A method and system for nondestructive testing of steel based on multimodal data

Abstract

This application provides a non-destructive testing method and system for steel based on multimodal data, relating to the field of materials testing technology. The method includes identifying the starting moment of the steel entering the plastic deformation stage based on the acoustic emission signal sequence during the stress process, and determining the time interval of the increasing activity of the acoustic emission signal; acquiring the metal magnetic memory signal and surface strain data within the time interval; further determining the location of the magnetic field gradient anomaly and the location of the strain concentration region of the steel; determining the spatial overlap region of the multimodal signals; performing multi-scale decomposition on the acoustic emission signals within the spatial overlap region of the multimodal signals, determining the cumulative degree of irreversible changes in the microstructure of the steel based on the abrupt change characteristics in the decomposed signals, and subsequently determining the test result of the steel. This application can realize dynamic monitoring of the transition of steel from the elastic stage to the plastic deformation stage during the stress process, accurately identifying changes in the internal microstructure of the steel and spatially locating them.

Description

Technical Field

[0001] This disclosure generally relates to the field of materials testing technology, and in particular to a method and system for non-destructive testing of steel based on multimodal data. Background Technology

[0002] As a core material in industrial production and infrastructure construction, the quality and safety of steel directly affect the stability and durability of engineering projects. This is especially true in critical sectors such as bridges, buildings, and railways, where the detection of internal defects in steel is paramount. Non-destructive testing (NDT) technology, as an important means of ensuring steel quality, has received widespread attention in recent years. However, accurately identifying subtle, hidden damage within steel remains a major challenge that the industry urgently needs to overcome.

[0003] However, current nondestructive testing methods often rely on single or limited testing techniques, resulting in low accuracy in identifying potential defects. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in related technologies, this application provides a non-destructive testing method and system for steel based on multimodal data, which can solve the technical problem of collaborative analysis and spatial matching of multi-source physical signals in the dynamic monitoring of the internal state of steel during stress.

[0005] Firstly, a nondestructive testing method for steel based on multimodal data is provided, the method comprising:

[0006] Acquire the acoustic emission signal sequence of steel during the stress process;

[0007] The starting time of the steel entering the plastic deformation stage is identified based on the acoustic emission signal sequence, and the time interval of the increase in the activity of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage is determined.

[0008] Within the time interval, acquire the metal magnetic memory signal and surface strain data of the steel in the same detection area;

[0009] The location of the magnetic field gradient anomaly in the steel is determined based on the metal magnetic memory signal, and the location of the strain concentration region in the steel is determined based on the surface strain data.

[0010] Spatially match the location of the high-energy signal during the plastic deformation stage, the location of the magnetic field gradient anomaly, and the location of the strain concentration region to determine the spatial overlap region of the multimodal signal;

[0011] The acoustic emission signal in the spatially overlapping region of the multimodal signal is decomposed into multiple scales, and the cumulative degree of irreversible changes in the microstructure of the steel is determined based on the abrupt change characteristics in the decomposed signal.

[0012] The detection result of the steel is determined based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the spatially overlapping regions of the multimodal signals.

[0013] The nondestructive testing method for steel based on multimodal data provided in this application dynamically monitors the state of steel by acquiring the acoustic emission signal sequence of steel during the stress process; and identifies the starting moment of the steel entering the plastic deformation stage based on the acoustic emission signal sequence. According to the amplitude and frequency center distribution characteristics of the acoustic emission signal, cluster analysis is performed to divide the signal into elastic stage class and plastic deformation stage class, thereby accurately capturing the transition point of steel from elastic to plastic. Furthermore, the time interval for the increase in acoustic emission signal activity within the acoustic emission signal sequence during the plastic deformation stage is determined to ensure the targeted nature of subsequent multimodal data acquisition. Within this time interval, the metal magnetic memory signal and surface strain data of the steel in the same detection area are acquired. Based on the metal magnetic memory signal, the location of the magnetic field gradient anomaly in the steel is determined. By calculating the magnetic field gradient in the spatial direction and synthesizing a comprehensive gradient anomaly amplitude sequence, the location of the magnetic field anomaly is accurately identified. Simultaneously, the location of the strain concentration region of the steel is determined based on the surface strain data. Subsequently, the locations of the high-energy signals, magnetic field gradient anomalies, and strain concentration regions during the plastic deformation stage are spatially matched to determine the spatially overlapping region of the multimodal signals, achieving spatial collaborative localization of multi-source signals. Finally, the acoustic emission signals within the spatially overlapping region of the multimodal signals are decomposed into multiple scales. Based on the abrupt change characteristics in the decomposed signals, the cumulative degree of irreversible changes in the steel's microstructure is determined. Based on the cumulative degree of irreversible changes in the steel's microstructure and the spatial overlap of the spatially overlapping region of the multimodal signals, the detection result of the steel is determined. This method effectively solves the technical challenge of dynamically monitoring the internal state of steel during stress by using collaborative analysis and spatial matching of multimodal signals, thereby improving the accuracy of nondestructive testing of steel damage.

[0014] Secondly, a non-destructive testing system for steel based on multimodal data is provided, the system comprising:

[0015] The first acquisition module is used to acquire the acoustic emission signal sequence of steel during the stress process;

[0016] The identification and determination module is used to identify the starting time of the steel entering the plastic deformation stage based on the acoustic emission signal sequence, and to determine the time interval of the increase in the activity level of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage.

[0017] The second acquisition module is used to acquire the metal magnetic memory signal and surface strain data of the steel in the same detection area within the time interval.

[0018] The first determining module is used to determine the location of the magnetic field gradient anomaly of the steel based on the metal magnetic memory signal, and to determine the location of the strain concentration region of the steel based on the surface strain data.

[0019] The matching and determination module is used to spatially match the location of the high-energy signal, the location of the magnetic field gradient anomaly, and the location of the strain concentration region during the plastic deformation stage to determine the spatial overlap region of the multimodal signals.

[0020] The decomposition and determination module is used to perform multi-scale decomposition of the acoustic emission signal in the spatial overlap region of the multimodal signal, and determine the cumulative degree of irreversible changes in the microstructure of the steel based on the abrupt change characteristics in the decomposed signal.

[0021] The second determining module is used to determine the detection result of the steel based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the spatially overlapping region of the multimodal signal. Attached Figure Description

[0022] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0023] Figure 1 A flowchart illustrating the steps of a nondestructive testing method for steel based on multimodal data, provided in this application embodiment;

[0024] Figure 2 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0025] Figure 3 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0026] Figure 4 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0027] Figure 5 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0028] Figure 6 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0029] Figure 7 A flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided for an embodiment of this application;

[0030] Figure 8This is a flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided as an embodiment of this application.

[0031] Figure 9 This is a flowchart illustrating a nondestructive testing method for steel based on multimodal data, provided as an embodiment of this application.

[0032] Figure 10 This is a structural block diagram of a non-destructive testing system for steel based on multimodal data, provided in an embodiment of this application. Detailed Implementation

[0033] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0034] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0035] Steel plays a major load-bearing role in engineering structures, and its service safety highly depends on the early and accurate perception of the evolution process of its internal microstructure. Existing non-destructive testing methods are mostly based on single physical signals for assessment. For example, they rely solely on changes in the energy or count rate of acoustic emission signals to determine damage activity, or they locate stress concentration areas solely through anomalies in magnetic memory signal gradients. They lack the ability to dynamically identify the critical mechanical threshold of the transition from elasticity to plasticity. They also cannot simultaneously reflect the multi-field coupled responses caused by irreversible microscopic mechanisms such as crystal slip, dislocation multiplication, and microvoid initiation. This leads to misjudgment of the damage initiation time, divergent spatial positioning, and ambiguous evolution stages, making it impossible to support high-reliability service condition assessment.

[0036] To address the aforementioned problems, this application provides a non-destructive testing method for steel based on multimodal data, such as... Figure 1 As shown in Figure 1, which is a flowchart illustrating a nondestructive testing method for steel based on multimodal data provided in this application, the method includes:

[0037] Step S20: Obtain the acoustic emission signal sequence of the steel during the stress process;

[0038] The acoustic emission signal sequence refers to the broadband stress wave signal generated by the instantaneous release of elastic strain energy in steel under external load due to irreversible changes in its internal microstructure (such as dislocation movement, grain boundary slip, microcrack initiation, and propagation). This signal is collected and digitized by a smart sensor (such as a broadband acoustic emission sensor) to form a time series data. The acoustic emission signal sequence may include parameters such as the arrival time, peak voltage amplitude, ring count, rise time, duration, and frequency center of each acoustic emission event, which are used to characterize the spatiotemporal distribution characteristics of energy release within the material. This acoustic emission signal sequence provides the original dynamic response basis for subsequent identification of the onset time of plastic deformation, division of active regions, and multi-scale decomposition.

[0039] For example, this application may involve synchronously acquiring stress wave signals generated during the stress process using multiple broadband acoustic emission sensors arranged on the surface of steel, and then performing analog-to-digital conversion and storage via a preamplifier (gain 40dB) and a high-speed data acquisition card (sampling rate ≥100kHz) to form an acoustic emission signal sequence. Alternatively, this application may employ a three-point time difference positioning method combined with the sensor's spatial coordinates and the longitudinal wave propagation velocity (5500m / s) to perform three-dimensional spatial positioning of valid events that meet the amplitude threshold (≥45dB) and HIT definition time (200μs), and output an acoustic emission signal sequence containing timestamps and spatial coordinates. No further limitations are imposed here.

[0040] Step S30: Identify the starting time of the steel entering the plastic deformation stage based on the acoustic emission signal sequence, and determine the time interval of the increase in the activity level of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage.

[0041] The starting moment of the plastic deformation stage can refer to the loading moment when the steel's macroscopic stress-strain curve first exceeds the yield strength (e.g., 250 MPa), corresponding to the critical node where dislocations begin to multiply and slip irreversibly at the microscopic scale. The time interval during which the activity of acoustic emission signals increases can refer to the period during the plastic deformation stage when the count rate of acoustic emission events and the proportion of high-energy events increase significantly and synchronously, reflecting the transition process of internal material damage activity from diffuse to localized accelerated evolution. This time interval starts at the beginning of plastic deformation and ends at the inflection point of the high-energy event count rate trend or a set increase threshold, which is used to limit the spatial and temporal window for subsequent synchronous acquisition of multimodal signals.

[0042] For example, this application can perform k-means clustering (with 2 clusters and initial centers set around 30dB and 70dB respectively) based on the amplitude and frequency center distribution characteristics of each event in the acoustic emission signal sequence. The cluster with lower amplitude and higher frequency center is identified as the elastic stage signal, and the cluster with higher amplitude and lower frequency center is identified as the plastic deformation stage signal. Then, through time series comparison, the first obvious switching point between the two types of signals is determined (the proportion of plastic events in 10 consecutive events first exceeds 0.65 and remains thereafter), which is recorded as the start time of the plastic deformation stage.

[0043] Optionally, for example, the starting time can be used as the dividing point, and the 3000th to 6000th sampling points thereafter can be used to form a signal segment for the plastic deformation stage; wavelet transform (decomposition into 5 levels) can be performed on the signal segment to extract high-energy events with energy peak values ​​exceeding 0.05mJ in the high-frequency components of the 3rd level; the number of high-energy events can be counted using every 100 sampling points as a sliding window, and the count rate change curve can be fitted. The rising segment with a slope significantly greater than 0.1 times / second can be identified as the time interval in which the acoustic emission signal activity increases.

[0044] Furthermore, this application may also define the activity level index as the product of the high-energy event count rate and the average peak amplitude. When the index rises from the initial value to more than 1.2 times, the corresponding time period is the time interval.

[0045] Step S40: Acquire the metal magnetic memory signal and surface strain data of the steel in the same detection area within the time interval;

[0046] The magnetic memory signal of a metal can refer to the measurable physical quantity in which the leakage magnetic field intensity components (Hx, Hy, Hz) on the surface of steel change with stress state under the combined action of the geomagnetic field and residual stress. This reflects the rearrangement and pinning effect of the magnetic domain structure inside the ferromagnetic material under plastic deformation. For example, a triaxial magnetic gradient sensor array (probe spacing 5mm) can be used to synchronously collect the magnetic field intensity components Hx, Hy, and Hz on the surface of the steel within the time interval defined above.

[0047] Surface strain data can refer to the distribution of normal strain and shear strain in a specific area of ​​the surface of steel under the same loading conditions, caused by macroscopic deformation. It is used to characterize the local mechanical response state of the material under external force. The same detection area refers to the physical area in the spatial coordinate system that has a higher overlap with the acoustic emission signal positioning result than a preset threshold (e.g., 80%), ensuring that the three types of signals have a common mechanical origin and spatial correlation basis.

[0048] Optionally, this application may also involve placing the magnetic memory sensor and strain measurement equipment (such as a digital image correlation DIC system or strain gauge array) in the vicinity of the acoustic emission sensor to ensure that the spatial overlap of the three is not less than 75%, thereby ensuring the spatial comparability of multimodal data.

[0049] Step S50: Determine the location of the magnetic field gradient anomaly in the steel based on the metal magnetic memory signal, and determine the location of the strain concentration area in the steel based on the surface strain data.

[0050] The magnetic field gradient anomaly location can refer to the coordinate point where the gradient amplitude obtained by differential analysis of the metal magnetic memory signal along the spatial direction (x and y directions) exceeds the preset anomaly range (e.g., K>0.8A / (m·mm)), reflecting the local magnetic field disturbance induced by the sudden change in stress / dislocation density inside the material. For example, it can be calculated as Kx=(Hx(i+1)-Hx(i)) / Δx and Ky=(Hy(j+1)-Hy(j)) / Δy (Δx=Δy=5mm) along the x and y directions of the metal magnetic memory signal, synthesize the comprehensive gradient anomaly amplitude K=√(Kx²+Ky²), and extract the coordinates of the target point with K value greater than 0.8A / (m·mm) as the magnetic field gradient anomaly location.

[0051] The location of the strain concentration region can refer to the geometric center coordinates of the region in the surface strain data where the strain value exceeds a preset threshold (e.g., 0.0015) and is spatially continuous. This represents a weak point where energy accumulates locally in the macroscopic mechanical response. Both the geometric center coordinates and the strain concentration region indicate the location of potential damage sources from magnetic and mechanical dimensions, respectively, providing independent but complementary spatial clues for subsequent spatial matching. For example, this application can use the surface strain field distribution data output by the finite element model to filter elements with strain values ​​≥0.0015, merge adjacent elements through neighborhood search to form a continuous strain concentration region, and calculate the geometric center coordinates of each region as the location of the strain concentration region.

[0052] Optionally, this application may also perform wavelet denoising (db4 wavelet, decomposition level 4, soft thresholding) on ​​the magnetic field gradient anomaly amplitude sequence before extracting the abnormal peak points, so as to suppress the influence of environmental magnetic field interference on location determination.

[0053] Step S60: Spatially match the location of the high-energy signal, the location of the magnetic field gradient anomaly, and the location of the strain concentration region during the plastic deformation stage to determine the spatial overlap region of the multimodal signals.

[0054] The location of the high-energy signal can refer to the acoustic emission signal whose energy peak exceeds a preset value (e.g., 0.05 mJ) during the plastic deformation stage identified above. The spatial three-dimensional coordinates of each acoustic emission signal are calculated using the classic three-point positioning algorithm based on time difference positioning. For example, for a certain high-energy signal A, its arrival time differences to the three smart sensors are Δt12=120 μs, Δt13=180 μs, and Δt23=60 μs, respectively. Combining the coordinates of the smart sensor array and the wave speed of 5500 m / s, the spatial location of the high-energy signal A is obtained by iteratively solving using the least squares method as (2.45 m, 1.32 m, 0.78 m).

[0055] Spatial matching can refer to mapping the locations of high-energy signals, magnetic field gradient anomalies, and strain concentration regions within the plastic deformation stage to the same coordinate system, and then selecting a set of spatially highly coupled points based on the dual-distance constraint criterion (i.e., the distance between a certain acoustic emission location and the nearest magnetic field gradient anomaly location, and the distance between a certain location and the geometric center of the nearest strain concentration region are both less than a preset threshold). The spatially overlapping region of multimodal signals is the minimum convex hull or spherical neighborhood enclosed by all spatially matching points that meet the matching conditions. It represents the suspected core damage region where the spatial consistency of the responses of the three types of physical signals—the locations of high-energy signals, magnetic field gradient anomalies, and strain concentration regions—is the highest within the plastic deformation stage.

[0056] Optionally, this application may map the high-energy signal location, magnetic field gradient anomaly location, and strain concentration region location to a global coordinate system; for each high-energy signal location, calculate its first distance to each magnetic field gradient anomaly location and its second distance to the geometric center of each strain concentration region; select the correspondence with the smallest first distance and second distance, and determine whether the high-energy signal location simultaneously satisfies the condition that the first distance is ≤0.08m and the second distance is ≤0.08m; if satisfied, mark it as a spatial matching point.

[0057] In another embodiment, this application may also define a comprehensive bias. Where dAB, dAC, and dBC are the Euclidean distances between each pair of the high-energy signal location, the magnetic field gradient anomaly location, and the strain concentration region location, respectively; the spatial center corresponding to the combination with D≤0.08m is taken as the center of the spatial overlap region of the multimodal signal.

[0058] Furthermore, this application can also use the minimum bounding sphere algorithm to determine a spherical region with a radius of 0.10m as the overlapping region based on the spatial distribution of all spatial matching points.

[0059] Step S70: Perform multi-scale decomposition on the acoustic emission signal in the overlapping region of the multimodal signal space, and determine the cumulative degree of irreversible changes in the microstructure of the steel based on the abrupt change characteristics in the decomposed signal.

[0060] Multi-scale decomposition can refer to using wavelet transform to perform multi-resolution analysis on acoustic emission signals in overlapping regions, separating detail components at different frequency bandwidths to reveal the multi-timescale dynamic characteristics hidden in the signal.

[0061] A mutation feature can refer to a transient extreme point in the detail component whose amplitude exceeds a preset multiple (e.g., 3 times) of the standard deviation of the corresponding component. Its density (frequency of occurrence per unit time) and amplitude (growth rate relative to the baseline value) can jointly reflect the activity and severity of irreversible changes in microstructure (such as dislocation pile-up and microcrack bursting). The cumulative degree of irreversible changes in microstructure is a quantitative representation of this mutation feature, used to connect macroscopic signal response and microscopic damage evolution process.

[0062] For example, the db4 wavelet basis can be used to perform discrete wavelet transform (DWT) on the acoustic emission signal in the overlapping region, decomposing it into 5 scale detail components D1–D5. The threshold for detecting abrupt change points for each component is set to 3 times the standard deviation of the corresponding component. The time location and amplitude of the abrupt change points are extracted. The density and average amplitude of the abrupt change points are statistically analyzed, and their rate of change relative to the initial state (such as the mean of the elastic stage) is calculated.

[0063] Alternatively, this application may use Empirical Mode Decomposition (EMD) instead of wavelet transform to obtain several intrinsic mode functions (IMFs), and perform the same mutation detection process on the high-frequency IMF components.

[0064] Step S80: Based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the multimodal signal spatial overlap region, determine the detection result of the steel.

[0065] Among them, the degree of spatial overlap can refer to the degree of spatial overlap of the signal intensity fields of acoustic emission signal, metallic magnetic memory signal and surface strain data in the overlapping region. It is obtained by integrating the ratio of the minimum and maximum values ​​of the normalized intensity field on the three-dimensional grid point by point. The degree of spatial overlap can reflect the consistency level of the physical mechanism of the response of the three types of signals (acoustic emission signal, metallic magnetic memory signal and surface strain data). The higher the value, the tighter the multi-field coupling and the more reliable the damage criterion.

[0066] The test results are qualitative conclusions output by combining the aforementioned two dimensions of indicators, including the current mechanical stage (elastic / plastic) of the steel, whether damage exists, the damage level (I / II / III), and the coordinates of the damage location, thus achieving a closed-loop determination from macroscopic loading state to microscopic damage mechanism.

[0067] For example, the method can be as follows: normalize the ultrasonic longitudinal wave, transverse wave and acoustic emission signals collected in the overlapping area, and interpolate to obtain a three-dimensional scalar field; calculate C=0.68; combine with the above-obtained D=0.42; check the table to determine: 0.38≤D<0.65 and C>0.72 is the stage of local plastic concentration development (level II), but currently C=0.68<0.72, so it is downgraded to the early uniform plastic deformation stage (level I), with slight damage; output the detection result: the steel is in the early uniform plastic deformation stage, the damage level is level one, and no immediate intervention is required.

[0068] This method uses the acoustic emission signal sequence as the main carrier of dynamic response, delineates the active rise interval of the signal with the start of the plastic deformation stage as the key anchor point, and simultaneously acquires the metal magnetic memory signal and surface strain data of the same detection area. The locations of magnetic field gradient anomalies and strain concentration areas are extracted from the magnetic and mechanical dimensions, respectively. By using a spatial matching mechanism, the locations of the three types of signals are mapped to a unified coordinate system and dual distance constraints are applied to determine the spatial overlap region of the multimodal signals. Then, wavelet multi-scale decomposition is performed on the acoustic emission signals in this region to extract the abrupt change features of detail components to quantify the degree of accumulation of irreversible changes in the microstructure. Finally, the degree of accumulation is fused with the degree of spatial overlap of the intensity fields of the three types of signals to achieve cross-scale collaborative determination of the development stage of plastic deformation of steel, the existence of early damage, the severity level, and the spatial location. This solves the problems of delayed damage identification, inaccurate positioning, and misjudgment of severity caused by the limitations of single-modal perception in the prior art.

[0069] In one embodiment, a method for identifying the start time of steel entering the plastic deformation stage based on an acoustic emission signal sequence is also provided, such as... Figure 2 As shown, it includes:

[0070] Step S201: Extract the amplitude and frequency center of each acoustic emission signal in the acoustic emission signal sequence;

[0071] The amplitude can refer to the absolute value of the peak voltage in an acoustic emission signal event, representing the strength of the energy released by the event; it can be the normalized voltage amplitude obtained after pre-amplification and analog-to-digital conversion, or it can be the physical magnitude amplitude (unit: mV or dB) of the original sensor output after calibration; in this embodiment, it is used to reflect the intensity of local energy release events such as the propagation of microcracks or grain slip inside the steel.

[0072] The frequency center can refer to the weighted average frequency of the power spectral density function of the acoustic emission signal. Its calculation method, for example, is to perform amplitude-weighted averaging of the power spectrum obtained by the Fast Fourier Transform (FFT) of the signal. It can be used to characterize key frequency domain features that represent differences in the physical mechanisms of acoustic emission sources. For example, the frequency center is higher when elastic vibration dominates, while it is lower when plastic deformation is accompanied by nonlinear processes such as friction and tearing. Specifically, in this embodiment, it can be used to distinguish the spectral response characteristics of acoustic emission signals under different deformation mechanisms. For example, this application may extract the time spectrum of each acoustic emission event based on the short-time Fourier transform and calculate the frequency center based on the power spectral density weighting; or it may obtain the energy distribution of each frequency band based on wavelet packet decomposition and then determine the frequency center using the energy centroid method.

[0073] For example, this application may involve collecting acoustic emission signal sequences of steel under uniaxial tensile load at a sampling rate of 100kHz; extracting waveform segments for each valid acoustic emission event (meeting an amplitude threshold of 45dB and a HIT defined time of 200μs); performing FFT operations on the segments to obtain their power spectrum; calculating the frequency center based on the power spectrum; synchronously recording the peak voltage amplitude of the event; thereby constructing a two-dimensional dataset containing the amplitude-frequency center of all events, with a total of N sample points, for subsequent cluster analysis.

[0074] Step S202: Based on the distribution characteristics of the amplitude and frequency center of each acoustic emission signal, perform cluster analysis to divide each acoustic emission signal into elastic stage signals and plastic deformation stage signals.

[0075] The distribution characteristics can refer to the scatter distribution pattern of amplitude and frequency centers on a two-dimensional plane, including cluster center position, intra-cluster dispersion, inter-cluster separation degree and overall contour trend, etc.; it can be used to reflect the macroscopic characterization of the statistical characteristics of acoustic emission signals under different physical mechanisms; in this embodiment, the distribution characteristics are used to support the feasibility basis for unsupervised division of two types of signals.

[0076] Cluster analysis can refer to a statistical learning method that automatically groups samples based on similarity measures without pre-defined labels; for example, k-means clustering, Gaussian mixture model, or hierarchical clustering; for example, this application uses k-means clustering and sets the number of clusters k=2 to achieve a binary division of elastic and plastic mechanism signals.

[0077] The elastic stage signal and the plastic deformation stage signal are two types of labels assigned by the clustering results, corresponding to the acoustic emission events generated by steel in the linear elastic response and nonlinear plastic response stages, respectively. Among them, the elastic stage signal has a joint distribution characteristic of relatively low amplitude and high frequency center, while the plastic deformation stage signal has a joint distribution characteristic of relatively high amplitude and low frequency center.

[0078] For example, this application may use the k-means clustering algorithm, with Euclidean distance as the similarity measure, to iteratively optimize the amplitude-frequency center two-dimensional dataset until the cluster centers converge to achieve cluster analysis; this application may also use a Gaussian mixture model to fit the probability density of the data distribution, and estimate the latent variables through the expectation-maximization (EM) algorithm to achieve cluster analysis, etc.

[0079] For example, this application may perform k-means clustering on the above two-dimensional dataset, with the initial cluster centers set near (30dB, 300kHz) and (70dB, 120kHz), respectively; after 5 rounds of iteration, convergence is achieved, resulting in two clusters: cluster A is centered at (42dB, 265kHz), and cluster B is centered at (78dB, 105kHz); combined with knowledge of materials mechanics and calibration experiments, cluster A corresponds to the elastic stage signal, and cluster B corresponds to the plastic deformation stage signal; all N events are assigned to the two classes according to the nearest cluster center principle to form a labeled time series.

[0080] Step S203: Determine the starting time of the plastic deformation stage based on the trend of the first significant appearance and persistence of plastic deformation stage-type signals in the time series.

[0081] Among them, the first significant and sustained trend can refer to the process characteristic of the plastic deformation stage signal changing from sporadic and occasional to stable and dominant in the time dimension, which is reflected in the fact that the proportion of this type of signal in the continuous sliding window exceeds the preset threshold and maintains a stable increase; it can be a robust judgment criterion to overcome the influence of single event misjudgment, noise interference and transient fluctuations; in this embodiment, it can be used to transform the category labels obtained by clustering into engineering criteria with clear time coordinates.

[0082] For example, this application can statistically determine the proportion of plastic deformation stage signals based on a sliding time window (10 consecutive acoustic emission events). When the proportion first exceeds 0.65 and the subsequent five consecutive windows are not lower than 0.60, the start time of the first qualifying window is recorded as the start time. Alternatively, this application can use a cumulative counting method, setting the cumulative number of plastic events to eight, with at least six of them appearing within the last 15 events, and then using the eighth event as the start time. Furthermore, this application can introduce a weighted moving average filter to smooth the category label sequence and detect the rising edge inflection point, using the corresponding time as the start time, without limitation.

[0083] For example, this application may sort the above-obtained labeled time series according to the original acquisition time; set the sliding window length to 10 events and the step size to 1; calculate the proportion of signals in the plastic deformation stage window by window; when the proportion of the i-th window is 0.70, and the proportions of the i+1 to i+4 windows are 0.68, 0.72, 0.75, and 0.73 respectively, it is determined that the condition of first significant occurrence and persistence is met; take the occurrence time t0 of the first event in the i-th window as the starting time of the plastic deformation stage; the loading stress corresponding to this time is about 252 MPa, which is slightly higher than the standard yield strength of 250 MPa, which is in line with the constitutive response law of the material.

[0084] This application constructs a two-dimensional feature space that reflects the differences in the internal deformation mechanism of steel by extracting the amplitude and frequency center of acoustic emission signals. Using cluster analysis, it achieves an unsupervised, physically driven division of elastic stage signals and plastic deformation stage signals. Furthermore, based on the trend of the first significant appearance and persistence of plastic deformation stage signals in the time series, the abstract category determination is transformed into an engineering critical point with a clear time coordinate. The synergy of these three methods utilizes the essential physical differences in amplitude and spectrum of acoustic emission signals while avoiding the susceptibility of single threshold criteria to noise interference. Thus, without relying on external stress / strain synchronization signals, it robustly and self-consistently identifies the starting moment of the steel's transition from elastic to plastic, providing a reliable time reference for subsequent focusing on the active range of plastic stage signals and conducting multimodal collaborative analysis.

[0085] In another optional embodiment, the application further includes the step of determining the time interval during which the activity of the acoustic emission signal increases in the acoustic emission signal sequence within the plastic deformation stage, such as... Figure 3 As shown:

[0086] Step S301: Using the start time of the plastic deformation stage as the boundary, the acoustic emission signal sequence is divided to obtain the signal segment of the plastic deformation stage.

[0087] The starting time of the plastic deformation stage is determined by the critical time point based on the trend of the first significant appearance and persistence of the plastic deformation stage class signal in the time series obtained by cluster analysis in this application.

[0088] The signal segment during the plastic deformation stage can refer to the continuous acoustic emission signal data segment from the start time to the end time of the loading test.

[0089] This application extracts the original acoustic emission signal sequence based on the starting time, ensuring that subsequent processing focuses only on the response signal after the steel has entered the plastic deformation state, eliminating interference signals in the elastic and unloading stages, thereby improving the time relevance and physical consistency of multimodal data acquisition.

[0090] Step S302: Perform wavelet transform decomposition on the signal segment of the plastic deformation stage to extract high-energy signals;

[0091] Among them, wavelet transform decomposition can refer to using discrete wavelet transform to perform multi-scale frequency band separation on signal segments in the plastic deformation stage, so as to enhance the signal-to-noise ratio and retain transient impact characteristics;

[0092] High-energy signals can refer to acoustic emission event waveform segments in the high-frequency detail components obtained by wavelet decomposition whose instantaneous energy amplitude exceeds a preset energy threshold.

[0093] For example, this application may use the energy peak value of the third level detail component after wavelet transform to make a threshold judgment, and identify events with energy values ​​greater than 0.05mJ as high-energy signals; this application may also use the maximum energy density value in the frequency band with the minimum energy entropy of each sub-frequency band after wavelet packet decomposition to make a dynamic threshold setting to identify high-energy signals, and there is no limitation here.

[0094] Step S303: Based on the count rate change trend of the high-energy signal, determine the time interval of the increase in the activity level of the acoustic emission signal.

[0095] Among them, the trend of the count rate can refer to the statistical law of monotonically increasing or accelerating growth of the number of high-energy signals occurring within a unit time window over time.

[0096] The time interval during which the activity level of acoustic emission signals increases can refer to the time span corresponding to the trend of count rate change meeting the preset increase judgment condition.

[0097] For example, this application can count the number of high-energy signals by using 100 sampling points as a sliding window, plot the count rate-time curve, and determine whether its slope is greater than 0.1 times / second by linear fitting. If it is satisfied, the time period between the fitting start point and the end point is determined as the rising interval. This application can also determine the starting time of the first window that meets the condition as the rising start point and the ending time of the last window that meets the condition as the rising end point based on the fact that the increase of the count rate in three consecutive windows exceeds 20% of the previous window and the absolute increment is ≥2 times / window. Furthermore, this application can also determine the starting and ending boundaries of the rising interval by combining local weighted regression (LOWESS) to identify the trend inflection point based on the ratio of the mean to the standard deviation of the first difference of the count rate sequence (i.e., the coefficient of variation) being greater than 1.5.

[0098] For example, in a uniaxial tensile test of a Q345B steel plate, this application determines the onset time of plastic deformation as 8.72 seconds after loading; the acoustic emission signal segment from 8.72s to 25.00s is taken as the signal segment of the plastic deformation stage; the signal segment is decomposed into 5 layers of db4 wavelets, and 127 events with energy peak values ​​>0.05mJ are extracted from the third layer detail component; the count rate is statistically analyzed with 100 points (corresponding to 1ms at a sampling rate of 100kHz), and it is found that the count rate increases continuously from 4.2 times / second from 11.35s to 14.8 times / second at 16.89s, with a linear fitting slope of 0.13 times / second, which meets the rise judgment condition; thus, the time interval for the increase in the activity of the acoustic emission signal is determined to be from 11.35s to 16.89s; within this interval, the metal magnetic memory sensor and digital image correlation (DIC) system are synchronously triggered to realize the spatiotemporal aligned acquisition of multimodal data during the active damage period.

[0099] This application ensures the physical purity of the object being analyzed by segmenting the signal based on the start time of the plastic deformation stage; it enhances the robustness of high-energy signal recognition by using wavelet transform decomposition, avoiding interference from background noise and low-amplitude elastic events; and it determines the active rising interval based on the trend of count rate changes rather than a fixed duration or absolute threshold, making the time window adaptive and adaptable to different loading rates, sample sizes, and material yield characteristics. This provides a reliable time basis for the accurate and synchronous acquisition of metal magnetic memory signals and surface strain data in the same detection area.

[0100] In yet another optional embodiment, this application further includes a step of determining the location of anomalies in the magnetic field gradient based on the metal magnetic memory signal, as shown in Figure 4, comprising:

[0101] Step S401: Calculate the magnetic field gradient of the metal magnetic memory signal along the spatial direction and synthesize the comprehensive gradient anomaly amplitude sequence.

[0102] The magnetic field gradient can refer to the rate of change of the magnetic field intensity component in the spatial coordinate direction of the metal magnetic memory signal, used to characterize the degree of local magnetic field distortion. The spatial direction includes two orthogonal dimensions along the detection surface: the horizontal (x-direction) and the vertical (y-direction). The magnetic field gradient is obtained by performing finite difference operations on the magnetic field intensity components Hx and Hy at each spatial sampling point in the metal magnetic memory signal, for example, Kx=(Hx(i+1)-Hx(i)) / Δx, Ky=(Hy(j+1)-Hy(j)) / Δy, where Δx and Δy are the distance between adjacent sensor probes. The comprehensive gradient anomaly amplitude sequence can be a scalar sequence obtained by vector synthesis of the gradient components in the x and y directions, for example, K=√(Kx²+Ky²), whose physical meaning is the comprehensive intensity of the change in magnetic field intensity per unit length. The comprehensive gradient anomaly amplitude sequence can be used to eliminate the directional dependence of a single-direction gradient on anomaly identification and improve the response consistency to multi-directional stress concentration areas. For example, this application can perform vector synthesis based on the finite difference gradients of the magnetic field intensity components in the x and y directions to obtain a comprehensive gradient anomaly amplitude sequence.

[0103] Optionally, this application further performs differential gradient calculation based on the smoothed filtering of the magnetic field intensity components in the x and y directions, and then weights and synthesizes a comprehensive amplitude sequence.

[0104] Furthermore, after performing wavelet denoising on the gradient components in the x and y directions, this application can also perform differential gradient calculation and amplitude synthesis to suppress high-frequency noise interference.

[0105] For example, this application uses a triaxial magnetic gradient sensor array (probe spacing 5mm) to synchronously acquire Hx and Hy component data of the steel surface; the Hx sequence is forward-differentiated along the x direction with a step size of 5mm to obtain the Kx sequence; the Hy sequence is forward-differentiated along the y direction with a step size of 5mm to obtain the Ky sequence; the square root of the sum of squares of the corresponding points of Kx and Ky is used to generate the K sequence; the amplitude of this sequence is concentrated in the undamaged area at 0.1–0.3A / (m·mm), while near dislocation pile-up or microcrack tip, the amplitude jumps to above 0.8A / (m·mm), forming obvious abnormal peaks.

[0106] Step S402: Extract the target comprehensive gradient anomaly amplitude within the anomaly range from the comprehensive gradient anomaly amplitude sequence, and use the spatial coordinates of the metal magnetic memory signal corresponding to the target comprehensive gradient anomaly amplitude as the magnetic field gradient anomaly location.

[0107] The abnormal range can refer to a preset amplitude threshold range used to determine whether the magnetic field gradient deviates significantly from the normal background level. Its setting can be based on the statistical gradient distribution of non-destructive samples of the same material and under the same working conditions. The target comprehensive gradient abnormal amplitude is the amplitude point in the comprehensive gradient abnormal amplitude sequence that is greater than the preset threshold, for example, K>0.8A / (m·mm). The magnetic field gradient abnormal position is the spatial coordinate of the original metal magnetic memory signal acquisition point corresponding to the target comprehensive gradient abnormal amplitude, that is, the (x, y) position of the point in the rectangular coordinate system of the detection surface. The magnetic field gradient abnormal position can be used for subsequent spatial matching with the position of acoustic emission high-energy signal and the position of strain concentration area. Its positioning accuracy directly affects the reliability of the determination of the spatial overlap area of ​​multimodal signals.

[0108] This application takes the set of peak points that continuously exceed the threshold in the comprehensive gradient anomaly amplitude sequence as the target amplitude and takes the coordinates of its geometric center as the magnetic field gradient anomaly location. This application also takes the coordinates of the starting point of the stable high amplitude segment that exceeds the threshold and lasts for a longer than a preset duration (such as 2 sampling periods) as the magnetic field gradient anomaly location. Furthermore, this application performs cluster analysis (such as DBSCAN) on all discrete points whose amplitude exceeds the threshold and takes the centroid coordinates of the cluster with the highest density as the magnetic field gradient anomaly location, without limitation.

[0109] For example, this application performs sliding window peak detection (window length 3 points, minimum peak interval 5 points) on the K sequence, identifying a total of 7 anomalous peaks with amplitude > 0.8 A / (m·mm), corresponding to coordinates (2.15m, 1.03m), (2.20m, 1.05m), (2.22m, 1.06m), (2.45m, 1.32m), (2.47m, 1.34m), (2.48m, 1.35m), and (2.70m, 1.68m). The first 3 points are clustered to obtain the first cluster centroid (2.19m, 1.05m), the middle 3 points are clustered to obtain the second cluster centroid (2.47m, 1.34m), and the last point is clustered separately. Finally, the two cluster centroids are selected as two independent magnetic field gradient anomaly locations for subsequent spatial matching.

[0110] This application enhances the sensitivity and directional robustness to local magnetic field distortion by calculating the differential gradients of the metal magnetic memory signal in the x and y directions and synthesizing a comprehensive gradient anomaly amplitude sequence. Furthermore, by setting the anomaly range and extracting the corresponding spatial coordinates, it achieves stable localization of the magnetic field gradient anomaly position from the original low signal-to-noise ratio magnetic signal. This provides an accurate and reproducible input data basis for spatially matching the position of the high-energy signal, the magnetic field gradient anomaly position, and the strain concentration region position during the plastic deformation stage in this application.

[0111] Optionally, this application also provides a method and steps for determining the location of strain concentration regions based on surface strain data, such as... Figure 5 As shown, it includes:

[0112] Step S501: Establish a finite element model of the steel and apply loads and boundary conditions corresponding to the actual stress conditions to solve the problem and obtain surface strain field distribution data.

[0113] Among them, the finite element model can refer to a mathematical model that discretizes the geometry of steel into a finite number of elements and nodes, and solves its displacement, stress and strain response under a given mechanical boundary through numerical methods; the load and boundary conditions can refer to the external force distribution (such as uniformly distributed pressure, concentrated force), constraint state (such as fixed end, simply supported end) and material constitutive parameters (such as elastic modulus, Poisson's ratio) set according to the actual loading conditions, and their values ​​are consistent with the measured or calibrated results.

[0114] In this embodiment, the finite element model is used to reproduce the surface deformation behavior of steel under actual stress. Its output surface strain field distribution data is the physical basis for subsequent identification of strain concentration regions. This data is organized in a two-dimensional or three-dimensional mesh, with each mesh element corresponding to a spatial coordinate and a set of strain components (such as ε). x ε γ The spatial resolution of the modeling mesh (εz or principal strain ε1) is directly related to the modeling mesh size, and the overall structure satisfies the continuity and balance constraints.

[0115] For example, this application may construct a quadrilateral or hexahedral element mesh based on the geometric dimensions and material parameters of the steel, use linear elastic or elastoplastic constitutive relations, apply full constraints on the left side and uniformly distributed loads on the right side to perform static solutions, and obtain surface strain field distribution data; this application may also construct a transient dynamic model based on measured load time history curves and support reaction force data and perform time-step solutions to obtain dynamic surface strain field distribution data, etc.

[0116] Step S502: Filter target strain field distribution data whose strain values ​​exceed a preset threshold from the surface strain field distribution data;

[0117] Among them, the preset threshold can refer to the empirical or statistical criteria used to distinguish between normal strain response and abnormal strain concentration. Its setting can be based on material yield strain, distribution of historical test data, simulation error range and multi-modal signal collaborative positioning requirements; the strain value can be any one of the strain component in a single direction, equivalent strain, maximum principal strain or strain gradient amplitude.

[0118] In this embodiment, a preset threshold is used to exclude local false high values ​​caused by mesh distortion, boundary disturbance or numerical noise, to ensure that the screening results reflect the true mechanical concentration behavior. This threshold is related to the steel material properties and loading level, and its value does not constitute a limitation on the protection range, but only serves as a triggering condition for the screening operation.

[0119] For example, this application can set a preset threshold of 80% of the material's yield strain, compare the maximum principal strain of all elements in the surface strain field distribution data point by point, and retain elements with strain values ​​≥0.0015, along with their coordinates and strain values. Alternatively, this application can be based on the statistical histogram of the surface strain field distribution data, selecting the strain value corresponding to the 95th percentile of the cumulative probability as a dynamic preset threshold to achieve adaptive filtering. Furthermore, this application can combine the time information of the acoustic emission active interval, extract the transient strain peak value within the corresponding time period, and use 70% of this peak value as a time-varying preset threshold for filtering.

[0120] Step S503: Cluster the selected adjacent target strain field distribution data to form continuous strain concentration regions, and calculate the geometric center coordinates of each strain concentration region as the location of the strain concentration region.

[0121] Among them, the strain concentration region can refer to a connected region composed of multiple spatially adjacent grid cells whose strain values ​​all exceed a preset threshold. Its shape reflects the local stress redistribution characteristics. The geometric center coordinates can refer to the arithmetic mean of the center coordinates of all cells in the connected region. It is used to characterize the spatial anchor point of the region and support subsequent spatial matching with the acoustic emission signal position and the magnetic field gradient anomaly position in a unified coordinate system.

[0122] In this embodiment, the clustering operation is achieved through neighborhood search. That is, for any target unit, if there is at least one other target unit in its eight-neighborhood (two-dimensional) or twenty-six-neighborhood (three-dimensional), it is included in the same connected region. This operation excludes isolated high-value points and ensures that the identified region has mechanical continuity and spatial extensibility. The calculation of the geometric center coordinates does not depend on the regularity of the region shape and is applicable to strain concentration areas of any topological structure.

[0123] For example, this application may employ a four-connected or eight-connected image processing algorithm to label the connected components of the two-dimensional surface strain field distribution data, identify all independent strain concentration regions, and calculate their geometric center coordinates respectively. Alternatively, this application may define adjacency relationships based on a spatial distance threshold (such as 0.5 times the element size), construct an adjacency graph for the three-dimensional strain field distribution data, and perform depth-first traversal to complete clustering and identify strain concentration regions. Furthermore, this application may introduce strain gradient constraints to further eliminate gradient abrupt change points within the connected regions, retain sub-regions with good gradient continuity, and then calculate the geometric center.

[0124] For example, this application can be used to apply a uniaxial tensile load of 250MPa to a Q345B steel plate with dimensions of 1000mm×500mm×10mm on a universal testing machine; establish its two-dimensional plane stress finite element model, divide it into 40,000 quadrilateral elements of 5mm×5mm; apply full constraint on the left and uniform strain on the right, and solve to obtain the surface principal strain distribution matrix; set a preset threshold of 0.0015, and filter out 127 elements with strain values ​​≥ the threshold; through eight-neighbor connectivity analysis, merge into 3 continuous strain concentration regions, whose geometric center coordinates are (247mm, 135mm), (682mm, 210mm), and (895mm, 342mm), respectively; the above coordinates will be mapped to the global coordinate system together with the acoustic emission positioning coordinates and magnetic memory anomaly coordinates for performing the spatial matching operation in Example 6.

[0125] This application obtains surface strain field distribution data by constructing a physically accurate finite element model. Combined with preset threshold screening and neighborhood clustering, it effectively distinguishes between real strain concentration areas and numerical artifacts or measurement noise. By using the geometric center coordinates as a unified spatial representation, it ensures the coordinate consistency and physical comparability of this area with the location of high-energy acoustic emission signals and the location of magnetic field gradient anomalies in subsequent spatial matching. This supports the accurate determination of the spatial overlap area of ​​multimodal signals in Example 1, and improves the reliability and engineering applicability of early damage location of steel.

[0126] In one optional embodiment, this application also provides an optional method embodiment for determining the spatially overlapping region of multimodal signals, such as... Figure 6 As shown, it includes:

[0127] Step S601: Map the location of the high-energy signal, the location of the magnetic field gradient anomaly, and the location of the strain concentration region to the same spatial coordinate system;

[0128] Among them, the same spatial coordinate system can refer to a unified three-dimensional rectangular coordinate system constructed to achieve spatial alignment of multi-source heterogeneous signals. Its origin can be set at the geometric center of the steel detection area, the x-axis is along the length of the steel, the y-axis is along the width, and the z-axis is perpendicular to the surface and points outward.

[0129] The same spatial coordinate system can be a reference coordinate system obtained by normalizing the original coordinate systems of each signal source through a coordinate transformation matrix;

[0130] In this embodiment, the location of the high-energy signal can be obtained by three-dimensional spatial coordinates (xe, ye, ze) through the three-point time difference positioning method, with its original coordinate system taking the geometric center of the acoustic emission sensor array as the origin; the location of the magnetic field gradient anomaly can be the coordinates (xm, ym, 0) directly output in the sensor physical coordinate system after being collected by the triaxial magnetic gradient sensor array and calculated by spatial difference, with its z coordinate defaulting to 0 (because the magnetic sensor is attached to the steel surface); the location of the strain concentration area can be the geometric center coordinates (xs, ys, 0) formed by clustering in the surface strain field distribution data obtained by solving the finite element model, with its original coordinate system consistent with the finite element modeling coordinate system.

[0131] For example, this application may map the three types of position data to a global coordinate system with the lower left corner of the steel as the origin and millimeters as the unit, based on the coordinate system calibration parameters and the rigid body transformation relationship.

[0132] Step S602: For each high-energy signal, calculate the first distance between the high-energy signal and the target magnetic field gradient anomaly location, and the second distance between the high-energy signal and the target strain concentration area location. The distance between the target magnetic field gradient anomaly location and the high-energy signal location is less than the distance between any other magnetic field gradient anomaly location and the high-energy signal location. The distance between the high-energy signal locations in the target strain concentration area location is less than the distance between any other strain concentration area location and the high-energy signal location.

[0133] The first distance can refer to the spatial distance between the location of the high-energy signal and the location with the smallest Euclidean distance among all the locations of magnetic field gradient anomalies.

[0134] The second distance can refer to the spatial distance between the location of the high-energy signal and the location of the strain concentration region with the smallest Euclidean distance among all the locations of the strain concentration regions.

[0135] The target magnetic field gradient anomaly location can be the single magnetic field gradient anomaly location with the smallest Euclidean distance to the current high-energy signal location in the set of all magnetic field gradient anomaly locations; the target magnetic field gradient anomaly location can be the spatial coordinates corresponding to the amplitude exceeding the anomaly range in the comprehensive gradient anomaly amplitude sequence of the metallic magnetic memory signal.

[0136] The location of the target strain concentration region can be the geometric center coordinates of the single strain concentration region with the smallest Euclidean distance to the current high-energy signal location in the set of all strain concentration region locations; the location of the target strain concentration region can be the geometric center of the continuous region obtained by clustering adjacent target strain field distribution data selected in the finite element model.

[0137] This application may be based, for example, on the Euclidean distance formula. For each high-energy signal location (xe, ye, ze), traverse all magnetic field gradient anomaly locations (xm1, ym1, 0), (xm2, ym2, 0)..., calculate the distance to each point and take the minimum value as the first distance.

[0138] This application may also employ a kd-tree index structure to accelerate nearest neighbor search, completing nearest neighbor matching between a single high-energy signal location and all magnetic field gradient anomaly locations within milliseconds.

[0139] Furthermore, this application can also combine a spatial hash table to grid the locations of magnetic field gradient anomalies, and after eliminating obviously distant candidate points in the coarse screening stage, perform precise distance calculation.

[0140] Step S603: The location of the high-energy signal whose first distance and second distance are both less than the preset distance threshold is determined as the spatial matching point;

[0141] The preset distance threshold can be an empirical or calibrated upper limit of distance used to determine whether the three types of signals have the possibility of spatial coupling; the preset distance threshold is 8mm; the preset distance threshold is a fixed value set based on the sensor spatial resolution, steel thickness, signal propagation attenuation characteristics and engineering fault tolerance requirements.

[0142] In this embodiment, a preset distance threshold is used to constrain the maximum acceptable deviation between the high-energy signal and the abnormal location of the magnetic field gradient and the location of the strain concentration area, reflecting the spatial local consistency of the three types of physical signals in the damage evolution process.

[0143] This application could, for example, compare the first distance and the second distance with preset distance thresholds respectively, and only if both simultaneously satisfy the threshold... and Only when the high-energy signal location is recorded as a spatial matching point; for example, this application may further introduce a time synchronization window (e.g., ±30ms) as an auxiliary criterion to improve matching robustness, based on satisfying the dual distance constraint; furthermore, this application may also adopt a weighted distance joint criterion to define a comprehensive matching score. ,when It is determined to be a spatial matching point.

[0144] Step S604: Based on the spatial distribution of all spatial matching points, determine the spatial overlap region of the multimodal signals.

[0145] The spatial overlap region of the multimodal signal can refer to the minimum connected geometric region enclosed by the set of spatial points composed of all spatial matching points after density clustering; the spatial overlap region of the multimodal signal can be a spherical, ellipsoidal, or convex hull region, etc.; for example, the spatial overlap region of the multimodal signal is a spherical region with a radius of 0.10m centered on the peak density of the spatial matching points.

[0146] For example, this application may employ the DBSCAN density clustering algorithm, using the three-dimensional coordinates of spatial matching points as input, setting the neighborhood radius ε=0.08m and the minimum number of points MinPts=3, to identify the core point cluster, and using its minimum circumscribed sphere as the spatial overlapping region of the multimodal signal; this application may also generate a three-dimensional spatial density distribution map based on kernel density estimation (KDE), and extract the volume enclosed by isosurfaces with a density greater than 0.75 times the peak value as the overlapping region; further, this application may also project the spatial matching points onto the xy plane for two-dimensional clustering, and then extend along the z direction with a fixed thickness (e.g., 2mm) to form a columnar overlapping region, which is not limited here.

[0147] For example, in a uniaxial tensile test of steel, the acoustic emission system locates five high-energy signal positions: (2.45, 1.33, 0.79), (2.48, 1.36, 0.81), (2.46, 1.34, 0.80), (2.47, 1.35, 0.80), and (2.49, 1.37, 0.82) (unit: m); three magnetic field gradient anomaly positions: (2.46, 1.35, 0), (2.81, 1.62, 0), and (1.93, 0.77, 0); and two strain concentration regions: (2.47, 1.35, 0). (3.12, 1.89, 0); After applying coordinate system one, the first and second distances from each high-energy signal location to the nearest magnetic field gradient anomaly location and the nearest strain concentration area location are calculated. It is found that the first four high-energy signal locations all satisfy the condition that the first distance is ≤0.042m and the second distance is ≤0.041m (the preset distance threshold is 0.08m), and are therefore identified as spatial matching points. DBSCAN clustering (ε=0.08m) is performed on these four matching points to identify a single core cluster with a minimum circumscribed sphere center of (2.47, 1.35, 0.80) and a radius of 0.045m, which is the multimodal signal spatial overlap region determined in this embodiment.

[0148] This application maps three types of heterogeneous signals to the same spatial coordinate system, uses the nearest neighbor strategy to obtain the first and second distances corresponding to the location of each high-energy signal, and uses the dual-distance joint constraint to screen out spatial matching points. Finally, it determines the spatial overlap region of multimodal signals based on the spatial distribution density of the matching points. This scheme avoids the computational explosion problem caused by full combination matching and strengthens the spatial coupling logic of the three types of signals in terms of damage physics mechanism through dual distance constraints. It ensures that the determined overlap region truly reflects the source of irreversible changes in the microstructure driven by plastic deformation inside the steel, and provides a reliable spatial anchoring basis for subsequent multi-scale decomposition and abrupt change feature analysis of acoustic emission signals in this region.

[0149] In another alternative embodiment, this application also provides a method embodiment for multi-scale decomposition of acoustic emission signals in a spatially overlapping region of multimodal signals, and for determining the cumulative degree of irreversible changes in the microstructure of steel based on abrupt change characteristics in the decomposed signals. This method embodiment includes the following steps:

[0150] Step S701: Perform wavelet multi-scale decomposition on the acoustic emission signal in the overlapping region of the multimodal signal space to obtain detail components at multiple scales;

[0151] Wavelet multi-scale decomposition can refer to using orthogonal or bioorthogonal wavelet basis functions to perform layer-by-layer low-pass and high-pass filtering and downsampling on a time-series signal, thereby separating approximate components and detail components that characterize different time-frequency local features. Optionally, wavelet multi-scale decomposition can be a multi-resolution analysis method based on discrete wavelet transform, which can scale the original signal according to a scale factor of 2. n The signal is decomposed step by step, which suppresses noise interference while preserving the signal energy distribution characteristics.

[0152] In this embodiment, wavelet multi-scale decomposition can be used to decompose acoustic emission signals focused on the overlapping region of multimodal signals into multiple detail components that reflect the contributions of different physical mechanisms. For example, high-frequency detail components mainly correspond to transient energy release caused by dislocation motion, mid-frequency detail components mainly correspond to stress disturbances during microcrack initiation and propagation, and low-frequency detail components mainly correspond to plastic flow characteristics in the macroscopic yielding stage.

[0153] This application may obtain detailed components at multiple scales by means of wavelet basis function selection and decomposition level setting; this application may also obtain detailed components at multiple scales by means of wavelet packet decomposition; furthermore, this application may also obtain detailed components at multiple scales by means of improving wavelet reconstruction strategy, and there is no limitation here.

[0154] Step S702: Detect mutation points in each detail component and record the density and amplitude of each mutation point;

[0155] Among them, the mutation point can refer to the position where the amplitude of the detail component changes significantly in the time series. Its mathematical representation is that the absolute value of the local first derivative exceeds the preset threshold, or the local standard deviation increases abruptly.

[0156] A mutation point can be a momentary energy jump, polarity reversal, or the starting point of a sustained oscillation that occurs at a certain moment in a signal. In acoustic emission signals, it corresponds to irreversible events such as dislocation pile-up breakthrough, grain boundary slip initiation, or microvoid nucleation in the microstructure.

[0157] In this embodiment, mutation points can be used as observable proxy indicators of irreversible changes in microstructure. Their density can reflect the frequency of irreversible events per unit time, and their amplitude can reflect the energy intensity released by a single event.

[0158] The mutation point density can refer to the number of samples identified as mutation points within a unit time window, such as counting the number of mutation points per second with a window of 0.1 seconds; the mutation point amplitude can refer to the deviation of the amplitude of the detail component signal at the mutation point from the mean of that component, such as taking the absolute deviation value or the normalized relative amplitude.

[0159] This application can identify abrupt change points, for example, using a local extremum detection method based on a standard deviation multiple threshold (e.g., 3 times the standard deviation); it can also identify abrupt change points using a sliding window energy ratio (current window energy / previous reference window energy) abrupt change criterion; further, it can identify abrupt change points using a wavelet mode maxima tracking method. This application can obtain the spatiotemporal location, density, and amplitude information of abrupt change points based on any of the above methods, without specific limitations.

[0160] Step S703: Determine the cumulative degree of irreversible changes in the microstructure of the steel based on the changes in the density and amplitude of each mutation point relative to the baseline value.

[0161] The benchmark value can refer to the statistical mean of the density and amplitude of abrupt change points obtained by the same wavelet multi-scale decomposition and abrupt change detection process for acoustic emission signals collected from the same detection area when the steel is in the elastic deformation stage and no obvious plastic damage has occurred. The benchmark value can be a pre-calibrated reference dataset, which can be a control sample with the same material, the same geometric size, and the same loading conditions but not in the plastic stage, or it can be the real-time statistical result of the elastic stage signal segment in the current detection process.

[0162] In this embodiment, the reference value can be used to eliminate measurement deviations caused by differences in the sensitivity of smart sensors, environmental noise levels, and system gain fluctuations, so that the density and amplitude of abrupt change points are comparable and stable.

[0163] The cumulative degree of irreversible changes in microstructure can be a comprehensive quantitative index that reflects the gradual evolution of internal crystal defects (such as dislocation density, void volume fraction, and local equivalent plastic strain increment) in steel as the loading process progresses. Its value is positively correlated with the growth rate of the density of abrupt change points relative to the baseline value and the growth rate of the amplitude of abrupt change points relative to the baseline value.

[0164] This application can determine the cumulative degree of irreversible changes in the microstructure of steel by, for example, by weighted summation of the growth rate of mutation point density and the growth rate of mutation point amplitude; it can also determine the cumulative degree of irreversible changes in the microstructure of steel by the deviation of the product of mutation point density and amplitude from a baseline product; further, it can determine the cumulative degree of irreversible changes in the microstructure of steel by the integral change trend of mutation point density and amplitude over time. This application obtains a cumulative degree index characterizing the intensity and rate of microstructural damage evolution based on any of the above methods.

[0165] For example, this application may collect acoustic emission signals in the overlapping region of multimodal signal space, with a sampling frequency of 100kHz, a total duration of 10s, and a total of 1,000,000 sampling points; use the db4 wavelet basis to perform a 5-level discrete wavelet transform to obtain 5 detail components from D1 to D5; calculate the standard deviation of each detail component, and set the mutation point detection threshold to 3 times the standard deviation of the corresponding component; in the D3 component, the mutation point density is detected to increase from the elastic stage baseline value of 5 points / second to 8 points / second, an increase of 60%, and the average amplitude of mutation points increases from the baseline value of 0.09 to 0.16, an increase of 78%; based on this, the weighted sum of the mutation point density growth rate and amplitude growth rate (with weights of 0.4 and 0.6 respectively) is calculated, and the cumulative degree of irreversible changes in microstructure is obtained as 0.70; this value is higher than the critical threshold of 0.35, indicating that the steel has entered the irreversible plastic evolution stage.

[0166] This application extracts multi-scale detail components through wavelet multi-scale decomposition, and uses the changes in the density and amplitude of abrupt change points relative to the benchmark value to jointly characterize the evolution state of irreversible changes in the microstructure. Based on this, it utilizes the response characteristics of multi-scale components to different physical mechanisms to map macroscopically observable signal abrupt changes to microscopic damage processes such as dislocation motion, microcrack propagation, and macroscopic yielding. Furthermore, it relies on benchmark value correction to eliminate system bias, ensuring that the cumulative degree index has reproducibility and comparability across equipment and operating conditions. Ultimately, this supports the reliable implementation of the overall technical logic of determining the detection result based on the cumulative degree and spatial overlap of irreversible changes in the microstructure, as defined in this application.

[0167] In one embodiment, this application also provides an optional method embodiment for determining the detection results of steel based on the cumulative degree of irreversible changes in the microstructure of the steel and the spatial overlap degree of the spatially overlapping region of the multimodal signal, such as... Figure 8 As shown, an embodiment of this method includes the following steps:

[0168] Step S801: Calculate the quantitative index of microstructure change based on at least one of the parameters among the local equivalent plastic strain increment, dislocation density, and void volume fraction in the cumulative degree of irreversible changes in the microstructure of steel.

[0169] Among them, the local equivalent plastic strain increment can refer to the increment of equivalent plastic strain caused by irreversible deformation mechanisms such as lattice slip within the local micro-element of the material during the stress process of steel. It can be obtained through finite element inversion, digital image correlation method, or empirical mapping relationship based on acoustic emission energy calibration; dislocation density can refer to the total length of movable dislocation lines per unit volume. It can be obtained through transmission electron microscopy (TEM) statistics, X-ray diffraction line width analysis (XRD), or inversion based on acoustic emission signal spectral characteristics; void volume fraction can refer to the ratio of the total volume of micropores inside the material to the total volume of the region. It can be obtained through scanning electron microscopy (SEM) fracture morphology analysis, ultrasonic attenuation coefficient inversion, or derivation based on multi-scale acoustic emission abrupt change characteristics modeling.

[0170] The above three parameters characterize three typical irreversible evolution mechanisms of steel microstructure during the plastic deformation stage: the local equivalent plastic strain increment reflects the degree of localization of macroscopic plastic flow, the dislocation density reflects the active level of crystal defect proliferation, and the void volume fraction reflects the initial state of micro-damage initiation and propagation.

[0171] In this embodiment, the local equivalent plastic strain increment, dislocation density, and void volume fraction are used as input parameters to construct a comprehensive quantitative expression of irreversible changes in microstructure; at least one of the three is actually used to form the basis for the calculation of the quantitative index of microstructure changes.

[0172] For example, this application may calculate the quantitative index of microstructure change based on the numerical value of the local equivalent plastic strain increment and a preset weighting coefficient; this application may also calculate the quantitative index of microstructure change through a weighted linear combination based on the joint functional relationship between dislocation density and void volume fraction; furthermore, this application may also calculate the quantitative index of microstructure change using a nonlinear mapping model based on the synergistic response relationship among the local equivalent plastic strain increment, dislocation density, and void volume fraction, without limitation.

[0173] Step S802: Calculate the degree of spatial overlap based on the overlap of the normalized intensity fields of at least two of the acoustic emission signal, ultrasonic longitudinal wave signal, and ultrasonic transverse wave signal in the spatial overlap region of the multimodal signal.

[0174] Among them, acoustic emission signal can refer to the high-frequency elastic wave signal generated by the instantaneous release of energy during the plastic deformation or microcrack propagation of steel; ultrasonic longitudinal wave signal can refer to the ultrasonic wave signal that propagates along the thickness direction of the material and the direction of particle vibration is consistent with the direction of propagation; ultrasonic transverse wave signal can refer to the ultrasonic wave signal that the direction of particle vibration is perpendicular to the direction of propagation.

[0175] A normalized intensity field can refer to a dimensionless spatial distribution field formed by linearly or nonlinearly normalizing the original intensity values ​​of various signals at spatial grid nodes. It can be a mapping of each signal intensity to the [0, 1] interval, or a scaling factor relative to the maximum intensity of the signal in the entire detection area.

[0176] Spatial overlap can refer to the degree to which the normalized intensity fields of at least two signals simultaneously exhibit significant responses at the same spatial location within a spatially overlapping region of multimodal signals. It can be characterized by the ratio of the integral overlapping region to the union region, or by statistically defining the proportion of spatial units that satisfy the condition that the intensities of both signals are higher than their respective thresholds.

[0177] In this embodiment, acoustic emission signal, ultrasonic longitudinal wave signal, and ultrasonic transverse wave signal are three types of independent observable signals with different physical origins and propagation characteristics: acoustic emission signal reflects the transient energy release of damage events, ultrasonic longitudinal wave signal is sensitive to volumetric defects, and ultrasonic transverse wave signal is sensitive to interface defects; the superposition of the intensity of the three at the same spatial location indicates that there is damage evolution behavior driven by multiple physical mechanisms in this region.

[0178] This application can, for example, calculate the spatial overlap degree by comparing the spatial integrals of the minimum and maximum intensities of the acoustic emission signal and the ultrasonic P-wave signal within the spatially overlapping region of the multimodal signals, based on the normalized intensity fields of the two signals. Alternatively, it can calculate the spatial overlap degree by statistically analyzing the proportion of spatial grid points in the spatially overlapping region of the ultrasonic P-wave and ultrasonic S-wave signals, where both signals have intensities exceeding their respective 0.6 normalization thresholds, relative to the total number of grid points. Furthermore, this application can also calculate the spatial overlap degree by using a three-dimensional voxel intersection operation within the spatially overlapping region of the multimodal signals, based on the normalized intensity fields of the acoustic emission signal, ultrasonic P-wave signal, and ultrasonic S-wave signal, to extract a set of connected voxels whose intensities are all in the high-response region, and calculating the spatial overlap degree based on their volume proportions. This application does not limit the calculation method for the spatial overlap degree in these cases.

[0179] Step S803: Determine the test results of the steel based on the quantitative index of microstructure changes and spatial overlap.

[0180] Among them, the microstructure change quantification index is a dimensionless value that reflects the degree of irreversible damage evolution at the microscale of steel, and its value range is, for example, [0, 1]; the spatial overlap is a dimensionless value that reflects the consistency of the response of macroscopic observable signals in the spatial dimension, and its value range is, for example, [0, 1].

[0181] In this embodiment, the quantitative index of microstructure change and the degree of spatial overlap together constitute the dual basis for judging the damage state of steel: the former provides support for the internal mechanism of damage occurrence, and the latter provides external signal verification of damage manifestation; both are indispensable and need to be judged in a coordinated manner to avoid misjudgment from a single dimension.

[0182] For example, this application can substitute the quantitative index of microstructure change and the spatial overlap into an empirical fusion formula to calculate the local damage index I_D = D × (1 + 2.5 × C), where D is the quantitative index of microstructure change and C is the spatial overlap. Then, based on the numerical range and spatial distribution characteristics of I_D, the plastic deformation development stage and corresponding damage level of the steel are divided. When I_D < 0.3 and C < 0.45, it is determined to be the elastic stage with no obvious damage. When 0.3 ≤ I_D < 0.7 and C ≥ 0.45, it is determined to be the early uniform plastic deformation stage with the first damage level. When I_D ≥ 0.7 and C > 0.72, it is determined to be the local plastic concentrated development stage or the microcrack initiation stage with the second or third damage level. Furthermore, the damage location is output by combining the spatial gradient and connectivity of I_D.

[0183] This application constructs a dual-track judgment framework driven by micro-mechanisms and verified by macro-signals by co-modeling the intrinsic damage mechanism characterized by the cumulative degree of irreversible changes in microstructure with the extrinsic damage manifestation characterized by the overlap of normalized intensity fields of at least two signals, such as acoustic emission, ultrasonic longitudinal waves, and ultrasonic transverse waves, in spatially overlapping regions. Based on this, by leveraging the weighted coupling relationship between the quantitative index of microstructure changes and the degree of spatial overlap, a comprehensive, cross-scale, calculable, and reproducible judgment of the development stage of plastic deformation and the early damage level of steel is achieved. This effectively suppresses the risk of misjudgment caused by single signal noise, modeling bias, or environmental interference, and improves the theoretical interpretability and engineering robustness of the detection results.

[0184] In yet another alternative embodiment, such as Figure 9 As shown, this application provides an optional method embodiment for determining the detection results of steel based on quantitative indicators of microstructural changes and spatial overlap, including:

[0185] Step S901: If the quantitative index of microstructure change is less than the first threshold and the spatial overlap is less than the second threshold, then the test result of the steel is determined to be that the steel is in the elastic stage and there is no obvious damage.

[0186] Among them, the microstructure change quantification index can refer to the dimensionless comprehensive characterization value calculated based on at least one of the parameters of local equivalent plastic strain increment, dislocation density and void volume fraction. It can be used to reflect the degree of accumulation of irreversible processes such as lattice slip, dislocation proliferation and microvoid evolution inside the steel. In this embodiment, the microstructure change quantification index is used as the core input variable for judging damage evolution. Its value directly corresponds to the severity level of damage to the internal microstructure of the material.

[0187] The first threshold is a critical criterion used to distinguish between elastic response and plastic initiation. Its value corresponds to the minimum measurable microstructure disturbance level at which the yielding behavior of steel begins to appear. In this embodiment, the first threshold serves as the starting point of the judgment logic. When the MCQI (Microstructure Change Quantification Index) is lower than this value, it indicates that the material has not yet undergone identifiable microstructure changes driven by plastic deformation.

[0188] Spatial overlap can be a scalar value calculated from the overlap of the normalized intensity fields of at least two of the acoustic emission signal, ultrasonic longitudinal wave signal, and ultrasonic transverse wave signal in the spatial overlap region of multimodal signals. It is used to characterize the degree of consistency of the spatial distribution of multi-source physical signals. In this embodiment, this index is used as a measure of the reliability of multimodal cooperation. The lower the value, the more independent the responses of each modal signal are and the absence of coupling characteristics.

[0189] The second threshold can be a lower bound of spatial consistency used to define whether the multimodal signal response has statistical significance, for example, 0.45. In this embodiment, the second threshold and the first threshold together constitute a dual constraint condition for the elastic stage determination. For example, the possibility of plastic deformation and early damage can only be ruled out when both MCQI (quantitative index of microstructural change) and SCD (spatial overlap) are lower than their respective thresholds.

[0190] This application determines whether steel is in the elastic stage by means of joint comparison of MCQI and SCD; this application determines whether steel is in the elastic stage by means of ...

[0191] For example, when MCQI=0.12 (less than the first threshold 0.15) and SCD=0.38 (less than the second threshold 0.45), the system determines that the steel is still in the elastic deformation range, there is no measurable irreversible change in the microstructure, and there is no spatial cooperative response of multimodal signals. The output detection result is the elastic stage, with no obvious damage.

[0192] Step S902: If the quantitative index of microstructure change is greater than or equal to the first threshold and less than the third threshold, and the spatial overlap is greater than or equal to the second threshold, then the test result of the steel is determined to be that the steel is in the early stage of uniform plastic deformation and the damage level is the first level.

[0193] The third threshold is an upper limit criterion used to define the transformation from uniform plastic deformation to local concentration. Its value corresponds to the maximum microstructural disturbance level at which the macroscopic strain distribution still maintains good uniformity after the material yields as a whole, for example, 0.38. In this embodiment, the third threshold and the first threshold together define the numerical range of the first-level damage, reflecting that the damage is still in the initial, diffuse, and relatively reversible development stage. The first level can refer to the level category with the lightest damage, which corresponds to the material that has not yet shown obvious stress concentration or geometric distortion on a macroscopic scale, and can be characterized by low dislocation density growth and a small number of uniformly distributed microvoids on a microscopic scale, with no significant degradation of overall mechanical properties. In this embodiment, the first level serves as the starting point for damage evolution, providing a baseline reference for subsequent monitoring.

[0194] This application determines the early uniform plastic deformation stage, for example, based on the joint condition that MCQI falls within the interval of [first threshold, third threshold) and SCD ≥ second threshold; this application determines the early uniform plastic deformation stage, for example, based on the coordinate judgment method that MCQI and SCD fall within a specified rectangular region in a two-dimensional plane; furthermore, this application determines the early uniform plastic deformation stage based on the judgment method that the intermediate value obtained after linear combination of MCQI and SCD falls within a preset interval and satisfies monotonicity constraints, without limitation.

[0195] For example, in this application, when MCQI=0.26 (satisfying 0.15≤0.26<0.38) and SCD=0.53 (≥0.45), the system determines that the steel has entered the initial plastic stage after yielding, the multimodal signals show preliminary coordination, but no local strengthening response has yet appeared, and the output detection result is the early uniform plastic deformation stage, and the damage level is the first level.

[0196] Step S903: If the quantitative index of microstructure change is greater than or equal to the third threshold and less than the fourth threshold, and the spatial overlap is greater than the fifth threshold, then the test result of the steel is determined to be that the steel is in the stage of local plasticity concentration development, and the damage level is the second level. The damage level of the second level is higher than that of the first level.

[0197] The third and fourth thresholds together constitute the numerical boundary of the second-level damage, which is set based on the critical microstructural disturbance level before necking occurs in the material under uniaxial tensile or bending loads. In this embodiment, this range reflects the comprehensive evolutionary characteristics of dislocation pile-up intensification, initial appearance of strain localization, magnetic memory gradient anomalies, and improved spatial matching degree of strain concentration regions. The fourth threshold is the lower limit for triggering localization response identification, for example, 0.65, which is higher than the third threshold, reflecting the transition from overall yielding to local instability. In this embodiment, the fourth threshold and the third threshold for spatial overlap form a strong coupling judgment condition, ensuring that damage is only detected when the multimodal signals are highly consistent in space and the microscopic indices reach a certain level. Only when a certain level is reached is it upgraded to the second level; the fifth threshold of spatial overlap is a spatial consistency threshold used to characterize the high degree of coordination of multimodal signals, for example, 0.72, which is higher than the second threshold, reflecting the physical nature of the signal response tending to focus from dispersion; the second level can refer to a level category where the degree of damage is higher than the first level. The corresponding material can be characterized by the superposition of local strain concentration and magnetic field gradient anomalies on a macroscopic scale, and on a microscopic scale, it can be characterized by high dislocation density, increased void nucleation and formation of microcrack precursors, and the mechanical properties begin to show nonlinear degradation; in this embodiment, this level marks the key node of the transformation of damage from latent to explicit, and it is necessary to start enhanced monitoring or load intervention.

[0198] This application determines the local plasticity concentration development stage, for example, by using a joint conditional judgment method where the MCQI falls within the interval of [third threshold, fourth threshold] and the SCD > fifth threshold; or by using a geometric judgment method where the MCQI and SCD fall within a specified oblique zone in a two-dimensional plane; further, this application determines the local plasticity concentration development stage by judging whether the damage index obtained after nonlinear mapping of the MCQI and SCD falls within the intermediate-risk interval and satisfies spatial gradient constraints. This application achieves accurate identification of the local plasticity concentration development stage based on any of the above methods, providing a key decision-making basis for structural health assessment.

[0199] For example, in this application, when MCQI=0.51 (satisfying 0.38≤0.51<0.65) and SCD=0.78 (>0.72), the system determines that the steel has undergone obvious local plastic flow, the high-energy acoustic emission event, the abnormal peak of the magnetic field gradient and the strain concentration area are highly overlapping in space, and the output detection result is the local plastic concentration development stage, and the damage level is the second level.

[0200] Step S904: If the quantitative index of microstructure change is greater than or equal to the fourth threshold, the test result of the steel is determined to be that the steel is in the microcrack initiation stage, the damage level is the third level, and the damage location is output. The damage level of the third level is higher than that of the second level.

[0201] The fourth threshold is an upper limit criterion used to define the critical point for microcrack initiation. Its value corresponds to the minimum microstructural disturbance level of internal material void connection, dislocation network collapse, and macrocrack precursor formation. In this embodiment, this threshold serves as the highest order of magnitude threshold for damage evolution, and the third-level judgment is activated only when the MCQI reaches this value. The damage location can refer to the geometric center coordinates of the spatially overlapping region of the multimodal signals, which is determined by the spatial distribution statistics of the spatial matching point set, such as using the arithmetic mean of the three-dimensional coordinates of all spatial matching points. In this embodiment, the damage location serves as a mandatory element of the third-level output, directly serving subsequent location repair or failure analysis.

[0202] For example, when the MCQI = 0.73 (≥ 0.65) is determined to be the third level of damage, and the geometric center coordinates of the overlapping area of ​​the multimodal signal space are (2.47m, 1.35m, 0.80m), the system can determine that microcrack precursors have appeared inside the steel, output the detection result as the microcrack initiation stage, the damage level as the third level, and simultaneously output the damage location (2.47m, 1.35m, 0.80m).

[0203] This application constructs a fourth-order progressive threshold judgment matrix by combining the quantitative index of microstructural changes with spatial overlap, ensuring that the detection results strictly follow the physical path of material damage evolution—starting from elastic response and sequentially covering four typical stages: early uniform plastic deformation, localized plastic concentration development, and microcrack initiation. Through the step-by-step design of the first to fourth thresholds, coupled with the differentiated constraints on spatial consistency imposed by the second and fifth thresholds, the physical interpretability and engineering operability of damage level classification are unified. Based on this, the damage location is forcibly output, completing a closed-loop technology chain from multi-source signal acquisition, feature extraction, spatial matching, multi-scale analysis to damage classification and location, effectively supporting accurate assessment of structural service status and risk warning decisions.

[0204] It should be noted that although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart may be performed in a different order.

[0205] like Figure 10As shown, this application also provides a non-destructive testing system 1000 for steel based on multimodal data. The system includes a first acquisition module 1001, an identification and determination module 1002, a second acquisition module 1003, a first determination module 1004, a matching and determination module 1005, a decomposition and determination module 1006, and a second determination module 1007.

[0206] The first acquisition module 1001 is used to acquire the acoustic emission signal sequence of steel during the stress process;

[0207] The identification and determination module 1002 is used to identify the starting time of the steel entering the plastic deformation stage based on the acoustic emission signal sequence, and to determine the time interval of the increase in the activity level of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage.

[0208] The second acquisition module 1003 is used to acquire the metal magnetic memory signal and surface strain data of steel in the same detection area within a time interval.

[0209] The first determining module 1004 is used to determine the location of the magnetic field gradient anomaly in the steel based on the metal magnetic memory signal, and to determine the location of the strain concentration area in the steel based on the surface strain data.

[0210] The matching and determination module 1005 is used to spatially match the location of high-energy signals, the location of magnetic field gradient anomalies, and the location of strain concentration areas during the plastic deformation stage to determine the spatial overlap area of ​​multimodal signals.

[0211] The decomposition and determination module 1006 is used to perform multi-scale decomposition of acoustic emission signals in the spatial overlap region of multi-modal signals, and determine the cumulative degree of irreversible changes in the microstructure of steel based on the abrupt change characteristics in the decomposed signals.

[0212] The second determining module 1007 is used to determine the detection result of the steel based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the multimodal signal spatial overlap region.

[0213] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A nondestructive testing method for steel based on multimodal data, characterized in that, The method includes: Acquire the acoustic emission signal sequence of steel during the stress process; The starting time of the steel entering the plastic deformation stage is identified based on the acoustic emission signal sequence, and the time interval of the increase in the activity of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage is determined. Within the time interval, acquire the metal magnetic memory signal and surface strain data of the steel in the same detection area; The location of the magnetic field gradient anomaly in the steel is determined based on the metal magnetic memory signal, and the location of the strain concentration region in the steel is determined based on the surface strain data. Spatially match the location of the high-energy signal during the plastic deformation stage, the location of the magnetic field gradient anomaly, and the location of the strain concentration region to determine the spatial overlap region of the multimodal signal; The acoustic emission signal in the overlapping region of the multimodal signal space is decomposed into multiple scales, and the cumulative degree of irreversible change in the microstructure of the steel is determined based on the abrupt change characteristics in the decomposed signal. The detection result of the steel is determined based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the spatially overlapping regions of the multimodal signals.

2. The method according to claim 1, characterized in that, The step of identifying the start time of the steel entering the plastic deformation stage based on the acoustic emission signal sequence includes: Extract the amplitude and frequency center of each acoustic emission signal in the acoustic emission signal sequence; Based on the distribution characteristics of the amplitude and frequency center of each acoustic emission signal, cluster analysis is performed to classify each acoustic emission signal into elastic stage signals and plastic deformation stage signals. The starting time of the plastic deformation stage is determined based on the first significant appearance and sustained trend of the plastic deformation stage-type signal in the time series.

3. The method according to claim 2, characterized in that, The determination of the time interval during which the activity of the acoustic emission signal increases in the acoustic emission signal sequence within the plastic deformation stage includes: Using the start time of the plastic deformation stage as the boundary, the acoustic emission signal sequence is divided to obtain the signal segment of the plastic deformation stage; Wavelet transform decomposition is performed on the signal segment of the plastic deformation stage to extract high-energy signals; Based on the count rate change trend of the high-energy signal, the time interval of the increase in the activity level of the acoustic emission signal is determined.

4. The method according to claim 1, characterized in that, The determination of the magnetic field gradient anomaly location based on the metal magnetic memory signal includes: The magnetic field gradient of the metal magnetic memory signal along the spatial direction is calculated, and a comprehensive gradient anomaly amplitude sequence is synthesized. Extract the target comprehensive gradient anomaly amplitude within the anomaly range from the comprehensive gradient anomaly amplitude sequence, and use the spatial coordinates of the metal magnetic memory signal corresponding to the target comprehensive gradient anomaly amplitude as the magnetic field gradient anomaly location.

5. The method according to claim 1, characterized in that, Determining the location of strain concentration regions based on the surface strain data includes: A finite element model of the steel is established, and loads and boundary conditions corresponding to the actual stress conditions are applied to solve the problem and obtain the surface strain field distribution data. Target strain field distribution data whose strain values ​​exceed a preset threshold are selected from the surface strain field distribution data; Clustering is performed on the selected adjacent target strain field distribution data to form continuous strain concentration regions, and the geometric center coordinates of each strain concentration region are calculated as the location of the strain concentration region.

6. The method according to claim 1, characterized in that, The step of spatially matching the location of the high-energy signal during the plastic deformation stage, the location of the magnetic field gradient anomaly, and the location of the strain concentration region to determine the spatially overlapping region of the multimodal signals includes: Map the location of the high-energy signal, the location of the magnetic field gradient anomaly, and the location of the strain concentration region to the same spatial coordinate system; For each of the high-energy signals, the first distance between the location of the high-energy signal and the target magnetic field gradient anomaly location, and the second distance between the location of the target strain concentration region and the location of the high-energy signal are calculated respectively. The distance between the target magnetic field gradient anomaly location and the location of the high-energy signal is less than the distance between any other magnetic field gradient anomaly location and the location of the high-energy signal. The distance between the location of the target strain concentration region and the location of the high-energy signal is less than the distance between any other strain concentration region location and the location of the high-energy signal. The locations of high-energy signals where both the first distance and the second distance are less than a preset distance threshold are determined as spatial matching points; Based on the spatial distribution of all spatial matching points, the spatial overlap region of the multimodal signal is determined.

7. The method according to claim 1, characterized in that, The step of performing multi-scale decomposition of the acoustic emission signal within the spatially overlapping region of the multimodal signal, and determining the cumulative degree of irreversible changes in the microstructure of the steel based on the abrupt change characteristics in the decomposed signal, includes: Wavelet multi-scale decomposition is performed on the acoustic emission signal in the overlapping region of the multimodal signal space to obtain detail components at multiple scales; Detect abrupt changes in each of the detailed components and record the density and magnitude of each abrupt change. The cumulative degree of irreversible changes in the microstructure of the steel is determined based on the changes in the density and amplitude of each mutation point relative to a baseline value.

8. The method according to claim 7, characterized in that, The determination of the detection result of the steel based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the spatially overlapping region of the multimodal signal includes: Based on at least one of the parameters among the local equivalent plastic strain increment, dislocation density, and void volume fraction in the cumulative degree of irreversible changes in the microstructure of the steel, a quantitative index of microstructure change is calculated. The degree of spatial overlap is calculated based on the overlap of the normalized intensity fields of at least two of the acoustic emission signal, ultrasonic longitudinal wave signal, and ultrasonic transverse wave signal within the spatial overlap region of the multimodal signal. The test results of the steel are determined based on the quantitative index of microstructural changes and the degree of spatial overlap.

9. The method according to claim 8, characterized in that, The determination of the steel's test results based on the quantitative index of microstructural changes and the spatial overlap includes: If the quantitative index of microstructure change is less than the first threshold and the spatial overlap is less than the second threshold, then the test result of the steel is determined to be that the steel is in the elastic stage and there is no obvious damage. If the quantitative index of microstructure change is greater than or equal to the first threshold and less than the third threshold, and the spatial overlap is greater than or equal to the second threshold, then the test result of the steel is determined to be that the steel is in the early uniform plastic deformation stage and the damage level is the first level. If the quantitative index of microstructure change is greater than or equal to the third threshold and less than the fourth threshold, and the spatial overlap is greater than the fifth threshold, then the test result of the steel is determined to be that the steel is in the stage of local plasticity concentration development, the damage level is the second level, and the damage level of the second level is higher than the first level. If the quantification index of microstructure change is greater than or equal to the fourth threshold, the detection result of the steel is determined to be that the steel is in the microcrack initiation stage, the damage level is the third level, and the damage location is output. The damage level of the third level is higher than that of the second level.

10. A non-destructive testing system for steel based on multimodal data, characterized in that, The system includes: The first acquisition module is used to acquire the acoustic emission signal sequence of steel during the stress process; The identification and determination module is used to identify the starting time of the steel entering the plastic deformation stage based on the acoustic emission signal sequence, and to determine the time interval of the increase in the activity level of the acoustic emission signal in the acoustic emission signal sequence within the plastic deformation stage. The second acquisition module is used to acquire the metal magnetic memory signal and surface strain data of the steel in the same detection area within the time interval. The first determining module is used to determine the location of the magnetic field gradient anomaly of the steel based on the metal magnetic memory signal, and to determine the location of the strain concentration region of the steel based on the surface strain data. The matching and determination module is used to spatially match the location of the high-energy signal, the location of the magnetic field gradient anomaly, and the location of the strain concentration region during the plastic deformation stage to determine the spatial overlap region of the multimodal signals. The decomposition and determination module is used to perform multi-scale decomposition of the acoustic emission signal in the spatial overlap region of the multimodal signal, and determine the cumulative degree of irreversible changes in the microstructure of the steel based on the abrupt change characteristics in the decomposed signal. The second determining module is used to determine the detection result of the steel based on the cumulative degree of irreversible changes in the microstructure of the steel and the degree of spatial overlap of the spatially overlapping region of the multimodal signal.

Images