribution, and phase of these eddy currents are influenced by the conductivity, permeability, and internal defects of the magnetic ring material. The induction coils, based on their spatial relative position to the magnetic ring under test and the number of turns, convert these changes in the eddy current field into a measurable voltage signal. For example, multiple induction coils can be arranged radially, axially, and circumferentially along the magnetic ring to form a two-dimensional or three-dimensional detection grid. The output voltage signal amplitude of each induction coil is typically in the microvolt to millivolt range and requires pre-amplification by a high-precision amplifier to improve the signal-to-noise ratio. The frequency components of the induced voltage signal correspond to the fundamental frequency and harmonics of the excitation signal. By analyzing the amplitude and phase of these frequency components, the response characteristics of the magnetic ring to electromagnetic fields of different frequencies can be obtained, thereby inferring its internal structure.

[0086] S302, magnetic field strength signal acquisition.

[0087] Hall effect sensors and magnetoresistive sensor arrays are used to accurately measure changes in the magnetic field strength of a magnetic ring under combined excitation. Hall effect sensors measure the magnitude and direction of the magnetic field strength, while magnetoresistive sensors detect the magnetic field by utilizing the property that their resistance changes with the magnetic field. These sensors can be arranged on or near the surface of the magnetic ring, forming a high-density array to obtain a detailed image of the magnetic field distribution on the ring surface. For example, eight Hall effect sensors are arranged at equal intervals around the circumference of the magnetic ring, while four rows of magnetoresistive sensors are arranged axially, thus forming a two-dimensional magnetic field detection plane with high spatial resolution. The voltage signals output by these sensors are proportional to the local magnetic field strength, and their amplitude varies from millivolts to volts depending on the sensor type and the magnetic field strength. By analyzing these magnetic field signals, it is possible to reveal whether there is local demagnetization, magnetization inhomogeneity, or magnetic field distortion due to defects within the magnetic ring. For example, cracks can cause local leakage of magnetic flux, thus creating an abnormal magnetic field gradient on the sensor array.

[0088] S303, temperature signal acquisition.

[0089] A temperature sensor array is used to monitor the surface temperature and local temperature rise of the magnetic ring in real time during the excitation process. When defects such as cracks or inclusions exist inside the magnetic ring, the eddy current density in the defect area may be abnormally concentrated under the excitation of an alternating electromagnetic field, or the hysteresis loss of the material may increase, leading to enhanced Joule heating and hysteresis heating effects in that area, thus generating a local temperature rise. For example, an infrared thermal imager or a contact thermocouple array can be used to acquire a real-time temperature distribution map of the magnetic ring surface at a frame rate of thirty frames per second or higher. The measurement accuracy of the temperature sensor can reach 0.1 degrees Celsius. By analyzing the temperature rise rate, the location of local hot spots, and the temperature gradient, the presence, location, and size of defects can be determined. For example, a crack with a size of 0.5 millimeters may cause the temperature around it to rise by 0.5 to 2 degrees Celsius, and the temperature rise rate is significantly higher than that of the defect-free area.

[0090] S4, high-precision preprocessing of acquired data.

[0091] The high-precision data acquisition and preprocessing module processes the raw signals acquired by the multimodal collaborative sensor array module to improve data quality and usability. This step is further broken down into sub-steps:

[0092] S401, signal filtering and noise reduction.

[0093] The raw acquired signal inevitably contains various noises, such as power frequency interference, random noise, and inherent noise from the sensor itself. This step applies digital filters to process the signal. For power frequency interference, notch filters or adaptive filtering algorithms such as the least mean square algorithm are used for suppression. For random noise, wavelet transform denoising or Kalman filtering algorithms are used, and noise and useful signals are effectively separated through multi-scale analysis or state estimation methods. For example, the wavelet basis function is set to the Daubechies series wavelet, the decomposition level is five, and the detail coefficients are thresholded to filter out high-frequency noise. The filtered signal has a higher signal-to-noise ratio, which helps improve the accuracy of subsequent feature extraction.

[0094] S402, Data Synchronization and Alignment.

[0095] Data from different sensor types and acquisition channels may exhibit temporal inconsistencies due to minute deviations in sampling clocks or data transmission delays. This step addresses this issue through timestamp alignment and interpolation algorithms. Each sensor acquires data with a precise timestamp, which is then corrected using a unified time reference. For channels with inconsistent sampling frequencies, linear interpolation, cubic spline interpolation, or Lagrange interpolation algorithms are used to unify all data onto a pre-defined common time axis, for example, resampling all data to a frequency of 1 kilohertz. The data alignment accuracy reaches, for example, the microsecond level, ensuring strict synchronization of multimodal data in the time dimension, laying the foundation for subsequent feature-level fusion.

[0096] S403, Baseline Calibration and Normalization.

[0097] Baseline calibration aims to eliminate the impact of sensor drift, environmental changes, and differences in initial system conditions on measurement results. Background effects are subtracted from the response data of the magnetic ring under test by using baseline data acquired under unexcited conditions or standard magnetic ring conditions. For example, the corresponding average baseline value is subtracted from the real-time measurement value of each channel. Normalization unifies data from different modes and dimensions to the same numerical scale or range to eliminate the impact of dimensional differences on subsequent data fusion and inversion algorithms. For example, all data is normalized to the range of zero to one, or Z-score standardization is used, i.e., the data is subtracted from the mean and divided by the standard deviation. The normalization formula is expressed as:

[0098] X_norm=(X-X_min) / (X_max-X_min)

[0099] Where X represents the original data, X_min represents the minimum value of the data, X_max represents the maximum value of the data, and X_norm represents the normalized data.

[0100] This process ensures that features from different modalities have the same weight when input into the fusion network, preventing certain modalities from dominating the fusion result due to their larger values.

[0101] S5, multimodal feature extraction and fusion.

[0102] The clean, synchronized, and normalized multimodal data output from the high-precision data acquisition and preprocessing module enters the multimodal data fusion and inversion module for in-depth feature mining and information integration. This step is further broken down into sub-steps:

[0103] S501, Time-domain and Frequency-domain Feature Extraction.

[0104] For induced voltage and magnetic field strength signals, statistical features such as peak value, root mean square value, variance, skewness, kurtosis, and zero-crossing rate are extracted from their time-domain waveforms. These time-domain features reflect the transient behavior and statistical distribution characteristics of the signal. Simultaneously, Fourier transform is performed on the signal to extract frequency-domain features such as fundamental frequency amplitude, harmonic component amplitude and phase, spectral entropy, and power spectral density. Frequency-domain features reveal the response intensity and energy distribution of the magnetic ring to excitations at different frequencies, which is significant for identifying frequency-dependent losses and spectral distortions caused by defects in the material. For temperature signals, in addition to extracting time-domain features such as peak temperature, average temperature, and temperature rise rate, their spatial gradient and time series autocorrelation characteristics can also be analyzed.

[0105] S502, Spatial Feature Extraction.

[0106] By utilizing the spatial arrangement information of the sensor array, the spatial distribution characteristics of each modal data can be extracted. For example, by interpolating and meshing the magnetic field strength measured by the Hall sensor array, a two-dimensional magnetic field distribution map of the magnetic ring surface can be generated. Based on this, the spatial distribution characteristics of the magnetic field gradient can be extracted, or the boundaries of magnetic field anomaly regions can be identified using edge detection algorithms. For induction coil arrays, spatial correlation characteristics, such as the diffusion path and distortion mode of the eddy current field, can be extracted by analyzing the differences in signals from adjacent coils or constructing a topological map. These spatial features can intuitively locate defects and provide geometric information about them.

[0107] S503, multimodal data fusion.

[0108] The extracted time-domain, frequency-domain, and spatial features are input into a multimodal data fusion network. This fusion network employs a deep learning architecture, such as a multi-branch convolutional neural network or a neural network incorporating an attention mechanism. Each branch is responsible for processing features of a specific modality; for example, one branch processes voltage signal features, another magnetic field strength features, and yet another temperature features. Within each branch, multiple convolutional and pooling layers extract deep, abstract features of that modality. At the output of each modal feature branch, an attention mechanism adaptively weights the importance of different modal features, achieving feature-level fusion. The attention mechanism can dynamically adjust the contribution of different modalities in the fusion process; for example, when temperature anomalies are significant, the weight of the temperature modality increases. The fused feature vector carries multi-dimensional information about the magnetic ring's electrical, magnetic, and thermal properties, providing a more comprehensive characterization of the ring's internal state. The mathematical representation of the fusion process is as follows:

[0109]

[0110] Where F_fused is the fused feature vector, M is the number of modes, α_m is the attention weight of the m-th mode, and f_m(F_m) is the feature representation of the m-th mode after feature extraction and transformation.

[0111] S6, Inversion of electromagnetic properties inside the magnetic ring and defect identification.

[0112] The fused multimodal feature vectors are fed into the inversion algorithm of the multimodal data fusion and inversion module to quantitatively evaluate the electromagnetic physical parameters inside the magnetic ring and identify defects. This step is further refined into sub-steps:

[0113] S601, Forward Model Establishment and Inversion Algorithm.

[0114] First, a forward electromagnetic model of the magnetic ring is established, describing the internal conductivity and permeability distributions, as well as the physical effects of geometric defects on external excitation and sensor response. The forward model can be constructed using the finite element method or finite difference method, simulating the theoretical sensor response under different internal parameters by numerically solving Maxwell's equations. The inversion algorithm aims to infer the internal electromagnetic parameter distribution of the magnetic ring from the actual measured sensor response and fused features. The inversion process can employ iterative optimization algorithms, such as Newton's method or Gauss-Newton method, updating the internal parameters by minimizing the error between the theoretical and actual responses. Alternatively, a deep learning-based inversion method can be used, directly mapping the fused features to the internal parameter distribution using a trained neural network. The convergence and stability of the inversion algorithm are guaranteed by regularization techniques, such as Tikhonov regularization, to prevent ill-conditioned problems during the inversion process. The inversion results are a three-dimensional distribution image of the internal conductivity and permeability of the magnetic ring, as well as the geometry and physical properties of the defect region. For example, inversion can yield a conductivity distribution map of a certain cross section inside the magnetic ring, where the conductivity value of the defect region will deviate significantly from that of the matrix material.

[0115] S602, Defect Characteristics and Identification.

[0116] Based on the inverted image of the electromagnetic parameter distribution inside the magnetic ring, the system further extracts and identifies defect features. Defect features include the defect's location, size (e.g., length, width, depth), shape, and the intensity of its influence on the surrounding electromagnetic field. For example, image processing algorithms can identify the boundaries of areas with abnormal conductivity and calculate their area and center coordinates. Defect identification utilizes pattern recognition algorithms, such as support vector machines, decision trees, or deep neural network classifiers, to match the extracted defect features with preset defect types. By training on a large number of known defective magnetic rings, the model can learn electromagnetic response feature patterns corresponding to different defect types, such as cracks, inclusions, porosity, and insufficient size. The accuracy of the identification process depends on the diversity of the training data and the model's generalization ability. For example, a deep neural network can identify the presence of a microcrack with a width of 0.1 mm based on the inverted image of conductivity inhomogeneity and provide its confidence level.

[0117] S7, multi-dimensional detection result analysis and output.

[0118] The final step of the detection method is to comprehensively analyze the results of the electromagnetic property inversion and defect identification inside the magnetic ring and present them in a user-friendly manner.

[0119] The system comprehensively evaluates the internal conductivity and permeability distribution maps of the magnetic ring obtained from the inversion process, as well as the identified defect information. For example, by comparing the electromagnetic parameter distribution with that of a standard magnetic ring, the system quantifies the uniformity index of the magnetic ring material, such as calculating the local coefficient of variation of permeability. For identified defects, the system assesses their risk level based on their type, size, and location, classifying defects as severe, general, or minor. The comprehensive evaluation results may include a determination of the overall quality level of the magnetic ring, such as whether it is acceptable, requires rework, or is scrapped. Regarding output, the inspection results are displayed on a high-resolution monitor in various visualization methods. For example, 3D rendering technology can be used to generate a color distribution map of the internal conductivity and permeability of the magnetic ring, where different colors represent different electromagnetic parameter values, and defect areas are highlighted or displayed with specific textures, intuitively presenting the internal structural information. Simultaneously, a detailed inspection report is generated, which includes the identification information of the magnetic ring under test, excitation parameters, statistics of the collected raw data, preprocessed data features, two-dimensional projection of multimodal fusion features, three-dimensional distribution of electromagnetic parameters obtained from inversion, a list of identified defects including defect type, coordinates, size, and confidence level, as well as the final quality assessment conclusion and subsequent processing suggestions. The data storage interface archives all inspection data and reports, facilitating quality traceability, process improvement, and big data analysis. Through this multi-dimensional, high-precision inspection and analysis, this invention can provide comprehensive quality control basis for the production and manufacturing of magnetic rings, effectively improving product reliability.

[0120] The present invention relates to a multi-dimensional non-destructive testing system for magnetic rings based on multi-frequency composite electromagnetic excitation. The system exhibits a tightly coordinated working mode with multi-level interactions and data flows among its functional modules. Under the command of the control and analysis unit, the composite electromagnetic excitation module generates a preset multi-frequency composite excitation signal and applies an electromagnetic field to the magnetic ring under test through an excitation coil array. Simultaneously, the multi-modal collaborative sensor array module captures the multi-source response signals of the magnetic ring under excitation in real time, including induced voltage, magnetic field strength, and temperature changes. These analog signals are then transmitted to a high-precision data acquisition and preprocessing module. This module performs analog-to-digital conversion, filtering and noise reduction, time synchronization, and normalization to convert the original analog signals into high-quality digital feature data. The preprocessed multi-modal digital data stream is transmitted to the multi-modal data fusion and inversion module via an internal data bus. In this module, deep extraction of time-domain, frequency-domain, and spatial features is first performed. Then, the multi-modal data fusion network achieves the collaborative integration of heterogeneous features, forming a unified, high-dimensional fused feature representation. This fused feature is then used as input to the electromagnetic inversion algorithm. Through a forward model and iterative optimization process, the conductivity, permeability distribution, and defect information within the magnetic ring are accurately inverted. All processing steps, including parameter configuration, task scheduling, data storage, and result querying, are uniformly managed and coordinated by the control and analysis unit. The final detection results, including a visualized 3D electromagnetic parameter distribution map, defect location and quantification information, and a comprehensive quality assessment report, are presented to the user through the result output and display module. The system's data flow is clear, bidirectional, and real-time. For example, the control and analysis unit can adjust excitation parameters based on real-time feedback from the multimodal data fusion and inversion module to optimize detection sensitivity or focus on defect detection in specific areas. The entire system ensures accurate alignment of data acquisition and processing through a precise clock synchronization mechanism and achieves efficient information flow and feedback control between modules through advanced algorithms, forming a closed-loop, intelligent detection ecosystem.

[0121] Example 3

[0122] In order to dynamically adjust the electromagnetic excitation parameters based on the preliminary test results during the testing process, in order to optimize the testing efficiency or improve the detection sensitivity for specific defects.

[0123] After the above data fusion and inversion steps are completed, an adaptive excitation parameter adjustment step is added, which specifically includes:

[0124] Preliminary Result Analysis: After completing a collaborative inversion calculation, the data processing and collaborative inversion module analyzes the preliminary information on the output magnetic property uniformity, material composition consistency, and internal defects. For example, it may identify a moderate anomaly in the magnetic property uniformity of a certain region, but the defect type is not yet clear; or although a specific type of defect (such as a microcrack) is preliminarily identified, the accuracy of its size or orientation is insufficient.

[0125] Optimization Target Setting: Based on the preliminary analysis results, the optimization target for this adaptive adjustment is set automatically or according to a preset strategy. For example, if microcracks are initially identified, the optimization target is set to "improve the detection sensitivity of microcracks"; if the magnetic properties of a certain area are unclear, the target is set to "improve the spatial resolution of the magnetic properties of that area".

[0126] Incentive strategy generation: Based on the optimization objective and a preset rule base or small machine learning decision model, a new combination of composite electromagnetic excitation parameters is generated.

[0127] For example, to improve the sensitivity of microcrack detection, the system may decide:

[0128] Fine-tuning the frequency range of the medium-to-high frequency eddy current excitation to a specific frequency band sensitive to this type of crack, and increasing the amplitude of the excitation current, can improve the eddy current penetration depth or surface resolution.

[0129] Adjusting the bias direction of the low-frequency magnetic field excitation or increasing the sampling density of the low-frequency sweep frequency can enhance the ability to capture magnetic field distortion near the crack.

[0130] Adjust the incentive sequence to make high-frequency incentives and low-frequency incentives more closely coordinated in time, forming a specific composite incentive pattern.

[0131] Excitation parameter update and re-detection: The newly generated excitation parameters are sent to the composite electromagnetic excitation module via the control interface. Based on the new parameters, the composite excitation field generation step is re-executed, and the collaborative data acquisition module is driven to perform data acquisition again.

[0132] Iterative optimization or completion: Newly acquired data will re-enter the data preprocessing step and the multimodal data fusion and integrated inversion analysis step. The system can perform multiple iterations as needed until the preset detection accuracy requirements or optimization goals are reached, or the maximum number of iterations is reached, and then proceed to the detection result output and display step.

[0133] In order to correlate the test results of magnetic rings with key process parameters in the production process, thereby providing data support for the optimization of production process and quality control.

[0134] After the test results are output and displayed, add a step to analyze the correlation between the test results and the production process, which includes:

[0135] Data Acquisition for Inspection and Process: Detailed inspection results (such as magnetic uniformity index, material composition deviation, number, type, location, and size of defects) for the latest completed magnetic rings are retrieved from the system's internal database. Simultaneously, through data exchange with external Manufacturing Execution System (MES), Enterprise Resource Planning (ERP), or Supervisory Control and Data Acquisition (SCADA) systems, key process parameters corresponding to the specific magnetic ring during production are obtained, such as raw material batch information, sintering temperature profile, sintering time, pressing pressure, atmosphere composition, cooling rate, and mold wear.

[0136] Data alignment and integration: The acquired testing and production process data are aligned with timestamps and associated with product batch numbers to create a one-to-one, multi-dimensional dataset. Data of different dimensions is standardized.

[0137] Feature engineering and model building: Extracting features from ensemble data that characterize process status and product quality. Building or calling pre-trained machine learning models (e.g., decision trees, random forests, support vector machines, neural networks, or causal inference-based models).

[0138] Correlation analysis and pattern recognition: Using machine learning models to analyze integrated datasets to identify potential correlations, causal relationships or statistical patterns between detection results (such as defect incidence rate, magnetic performance deviation) and specific process parameters.

[0139] For example, the system may identify: "When the sintering temperature fluctuates by more than X degrees Celsius, the probability of the magnetic permeability uniformity of the magnetic ring decreasing by Y% increases by Z%"; or "Inclusion defects in a specific batch of raw materials that can cause abnormal dielectric constants".

[0140] Production optimization suggestions and early warnings: Based on the correlation analysis results, the system automatically generates targeted production optimization suggestions, such as "It is recommended to control the sintering temperature within a narrower range," or "A certain batch of raw materials has a high risk; it is recommended to strengthen testing." Simultaneously, the system can issue early warnings for abnormal process parameters that may lead to serious quality problems.

[0141] Report generation and feedback: Generate customized reports that include correlation analysis results, pattern recognition insights, and optimization suggestions, and present and provide feedback to quality engineers, production managers, and other relevant personnel.

[0142] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation, characterized in that, include: The step of applying composite electromagnetic excitation is used to apply multi-frequency composite electromagnetic excitation to the magnetic ring to be tested through the composite electromagnetic excitation module. The multi-frequency composite electromagnetic excitation includes low-frequency magnetic field excitation, medium- and high-frequency eddy current excitation and broadband dielectric excitation. The step of capturing response signals is used to capture in real time the multidimensional response signals of the magnetic ring under test under the multi-frequency composite electromagnetic excitation through a multimodal collaborative sensor array module. The multidimensional response signals include magnetic field changes, eddy current signals, impedance spectra, and dielectric responses. The data preprocessing step is used to perform high-precision acquisition, time synchronization, filtering and noise reduction, and data normalization on the multidimensional response signal through a high-precision data acquisition and preprocessing module. The data fusion and inversion step is used to extract the multimodal features of the magnetic ring based on the preprocessed multidimensional response signal through the multimodal data fusion and inversion module, and to fuse and invert the multimodal features through a deep learning network to output the magnetic property uniformity, material composition consistency and potential defect information of the magnetic ring. The system control steps are used to coordinate the collaborative operation of the composite electromagnetic excitation module, the multimodal collaborative sensor array module, the high-precision data acquisition and preprocessing module, and the multimodal data fusion and inversion module through the system control and human-machine interaction module, and to provide a human-machine interaction interface.

2. The method for multi-dimensional non-destructive testing of magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 1, characterized in that, The step of applying the composite electromagnetic excitation includes: A low-frequency magnetic field excitation is applied to the magnetic ring to be tested by a low-frequency magnetic field excitation unit; A medium-to-high frequency eddy current excitation is applied to the magnetic ring under test by a medium-to-high frequency eddy current excitation unit; A broadband dielectric excitation is applied to the magnetic ring to be tested by a broadband dielectric excitation unit; The low-frequency magnetic field excitation unit consists of a pair of Helmholtz coils; the medium- and high-frequency eddy current excitation unit consists of a flat spiral coil; and the broadband dielectric excitation unit consists of a pair of parallel plate capacitors.

3. The method for multi-dimensional non-destructive testing of magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 2, characterized in that, The step of capturing the response signal includes: The uniformity signal of magnetic field distribution under low-frequency magnetic field excitation is captured by a three-axis fluxgate sensor array; The high-sensitivity Hall sensor array captures the secondary magnetic field change signal induced by medium- and high-frequency eddy currents. The complex impedance spectrum signal under broadband dielectric excitation was captured by a broadband dielectric response sensor array. The triaxial fluxgate sensor array comprises multiple sensors evenly distributed along the circumference of the magnetic ring, each capable of measuring the magnetic field strength in three orthogonal directions; the high-sensitivity Hall sensor array comprises multiple sensors evenly distributed along the radial and axial directions of the magnetic ring, each Hall sensor used to measure local magnetic field strength changes; and the broadband dielectric response sensor array consists of multiple high-frequency probes.

4. The multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 3, characterized in that, The data preprocessing steps include: The multimodal collaborative sensor array module captures data from each sensor synchronously using a high-resolution analog-to-digital converter. Hardware-level timestamp synchronization of data from various sensors is achieved through a general-purpose timer chip and a high-precision clock source. The acquired signal is noise-suppressed using a digital low-pass filter, a band-pass filter, and an adaptive Wiener filter. Sensor data with different dimensions and amplitudes are subjected to minimum-maximum normalization or Z-score normalization.

5. The multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 4, characterized in that, The data fusion and inversion steps include: Extract the magnetic field intensity gradient, harmonic content, and hysteresis loop parameters from the magnetic field change signal; The real part, imaginary part, phase angle, and amplitude attenuation curve of the eddy current impedance are extracted from the eddy current signal using wavelet transform and Fourier transform. The complex dielectric constant, dielectric loss tangent, and relaxation time distribution are extracted from the dielectric response signal.

6. The multi-dimensional non-destructive testing method for magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 5, characterized in that, The data fusion and inversion steps also include: The extracted magnetic field feature vector, eddy current feature matrix, and dielectric spectrum feature vector are input into the deep learning network for fusion. The deep learning network is a deep convolutional recurrent network based on an attention mechanism. The deep convolutional recurrent network structure includes multiple convolutional layers, each followed by batch normalization and ReLU activation functions, then multiple bidirectional long short-term memory network layers, and finally a multi-head attention layer.

7. The method for multi-dimensional non-destructive testing of magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 6, characterized in that, The data fusion and inversion steps also include: The output layer of the deep convolutional recurrent network is connected to a multilayer perceptron to invert the two-dimensional distribution of magnetic domains, residual magnetic flux density, and coercivity inside the magnetic ring. Invert the three-dimensional distribution of electrical conductivity, magnetic permeability and dielectric constant of the material; Identify the location, size, and shape of defects such as cracks, pores, and inclusions.

8. The method for multi-dimensional non-destructive testing of magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 7, characterized in that, The method further includes: The results display and diagnostic report generation module displays the distribution of magnetic properties, material parameters, and defect locations inside the magnetic ring in a 3D visualization interface, and generates a customized report that includes detection parameters, raw data curves, inversion results, defect details, and quality assessment level.

9. A multi-dimensional non-destructive testing system for magnetic rings based on multi-frequency composite electromagnetic excitation, applicable to the method described in any one of claims 1-8, characterized in that, include: A composite electromagnetic excitation module is used to apply multi-frequency composite electromagnetic excitation to the magnetic ring to be tested. The multi-frequency composite electromagnetic excitation includes low-frequency magnetic field excitation, medium- and high-frequency eddy current excitation and broadband dielectric excitation. A multimodal collaborative sensor array module is used to capture in real time the multidimensional response signal of the magnetic ring under test under the multi-frequency composite electromagnetic excitation. The multidimensional response signal includes magnetic field change, eddy current signal, impedance spectrum and dielectric response. A high-precision data acquisition and preprocessing module is used to perform high-precision acquisition, time synchronization, filtering and noise reduction, and data normalization on the multidimensional response signal. The multimodal data fusion and inversion module is used to extract multimodal features of the magnetic ring based on the preprocessed multidimensional response signal, and to fuse and invert the multimodal features through a deep learning network to output information on the magnetic property uniformity, material composition consistency and potential defects of the magnetic ring. The system control and human-machine interaction module is used to coordinate and control the collaborative work of the composite electromagnetic excitation module, the multimodal collaborative sensor array module, the high-precision data acquisition and preprocessing module, and the multimodal data fusion and inversion module, and to provide a human-machine interaction interface.

10. The multi-dimensional non-destructive testing system for magnetic rings based on multi-frequency composite electromagnetic excitation according to claim 9, characterized in that, The composite electromagnetic excitation module includes: A low-frequency magnetic field excitation unit is used to apply a low-frequency magnetic field excitation to the magnetic ring to be tested. The low-frequency magnetic field excitation unit is composed of a pair of Helmholtz coils. A medium-high frequency eddy current excitation unit is used to apply medium-high frequency eddy current excitation to the magnetic ring to be tested. The medium-high frequency eddy current excitation unit is composed of a flat spiral coil. A broadband dielectric excitation unit is used to apply broadband dielectric excitation to the magnetic ring to be tested. The broadband dielectric excitation unit consists of a pair of parallel plate capacitors.

Images