Non-contact laser ultrasonic internal defect detection method and system for high-entropy alloy
By employing a non-contact laser ultrasonic testing method, combined with adaptive laser parameters and multi-algorithm processing, the problems of surface damage and low accuracy in the testing of high-entropy alloys have been solved, achieving high adaptability and automated batch testing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INNER MONGOLIA UNIV OF SCI & TECH
- Filing Date
- 2026-05-11
- Publication Date
- 2026-06-30
AI Technical Summary
In existing technologies, traditional contact ultrasonic testing methods are prone to damaging the surface of high-entropy alloys and have low detection accuracy. In addition, non-contact laser ultrasonic testing has poor adaptability and poor noise suppression effect, and cannot achieve automated online testing of batch samples.
A non-contact laser ultrasonic testing method is adopted. Impurities are removed through surface pretreatment, laser pulse parameters are adaptively adjusted, and signal processing is performed by combining wavelet packet transform, principal component analysis, local outlier factor and support vector machine algorithms. Defect identification and analysis are performed by combining an improved ultrasonic wave propagation model to achieve automated online detection.
No coupling agent is required, and it is compatible with different types of high-entropy alloys, improving detection accuracy and compatibility. It enables automated batch online detection of internal defects in high-entropy alloys, reducing manual intervention and errors.
Smart Images
Figure CN122306709A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of alloy defect detection technology, specifically to a method and system for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound. Background Technology
[0002] High-entropy alloys are prone to defects such as porosity, cracks, and inclusions due to their manufacturing processes (such as melting, forging, and sintering). These internal defects can seriously affect the mechanical properties and service safety of high-entropy alloy components. Therefore, it is crucial to conduct accurate and efficient detection of internal defects in high-entropy alloys.
[0003] Currently, the detection of internal defects in high-entropy alloys mainly adopts the traditional contact ultrasonic testing method. This method requires the addition of a coupling agent between the test sample and the ultrasonic probe, which not only easily damages the surface of the high-entropy alloy (high-entropy alloys have high hardness and the surface is easy to scratch), but also affects the detection accuracy due to the uniformity of the coupling agent application. At the same time, the signal processing algorithm of traditional contact ultrasonic testing is simple and it is difficult to distinguish the interference signal caused by the lattice distortion of high-entropy alloys from the defect signal, resulting in low detection accuracy and inability to adapt to high-entropy alloy samples of different types and thicknesses.
[0004] While existing non-contact laser ultrasonic testing technology has solved the damage problem of contact testing, it still has many shortcomings: First, the laser pulse parameters cannot be adaptively adjusted according to the thermal conductivity characteristics of high-entropy alloys, resulting in poor adaptability; second, the signal preprocessing and anomaly recognition algorithms are simple, with poor noise suppression and low defect recognition accuracy; third, the coordination between system modules is poor, signal transmission delay is large, and automated online testing of batch samples cannot be achieved; fourth, the verification and correction of test results rely on manual intervention, which is inefficient and prone to errors.
[0005] Therefore, developing a non-contact laser-ultrasonic method and system for detecting internal defects in high-entropy alloys that is highly adaptable, accurate, automated, and capable of batch online detection has become an urgent technical problem to be solved in the field of high-entropy alloy detection. Summary of the Invention
[0006] To address the shortcomings of existing technologies, this invention discloses a non-contact laser-ultrasound method and system for detecting internal defects in high-entropy alloys, in order to solve the problems mentioned in the background section.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a non-contact laser-ultrasound method for detecting internal defects in high-entropy alloys, comprising the following steps:
[0008] S1. Surface pretreatment of high-entropy alloy test samples removes impurities and oxide layers from the sample surface without the need for coupling agent, which is suitable for the characteristics of high-entropy alloys, such as high hardness and easy damage.
[0009] S2. A pulsed laser is used to emit laser pulses onto the surface of the pretreated high-entropy alloy sample through the thermoelastic effect, which excites the generation of ultrasonic waves inside the sample. The laser pulse parameters are adaptively adjusted according to the thermal conductivity of the high-entropy alloy.
[0010] S3. A laser receiving device is used to receive ultrasonic signals reflected from the sample surface in a non-contact manner. The ultrasonic signals include defect reflection signals, defect-free region reflection signals, and lattice distortion interference signals.
[0011] S4. The received ultrasonic signal is preprocessed using a wavelet packet transform (WPT) and principal component analysis (PCA) fusion algorithm to achieve noise suppression and redundant information removal.
[0012] S5. Based on the preprocessed effective signal, the defect feature parameters are extracted and the abnormal signal is accurately identified by the fusion algorithm of Local Outlier Factor (LOF) and Support Vector Machine (SVM), so as to effectively distinguish the defect signal from the lattice distortion interference signal.
[0013] S6. Combining the mechanical parameters and lattice distortion coefficient of the high-entropy alloy, the defect signal is analyzed using an improved ultrasonic propagation model to accurately determine the type, size, depth, and location of internal defects in the high-entropy alloy; the lattice distortion coefficient δ is obtained through three methods—theoretical calculation, experimental measurement, or online inversion—based on the composition and microstructure of the high-entropy alloy, and its value ranges from 0 to 0.1.
[0014] S7. Cross-validation and error correction algorithms are used to verify and correct the defect detection results to ensure that the detection accuracy meets the preset requirements.
[0015] Preferably, in step S1, the surface pretreatment specifically includes: wiping the surface of the high-entropy alloy sample with anhydrous ethanol and inert gas to remove surface oil, dust and oxide layer, and then drying it with inert gas to avoid oxidation of the sample surface; after pretreatment, the surface roughness of the sample is in the micron range, which ensures effective excitation of the laser pulse, avoids additional interference to the propagation of ultrasonic signals due to surface roughness, and eliminates detection errors and sample surface damage caused by the use of coupling agent.
[0016] Preferably, in step S2, the laser pulse excitation process is as follows: the laser pulse emitted by the pulsed laser is focused by a focusing lens and applied to the surface of the high-entropy alloy sample. The wavelength of the laser pulse is in the range of 400nm~1550nm, the pulse width is in the range of 1ns~10ns, and the peak energy is in the range of 0.5mJ~5mJ. Through the thermoelastic effect, the thin layer on the sample surface is instantly heated and expanded to generate stress waves. The stress waves propagate into the sample to form longitudinal and transverse waves. The longitudinal waves are used for defect depth detection, and the transverse waves are used for defect size and location detection. The laser pulse parameters are adaptively adjusted by the thermal conductivity coefficient of the high-entropy alloy, and the adjustment formula is:
[0017] ;
[0018] Where E is the peak energy of the laser pulse; k1 is the adaptive adjustment coefficient (ranging from 0.1 to 0.5). The wavelength of the laser pulse; The thermal conductivity coefficient of the high-entropy alloy is dynamically obtained based on the composition and microstructure of the high-entropy alloy, enabling precise matching between the laser pulse and the properties of the high-entropy alloy.
[0019] Preferably, in step S3, the laser receiving device is a laser interferometer with a receiving frequency in the range of 0.5MHz to 5MHz. During the receiving process, the receiving angle is adjusted by a two-dimensional deflection mirror to ensure accurate reception of ultrasonic reflection signals from different positions on the sample surface. The received signal is amplified by a preamplifier and then transmitted to the signal processing module. The gain of the preamplifier is in the range of 40dB to 80dB. At the same time, a shielding structure is used to reduce electromagnetic interference and reduce signal fluctuations caused by high-entropy alloy lattice distortion.
[0020] Preferably, in step S4, the preprocessing of the wavelet packet transform (WPT) and principal component analysis (PCA) fusion algorithm specifically includes:
[0021] S41. Wavelet packet transform is used to perform multi-scale decomposition on the received ultrasonic signal. The decomposition formula is as follows:
[0022] ;
[0023] in, This represents the low-frequency component (effective signal) of the j-th level wavelet packet decomposition. denoted as the high-frequency component (noise) of the j-th wavelet packet decomposition; h(n) represents the low-pass filter coefficients, g(n) represents the high-pass filter coefficients; and j represents the decomposition level (ranging from 3 to 8 levels). For the decomposition node, Z represents the range of values for the discrete index n, which is all integers;
[0024] S42. Perform threshold denoising on the decomposed low-frequency components to remove residual noise. The threshold calculation formula is as follows:
[0025] ;
[0026] Among them, is the denoising threshold; is the standard deviation of the noise, which is estimated from the noise component; N is the number of signal sampling points;
[0027] S43. Perform PCA dimensionality reduction on the denoised signal to remove redundant information in the signal and retain the effective feature components. The core formula of PCA dimensionality reduction is:
[0028] ;
[0029] Among them, Y is the effective signal matrix after dimensionality reduction (dimension ); X sig is the original signal matrix after denoising (dimension ); is the mean matrix of the original signal matrix; W is the eigenvector matrix (dimension ), which is composed of the eigenvectors corresponding to the first k3 largest eigenvalues of the covariance matrix of the original signal matrix (k3 < n); m is the number of signal sampling points, n is the dimension of the original signal, and k3 is the dimension of the signal after dimensionality reduction.
[0030] Preferably, in step S5, the feature extraction and anomaly recognition by the fusion algorithm of local outlier factor (LOF) and support vector machine (SVM) specifically include:
[0031] S51. Extract the characteristic parameters of the effective signal after preprocessing. The characteristic parameters include signal amplitude, frequency, phase, waveform mutation point, and signal attenuation coefficient. Perform dimensionless normalization on each characteristic parameter to eliminate the dimensional and numerical magnitude differences of different characteristics; construct a characteristic parameter vector based on the normalized characteristic parameters , where is the normalized amplitude eigenvalue, is the normalized frequency eigenvalue, is the normalized phase eigenvalue, is the normalized number of waveform mutation point eigenvalues, is the normalized signal attenuation coefficient eigenvalue;
[0032] S52. Use the LOF algorithm to perform preliminary anomaly screening on the characteristic parameter vector. The formula of the LOF algorithm is:
[0033] ;
[0034] Among them, is the local outlier factor of the sample The sample For feature parameter vectors; The preset number of neighborhood samples ranges from 5 to 20. For the sample of The set of all samples within the neighborhood; For the sample Locally achievable density;
[0035] ;
[0036] in, For the sample To its If the average reachability distance of all samples in the neighborhood, calculated by the LOF algorithm, exceeds a preset value, then a candidate abnormal sample is identified.
[0037] S53. Input the candidate anomaly samples into the SVM classifier for accurate classification, distinguishing between defect signals and lattice distortion interference signals. The SVM classifier uses the radial basis function (RBF) kernel function, and the kernel function formula is:
[0038] ;
[0039] in, For the sample With sample The kernel function value; This is a parameter for the kernel function, with a value range of 0.1 to 10; For the sample With sample The Euclidean distance; the classification decision function of the SVM classifier is:
[0040] ;
[0041] in, For the classification results, 1 represents a defect signal and -1 represents an interference signal; These are the support vector coefficients; is the sample label; l is the number of support vectors; b is the classification threshold.
[0042] Preferably, in step S6, the improved ultrasonic propagation model, combined with the high-entropy alloy lattice distortion coefficient, achieves accurate calculation of defect parameters, specifically including:
[0043] S61. Obtain the mechanical parameters and lattice distortion coefficient of the high-entropy alloy, wherein the mechanical parameters include the elastic modulus. Poisson's ratio and the propagation speed of ultrasound v, lattice distortion coefficient The value ranges from 0 to 0.1. The ultrasonic propagation velocity is calculated based on the elastic modulus and Poisson's ratio using the following formula:
[0044] ;
[0045] in, The density of a high-entropy alloy;
[0046] S62. Combining the lattice distortion coefficient to correct the ultrasonic wave propagation time difference, the defect depth is calculated using the following correction formula:
[0047] ;
[0048] in, For the depth of the defect; The original propagation time difference of the ultrasound wave; This is the time difference correction caused by lattice distortion, derived from the lattice distortion coefficient. It was estimated; This is the lattice distortion correction factor;
[0049] S63. Based on the amplitude attenuation characteristics of the defect reflection signal and combined with the ultrasonic attenuation model, the defect size is calculated using the following formula:
[0050] ;
[0051] in, The amplitude of the defect reflection signal; The amplitude of the reflected signal from the defect-free region; The attenuation coefficient of ultrasound waves inside a high-entropy alloy; This is the coefficient representing the effect of lattice distortion on attenuation. The defect shape correction factor is 0.8~0.9 for cracks, 0.6~0.7 for porosity, and 0.7~0.8 for inclusions, and then the defect size is deduced.
[0052] S64. Based on the laser excitation position coordinates Coordinates of the ultrasound receiving location By combining the ultrasonic wave propagation path, the specific location of the defect can be determined. The coordinates are calculated using the triangulation method, and the formula is:
[0053] ,
[0054] in, The time from laser excitation to defect reflection. The time it takes for the defect to be reflected to the ultrasonic receiver.
[0055] Preferably, in step S7, the cross-validation and error correction algorithm specifically includes:
[0056] S71: K-fold cross-validation is used to verify the detection results. K ranges from 5 to 10. The detection samples are divided into K groups. K-1 groups are selected in turn as the training set and 1 group is selected as the test set. The detection error of each test set is calculated.
[0057] S72: Calculate the average value of the detection errors of K groups as the overall detection error. The error correction formula is as follows:
[0058] ,
[0059] ;
[0060] in, The corrected defect depth; The depth of the defect before correction; This is the error correction amount; The average detection error of K-fold cross-validation;
[0061] S73: Determine whether the corrected detection accuracy is not lower than the preset threshold. If not, return to step S4 to adjust the number of wavelet packet decomposition layers and PCA dimensionality reduction, or step S5 to adjust the number of LOF neighborhoods and SVM kernel function parameters, and re-perform detection until the detection accuracy meets the requirements. The preset threshold is dynamically set according to the application scenario of high-entropy alloys.
[0062] This invention provides a non-contact laser-ultrasound internal defect detection system for high-entropy alloys, comprising:
[0063] The sample pretreatment module is used to pretreat the surface of high-entropy alloy test samples. It uses anhydrous ethanol and inert gas to wipe the samples to remove surface impurities and oxide layers without the need to add coupling agent, ensuring that the surface roughness of the sample meets the test requirements.
[0064] The laser excitation module is used to emit laser pulses onto the surface of a pretreated high-entropy alloy sample through a pulsed laser via thermoelastic effect, thereby exciting the generation of ultrasonic waves inside the sample. It includes a pulsed laser, a focusing lens, a two-dimensional deflecting mirror, and a parameter adaptive adjustment unit. The parameter adaptive adjustment unit is based on the thermal conductivity coefficient of the high-entropy alloy. The pulsed laser can achieve continuous adjustment of wavelength, pulse width, and peak energy to adapt to high-entropy alloy samples of different types and thicknesses.
[0065] An ultrasonic receiving module is used to receive ultrasonic signals reflected from the surface of a sample non-contactly using a laser interferometer. It includes a laser interferometer, a preamplifier, and a shielding unit. The preamplifier amplifies the signal, the shielding unit reduces electromagnetic interference, and the processed signal is transmitted to a signal processing module.
[0066] The signal processing module is used to preprocess, extract features, identify anomalies, and analyze defects in the received ultrasonic signals. It has built-in wavelet packet decomposition unit, PCA dimensionality reduction unit, LOF algorithm unit, SVM classification unit, and improved ultrasonic propagation model unit, which correspond to the processing steps S4 to S6, respectively, and can realize dynamic adjustment of algorithm parameters. The SVM classification unit has built-in pre-trained classification model and can update model parameters according to the type of high-entropy alloy.
[0067] The motion control module is used to drive the high-entropy alloy sample to perform precise motion and realize comprehensive scanning and detection of the sample. It includes a stepping displacement platform and platform controller, which can adjust the motion speed and displacement accuracy to ensure that the laser excitation module and ultrasonic receiving module cover all detection areas of the sample.
[0068] The detection and verification module is used to verify and correct defect detection results using cross-validation and error correction algorithms. It has a built-in K-fold cross-validation unit and error correction unit, corresponding to the processing in step S7, and provides feedback on the verification results and triggers parameter adjustments. It can automatically calculate the detection error and complete the defect parameter correction without manual intervention.
[0069] The main control module is electrically connected to each of the above modules and is used to control the coordinated operation of each module and adjust the working parameters of each module. It has a built-in data storage unit and display unit to store signal data, characteristic parameters and detection results, and to display the detection process, signal waveforms and defect parameters. The main control module supports both manual adjustment and automatic control modes, enabling automated detection of batch samples. The modules are connected through a high-speed data bus to ensure the real-time performance and accuracy of signal transmission, adapting to the online detection requirements in the mass production of high-entropy alloys.
[0070] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0071] 1. In this invention, a non-contact laser ultrasonic detection method is adopted, which eliminates the need for coupling agent and avoids damage to the surface of high-entropy alloy samples. The laser pulse parameters can be adaptively adjusted according to the thermal conductivity characteristics of high-entropy alloys, effectively improving the detection adaptability to high-entropy alloy samples of different types and with different microstructures.
[0072] 2. In this invention, a multi-algorithm fusion approach is used to perform preprocessing, feature extraction, and anomaly identification on ultrasonic signals. This approach can efficiently suppress noise and eliminate redundant information, and can accurately distinguish between defect signals and lattice distortion interference signals, significantly improving the accuracy of internal defect identification and overall detection precision of high-entropy alloys.
[0073] 3. This invention constructs an integrated detection system with multiple modules working collaboratively, realizing automated and comprehensive scanning and batch online detection of high-entropy alloy samples. The verification and correction of detection results can be completed automatically, and the detection parameters can be dynamically adjusted according to the accuracy requirements without manual intervention, thereby improving the automation level and detection efficiency of defect detection, while ensuring the stability and reliability of the detection results. Attached Figure Description
[0074] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.
[0075] In the attached diagram:
[0076] Figure 1 This is a flowchart of the non-contact laser-ultrasound method for detecting internal defects in high-entropy alloys according to the present invention;
[0077] Figure 2 This is a modular framework diagram of the non-contact laser ultrasonic high-entropy alloy internal defect detection system of the present invention. Detailed Implementation
[0078] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.
[0079] Example: This invention provides a non-contact laser ultrasonic internal defect detection method and system for CoCrFeMnNi series high-entropy alloy plates with a thickness of 5~10mm, which may have internal defects such as pores and microcracks. The detection accuracy requirement is not less than 94%, which is suitable for batch online detection needs.
[0080] like Figure 1 The image shows a non-contact laser-ultrasound method for detecting internal defects in high-entropy alloys. The specific implementation steps are as follows:
[0081] S1. Surface pretreatment: The surface of the CoCrFeMnNi high-entropy alloy plate is wiped with anhydrous ethanol and argon to remove surface oil, dust and oxide layer. After wiping, it is dried with argon to avoid surface oxidation. The surface roughness of the sample is controlled in the range of 0.5~2μm after pretreatment, and no coupling agent is required.
[0082] S2. Laser Pulse Excitation: A pulsed laser is used, and the parameters are adjusted adaptively based on the thermal conductivity of the CoCrFeMnNi high-entropy alloy. (W / m Kelvin), Adjust laser pulse parameters: wavelength The pulse width is 5ns, and the peak energy is obtained through the formula. Calculate, take The peak energy obtained is 1.6~2.1mJ; after the laser pulse is focused by the focusing lens, it acts on the sample surface and excites longitudinal and transverse waves inside the sample through the thermoelastic effect.
[0083] S3. Signal Reception: A laser interferometer is used as the laser receiving device, with the receiving frequency set to 2MHz. The receiving angle is adjusted by a two-dimensional deflection mirror to ensure the reception of ultrasonic reflection signals from different positions on the sample surface. The received signal is amplified by a preamplifier (gain set to 60dB), and then transmitted to the signal processing module after electromagnetic interference is reduced by a shielding unit.
[0084] S4. Signal Preprocessing: Preprocessing is performed using a wavelet packet transform + PCA fusion algorithm. The wavelet packet decomposition level j=5 levels, and the decomposition nodes... ; Noise reduction threshold Calculated using the following formula:
[0085] ,
[0086] Among them, the standard deviation of noise Number of signal sampling points ,get Signal dimensionality after PCA dimensionality reduction The original signal dimension n=8, achieving noise suppression and redundant information removal.
[0087] S5. Feature Extraction and Anomaly Identification: Extract the amplitude, frequency, phase, waveform abrupt change points, and signal attenuation coefficient of the preprocessed effective signal. For this CoCrFeMnNi high-entropy alloy, 100 sets of feature values of defect-free samples were pre-collected to determine the maximum and minimum values of each feature dimension: amplitude Max(x1) = 100mV, Min(x1) = 5mV; frequency Max(x2) = 2MHz, Min(x2) = 1.8MHz; phase Max(x3) = 2π, Min(x3) = 0; number of waveform abrupt change points Max(x4) = 2, Min(x4) = 0; attenuation coefficient Max(x5) = 0.6dB / mm, Min(x5) = 0.4dB / mm. Perform min-max normalization on each feature value of the tested sample, mapping all features to the [0,1] interval, and construct a feature parameter vector based on the normalized feature parameters. The LOF algorithm is used to initially screen out abnormal samples, and the number of neighboring samples is set. The calculated values are compared with a set threshold to identify candidate anomaly samples; these candidate anomaly samples are then input into an SVM classifier, with kernel function parameters... It classifies defect signals and lattice distortion interference signals by using a built-in pre-trained classification model (trained for CoCrFeMnNi system high-entropy alloys).
[0088] S6. Defect Analysis: Obtain the mechanical parameters of this CoCrFeMnNi high-entropy alloy: elastic modulus Poisson's ratio ,density Lattice distortion coefficient The specific method of obtaining it is as follows:
[0089] This CoCrFeMnNi high-entropy alloy is a quinary alloy with equal atomic ratios. The main atoms are Co (atomic radius 0.125 nm), Cr (atomic radius 0.128 nm), Fe (atomic radius 0.124 nm), Mn (atomic radius 0.127 nm), and Ni (atomic radius 0.125 nm). The lattice distortion coefficient is calculated using the atomic radius mismatch formula commonly used in the field of high-entropy alloys.
[0090] ,
[0091] In this case, n=5 is the number of principal components. Let be the atomic fraction of the i-th principal component (0.2 for equal atomic ratios). Let be the atomic radius of the i-th principal component. Let $\frac{ ... =0.1258nm, lattice distortion coefficient δ=0.05, consistent with experimental results.
[0092] The speed of ultrasonic wave propagation is calculated using the following formula:
[0093] =5800m / s,
[0094] The defect depth is calculated using the following formula:
[0095] ,
[0096] in ; through formula Inversely derive the defect size, where the attenuation coefficient is... When the defect is porosity Determining the location of defects using the triangulation method formula .
[0097] S7. Detection Verification and Correction: K-fold cross-validation (K=8) is used to verify the detection results, and the average detection error is calculated. The defect depth is corrected using the following formula:
[0098] ,
[0099] ;
[0100] in, The corrected defect depth; The depth of the defect before correction; This is the error correction amount; The average detection error of K-fold cross-validation;
[0101] Determine whether the corrected detection accuracy is not lower than the preset threshold. If not, return to step S4 to adjust the number of wavelet packet decomposition layers and PCA dimensionality reduction, or step S5 to adjust the number of LOF neighbors and SVM kernel function parameters, and re-perform detection until the detection accuracy meets the requirements. The preset threshold is dynamically set according to the application scenario of high-entropy alloys.
[0102] like Figure 2 The image shows a non-contact laser-ultrasound internal defect detection system for high-entropy alloys. The specific working process is as follows:
[0103] CoCrFeMnNi high-entropy alloy plates are placed in batches on the stepping displacement platform of the motion control module. The main control module switches to automatic control mode and sets batch detection parameters. The sample pretreatment module pretreats the surface of the plates. After the pretreatment is completed, the main control module controls the laser excitation module, ultrasonic receiving module, signal processing module, and detection and verification module to work together.
[0104] The pulsed laser in the laser excitation module enables continuous adjustment of wavelength, pulse width, and peak energy to adapt to high-entropy alloy plates of different thicknesses (5~10mm) in this batch; the parameter adaptive adjustment unit automatically adjusts the laser pulse parameters according to the thermal conductivity of each plate to ensure the excitation effect.
[0105] The ultrasonic signal received by the ultrasonic receiving module is amplified and processed to resist interference before being transmitted to the signal processing module. The SVM classification unit of the signal processing module calls the classification model trained for CoCrFeMnNi system high-entropy alloys to complete feature extraction, anomaly identification and defect analysis. The model parameters can be updated according to the subtle differences in the composition of the plate.
[0106] The detection and verification module automatically calculates the detection error of each board and corrects the defect parameters without manual intervention; the motion control module drives the board to move precisely, realizing comprehensive scanning detection. The detection data is transmitted to the storage unit of the main control module in real time, and the display unit displays the detection process, signal waveforms and defect parameters in real time.
[0107] The modules are connected via a high-speed data bus with a signal transmission delay of ≤1ms, ensuring the real-time performance and accuracy of the detection. After batch detection is completed, the main control module automatically generates a detection report, realizing automated online detection of batch samples and adapting to the needs of high-entropy alloy mass production.
[0108] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound, characterized in that, Includes the following steps: S1. Perform surface pretreatment on the high-entropy alloy test sample to remove impurities and oxide layer from the sample surface; S2. A pulsed laser is used to emit laser pulses onto the surface of the pretreated high-entropy alloy sample through the thermoelastic effect, which excites the generation of ultrasonic waves inside the sample. The laser pulse parameters are adaptively adjusted according to the thermal conductivity of the high-entropy alloy. S3. A laser receiving device is used to receive ultrasonic signals reflected from the sample surface in a non-contact manner. The ultrasonic signals include defect reflection signals, defect-free region reflection signals, and lattice distortion interference signals. S4. Wavelet packet transform and principal component analysis fusion algorithm are used to preprocess the received ultrasonic signal to achieve noise suppression and redundant information removal. S5. Based on the preprocessed effective signal, the defect feature parameters are extracted and the abnormal signal is accurately identified by the fusion algorithm of local outlier factor and support vector machine, so as to effectively distinguish between defect signal and lattice distortion interference signal. S6. Combining the mechanical parameters and lattice distortion coefficient of the high-entropy alloy, the defect signal is analyzed using an improved ultrasonic propagation model to determine the type, size, depth, and location of internal defects in the high-entropy alloy. S7. Cross-validation and error correction algorithms are used to verify and correct the defect detection results to ensure that the detection accuracy meets the preset requirements.
2. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 1, characterized in that: In step S1, the surface pretreatment specifically includes: wiping the surface of the high-entropy alloy sample with anhydrous ethanol and inert gas to remove surface oil, dust and oxide layer, and then drying it with inert gas; after pretreatment, the surface roughness of the sample is in the micron range.
3. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 2, characterized in that: In step S2, the laser pulse excitation process is as follows: the laser pulse emitted by the pulsed laser is focused by a focusing lens and acts on the surface of the high-entropy alloy sample; through the thermoelastic effect, the thin layer on the sample surface is instantly heated and expanded to generate stress waves, which propagate into the sample to form longitudinal and transverse waves; the laser pulse parameters are adaptively adjusted by the thermal conductivity coefficient of the high-entropy alloy, and the adjustment formula is: ； Where E is the peak energy of the laser pulse; k1 is the adaptive adjustment coefficient; The wavelength of the laser pulse; The thermal conductivity coefficient of the high-entropy alloy is dynamically obtained based on its composition and microstructure.
4. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 3, characterized in that: In step S3, the laser receiving device is a laser interferometer, and the receiving angle is adjusted by a two-dimensional deflection mirror during the receiving process; the received signal is amplified by a preamplifier and then transmitted to the signal processing module.
5. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 4, characterized in that: In step S4, the preprocessing of the wavelet packet transform and principal component analysis fusion algorithm specifically includes: S41. Wavelet packet transform is used to perform multi-scale decomposition on the received ultrasonic signal. The decomposition formula is as follows: ； in, For the low-frequency component of the j-th level wavelet packet decomposition, Let be the high-frequency components of the j-th level wavelet packet decomposition; h(n) be the low-pass filter coefficients, g(n) be the high-pass filter coefficients; and j be the decomposition level. For the decomposition node, Z represents the range of values for the discrete index n, which is all integers; S42. Perform threshold denoising on the decomposed low-frequency components to remove residual noise. The threshold calculation formula is as follows: ； in, The noise reduction threshold; The noise standard deviation is estimated from the noise components; N is the number of signal sampling points. S43. Perform PCA dimensionality reduction on the denoised signal to remove redundant information and retain effective feature components. The core formula for PCA dimensionality reduction is: ； Where Y is the effective signal matrix after dimensionality reduction; X sig This is the original signal matrix after denoising; is the mean matrix of the original signal matrix; W is the eigenvector matrix, which is composed of the eigenvectors corresponding to the first k3 largest eigenvalues of the covariance matrix of the original signal matrix; m is the number of signal sampling points, n is the dimension of the original signal, and k3 is the dimension of the signal after dimensionality reduction.
6. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 5, characterized in that: In step S5, the feature extraction and anomaly identification using the fusion algorithm of local outlier factor and support vector machine specifically includes: S51. Extract the feature parameters of the preprocessed effective signal. The feature parameters include signal amplitude, frequency, phase, waveform abrupt change points, and signal attenuation coefficient. Perform dimensionless normalization processing on each feature parameter to eliminate the differences in dimensions and numerical magnitudes of different features. Construct a feature parameter vector based on the normalized feature parameters. ,in These are the normalized amplitude eigenvalues. These are the normalized frequency eigenvalues. These are the normalized phase eigenvalues. This represents the characteristic value of the number of abrupt change points in the normalized waveform. The normalized characteristic value of the signal attenuation coefficient; S52. The LOF algorithm is used to perform preliminary anomaly screening on the feature parameter vector. The formula for the LOF algorithm is: ； in, For the sample Local outlier, sample For feature parameter vectors; The preset number of neighborhood samples; For the sample of The set of all samples within the neighborhood; For the sample Locally achievable density; ； in, For the sample To its The average reachability distance of all samples within the neighborhood; S53. Input the candidate anomaly samples into the SVM classifier for accurate classification, distinguishing between defect signals and lattice distortion interference signals. The SVM classifier uses the radial basis function kernel function, and the kernel function formula is: ； in, For the sample With sample The kernel function value; These are kernel function parameters; For the sample With sample The Euclidean distance; the classification decision function of the SVM classifier is: ； in, For the classification results, 1 represents a defect signal and -1 represents an interference signal; These are the support vector coefficients; is the sample label; l is the number of support vectors; b is the classification threshold.
7. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 6, characterized in that: In step S6, the improved ultrasonic propagation model, combined with the high-entropy alloy lattice distortion coefficient, achieves accurate calculation of defect parameters, specifically including: S61. Obtain the mechanical parameters and lattice distortion coefficient of the high-entropy alloy, wherein the mechanical parameters include the elastic modulus. Poisson's ratio and the propagation speed of ultrasound v, lattice distortion coefficient The ultrasonic wave propagation velocity is calculated based on the elastic modulus and Poisson's ratio using the following formula: ； in, The density of a high-entropy alloy; S62. Combining the lattice distortion coefficient to correct the ultrasonic wave propagation time difference, the defect depth is calculated using the following correction formula: ； in, For the depth of the defect; The original propagation time difference of the ultrasound wave; This is the time difference correction caused by lattice distortion, derived from the lattice distortion coefficient. It was estimated; This is the lattice distortion correction factor; S63. Based on the amplitude attenuation characteristics of the defect reflection signal and combined with the ultrasonic attenuation model, the defect size is calculated using the following formula: ； in, The amplitude of the defect reflection signal; The amplitude of the reflected signal from the defect-free region; The attenuation coefficient of ultrasound waves inside a high-entropy alloy; This is the coefficient representing the effect of lattice distortion on attenuation. This is a defect shape correction factor, which is used to infer the defect size. S64. Based on the laser excitation position coordinates Coordinates of the ultrasound receiving location By combining the ultrasonic wave propagation path, the specific location of the defect can be determined. The coordinates are calculated using the triangulation method, and the formula is: in, The time from laser excitation to defect reflection. The time it takes for the defect to be reflected to the ultrasonic receiver.
8. The method for detecting internal defects in high-entropy alloys based on non-contact laser-ultrasound as described in claim 7, characterized in that: In step S7, the cross-validation and error correction algorithm specifically includes: S71: K-fold cross-validation is used to verify the detection results. The detection samples are divided into K groups, and K-1 groups are selected in turn as the training set and 1 group as the test set. The detection error of each test set is calculated. S72: Calculate the average value of the detection errors of K groups as the overall detection error. The error correction formula is as follows: ， ， in, The corrected defect depth; The depth of the defect before correction; This is the error correction amount; The average detection error of K-fold cross-validation; S73: Determine whether the corrected detection accuracy is not lower than the preset threshold. If not, return to step S4 to adjust the number of wavelet packet decomposition layers and PCA dimensionality reduction, or step S5 to adjust the number of LOF neighborhoods and SVM kernel function parameters, and re-perform detection until the detection accuracy meets the requirements. The preset threshold is dynamically set according to the application scenario of high-entropy alloys.
9. A non-contact laser-ultrasonic high-entropy alloy internal defect detection system for implementing the method of any one of claims 1-8, characterized in that, include: The sample pretreatment module is used to pretreat the surface of high-entropy alloy test samples by wiping with anhydrous ethanol and inert gas to remove surface impurities and oxide layers. The laser excitation module is used to emit laser pulses onto the surface of a pretreated high-entropy alloy sample through a pulsed laser via thermoelastic effect, thereby exciting the generation of ultrasonic waves inside the sample. It includes a pulsed laser, a focusing lens, a two-dimensional deflection mirror, and a parameter adaptive adjustment unit, wherein the parameter adaptive adjustment unit is based on the thermal conductivity coefficient of the high-entropy alloy. An ultrasonic receiving module is used to receive ultrasonic signals reflected from the surface of a sample non-contactly using a laser interferometer. It includes a laser interferometer, a preamplifier, and a shielding unit. The preamplifier amplifies the signal, the shielding unit reduces electromagnetic interference, and the processed signal is transmitted to a signal processing module. The signal processing module is used to preprocess, extract features, identify anomalies, and analyze defects in the received ultrasonic signals. It includes a wavelet packet decomposition unit, a PCA dimensionality reduction unit, a LOF algorithm unit, an SVM classification unit, and an improved ultrasonic propagation model unit. The motion control module is used to drive the high-entropy alloy sample to perform precise motion and realize comprehensive scanning and detection of the sample, including a stepping displacement platform and a platform controller. The detection and verification module is used to verify and correct the defect detection results using cross-validation and error correction algorithms. It has a built-in K-fold cross-validation unit and error correction unit. The main control module is electrically connected to each of the above modules and is used to control the coordinated operation of each module and adjust the working parameters of each module. It has a built-in data storage unit and display unit to store signal data, characteristic parameters and detection results, and to display the detection process, signal waveforms and defect parameters.