Stress assessment methods, apparatus, equipment and storage media for ferromagnetic components

By extracting multidimensional features and evaluating multi-level classification models from the original magnetic memory signals of ferromagnetic components, the subjectivity and accuracy issues of existing detection methods are resolved, and the stress assessment of ferromagnetic components is automated and precise.

CN123069162APending Publication Date: 2026-11-10HEBEI UNIV OF SCI & TECH +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611560412.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-10
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing methods for detecting ferromagnetic components rely on manual experience, resulting in subjective and inaccurate test results that fail to meet the needs for rapid screening and precise diagnosis on engineering sites.

Method used

By acquiring the original magnetic memory signal of the ferromagnetic component, preprocessing it, extracting the multidimensional magnetic memory feature vector, and inputting it into a multi-level classification stress assessment model, including anomaly detection model, defect location identification model, and stress concentration degree assessment model, a hierarchical intelligent assessment process from coarse to fine is realized.

Benefits of technology

It improves the accuracy and efficiency of detection, can automatically identify stress concentration areas and defect locations of ferromagnetic components, reduces the subjectivity of manual judgment, and adapts to the detection needs under complex working conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069162A_ABST
    Figure CN123069162A_ABST
Patent Text Reader

Abstract

This invention provides a method, apparatus, device, and storage medium for stress assessment of ferromagnetic components, relating to the field of nondestructive testing technology for metals. It includes: preprocessing the original magnetic memory signal; extracting multidimensional features from the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector; and inputting the multidimensional magnetic memory feature vector into a pre-trained multi-level classification stress assessment model to obtain the assessment result of the ferromagnetic component to be assessed. The multi-level classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration assessment model. Only when the anomaly detection model determines that a defect exists in the ferromagnetic component to be assessed, the defect location is determined based on the defect location identification model. After the defect location is determined by the defect location identification model, the stress concentration level of the ferromagnetic component to be assessed is determined based on the stress concentration assessment model. This invention can improve the accuracy of assessment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology for metals, and in particular to a method, apparatus, equipment, and storage medium for stress assessment of ferromagnetic components. Background Technology

[0002] When ferromagnetic components are subjected to tension, bending, fatigue, or localized contact loads during service, their internal magnetic domain structure undergoes irreversible rearrangement, forming localized magnetic anomalies in stress concentration areas. By detecting the changes in the magnetic memory signal on the component surface, non-destructive testing can be performed on early stress concentration areas of ferromagnetic components, providing a basis for structural safety assessment and early damage identification.

[0003] Metal Magnetic Memory (MMM) testing technology is a non-destructive testing method based on the distribution pattern of spontaneous leakage magnetic fields formed by ferromagnetic materials under the combined action of working load and geomagnetic field. In a geomagnetic environment, the mechanomagnetic effect generated by the load on ferromagnetic materials affects the distribution of the spatial leakage magnetic field. By detecting the distribution of the magnetic memory signal of the leakage magnetic field, it can be used to evaluate the early loss state of ferromagnetic materials.

[0004] However, currently, the testing process requires engineers to make a comprehensive judgment based on the shape of the magnetic memory curve, the peak position, and gradient changes. The test results have a certain degree of subjectivity, which leads to inaccurate test results. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and storage medium for stress assessment of ferromagnetic components to solve the problem of inaccurate detection by current testing methods.

[0006] In a first aspect, embodiments of the present invention provide a method for stress evaluation of ferromagnetic components, comprising: Acquire the original magnetic memory signal of the ferromagnetic component to be evaluated, and preprocess the original magnetic memory signal; Multidimensional feature extraction is performed on the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector. The multidimensional magnetic memory feature vector is input into a pre-trained multi-level classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated. The multi-layer classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration degree assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. After the defect location identification model determines the defect location of the ferromagnetic component to be assessed, the stress concentration degree of the ferromagnetic component to be assessed is determined based on the stress concentration degree assessment model.

[0007] In one possible implementation, the original magnetic memory signal is a normal magnetic memory signal; The step of extracting multidimensional features from the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector includes: Calculate the gradient signal of the preprocessed normal magnetic memory signal; Multidimensional feature extraction is performed on the preprocessed normal magnetic memory signal and the gradient signal to obtain multidimensional magnetic memory feature vectors.

[0008] In one possible implementation, the step of performing multidimensional feature extraction on the preprocessed normal magnetic memory signal and the gradient signal to obtain a multidimensional magnetic memory feature vector includes: Extract the magnetic memory signal features of the preprocessed normal magnetic memory signal; wherein, the magnetic memory signal features include signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, and signal morphology features; Extract the gradient enhancement features of the gradient signal; wherein, the gradient enhancement features include gradient statistical features, gradient energy features, gradient rate of change features, and gradient curvature features; The magnetic memory signal features and the gradient enhancement features are fused to obtain the multidimensional magnetic memory feature vector.

[0009] In one possible implementation, inputting the multidimensional magnetic memory feature vector into a pre-trained multi-level classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated includes: An anomaly detection feature subset is input into the anomaly detection model. If it is determined that the ferromagnetic component to be evaluated has a defect, the defect location feature subset is input into the defect location identification model. After determining the defect location of the ferromagnetic component to be evaluated, the stress concentration feature subset is input into the stress concentration degree evaluation model to obtain the stress concentration degree of the ferromagnetic component to be evaluated. The anomaly detection feature subset, the defect location feature subset, and the stress concentration feature subset are selected from the multidimensional magnetic memory feature vector based on the contribution of each candidate feature to the gain value generated by node splitting during the training process of the anomaly detection model, the defect location identification model, and the stress concentration degree evaluation model. The candidate feature is any one of the features in the multidimensional magnetic memory feature vector.

[0010] In one possible implementation, the anomaly detection feature subset includes signal time-domain amplitude statistical features, signal energy features, signal wavelet domain features, gradient statistical features, and gradient energy features; The defect location feature subset includes signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, signal morphological features, gradient statistical features, and gradient change rate features; The stress concentration feature subset includes signal time-domain amplitude statistical features, signal energy features, gradient statistical features, gradient rate of change features, and gradient curvature features.

[0011] In one possible implementation, the anomaly detection feature subset includes the maximum amplitude of the magnetic anomaly response, the minimum amplitude of the magnetic anomaly response, the standard deviation of the magnetic anomaly response, the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, and the gradient energy of the gradient signal. The defect location feature subset includes the peak-to-peak value of the magnetic anomaly response, the overall response intensity of the magnetic anomaly region, the dominant frequency of the magnetic signal spectrum, the centroid of the magnetic signal spectrum, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the peak-to-valley distance of the magnetic signal waveform, the peak-to-valley amplitude ratio of the magnetic signal waveform, the number of zero crossovers of the magnetic signal waveform, the maximum gradient value of the gradient signal, the standard deviation of the gradient signal, the gradient peak position of the gradient rate of change, the gradient peak-to-valley spacing of the gradient rate of change, the degree of change of the gradient peak of the gradient rate of change, and the gradient integral area of ​​the gradient rate of change. The stress concentration feature subset includes the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, the degree of change of the gradient peak value of the gradient rate of change, the gradient integral area of ​​the gradient rate of change, the maximum gradient curvature, and the standard deviation of the gradient curvature.

[0012] In one possible implementation, the preprocessing of the original magnetic memory signal includes: The original magnetic memory signal is sequentially subjected to interpolation resampling, filtering, and standardization processing.

[0013] Secondly, embodiments of the present invention provide a stress evaluation device for ferromagnetic components, comprising: The acquisition module is used to acquire the original magnetic memory signal of the ferromagnetic component to be evaluated and to preprocess the original magnetic memory signal. The extraction module is used to extract multi-dimensional features from the preprocessed raw magnetic memory signal to obtain a multi-dimensional magnetic memory feature vector. The evaluation module is used to input the multidimensional magnetic memory feature vector into a pre-trained multi-layer classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated. The multi-layer classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration degree assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. Only when the defect location identification model determines the defect location of the ferromagnetic component to be assessed, the stress concentration degree of the ferromagnetic component to be assessed is determined based on the stress concentration degree assessment model.

[0014] Thirdly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect or any possible implementation thereof.

[0015] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect or any possible implementation thereof.

[0016] Fifthly, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect or any possible implementation thereof.

[0017] In this embodiment of the invention, to improve detection accuracy and efficiency, a multidimensional magnetic memory feature vector is extracted from the preprocessed original magnetic memory signal. This multidimensional magnetic memory feature vector is then input into a pre-trained multi-level classification stress assessment model to obtain the evaluation result of the ferromagnetic component to be evaluated. To improve the automated recognition level of the magnetic memory signal, a multi-level classification stress assessment model is constructed, which simultaneously includes an anomaly detection model, a defect location identification model, and a stress concentration assessment model. This organically integrates anomaly detection, defect location, and stress concentration assessment, forming a hierarchical intelligent assessment process from coarse to fine, which can improve the evaluation accuracy of the magnetic memory signal. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the implementation of the stress assessment method for ferromagnetic components provided in this embodiment of the invention. Figure 2 This is a schematic diagram of the stress evaluation device for ferromagnetic components provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0019] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0020] As introduced in the background section, magnetic memory testing technology for metals has been widely applied in the health monitoring of ferromagnetic components such as pressure vessels, long-distance pipelines, bridge steel structures, rail transit equipment, and lifting machinery. Current research mainly focuses on magnetic memory signal analysis, feature parameter extraction, stress concentration identification, and defect detection, and analyzes stress concentration areas using parameters such as magnetic signal peak value, zero crossing point, gradient value, and statistical characteristics. However, most studies still adopt a single-stage detection mode, typically relying directly on magnetic signal characteristics to complete defect identification or stress concentration classification, lacking a hierarchical intelligent analysis mechanism that conforms to engineering inspection processes.

[0021] However, the inventors discovered that current detection results still rely heavily on human experience. Engineers need to make comprehensive judgments by combining the shape of the magnetic memory curve, peak position, and gradient changes, resulting in a degree of subjectivity in the detection results. Furthermore, current methods typically use a single model to directly complete defect identification or stress concentration classification, lacking a hierarchical decision-making process from coarse to fine. When there are many objects to be inspected, the working conditions are complex, or the defect types are diverse, the model is prone to misjudgment, making it difficult to meet the needs of combining rapid screening and accurate diagnosis in engineering settings.

[0022] Furthermore, most current identification methods directly utilize magnetic memory signals for classification, without fully integrating the logical relationships between detection tasks. For example, the existence of stress anomalies, the location of the anomaly area as a defect, and the level of stress concentration are essentially decision problems at different levels. However, existing methods typically use a single classification model to complete all judgments at once, resulting in high model complexity, limited generalization ability, and a high likelihood of inaccurate identification.

[0023] See Figure 1 The flowchart illustrating the implementation of the stress evaluation method for ferromagnetic components provided in this embodiment of the invention is described in detail below: S110. Obtain the original magnetic memory signal of the ferromagnetic component to be evaluated, and preprocess the original magnetic memory signal.

[0024] In some embodiments, the original magnetic memory signal may be subjected to interpolation resampling, filtering, and normalization processing in sequence.

[0025] In this embodiment, since the number of magnetic memory signal sampling points may vary under different operating conditions, interpolation resampling method can be used to unify different signals to a fixed number of sampling points.

[0026] For example, a cubic interpolation method can be used to uniformly resample the signal to 256 sampling points. After resampling, a filter is used to smooth the magnetic memory signal, reducing the influence of random noise and preserving the waveform variation characteristics as much as possible.

[0027] Subsequently, the Z-score normalization method can be used to eliminate the differences in dimensions and amplitudes between different signals, thereby improving the stability of the data.

[0028] S120. Perform multidimensional feature extraction on the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector.

[0029] In some embodiments, the original magnetic memory signal may be a normal magnetic memory signal.

[0030] In this embodiment, the normal magnetic memory signal Hp(y) can be obtained by solving the magnetic field of the preprocessed original magnetic memory signal.

[0031] In some embodiments, in order to enhance the local abrupt change information in the stress concentration region, the gradient signal of the preprocessed normal magnetic memory signal can be calculated first, and then multidimensional feature extraction can be performed on the preprocessed normal magnetic memory signal and the gradient signal to obtain the multidimensional magnetic memory feature vector.

[0032] In this embodiment, the gradient signal of the preprocessed normal magnetic memory signal Hp(y) is calculated, and the gradient signal can be expressed as: ; Where K is the magnetic memory gradient signal, Hp(y) is the normal magnetic memory signal, and x is the detection path coordinate.

[0033] Since magnetic memory signals in stress concentration regions typically exhibit local abrupt changes, slope variations, and gradient anomalies, gradient calculation of the normal magnetic memory signal can effectively enhance the information on local variations in the magnetic signal, highlight the boundary features of magnetic anomaly regions, and improve the sensitivity and resolution of stress concentration areas. Compared to analysis using only the original magnetic memory signal, gradient features can more accurately characterize the spatial variation of the magnetic signal and form complementary features with the original magnetic memory signal. This provides richer and more sensitive feature information for subsequent multidimensional feature extraction and multi-level classification stress assessment models, thereby improving the accuracy and stability of stress anomaly detection, defect location, and stress concentration assessment under complex working conditions.

[0034] In this embodiment, the magnetic memory signal features of the preprocessed normal magnetic memory signal can be extracted first. Then, the gradient enhancement features of the gradient signal can be extracted. Finally, the magnetic memory signal features and the gradient enhancement features can be fused to obtain a multidimensional magnetic memory feature vector.

[0035] Among them, the magnetic memory signal features are used to describe the magnetic field amplitude distribution and overall response law of the detection area. They are features extracted based on the preprocessed normal magnetic memory signal Hp(y), including signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, and signal morphology features. Gradient enhancement features are used to enhance the local change information of the magnetic memory signal and highlight the abnormal change features at stress concentration areas and defect boundaries. These features include gradient statistical features, gradient energy features, gradient change rate features, and gradient curvature features.

[0036] In this embodiment, the time-domain amplitude statistical characteristics of the signal are used to describe the overall amplitude level and waveform distribution of the magnetic memory signal, including: the maximum value of the magnetic anomaly response. Minimum value mean Standard deviation Root mean square value Peak-to-peak value Skewness, Kurt, crest factor CF, waveform factor SF, impulse factor IF, and margin factor MF.

[0037] Maximum value of magnetic anomaly response The maximum amplitude used to reflect the magnetic anomaly response within the detection area is: .

[0038] Minimum value of magnetic anomaly response Used to characterize the amplitude of negative changes in magnetic signals, .

[0039] Mean of magnetic anomaly response Used to describe the overall magnetic field offset level of the detection area, .

[0040] Standard deviation of magnetic anomaly response Used to evaluate the degree of amplitude fluctuation of magnetic signals, .

[0041] Root mean square value of magnetic anomaly response Used to reflect the overall energy level of the magnetic signal, .

[0042] Peak-to-peak value of magnetic anomaly response Used to characterize the maximum range of variation in a magnetic anomaly region, .

[0043] The skewness (Skew) of the magnetic anomaly response describes the degree of asymmetry in the amplitude distribution of the magnetic signal and reflects the direction of magnetic field shift in the anomaly region. .

[0044] The kurtosis (Kurt) of a magnetic anomaly response is used to describe the degree of spikes in the magnetic signal and can reflect the characteristics of local anomalous changes. .

[0045] The peak factor CF of the magnetic anomaly response is used to evaluate the change in the peak value of the magnetic signal relative to the overall energy level, and is: .

[0046] The waveform factor SF of the magnetic anomaly response is used to describe the waveform shape characteristics of the magnetic signal, and is as follows: .

[0047] The impulse factor IF of the magnetic anomaly response is used to reflect the transient peak response in the magnetic anomaly region, and is: .

[0048] The margin factor MF of the magnetic anomaly response is used to evaluate the difference between the peak value and the average amplitude of the magnetic signal. .

[0049] In this embodiment, the signal energy characteristics include the overall energy E of the magnetic anomaly region. H Overall response intensity A in the magnetic anomaly region H .

[0050] The overall energy E of the magnetic anomaly region can be calculated based on the normal magnetic memory signal. H for: .

[0051] Overall response intensity A of the magnetic anomaly region H The signal integration area can then be calculated, as follows: .

[0052] Overall response intensity A of the magnetic anomaly region H Used to evaluate the overall response intensity and spatial distribution range of magnetic anomaly regions.

[0053] In this embodiment, the frequency domain characteristics of the signal include the dominant frequency f of the magnetic signal spectrum. m Maximum spectral amplitude A of magnetic signal f Total energy E of magnetic signal spectrum f Magnetic signal spectrum centroid C f and the ratio of magnetic signal frequency band energy R i .

[0054] The dominant frequency f of the magnetic signal spectrum m The frequency representing the main change in the magnetic signal is: ; Maximum spectral amplitude A of magnetic signalf The intensity of the main frequency components is represented as follows: ; Total energy E of magnetic signal spectrum f Used to evaluate the frequency domain response intensity, as follows: ; Magnetic signal spectrum centroid C f Used to describe the location of concentrated spectral energy, as follows: ; Magnetic signal frequency band energy ratio R i The energy proportions of low-frequency, mid-frequency, and high-frequency components are used to describe the magnetic response characteristics at different scales, as follows: .

[0055] In this embodiment, the wavelet domain features of the signal are obtained using three-level discrete wavelet decomposition. ; Multi-scale analysis was performed on the normal magnetic memory signal to extract wavelet energy features and wavelet entropy features at different frequency bands.

[0056] Wavelet domain features of the signal include the wavelet energy E of magnetic signals at different scales. i Entropy of magnetic signals at different scales, and wavelet energy E of magnetic signals at different scales. i The wavelet entropy (Entropy) of magnetic signals at different scales is used to describe the intensity of changes in magnetic signals at different scales, and is used to evaluate the complexity of the frequency band distribution of magnetic signals.

[0057] Wavelet energy E of magnetic signals at different scales i for: C i These are wavelet coefficients at different scales.

[0058] The wavelet entropy of magnetic signals at different scales is: , .

[0059] In this embodiment, signal morphology features are used to describe the spatial distribution pattern and waveform structure changes of the magnetic anomaly region, including the peak-to-valley distance L of the magnetic signal waveform. pv The peak-to-valley amplitude ratio R of the magnetic signal waveform pv The zero-crossing position x0 of the magnetic signal waveform and the number of zero-crossings N of the magnetic signal waveform. z The signal entropy S of the magnetic signal waveform H and the spectral entropy S of the magnetic signal waveform f .

[0060] Specifically, the peak-to-valley distance L of the magnetic signal waveform. pv The width of the magnetic anomaly region is used to describe the spatial width of the region. ; The peak-to-valley amplitude ratio R of the magnetic signal waveform pv Used to describe waveform symmetry, as follows: ; The zero-crossing position x0 of the magnetic signal waveform is used to describe the corresponding position as a zero-point feature, as follows: ; The number of zero-crossings N in the waveform of the magnetic signal z Used to describe the complexity of changes in magnetic signals.

[0061] The signal entropy S of the magnetic signal waveform H Used to evaluate the randomness of a signal, as follows: ; Spectral entropy S of magnetic signal waveform f Used to describe the complexity of the spectrum: .

[0062] In this embodiment, the gradient statistical features include: the maximum gradient value of the gradient signal. Minimum gradient of the gradient signal The gradient mean of the gradient signal Standard deviation of gradient signal Root mean square of gradient signal Gradient skewness of the gradient signal (Skew) K and the kurtosis of the gradient signal Kurt K .

[0063] Among them, the maximum gradient of the gradient signal for .

[0064] Minimum gradient of the gradient signal for .

[0065] Gradient mean of the gradient signal for .

[0066] Standard deviation of gradient signal for .

[0067] root mean square of gradient signal for .

[0068] Gradient skewness of the gradient signal (Skew) K for .

[0069] Kurt gradient of gradient signal K for .

[0070] Gradient statistical features can enhance the trend of magnetic anomaly changes in stress concentration areas and improve the sensitivity to local anomaly areas.

[0071] In this embodiment, the gradient energy feature is used to evaluate the intensity of the change in magnetic signal along the detection direction, including the gradient energy E of the gradient signal. K ,for: .

[0072] In this embodiment, the gradient rate of change feature is used to characterize the location features of the defect boundary and stress concentration region, including the location of the gradient peak x of the gradient rate of change. Kmax The gradient peak-to-valley spacing L of the gradient rate of change K The degree of change of the peak gradient of the rate of change ΔK and the area A of the gradient integral of the rate of change. K .

[0073] Wherein, the peak position of the gradient rate of change is x Kmax Used to locate the area with the greatest changes in magnetic anomalies.

[0074] The gradient peak-to-valley spacing L of the gradient rate of change K for: ; The degree of change of the peak gradient of the rate of change is ΔK. ; The area A of the gradient integral of the rate of change of gradient K for .

[0075] In this embodiment, the gradient curvature feature is calculated based on the changing trend of the normal magnetic memory signal curve to determine the second-order variation characteristics of the signal. .

[0076] Gradient curvature features are used to describe the degree and trend of local bending of magnetic signals, and can further distinguish the morphological differences of magnetic anomaly regions under different stress concentration states, including: the maximum gradient curvature C. max Minimum value of gradient curvature C min and the standard deviation of gradient curvature C std .

[0077] Wherein, the maximum gradient curvature C max for ; Minimum value of gradient curvature C min for ; Gradient curvature standard deviation C std for .

[0078] Finally, by fusing the magnetic memory signal features and gradient enhancement features, a multidimensional magnetic memory feature vector is obtained. Thus, the multidimensional magnetic memory feature vector simultaneously contains amplitude response information, spatial variation information, local abrupt change information, and multi-scale spectral information of the magnetic anomaly region, which can improve the accuracy and stability of stress anomaly identification, defect location, and stress concentration assessment under complex working conditions.

[0079] S130. Input the multidimensional magnetic memory feature vector into the pre-trained multi-level classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated.

[0080] The multi-level classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. After the defect location identification model determines the defect location of the ferromagnetic component to be assessed, the stress concentration of the ferromagnetic component to be assessed is determined based on the stress concentration assessment model.

[0081] In some embodiments, since the magnetic memory response laws that are of interest to anomaly detection, defect location identification, and stress concentration assessment are different, different feature subsets can be constructed for different models to facilitate accurate subsequent assessment.

[0082] In this embodiment, the anomaly detection feature subset, defect location feature subset, and stress concentration feature subset are selected from the multidimensional magnetic memory feature vector based on the contribution of each candidate feature to the gain value generated by node splitting during the training process of the anomaly detection model, defect location identification model, and stress concentration degree evaluation model. The candidate feature is any one of the features in the multidimensional magnetic memory feature vector.

[0083] Specifically, firstly, the multidimensional magnetic memory feature vector is used as input to normalize all candidate features, thereby eliminating the influence of dimensional differences between different feature quantities. Then, for the anomaly detection model, defect location identification model, and stress concentration assessment model, the contribution of each candidate feature is analyzed. The anomaly detection model, defect location identification model, and stress concentration assessment model are all constructed using the XGBoost model. Finally, based on the gain value generated by node splitting during training, the candidate features are ranked according to their contribution, and those with a cumulative contribution rate exceeding the feature retention threshold are retained.

[0084] For example, the gain value generated by node splitting during XGBoost model training is used as a feature importance evaluation metric: ; Among them: I i Gain represents the importance evaluation value of the i-th candidate feature. ij The i-th candidate feature represents the contribution of the i-th candidate feature to the loss function reduction during the splitting of the j-th decision tree node; n represents the total number of candidate features.

[0085] Then, based on the importance evaluation results of each candidate feature, the candidate features are sorted in descending order of contribution, and the cumulative feature contribution rate is calculated: ; Where: C m This represents the cumulative contribution rate of the first m candidate features; N represents the total number of candidate features.

[0086] Finally, the cumulative contribution rate of 85% can be used as the feature retention threshold. The main features are selected according to the importance ranking of each candidate feature, and then the corresponding anomaly detection feature subset, defect location feature subset, and stress concentration feature subset are formed.

[0087] Furthermore, features can be further filtered based on their correlation and the type of magnetic memory response they represent, removing redundant features with similar or highly correlated physical meanings, and ultimately forming feature subsets for different tasks.

[0088] Through the above screening, the anomaly detection feature subset includes signal time-domain amplitude statistics, signal energy characteristics, signal wavelet domain characteristics, gradient statistics, and gradient energy characteristics, used to describe the overall magnetic response anomaly degree of the detected object. The defect location feature subset includes signal time-domain amplitude statistics, signal energy characteristics, signal frequency domain characteristics, signal wavelet domain characteristics, signal morphology characteristics, gradient statistics, and gradient change rate characteristics, used to describe the spatial distribution differences of the magnetic field caused by different defect locations. The stress concentration feature subset includes signal time-domain amplitude statistics, signal energy characteristics, gradient statistics, gradient change rate characteristics, and gradient curvature characteristics, used to describe the strength of the magnetoelastic response in the defect region and the local magnetic field variation law.

[0089] For example, the anomaly detection feature subset includes the maximum amplitude H of the magnetic anomaly response. max The minimum amplitude H of the magnetic anomaly response min The standard deviation H of the magnetic anomaly response std The root mean square value H of the magnetic anomaly response rms Peak-to-peak value ΔH of magnetic anomaly response p Overall energy E of the magnetic anomaly region H Overall response intensity A in the magnetic anomaly region H Wavelet energy E of magnetic signals at different scales i Wavelet entropy of magnetic signals at different scales, and maximum gradient K of gradient signals. max The root mean square of the gradient of the gradient signal K rms and the gradient energy E of the gradient signal K H max H min ΔH p H is used to describe the range of magnetic signal amplitude variation. std H rms E is used to describe the degree of fluctuation in magnetic signals. H A H K is used to evaluate the overall response intensity of a magnetic anomaly region. max K rms E K E is used to reflect gradient changes in abnormal regions. i Entropy is used to describe multi-scale magnetic response characteristics.

[0090] The defect location feature subset includes the peak-to-peak value ΔH of the magnetic anomaly response. p Overall response intensity A in the magnetic anomaly region H Magnetic signal spectrum, dominant frequency f m Magnetic signal spectrum centroid C f Wavelet energy E of magnetic signals at different scales i Wavelet entropy of magnetic signals at different scales, and peak-to-valley distance L of magnetic signal waveform morphology.pv The peak-to-valley amplitude ratio R of the magnetic signal waveform pv The number of zero-crossings N in the waveform of the magnetic signal z The maximum gradient of the gradient signal Standard deviation of gradient signal The location of the peak gradient of the rate of change of gradient x Kmax The gradient peak-to-valley spacing L of the gradient rate of change K The degree of change of the peak gradient of the rate of change ΔK and the area A of the gradient integral of the rate of change. K L pv R pv N z Used to describe the spatial morphology of magnetic anomaly waveforms. x Kmax L K ΔK, A K Used to reflect the location of gradient changes and defect boundary information; frequency domain and wavelet domain features are used to distinguish the differences in multi-scale magnetic response caused by different defect locations.

[0091] The stress concentration feature subset includes the root mean square value H of the magnetic anomaly response. rms Peak-to-peak value ΔH of magnetic anomaly response p Overall energy E of the magnetic anomaly region H Overall response intensity A in the magnetic anomaly region H The maximum gradient of the gradient signal Root mean square of gradient signal The degree of change of the peak gradient of the rate of change ΔK and the area A of the gradient integral of the rate of change. K Maximum gradient curvature C max and the standard deviation of gradient curvature C std ΔH p H rms E H Used to describe the degree of magnetic response enhancement; K max K rms E K ΔK is used to characterize the change in magnetic gradient caused by stress concentration; C max C std Used to describe local curvature changes in magnetic signals.

[0092] In some embodiments, after determining the feature subset of each model, a nonlinear mapping relationship between the feature subset of each model and the detection target can be established through a machine learning classification algorithm.

[0093] In this embodiment, the XGBoost classification model, which has strong nonlinear fitting and feature selection capabilities, can be used for training, meaning that all three models use the XGBoost classification model.

[0094] An anomaly detection model is used to determine whether an object is in an abnormal state, thus achieving preliminary screening of magnetic memory signals. The model's input features are a subset of anomaly detection features. ; Each sample is assigned an anomaly detection label, which is marked according to the specimen's condition: undamaged specimens are marked as 0; specimens with defects or abnormal areas are marked as 1.

[0095] By training an XGBoost binary classification model, a mapping relationship between the anomaly detection feature subset and the abnormal state of the specimen is established.

[0096] The main function of this model is not only to quickly determine whether the object being inspected has any anomalies, but also to filter out samples in a normal state, reducing the computational load on subsequent models. This prevents defect-free samples from entering the subsequent defect localization and stress concentration assessment processes, thereby improving overall inspection efficiency.

[0097] When the anomaly detection model outputs a normal state, the multi-layer classification stress assessment model will directly output the detection result. When the anomaly detection model outputs an abnormal state, the defect location identification model will continue to be used.

[0098] The defect location identification model is used to further determine the location type of the abnormal area, achieving spatial defect localization. Since the magnetic field propagation path and gradient change patterns differ depending on the defect's surface, back side, and interior, a subset of defect location features is used to train the defect location identification model. ; Each sample is assigned a defect location label, which is categorized based on the spatial location of the defect: surface defects, back defects, and internal defects.

[0099] The corresponding output result is: ; By training an XGBoost multi-classification model, a nonlinear mapping relationship is established between the feature subset of defect locations and the defect locations.

[0100] The main function of the defect location identification model is to identify the location of the anomaly source by utilizing the spatial variation law of magnetic memory signal, distinguish the differences in magnetic field response caused by defects at different locations, and provide a basis for location classification for subsequent stress concentration assessment.

[0101] After the defect location identification model completes the defect location determination, it inputs the stress concentration feature subset of the corresponding sample into the stress concentration degree evaluation model for further evaluation.

[0102] The stress concentration assessment model is used to evaluate the stress concentration level in defect areas, extending the assessment from defect identification to structural safety status evaluation. The stress concentration feature subset of this model is: ; The input matrix for the stress concentration assessment model is: .

[0103] To establish the correspondence between the characteristic subset of stress concentration and the degree of stress concentration, a stress concentration factor is used. As an evaluation indicator ; in: This represents the local maximum stress in the defect area; The nominal stress of the specimen.

[0104] Based on the finite element analysis results or experimental calculations of existing samples, obtain the K corresponding to each defect sample. t The values ​​were determined, and statistical distribution methods were used to classify the samples into grades.

[0105] First, for all samples, K t The values ​​are sorted in ascending order, and the 33% and 66% quantiles are calculated separately: ; Among them, K t1 and K t2 These are used as thresholds for classifying low, medium, and high stress concentration levels, respectively.

[0106] In this embodiment, based on defect sample K t result: ; Therefore, samples containing defects are divided into three levels: The above method is used to reduce the continuously varying stress concentration factor K. t The labels are converted into discrete levels to provide a supervised learning objective for training the stress concentration assessment model.

[0107] The training process for the stress concentration assessment model includes: An XGBoost multi-classification model is used to establish a stress concentration assessment model. The input is a subset of stress concentration features, and the output is Y3, which is the corresponding stress concentration level label.

[0108] By learning the characteristic subset of stress concentration and K tThe nonlinear mapping relationship between stress concentration levels enables intelligent prediction of stress concentration levels from magnetic memory signal response characteristics. After training, for an unknown test specimen, only the feature parameters extracted from its magnetic memory signal need to be input to output the corresponding stress concentration level.

[0109] The stress concentration assessment model indirectly reflects the stress concentration level in the defect area by utilizing changes in the amplitude of magnetic memory signals, the degree of gradient abrupt changes, and local waveform variations. This avoids the need for traditional methods that rely on local stress testing or finite element calculations to obtain K. t This overcomes the limitations of traditional methods and enables rapid risk level assessment based on magnetic memory detection signals. It also allows for a further extension from defect identification to structural safety status evaluation.

[0110] The multi-level classification stress assessment model operates in a cascaded manner, following the sequence of anomaly detection, defect location, and stress concentration assessment, to achieve step-by-step analysis from anomaly identification and defect location to structural safety evaluation.

[0111] In this embodiment, a sample library needs to be constructed to facilitate the training of the multi-layer classification stress assessment model. A magnetic memory signal dataset can be established using finite element simulation, and a magnetic coupling finite element model of ferromagnetic materials can be built. Magnetic memory detection signals are obtained by setting different defect morphologies, defect sizes, load levels, and lift-off values. Subsequently, the finite element model is validated using magnetic memory tensile test results, and the sample library is constructed.

[0112] The defect morphologies include rectangular grooves, holes, and pits. Defect locations include surface defects, back defects, and internal defects. An axial tensile load is used, and the lift-off value is used to simulate the change in distance between the probe and the specimen surface. Furthermore, for ease of training, sample labels are assigned to each sample, including a primary anomaly detection label, a secondary defect location label, and a tertiary stress concentration label. The primary label distinguishes between normal and abnormal samples, the secondary label characterizes defect locations (surface defects, back defects, and internal defects), and the tertiary label is based on the stress concentration factor K. t Establish stress concentration level classifications to characterize different degrees of stress concentration.

[0113] In this embodiment, anomaly detection feature subsets, defect location feature subsets, and stress concentration feature subsets can first be extracted from the multidimensional magnetic memory feature vector. Then, the anomaly detection feature subsets are input into the anomaly detection model. If a defect is determined in the ferromagnetic component to be evaluated, the defect location feature subset is input into the defect location identification model. After determining the defect location of the ferromagnetic component to be evaluated, the stress concentration feature subset is input into the stress concentration degree evaluation model to obtain the stress concentration degree of the ferromagnetic component to be evaluated.

[0114] In some embodiments, the original magnetic memory signal, the preprocessed original magnetic memory signal, and characteristic parameters can be visualized, and a test report can be generated. The test results can be exported as text or tabular files for test recording, equipment status assessment, and subsequent analysis.

[0115] To improve detection accuracy and efficiency, this invention extracts multidimensional magnetic memory feature vectors from the preprocessed raw magnetic memory signal and inputs these vectors into a pre-trained multi-level classification stress assessment model to obtain the evaluation results for the ferromagnetic component to be evaluated. To enhance the automated recognition level of the magnetic memory signal, a multi-level classification stress assessment model is constructed, simultaneously including an anomaly detection model, a defect location identification model, and a stress concentration assessment model. This organically integrates anomaly detection, defect location, and stress concentration assessment, forming a hierarchical intelligent evaluation process from coarse to fine, thereby improving the evaluation accuracy of the magnetic memory signal.

[0116] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0117] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0118] Figure 2 A schematic diagram of the stress evaluation device for ferromagnetic components provided in an embodiment of the present invention is shown. For ease of explanation, only the parts relevant to the embodiment of the present invention are shown, and are described in detail below: like Figure 2 As shown, the stress assessment device for ferromagnetic components includes: The acquisition module 210 is used to acquire the original magnetic memory signal of the ferromagnetic component to be evaluated and to preprocess the original magnetic memory signal. Extraction module 220 is used to extract multidimensional features from the preprocessed original magnetic memory signal to obtain multidimensional magnetic memory feature vector; Evaluation module 230 is used to input multidimensional magnetic memory feature vectors into a pre-trained multi-layer classification stress evaluation model to obtain the evaluation results of the ferromagnetic component to be evaluated. The multi-level classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. After the defect location identification model determines the defect location of the ferromagnetic component to be assessed, the stress concentration of the ferromagnetic component to be assessed is determined based on the stress concentration assessment model.

[0119] In one possible implementation, the original magnetic memory signal is a normal magnetic memory signal; Extraction module 220 is used to calculate the gradient signal of the preprocessed normal magnetic memory signal; Multidimensional feature extraction is performed on the preprocessed normal magnetic memory signal and gradient signal to obtain multidimensional magnetic memory feature vectors.

[0120] In one possible implementation, the extraction module 220 is used to extract the magnetic memory signal features of the preprocessed normal magnetic memory signal; wherein, the magnetic memory signal features include signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, and signal morphological features; Gradient enhancement features of the gradient signal are extracted; these features include gradient statistical features, gradient energy features, gradient rate of change features, and gradient curvature features. By fusing magnetic memory signal features and gradient enhancement features, a multidimensional magnetic memory feature vector is obtained.

[0121] In one possible implementation, the evaluation module 230 is used to input a subset of anomaly detection features into an anomaly detection model, and if it is determined that there is a defect in the ferromagnetic component to be evaluated, then input a subset of defect location features into a defect location identification model. After determining the location of the defect in the ferromagnetic component to be evaluated, the stress concentration feature subset is input into the stress concentration degree evaluation model to obtain the stress concentration degree of the ferromagnetic component to be evaluated. Among them, the anomaly detection feature subset, defect location feature subset, and stress concentration feature subset are selected from the multidimensional magnetic memory feature vector based on the contribution of each candidate feature to the gain value generated by node splitting during the training process of the anomaly detection model, defect location identification model, and stress concentration degree assessment model. The candidate feature is any one of the features in the multidimensional magnetic memory feature vector.

[0122] In one possible implementation, the anomaly detection feature subset includes signal time-domain amplitude statistics, signal energy features, signal wavelet domain features, gradient statistics, and gradient energy features; The defect location feature subset includes signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, signal morphological features, gradient statistical features, and gradient change rate features; The stress concentration feature subset includes the signal time-domain amplitude statistical features, signal energy features, gradient statistical features, gradient rate of change features, and gradient curvature features.

[0123] In one possible implementation, the anomaly detection feature subset includes the maximum amplitude of the magnetic anomaly response, the minimum amplitude of the magnetic anomaly response, the standard deviation of the magnetic anomaly response, the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, and the gradient energy of the gradient signal. The defect location feature subset includes the peak-to-peak value of the magnetic anomaly response, the overall response intensity of the magnetic anomaly region, the dominant frequency of the magnetic signal spectrum, the centroid of the magnetic signal spectrum, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the peak-to-valley distance of the magnetic signal waveform, the peak-to-valley amplitude ratio of the magnetic signal waveform, the number of zero crossovers of the magnetic signal waveform, the maximum gradient value of the gradient signal, the standard deviation of the gradient signal, the gradient peak position of the gradient rate of change, the gradient peak-to-valley spacing of the gradient rate of change, the degree of change of the gradient peak of the gradient rate of change, and the gradient integral area of ​​the gradient rate of change. The stress concentration feature subset includes the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, the degree of change of the gradient peak value of the gradient rate of change, the gradient integral area of ​​the gradient rate of change, the maximum gradient curvature, and the standard deviation of the gradient curvature.

[0124] In one possible implementation, the acquisition module 210 is used to sequentially perform interpolation resampling, filtering, and normalization processing on the original magnetic memory signal.

[0125] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Figure 3 As shown, the electronic device 3 of this embodiment includes a processor 30 and a memory 31. The memory 31 stores a computer program 32. When the processor 30 executes the computer program 32, it implements the steps in the various method embodiments described above. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module / unit in the various device embodiments described above.

[0126] For example, computer program 32 may be divided into one or more modules / units, which are stored in memory 31 and executed by processor 30 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of computer program 32 in electronic device 3.

[0127] Electronic device 3 may include, but is not limited to, processor 30 and memory 31. Those skilled in the art will understand that... Figure 3This is merely an example of electronic device 3 and does not constitute a limitation on electronic device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device 3 may also include input / output devices, network access devices, buses, etc.

[0128] The processor 30 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0129] The memory 31 can be an internal storage unit of the electronic device 3, such as a hard disk or memory of the electronic device 3. The memory 31 can also be an external storage device of the electronic device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 3. Furthermore, the memory 31 can include both internal and external storage units of the electronic device 3. The memory 31 is used to store the computer program 32 and other programs and data required by the electronic device 3. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0130] For the sake of simplicity and clarity, only the above-described functional modules / units are used as examples. In practical applications, the functions described above can be assigned to different functional modules / units as needed. These modules / units can be implemented in hardware, software, or a combination of both.

[0131] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0132] This invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the methods described in the above-described method embodiments.

[0133] Computer programs include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0134] In the above embodiments, the descriptions of each embodiment have their own emphasis. Parts not detailed or described in a particular embodiment can be referred to in the relevant descriptions of other embodiments. Unless otherwise specified or in conflict with logic, the terminology and / or descriptions between different embodiments are consistent and can be referenced interchangeably. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0135] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for stress evaluation of ferromagnetic components, characterized in that, include: Acquire the original magnetic memory signal of the ferromagnetic component to be evaluated, and preprocess the original magnetic memory signal; Multidimensional feature extraction is performed on the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector. The multidimensional magnetic memory feature vector is input into a pre-trained multi-level classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated. The multi-layer classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration degree assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. After the defect location of the ferromagnetic component to be assessed is determined by the defect location identification model, the stress concentration degree of the ferromagnetic component to be assessed is determined based on the stress concentration degree assessment model.

2. The stress assessment method for ferromagnetic components according to claim 1, characterized in that, The original magnetic memory signal is a normal magnetic memory signal; The step of extracting multidimensional features from the preprocessed original magnetic memory signal to obtain a multidimensional magnetic memory feature vector includes: Calculate the gradient signal of the preprocessed normal magnetic memory signal; Multidimensional feature extraction is performed on the preprocessed normal magnetic memory signal and the gradient signal to obtain multidimensional magnetic memory feature vectors.

3. The stress assessment method for ferromagnetic components according to claim 2, characterized in that, The process of extracting multidimensional features from the preprocessed normal magnetic memory signal and the gradient signal to obtain a multidimensional magnetic memory feature vector includes: Extract the magnetic memory signal features of the preprocessed normal magnetic memory signal; wherein, the magnetic memory signal features include signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, and signal morphology features; Extract the gradient enhancement features of the gradient signal; wherein, the gradient enhancement features include gradient statistical features, gradient energy features, gradient rate of change features, and gradient curvature features; The magnetic memory signal features and the gradient enhancement features are fused to obtain the multidimensional magnetic memory feature vector.

4. The stress assessment method for ferromagnetic components according to claim 1, characterized in that, The step of inputting the multidimensional magnetic memory feature vector into a pre-trained multi-layer classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated includes: An anomaly detection feature subset is input into the anomaly detection model. If it is determined that the ferromagnetic component to be evaluated has a defect, the defect location feature subset is input into the defect location identification model. After determining the defect location of the ferromagnetic component to be evaluated, the stress concentration feature subset is input into the stress concentration degree evaluation model to obtain the stress concentration degree of the ferromagnetic component to be evaluated. The anomaly detection feature subset, the defect location feature subset, and the stress concentration feature subset are selected from the multidimensional magnetic memory feature vector based on the contribution of each candidate feature to the gain value generated by node splitting during the training process of the anomaly detection model, the defect location identification model, and the stress concentration degree evaluation model. The candidate feature is any one of the features in the multidimensional magnetic memory feature vector.

5. The stress evaluation method for ferromagnetic components according to claim 4, characterized in that, The anomaly detection feature subset includes signal time-domain amplitude statistical features, signal energy features, signal wavelet domain features, gradient statistical features, and gradient energy features; The defect location feature subset includes signal time-domain amplitude statistical features, signal energy features, signal frequency-domain features, signal wavelet domain features, signal morphological features, gradient statistical features, and gradient change rate features; The stress concentration feature subset includes signal time-domain amplitude statistical features, signal energy features, gradient statistical features, gradient rate of change features, and gradient curvature features.

6. The stress assessment method for ferromagnetic components according to claim 5, characterized in that, The anomaly detection feature subset includes the maximum amplitude of the magnetic anomaly response, the minimum amplitude of the magnetic anomaly response, the standard deviation of the magnetic anomaly response, the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, and the gradient energy of the gradient signal. The defect location feature subset includes the peak-to-peak value of the magnetic anomaly response, the overall response intensity of the magnetic anomaly region, the dominant frequency of the magnetic signal spectrum, the centroid of the magnetic signal spectrum, the wavelet energy of magnetic signals at different scales, the wavelet entropy of magnetic signals at different scales, the peak-to-valley distance of the magnetic signal waveform, the peak-to-valley amplitude ratio of the magnetic signal waveform, the number of zero crossovers of the magnetic signal waveform, the maximum gradient value of the gradient signal, the standard deviation of the gradient signal, the gradient peak position of the gradient rate of change, the gradient peak-to-valley spacing of the gradient rate of change, the degree of change of the gradient peak of the gradient rate of change, and the gradient integral area of ​​the gradient rate of change. The stress concentration feature subset includes the root mean square value of the magnetic anomaly response, the peak-to-peak value of the magnetic anomaly response, the overall energy of the magnetic anomaly region, the overall response intensity of the magnetic anomaly region, the maximum gradient value of the gradient signal, the root mean square gradient value of the gradient signal, the degree of change of the gradient peak value of the gradient rate of change, the gradient integral area of ​​the gradient rate of change, the maximum gradient curvature, and the standard deviation of the gradient curvature.

7. The stress assessment method for ferromagnetic components according to any one of claims 1-6, characterized in that, The preprocessing of the original magnetic memory signal includes: The original magnetic memory signal is sequentially subjected to interpolation resampling, filtering, and standardization processing.

8. A stress evaluation device for ferromagnetic components, characterized in that, include: The acquisition module is used to acquire the original magnetic memory signal of the ferromagnetic component to be evaluated and to preprocess the original magnetic memory signal. The extraction module is used to extract multi-dimensional features from the preprocessed raw magnetic memory signal to obtain a multi-dimensional magnetic memory feature vector. The evaluation module is used to input the multidimensional magnetic memory feature vector into a pre-trained multi-layer classification stress evaluation model to obtain the evaluation result of the ferromagnetic component to be evaluated. The multi-layer classification stress assessment model includes an anomaly detection model, a defect location identification model, and a stress concentration degree assessment model. Only when the anomaly detection model determines that there is a defect in the ferromagnetic component to be assessed, the defect location of the ferromagnetic component to be assessed is determined based on the defect location identification model. After the defect location of the ferromagnetic component to be assessed is determined by the defect location identification model, the stress concentration degree of the ferromagnetic component to be assessed is determined based on the stress concentration degree assessment model.

9. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 7.

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