Material performance detection method, device and system based on multi-source micro-magnetic signal data fusion
By using a multi-source micromagnetic signal data fusion method and weighted averaging of MBN and EMAT signals, the problem of large errors in TRIP steel testing was solved, and accurate detection of strain, stress and microstructure content was achieved, making it suitable for batch non-destructive testing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAZHONG UNIV OF SCI & TECH
- Filing Date
- 2023-08-25
- Publication Date
- 2026-06-09
Smart Images
Figure CN117074510B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of material performance testing, and more specifically, relates to a material performance testing method, device and system based on multi-source micro-magnetic signal data fusion. Background Technology
[0002] Steel is widely used in the automotive industry. Deformation of non-TRIP steel during processing can pose safety hazards, while the complex stress and strain conditions during processing of TRIP steel can affect the transformation of retained austenite to martensite, leading to significant differences in microstructure and mechanical properties. Therefore, it is necessary to inspect parts during processing. Currently, the most common methods are X-ray diffraction (XRD), scanning electron microscopy, or metallographic observation. However, these traditional methods can damage parts and are not suitable for batch testing. There is an urgent need for rapid, non-destructive testing methods to deeply study the microstructure and mechanical properties of manganese steel parts.
[0003] Magnetic Barkhausen noise (MBN) detection technology utilizes the complex phenomena generated during the irreversible magnetization process of ferromagnetic materials to reflect their internal structure and external stress. Electromagnetic acoustic transducer (EMAT) detection technology uses electromagnetic coupling to excite and receive ultrasonic waves, and has been widely used in stress detection and other fields in industrial production.
[0004] Because the stress and strain experienced by steel interact with each other and both affect MBN and EMAT detection, and because the microstructure (martensite) content of TRIP steel changes with stress and strain, it also alters the MBN signal intensity. Therefore, using MBN or EMAT detection alone has limitations: MBN signals have a wide applicability, capable of detecting common stress, strain, and microstructure content; EMAT signals have high accuracy, but their current applicability is less than that of MBN signals, and their relationship with microstructure changes is weaker, currently only used for stress and strain detection. Stress, strain, and microstructure content parameters all affect the signal, and these parameters interact with each other. Stress leads to strain, and for TRIP steel, changes in strain affect the microstructure (martensite) content. To reduce errors, a composite detection and data fusion method for MBN and EMAT signals is urgently needed to detect stress, strain, and other parameters in steel materials. Summary of the Invention
[0005] In view of the above-mentioned defects or improvement needs of the existing technology, the present invention provides a material performance testing method, device and system for multi-source micro magnetic signal data fusion. Its purpose is to achieve accurate detection of multiple performance parameters of steel materials by fusing MBN signals and EMAT signals.
[0006] To achieve the above objectives, according to a first aspect of the present invention, a method for detecting material properties through multi-source micromagnetic signal data fusion is proposed, comprising the following steps:
[0007] The MBN and EMAT signals of steel under different performance parameters were detected by magnetic Barkhausen noise method and electromagnetic ultrasonic method, respectively, to obtain a dataset; the performance parameters include strain, stress and microstructure content.
[0008] The dataset is preprocessed, and the proportion of each performance parameter in the MBN signal and EMAT signal is obtained by multivariate nonlinear regression analysis, thereby obtaining the corresponding data of each performance parameter acting on the MBN signal or EMAT signal individually.
[0009] The data corresponding to the MBN signal or EMAT signal, which are individually applied by each performance parameter, are linearized. Then, the performance parameters are substituted to obtain the theoretical signal. Based on the error between the theoretical signal and the actual signal, the weights of the MBN signal and the EMAT signal are determined. The MBN signal and the EMAT signal are then weighted and fused according to the weights to obtain the fused signal. The correspondence between the performance parameters and the fused signal is then determined.
[0010] Based on the correspondence between performance parameters and fused signals, the strain, stress, and microstructure content of the steel under test can be detected.
[0011] As a further preferred option, preprocessing of the dataset includes: denoising the MBN and EMAT signals, and normalizing the strain, stress, and tissue content data.
[0012] As a further preferred method, the noise reduction of the MBN signal and the EMAT signal includes: using bandpass filtering to reduce noise in the MBN signal and the EMAT signal, using the superposition averaging method to improve the signal-to-noise ratio and reduce white noise in the filtered signal, and performing smoothing processing through wavelet denoising.
[0013] As a further preferred option, normalizing the strain, stress, and microstructure content data includes performing a linear transformation on the strain, stress, and microstructure content data to map them into a closed interval [0, 1].
[0014] As a further preferred option, the corresponding data of the MBN signal or EMAT signal are linearized by applying the extended Kalman filter algorithm to each performance parameter individually.
[0015] As a further preferred method, the weights of the MBN signal and the EMAT signal are determined as follows:
[0016] Preset multiple sets of weight values for MBN and EMAT signals;
[0017] For any set of weight values: select performance parameters, substitute them into the corresponding data after linearization, and obtain the theoretical MBN signal and the theoretical EMAT signal; at the same time, based on the performance parameters, detect the actual MBN signal and the actual EMAT signal; then, based on the preset weight values, calculate the theoretical fused signal and the actual fused signal, and then determine the error based on the theoretical fused signal and the actual fused signal.
[0018] The set of weight values corresponding to the minimum error is taken as the final weight.
[0019] As a further preferred option, the error between the theoretical fused signal and the actual fused signal is determined based on the mean square error of MSE.
[0020] As a further preferred option, the corresponding data for each performance parameter acting individually on the MBN signal or EMAT signal include: the corresponding relationship curves of stress-MBN signal, strain-MBN signal, tissue content-MBN signal, stress-EMAT signal, and strain-EMAT signal.
[0021] According to a second aspect of the present invention, a composite detection device is provided for implementing the above-described multi-source micromagnetic signal data fusion method for material property detection, comprising a signal excitation module, a probe module, a signal acquisition module, and a signal processing module, wherein:
[0022] The signal excitation module is used to generate excitation signals;
[0023] The probe module is used to receive the excitation signal to detect the steel and transmit the induced signal to the signal acquisition module; the probe module includes an MBN probe and an EMAT probe, which are used for magnetic Barkhausen noise method and electromagnetic ultrasonic method detection, respectively, and the MBN probe and EMAT probe can be flexibly replaced and connected to the signal excitation module;
[0024] The signal acquisition module is used to acquire the MBN signal and EMAT signal detected by the probe module;
[0025] The signal processing module is used to obtain a fused signal based on the MBN signal and the EMAT signal, and to obtain the performance parameters of the steel according to the correspondence between the performance parameters and the fused signal.
[0026] According to a third aspect of the present invention, a material performance testing system based on multi-source micromagnetic signal data fusion is provided, comprising a processor for executing the above-described material performance testing method based on multi-source micromagnetic signal data fusion.
[0027] In summary, compared with the prior art, the above-described technical solutions conceived by this invention mainly possess the following technical advantages:
[0028] 1. This invention first obtains the correspondence between each performance parameter acting individually on the MBN signal and the EMAT signal. Then, based on the error between the theoretical signal and the actual signal, the MBN signal and the EMAT signal are weighted and fused to determine the correspondence between each performance parameter and the fused signal. This achieves the fusion of the MBN and EMAT signals, and the fused signal value can be used to accurately detect multiple parameters such as strain, stress, and microstructure content of the steel material under test.
[0029] 2. This invention uses multiple methods to jointly reduce noise in the data and normalizes each performance parameter to eliminate the difficulty in comparison between parameters due to different dimensions. Then, through multivariate nonlinear regression analysis, the corresponding data of stress / strain / structure content with MBN signal and stress / strain with EMAT signal are obtained respectively, thereby eliminating the error caused by the simultaneous action of multiple factors. Attached Figure Description
[0030] Figure 1 This is a flowchart of a material performance testing method based on multi-source micromagnetic signal data fusion according to an embodiment of the present invention;
[0031] Figure 2 This is a schematic diagram of data normalization processing and multivariate nonlinear regression analysis in an embodiment of the present invention;
[0032] Figure 3 This is a schematic diagram of the composite detection device according to an embodiment of the present invention;
[0033] Figure 4 This is a flowchart illustrating the signal fusion process according to an embodiment of the present invention.
[0034] Figure 5 (a) and (b) are schematic diagrams of the sample size and structure in the embodiments of the present invention; Detailed Implementation
[0035] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.
[0036] This invention provides a method for detecting material properties through multi-source micromagnetic signal data fusion, such as... Figure 1 As shown, it includes the following steps:
[0037] S1. Preparation of steel sample to be tested: pretreatment to remove surface stress of the material so that the material meets the basic conditions for testing.
[0038] S2. The relationship between the stress / strain / microstructure and the test values of steel is obtained by using a combination of Barkhausen and electromagnetic ultrasonic testing methods: The sample is subjected to in-situ tensile testing, and the voltage peak value under different performance parameters is detected by magnetic Barkhausen noise method and electromagnetic ultrasonic method respectively to obtain MBN signal and EMAT signal. The performance parameters include strain, stress and microstructure (martensite) content; During the test, for each set of performance parameters, the Barkhausen (electromagnetic ultrasonic) peak value needs to be measured in three specific areas located at the center of the sample, and the average value is taken to reduce the error.
[0039] S3. Preprocessing of multi-source signals: such as Figure 2 As shown, the data corresponding to the MBN and EMAT signals are first subjected to noise reduction processing. Then, the units of stress, strain, and microstructure (martensite) content are normalized. Multivariate nonlinear regression analysis is used to obtain the proportion of stress, strain, and microstructure content's influence on the MBN signal. Similarly, the proportion of stress and strain's influence on the EMAT signal is obtained. This yields the corresponding data for stress / strain / microstructure content acting alone on the MBN signal, and the corresponding relationship for stress / strain acting alone on the EMAT signal. That is, the corresponding data curves for stress-MBN signal, strain-MBN signal, microstructure content-MBN signal, stress-EMAT signal, and strain-EMAT signal (unknown points on the curves are connected by fitting). It should be noted that the signals in the corresponding data curves are actually sub-signals; that is, for any stress in the material, its strain and microstructure content are constant, and the theoretical MBN signal is the sum of the MBN sub-signal intensities corresponding to stress / strain / microstructure content.
[0040] In some embodiments, signal denoising employs bandpass filtering. The filtered signal is then subjected to a superposition averaging method to improve the signal-to-noise ratio and reduce white noise, followed by smoothing through wavelet denoising. Normalization of different parameters aims to separate multiple influencing parameters that simultaneously affect the original data. Function transformation is used to eliminate the dimensions of the three parameters—stress (MPa), strain (%), and tissue content (%)—and map them to a closed interval [0, 1], thereby eliminating the influence of dimensions on the final result and making stress, strain, and tissue content comparable.
[0041] S4. Fusion of multi-source signal data improves the accuracy of performance parameter detection: such as... Figure 4As shown, the five sets of correlation curves—corresponding to stress / strain / microstructure content and MBN signal, and stress / strain and EMAT signal—all exhibit nonlinear correlations. For example, the increase in MBN signal is nonlinear with increasing stress. Therefore, based on the principle of the extended Kalman filter algorithm, Taylor series expansion is used to linearize the nonlinear correlations. Subsequently, the error between the theoretical and actual signals is determined, and the weights of the MBN and EMAT signals are determined based on this error. Finally, information fusion is completed through weighted fusion using the Kalman filter algorithm, determining the correspondence between performance parameters and the fused signal.
[0042] Specifically, preset weight values k1 and k2 are used for information fusion (i.e., fused signal = k1 * MBN signal + k2 * EMAT signal, k1 + k2 = 1). Then, performance parameters (i.e., stress / strain / microstructure content, which are unmeasured fitting points on the corresponding curves) are selected. Based on the linearized relationship between each performance parameter and the MBN / EMAT signals, theoretical MBN and theoretical EMAT signals are calculated. Simultaneously, based on the stress / strain / microstructure content, actual MBN and actual EMAT signals are obtained through actual detection. Then, based on the preset weight values, theoretical and actual fused signals are calculated, and the error between the two is determined based on the mean square error of MSE. Multiple sets of weight values are preset, and the above process is repeated, with the weight value having the smallest mean square error of MSE used as the final weight.
[0043] More specifically, the weighting parameters k1 and k2 for stress, strain, and microstructure content are determined separately; taking stress as an example, in some embodiments, the error is calculated as follows:
[0044] (1 / n)∑[k1×|ML-aMS|+k2×|EL-bES|] 2
[0045] Wherein, ML and EL are the theoretical MBN and theoretical EMAT sub-signals corresponding to the stress obtained from the corresponding relationship curve obtained in step S3, respectively; MS and ES are the actual MBN and actual EMAT signals obtained from the detection, respectively; a and b are the proportions of the stress obtained in step S3 in the MBN and EMAT signals, respectively; and n is the number of stress points selected.
[0046] Then, based on the Kalman filter algorithm, the two signals are weighted and averaged according to their final weights to obtain a fused signal. This determines the correspondence between performance parameters and the fused signal, namely, the correspondence between stress and the fused signal, strain and the fused signal, and the microstructure content and the MBN signal (EMAT signal weight is 0). Based on this correspondence, during formal testing, after acquiring the MBN and EMAT signals of the steel to be tested, the fused signal is calculated according to the final weights, thus obtaining the stress, strain, and microstructure content of the steel.
[0047] The above-mentioned testing method is applicable to the performance parameter testing of both TRIP steel and non-TRIP steel. Since there is no change in the content of martensite during the tensile process of non-TRIP steel, when testing non-TRIP steel samples, only the MBN signal and EMAT signal of the steel under different stress / strain can be detected, simplifying the calculation process.
[0048] The above detection methods can be performed using a composite detection device, such as... Figure 3 As shown, the composite detection device includes a signal excitation module, a probe module, a signal acquisition module, and a signal processing module, which can simultaneously perform detection using two methods.
[0049] The signal excitation module includes a signal generator and a power amplifier. Considering the miniaturization and portability of the system, the signal generator adopts a signal generator based on direct digital synthesis technology, and the power amplifier uses a mono power amplifier board to enhance and amplify the sine signal.
[0050] The probe module is used to receive the excitation signal to detect the steel and transmit the induced signal to the signal acquisition module. The probe module includes an MBN probe and an EMAT probe, which are used for magnetic Barkhausen noise method and electromagnetic ultrasonic method detection, respectively. In use, the MBN probe and EMAT probe can be flexibly replaced and connected to the power amplifier to perform detection by the two methods.
[0051] In some embodiments, the MBN probe includes an excitation coil and a magnetic sensor. The excitation coil uses silicon steel, a soft magnetic material, as its core material, and enameled wire for generating the excitation magnetic field is wound around the core arm. The magnetic sensor uses a detection coil and a manganese-zinc ferrite core. The EMAT probe includes a permanent magnet and an induction coil. Alternating current passes through the induction coil and generates a changing magnetic field. This magnetic field penetrates the detection material and generates eddy currents therein. The direction of the charged particles generating the eddy currents is opposite to the direction of the current in the coil. The permanent magnet typically generates a continuous magnetic field pointing towards the surface of the object being detected. A Lorentz force is applied to the charged particles flowing in the magnetic field and oriented along the surface area of the object being detected. The Lorentz force causes some mechanical displacement in the eddy current region that forms sound waves.
[0052] The signal acquisition module includes a signal amplifier and a data acquisition card. The signal amplifier has a good amplification effect for high-frequency Barkhausen and electromagnetic ultrasonic signals. The data acquisition module is based on the data acquisition card and its compatible development environment. Customized data acquisition software is developed using corresponding functions to achieve communication with the data acquisition card and acquire data.
[0053] The signal processing module acquires the variation characteristics of the MBN signal and EMAT signal, and obtains the numerical relationship between the signal and the measured performance (stress, etc.) by calculating the average value of the signal peaks over multiple cycles, thus obtaining the performance parameters of the steel.
[0054] The following are specific examples:
[0055] S1. A certain TRIP steel was pretreated as a sample. The sample and dimensions are as follows: Figure 5 As shown.
[0056] S2. Conduct in-situ tensile tests and record the peak values of MBN / EMAT signals under multiple sets of stress / strain / tissue content.
[0057] S3, signal denoising and filtering, and normalization processing, and then the corresponding data are obtained through multivariate nonlinear regression.
[0058] Specifically, the original corresponding values of the MBN signal data obtained in step S2 are shown in Table 1. The selected TRIP steel has a martensite content of 0 within the stress range cut off in the table. The method for separating parameters is as follows: set the maximum detection value to 1, and simplify the different unit parameters to dimensionless constants in [0,1]. The simplified results are shown in Table 2.
[0059] Table 1 Original Data
[0060]
[0061] Table 2 Normalized Data
[0062]
[0063] By combining the changes in the MBN signal, the influence ratios of stress, strain, and microstructure (martensite) content can be obtained through multivariate nonlinear regression method, which are 0.1, 0.9, and 0, respectively. The MBN signal has an original peak value of 40mV.
[0064] At this point, the error caused by the simultaneous action of multiple factors has been eliminated. At the same time, the original peak value of the MBN signal is removed and distributed according to the proportion. The corresponding data of the stress-MBN signal and strain-MBN signal are shown in Table 3 and Table 4.
[0065] Table 3 Stress-MBN Peak Value
[0066]
[0067] Table 4 Strain-MBN Peak Value
[0068]
[0069] S4. The correspondence is linearized by using the extended Kalman filter algorithm, and information fusion is completed by weighted fusion.
[0070] The multivariate nonlinear regression method in S3 has shown that the influence ratios of stress, strain, and microstructure (martensite) content on the MBN signal are 0.1, 0.9, and 0, respectively, and the original peak value of the MBN signal is 40mV. In this embodiment, the EMAT signal is the same.
[0071] The calculation of the fused signal requires the establishment of two weight parameters k1 and k2, the sum of which is 1, representing the weights of the MBN signal and the EMAT signal in the fused signal, respectively.
[0072] The theoretical value of the MBN / EMAT signal is the signal value on the curve when parameters such as stress are known. These values are obtained by fitting multiple sets of scattered data measured in S2. Therefore, the theoretical value of the fused signal is k1*theoretical value of MBN signal ML + k2*theoretical value of EMAT signal EL. The actual value of the MBN / EMAT signal is (detected value - 40). For stress, the corresponding actual value of the fused signal is 0.1*k1*MBN measurement signal MS + 0.1*k2*EMAT measurement signal ES. The expression for the error in estimating stress using the fused signal is (1 / n)∑[k1×|ML-0.1MS|+k2×|EL-0.1ES|]. 2 When calculating the strain error, simply replace 0.1 with 0.9 in the formula. The MSE (Mean Square Error) method adds up the total errors under the same state (a certain stress and its corresponding strain and microstructure content). Minimizing the sum of errors obtained from multiple states yields suitable k1 and k2. Subsequently, a weighted average fusion is achieved using a Kalman filter algorithm. Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for detecting material properties through multi-source micromagnetic signal data fusion, characterized in that, Includes the following steps: The MBN and EMAT signals of steel under different performance parameters were detected by magnetic Barkhausen noise method and electromagnetic ultrasonic method, respectively, to obtain a dataset; the performance parameters include strain, stress and microstructure content. The dataset is preprocessed, and the proportion of each performance parameter in the MBN signal and EMAT signal is obtained by multivariate nonlinear regression analysis, thereby obtaining the corresponding data of each performance parameter acting on the MBN signal or EMAT signal individually. The corresponding data of the MBN signal or EMAT signal are linearized by applying each performance parameter individually, and then the performance parameters are substituted to obtain the theoretical signal. The weights of the MBN and EMAT signals are determined based on the error between the theoretical and actual signals. The MBN signal and EMAT signal are weighted and averaged to obtain a fused signal, and then the correspondence between the performance parameters and the fused signal is determined. Based on the correspondence between performance parameters and fused signals, the strain, stress, and microstructure content of the steel under test can be detected.
2. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 1, characterized in that, Preprocessing of the dataset includes: noise reduction of MBN and EMAT signals, and normalization of strain, stress, and microstructure content data.
3. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 2, characterized in that, The noise reduction of MBN and EMAT signals includes: using bandpass filtering to reduce noise in MBN and EMAT signals, using superposition averaging to improve the signal-to-noise ratio and reduce white noise in the filtered signals, and smoothing the signals by wavelet denoising.
4. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 2, characterized in that, Normalizing strain, stress, and microstructure content data includes performing a linear transformation on the strain, stress, and microstructure content data to map them to the closed interval [0, 1].
5. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 1, characterized in that, The extended Kalman filter algorithm is used to linearize the corresponding data of the MBN signal or EMAT signal by applying each performance parameter individually.
6. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 1, characterized in that, The method for determining the weights of the MBN and EMAT signals is as follows: Preset multiple sets of weight values for MBN and EMAT signals; For any set of weight values: select performance parameters, substitute them into the corresponding data after linearization, and obtain the theoretical MBN signal and the theoretical EMAT signal; at the same time, based on the performance parameters, detect the actual MBN signal and the actual EMAT signal; then, based on the preset weight values, calculate the theoretical fused signal and the actual fused signal, and then determine the error based on the theoretical fused signal and the actual fused signal. The set of weight values corresponding to the minimum error is taken as the final weight.
7. The material property testing method based on multi-source micromagnetic signal data fusion as described in claim 6, characterized in that, The error between the theoretical fused signal and the actual fused signal is determined based on the mean square error of MSE.
8. The material property testing method based on multi-source micromagnetic signal data fusion as described in any one of claims 1-7, characterized in that, The corresponding data for each performance parameter acting individually on the MBN signal or EMAT signal include: the relationship curves between stress-MBN signal, strain-MBN signal, tissue content-MBN signal, stress-EMAT signal, and strain-EMAT signal.
9. A composite testing device for implementing the material property testing method of multi-source micromagnetic signal data fusion as described in any one of claims 1-8, characterized in that, It includes a signal excitation module, a probe module, a signal acquisition module, and a signal processing module, wherein: The signal excitation module is used to generate excitation signals; The probe module is used to receive the excitation signal to detect the steel and transmit the induced signal to the signal acquisition module; the probe module includes an MBN probe and an EMAT probe, which are used for magnetic Barkhausen noise method and electromagnetic ultrasonic method detection, respectively, and the MBN probe and EMAT probe can be flexibly replaced and connected to the signal excitation module; The signal acquisition module is used to acquire the MBN signal and EMAT signal detected by the probe module; The signal processing module is used to obtain a fused signal based on the MBN signal and the EMAT signal, and to obtain the performance parameters of the steel according to the correspondence between the performance parameters and the fused signal.
10. A material property testing system based on multi-source micromagnetic signal data fusion, characterized in that, Includes a processor, the processor being configured to perform a material property testing method based on multi-source micromagnetic signal data fusion as described in any one of claims 1-8.