Magnetic shoe dark crack defect detection method based on acoustic vibration signal decomposition

Abstract

The invention belongs to the technical field of magnetic shoe internal dark crack detection, and particularly discloses a magnetic shoe dark crack defect detection method based on acoustic vibration signal decomposition, which comprises the following steps: processing an acquired acoustic vibration signal by using a filter, then acquiring two maximum-amplitude frequency values in a frequency range of 9KHz to 20KHz according to the frequency spectrum information of the magnetic shoe acoustic vibration signal, and calculating the magnetic shoe dark crack defect according to the two maximum-amplitude frequency values. The two frequency values are used as center frequency values of a VMD algorithm to perform two-mode decomposition, center frequency, energy, a zero-crossing rate, a spectral centroid and a kurtosis spectrum are calculated from the two decomposed mode components to serve as feature vectors, and then respective weight coefficients are distributed according to the expression degree of the feature vectors to the magnetic shoe acoustic vibration signals; and then the feature vector is substituted into a feature vector space to complete feature recognition, so that a set of detection algorithm with high detection accuracy and high detection efficiency is formed, and whether a crack exists in the magnetic shoe or not is judged through frequency spectrum information of a magnetic shoe acoustic vibration signal.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic tile internal hidden crack detection, and in particular to a magnetic tile hidden crack defect detection method based on acoustic vibration signal decomposition. Background Art

[0002] Magnetic tile is a tile-shaped magnetic material mainly made of composite materials such as Ru-Fe-B. As a key component for generating a constant magnetic field in the motor, magnetic tile is widely used in electronic manufacturing industries such as automobiles, air conditioners, televisions, microwave ovens, fitness equipment, toys, transformers, etc. Due to the complex manufacturing process, the pressing force of the die during the production of magnetic tiles, the sintering temperature and other unstable factors may cause defects in the magnetic tiles. Defects of varying degrees will affect the performance of the magnetic tile, weaken the strength of the magnetic tile, and even cause the magnetic tile to break, causing motor failure and safety accidents. Therefore, product quality inspection is required before the magnetic tile is packaged. The defects of magnetic tiles are divided into external defects and internal defects. The internal defects of magnetic tiles are characterized by invisibility, random location, degree uncertainty and complexity of causes, which are difficult to detect and avoid in time in actual production. At present, the internal defects of magnetic tiles are mainly detected by artificial listening. However, the human ear detection method is gradually unable to meet the production efficiency required by enterprises. In addition, high-intensity detection can easily cause auditory fatigue of the inspectors, resulting in missed detection and misjudgment, which in turn affects the detection quality.

[0003] In order to avoid a series of problems caused by artificial listening and improve detection efficiency, acoustic vibration detection is selected as the detection method. The difficulty of acoustic vibration detection lies in filtering out the high and low frequency noise of the environment and extracting the acoustic vibration signal of the magnetic tile that can reflect the internal situation of the magnetic tile.

[0004] For example, the patent CN112464923A entitled "A method for detecting internal defects of magnetic tiles based on improved variational modal decomposition" constructs a fitness function to evaluate the influence of VMD parameters on the decomposition effect of magnetic tile acoustic vibration signals, uses SSA optimization to obtain the optimal parameter settings, decomposes the signal and extracts the maximum peak frequency of the main mode as the defect detection feature, and finally uses KNN classification to determine whether there are internal defects in the magnetic tiles. However, this method has the following shortcomings: (1) It does not process environmental noise, which increases the amount of calculation and may cause failure in extracting effective signals; (2) It decomposes too many modes, which affects the detection efficiency. In fact, only the characteristic modal signal needs to be extracted; (3) Only a single frequency peak index is used as a feature quantity, and multi-dimensional related feature quantities should be added. (4) Defective samples cannot be fully covered due to the large differences in dark crack positions and lengths. Only good magnetic tile signals should be classified to improve detection reliability and efficiency. Summary of the invention

[0005] The purpose of the present invention is: to address the above-mentioned problems, the present invention provides a method for detecting hidden crack defects in magnetic tiles based on acoustic vibration signal decomposition, by performing a series of operations such as filtering, spectrum transformation, modal decomposition, and feature extraction on the magnetic tile acoustic vibration signal, and then substituting the feature vector into the feature vector space to complete feature recognition, thereby forming a set of detection algorithms with high detection accuracy and high detection efficiency, and realizing the judgment of whether there are cracks inside the magnetic tile through the spectrum information of the magnetic tile acoustic vibration signal.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A method for detecting hidden crack defects in magnetic tiles based on acoustic vibration signal decomposition, the method comprising:

[0008] Acquire the acoustic vibration signal generated when the magnetic tile collides with the exciting block;

[0009] Filtering the acoustic vibration signal to obtain a filtered signal that retains the characteristic frequency of the acoustic vibration signal of the magnetic tile;

[0010] The magnetic tile acoustic vibration signal is subjected to spectrum transformation, modal decomposition, and feature extraction to obtain a feature vector of the magnetic tile acoustic vibration signal;

[0011] The feature vector is brought into the feature vector space for feature judgment to obtain the magnetic tile dark crack defect information.

[0012] Furthermore, in the process of acquiring the acoustic vibration signal, a collision threshold is set, and when the threshold is reached, the magnetic shoe acoustic vibration signal is continuously extracted from the filtered signal.

[0013] Furthermore, the spectrum transformation specifically includes:

[0014] Perform FFT transformation on the acoustic vibration signal of the magnetic tile, and find the two frequency points (f1, f2) with the largest amplitude from the spectrum characteristics. The two frequency points satisfy the following formula:

[0015] 9.0KHz <fi<20KHz i=1,2 (1)。

[0016] Furthermore, the modal decomposition specifically includes:

[0017] Taking two frequency points (f1, f2) as the initial values ​​of modal frequencies, the spectrum signal of the magnetic tile acoustic vibration signal is subjected to two-mode VMD decomposition to obtain two modal components (Y1, Y2).

[0018] Furthermore, feature extraction includes:

[0019] Perform IFFT transformation on the two modal components (Y1, Y2) to obtain two modal decomposed time domain signals (S1, S2);

[0020] According to the two modal components (Y1, Y2) and the time domain signals (S1, S2), the center frequency, maximum energy, zero-crossing rate, spectral centroid and kurtosis spectrum of the two modes are calculated as the characteristic vectors of the magnetic tile acoustic vibration signal.

[0021] Furthermore, the center frequency is specifically obtained by the following formula:

[0022]

[0023] Where ω i is the center frequency;

[0024] The maximum energy is obtained by the following formula:

[0025] E m =max(abs(Y i (ω))) i=1,2 (3)

[0026] In the formula, E m is the maximum energy;

[0027] The zero-crossing rate is obtained by the following formula:

[0028]

[0029] In the formula, Z cr is the zero-crossing rate;

[0030] The spectral centroid is obtained by the following formula:

[0031]

[0032] Where SC is the spectral centroid;

[0033] The kurtosis spectrum is obtained by the following formula:

[0034]

[0035] Where K4 is the kurtosis spectrum.

[0036] Furthermore, before the judgment, according to the modal decomposition results of the acoustic vibration signal of the good magnetic tile, the frequency points of the mode are concentrated near three frequency points. Therefore, the two-mode decomposition of the good magnetic tile has three cases through the free combination of the three frequency points. For the three cases, the mean of the eigenvectors is calculated as the cluster center (C i(j) , i=1,2,3,j=1,2,3), and then set the decision threshold θi(i=1,2,3).

[0037] Furthermore, the judgment process is as follows:

[0038] Calculate the eigenvectors of the acoustic vibration signal of the magnetic tile to be tested and the three cluster centers C respectively iThe distance Ri, i = 1, 2, 3; the good product judgment is carried out through the following formula to obtain the dark crack defect information of the magnetic tile:

[0039] Ri<θi i=1,2,3 (7)

[0040] If the formula (6) is satisfied, it is judged as a good product, otherwise it is judged as a defective product, that is, the magnetic tile contains a hidden crack defect.

[0041] Furthermore, the distance Ri is obtained by the following formula:

[0042]

[0043] In the formula, F j (j=1,…,10) is the eigenvector of the magnetic tile to be tested, and α is the weight coefficient.

[0044] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0045] The magnetic tile hidden crack defect detection method based on acoustic vibration signal decomposition of the present invention filters out high and low frequency signals from the environment through a filtering algorithm, and retains only the high frequency characteristics reflecting the internal conditions of the magnetic tile as much as possible. This not only greatly reduces the calculation amount of the algorithm, but also avoids the characteristic frequency of the magnetic tile acoustic vibration signal being covered by other frequencies.

[0046] The present invention processes the acoustic vibration signal of the magnetic tile to find the initial value of the characteristic frequency, which not only reduces the dependence of the decomposition algorithm on the algorithm parameters, but also reduces the complexity of the algorithm, thereby reducing the processing time of the algorithm and further improving the efficiency of the algorithm in processing the acoustic vibration signal of the magnetic tile.

[0047] The present invention analyzes the acoustic vibration signals of the magnetic tiles and calculates characteristic values ​​such as characteristic frequency, maximum energy, zero-crossing rate, spectral centroid and kurtosis spectrum to measure whether there are dark cracks in the magnetic tiles, thereby further improving the accuracy of the algorithm in judging the acoustic vibration signals of the magnetic tiles. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of a method for detecting hidden crack defects in magnetic tiles based on acoustic vibration signal decomposition according to the present invention;

[0049] Figure 2 This is an analysis diagram of acoustic vibration signals of good magnetic tiles in the magnetic tile hidden crack defect detection method based on acoustic vibration signal decomposition of the present invention;

[0050] Figure 3 It is an analysis diagram of acoustic vibration signals of dark-cracked magnetic tiles in the magnetic tile dark crack defect detection method based on acoustic vibration signal decomposition of the present invention;

[0051] Figure 4It is a characteristic spectrum diagram of the unfiltered acoustic vibration signal of a good magnetic tile in the magnetic tile dark crack defect detection method based on acoustic vibration signal decomposition of the present invention;

[0052] Figure 5 It is a schematic diagram of the characteristic space of the acoustic vibration of the magnetic tile in the magnetic tile hidden crack defect detection method based on the decomposition of the acoustic vibration signal of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0054] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0055] Example

[0056] This embodiment provides a method for detecting hidden crack defects in a magnetic tile based on acoustic vibration signal decomposition. Figure 1 As shown, specifically including:

[0057] (1) Obtain the acoustic vibration signal generated when the magnetic tile collides with the excitation block:

[0058] To obtain the acoustic vibration signal generated when the magnetic tile collides with the excitation block, pay attention to the distance from the falling point of the magnetic tile to the excitation block. If the distance is too large, the magnetic tile will be damaged, and if the distance is too small, the acoustic vibration signal cannot be fully excited.

[0059] When the moment of magnetic tile collision is detected, the effective magnetic tile acoustic vibration signal (4096 points) is extracted from the filtered signal: a suitable value is taken as the threshold, which is slightly larger than when there is no collision. When the threshold is reached, 4086 points are taken in succession, and then the 10 points before the threshold are added to form 4096 points.

[0060] (2) Filtering of acoustic vibration signal:

[0061] The acoustic vibration signal is filtered to remove DC and high- and low-frequency environmental noise, thereby obtaining a filtered signal that retains the characteristic frequency of the magnetic tile acoustic vibration signal.

[0062] (3) Perform FFT transformation on the acoustic vibration signal: Perform FFT transformation on the filtered acoustic vibration signal.

[0063] (4) Obtain the maximum characteristic frequency: After performing FFT transformation on the acoustic vibration signal of the magnetic tile, find the two frequency points (f1, f2) with the largest amplitude from the spectrum characteristics: From the spectrum characteristics, traverse and compare the frequencies in the range of 9KHz to 20KHz to obtain the frequency f1 with the largest amplitude, then clear the amplitude of the frequency band with this frequency as the center frequency and a bandwidth of 2KHz, and then traverse and compare the frequencies in the range of 10KHz to 20KHz again to obtain the frequency f2 with the largest amplitude.

[0064] (5) Perform bimodal decomposition on the acoustic vibration signal: Taking (f1, f2) as the initial value of the modal frequency, perform bimodal VMD decomposition on the spectrum signal of the magnetic tile acoustic vibration signal to obtain two modal components (Y1, Y2), specifically including:

[0065] Since the initial values ​​of the modal frequencies (f1, f2) have been determined, the penalty parameter alpha of VMD has little effect. It is empirically set to twice the signal length and the number of decomposition levels K is set to 2.

[0066] The value of the maximum number of iterations T is set to 30, and the parameters of each iteration are calibrated until two non-aliased modal components (Y1, Y2) are decomposed and output;

[0067] (6) Modal component IFFT transformation: The two modal components (Y1, Y2) are subjected to a 4096-point IFFT transformation to obtain two modal decomposed time domain signals (S1, S2).

[0068] (7) Calculate the required characteristic quantities: Based on the modal components (Y1, Y2) and the time domain signals (S1, S2), calculate the center frequency, maximum energy, zero-crossing rate, spectral centroid and kurtosis spectrum of the two modes as the characteristic vectors of the magnetic tile acoustic vibration signal:

[0069]

[0070] Calculate the center frequency ω i ;

[0071] The maximum energy is obtained by the following formula:

[0072] E m =max(abs(Y i (ω))) i=1,2 (3)

[0073] In the formula, E m is the maximum energy;

[0074] The zero-crossing rate is obtained by the following formula:

[0075]

[0076] In the formula, Zcr is the zero-crossing rate;

[0077] The spectral centroid is obtained by the following formula:

[0078]

[0079] Where SC is the spectral centroid;

[0080] The kurtosis spectrum is obtained by the following formula:

[0081]

[0082] Where K4 is the kurtosis spectrum.

[0083] (8) Calculate the weighted distance of the feature vector in the feature space:

[0084] Before the judgment, according to the modal decomposition results of the acoustic vibration signal of the good magnetic tile, the modal frequency points are concentrated near three frequency points. Therefore, the two-mode decomposition of the good magnetic tile has three situations through the free combination of the three frequency points. For the three situations, the mean of the eigenvectors is calculated as the cluster center (C i(j) , i=1,2,3, j=1,2,3), and then set the decision threshold θi(i=1,2,3); calculate the eigenvector of the acoustic vibration signal of the magnetic tile to be detected and the three cluster centers (C i ) distance Ri (i=1,2,3), it is judged as good product (no dark crack defect in the magnetic tile) when one of the following conditions is met:

[0085] Ri<θi i=1,2,3 (7)

[0086] If the formula (6) is satisfied, it is judged as a good product, otherwise it is judged as a defective product, that is, the magnetic tile contains a hidden crack defect.

[0087] Specifically, after obtaining the eigenvector of the detection magnetic tile, first multiply its five eigenvalues ​​by their respective weight coefficients, and then calculate according to the formula:

[0088]

[0089] In the formula, in the formula, F j (j=1,…,10) is the eigenvector of the magnetic tile to be tested, α is the weight coefficient, the weight coefficient is determined by the importance of each eigenvalue in the eigenvector to the judgment result, and its size is determined by the correlation and identity of the eigenvalues ​​calculated from the spectrum of good magnetic tiles, and the sum of the weight coefficients is 1.

[0090] like Figure 2 As shown in the figure, the signal analysis diagram of the acoustic vibration signal processing of good magnetic tiles is further described:

[0091] The filtered time domain image is as follows Figure 2 (a) is shown; the spectrum after FFT transformation is shown Figure 2 As shown in (b), the interference frequency of the signal below 9KHz and above 20KHz is relatively small, indicating that the filtering effect meets the requirements; the modal component decomposition is also relatively complete, and the spectrum of the modal component is as follows Figure 2 (c) and Figure 2 (d) as shown.

[0092] like Figure 3 As shown in the figure, the signal analysis diagram of the acoustic vibration signal processing of the dark cracked magnetic tile is further described:

[0093] The filtered time domain image is as follows Figure 3 (a) is shown; the spectrum after FFT transformation is shown Figure 3 As shown in (b), the interference frequency of the signal below 10KHz and above 20KHz is relatively small, indicating that the filtering effect meets the requirements; the modal component decomposition is also relatively complete, and the spectrum of the modal component is as follows Figure 3 (c) and Figure 3 (d) as shown.

[0094] like Figure 4 As shown in the figure, the four spectral characteristics of the acoustic vibration signal of good magnetic tiles are further described:

[0095] A large number of good magnetic tiles of this model in different batches can be divided into four categories. Since dark cracked magnetic tiles also have similar low-frequency parts, the low-frequency part signal is not characteristic. At least two of the high-frequency characteristic frequencies of the four types of good magnetic tiles fall between 10~19KHz and 16~19KHz. The maximum characteristic frequency and modal decomposition are also processed according to the characteristics of the characteristic frequency.

[0096] like Figure 5 As shown in the figure, the feature vector space of the acoustic vibration signal of good magnetic tiles is further described as follows:

[0097] According to the characteristic vector characteristics of the acoustic vibration signals of a large number of good magnetic tiles of different batches of this model, such as Figure 4 As shown, construct Figure 5 The three cluster centers C shown i and the decision threshold θi to cover the three most likely frequency points where the peak of the magnetic tile acoustic vibration signal spectrum appears; the specific method is:

[0098] 10,000 good magnetic tile samples were selected to obtain acoustic vibration signals, and the method of this embodiment was used to operate. From the modal decomposition results, it was found that the frequency points of the modes were basically concentrated near three frequency points: from low frequency to high frequency, recorded as frequency point 1, frequency point 2, and frequency point 3; and the other four eigenvalues ​​corresponding to each of the three frequency points also showed consistency in the acoustic vibration signals of a large number of good magnetic tiles; there were three cases of two-mode decomposition of good magnetic tiles: there were 1547 good products in the first category (frequency points 1, 2), 321 good products in the second category (frequency points 2, 3), and 132 good products in the third category (frequency points 1, 3), and then the eigenvector means of each of the three cases were calculated as the cluster center C i , and then calculate the appropriate decision threshold θi. The three cluster centers are set with appropriate radius as the decision threshold θi according to the allowed missed detection rate and false detection rate, so as to cover the three most likely frequency points of the spectrum peak of the magnetic tile acoustic vibration signal. Then test as many dark-cracked magnetic tiles as possible. If the test results are basically dark-cracked magnetic tiles, the cluster center and decision threshold are successfully constructed.

[0099] As shown in Table 1, the test results of the magnetic tile test samples are further described:

[0100] Table 1 Test results of magnetic tile test samples

[0101] Total number of tests Number of false positives Accuracy Good quality magnetic tile 900 25 97.23％ Defective tiles 100 1 99％ total 1000 26 97.4％

[0102] 1,000 tiles that are not produced in the same batch are selected, including 900 good tiles and 100 dark-cracked tiles. Then, the 1,000 tiles are identified by performing steps S1-S9 in the tile sorting workshop. After the experiment, the recognition rate of the entire detection algorithm for 1,000 tile samples is shown in the following table. It can be concluded from the table that the detection algorithm has an accuracy rate of 97.7% in identifying tiles in the actual sorting environment, and can effectively detect whether there are defects inside the tiles.

[0103] The principles and implementation methods of the present invention are described in this article using specific embodiments. The description of the above embodiments is only used to help understand the method and core ideas of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition, characterized in that: The method comprises: Acquire the acoustic vibration signal generated when the magnetic tile collides with the exciting block; Filtering the acoustic vibration signal to obtain a filtered signal that retains the characteristic frequency of the acoustic vibration signal of the magnetic tile; The magnetic tile acoustic vibration signal is subjected to spectrum transformation, modal decomposition, and feature extraction to obtain a feature vector of the magnetic tile acoustic vibration signal; The feature vector is brought into the feature vector space for feature judgment to obtain the magnetic tile dark crack defect information.

2. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 1 is characterized in that: In the process of acquiring the acoustic vibration signal, a collision threshold is set, and when the threshold is reached, the magnetic tile acoustic vibration signal is continuously extracted from the filtered signal.

3. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 1 is characterized in that: The spectrum conversion specifically includes: Perform FFT transformation on the acoustic vibration signal of the magnetic tile, and find the two frequency points (f1, f2) with the largest amplitude from the spectrum characteristics. The two frequency points satisfy the following formula: 9.0KHz<fi<20KHz i=1, 2 (1).

4. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 3 is characterized in that: The modal decomposition specifically includes: Taking two frequency points (f1, f2) as the initial values ​​of modal frequencies, the spectrum signal of the magnetic tile acoustic vibration signal is subjected to two-mode VMD decomposition to obtain two modal components (Y1, Y2).

5. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 4 is characterized in that: The feature extraction comprises: Perform IFFT transformation on the two modal components (Y1, Y2) to obtain two modal decomposed time domain signals (S1, S2); According to the two modal components (Y1, Y2) and the time domain signals (S1, S2), the center frequency, maximum energy, zero-crossing rate, spectral centroid and kurtosis spectrum of the two modes are calculated as the characteristic vectors of the magnetic tile acoustic vibration signal.

6. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 5 is characterized in that: The center frequency is specifically obtained by the following formula: Where ω i is the center frequency; The maximum energy is obtained by the following formula: E m =max(abs(Y i (ω))) i=1,2 (3) In the formula, E m is the maximum energy; The zero-crossing rate is obtained by the following formula: In the formula, Z cr is the zero-crossing rate; The spectral centroid is obtained by the following formula: Where SC is the spectral centroid; The kurtosis spectrum is obtained by the following formula: Where K4 is the kurtosis spectrum.

7. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 1 is characterized in that: Before the judgment, according to the modal decomposition results of the acoustic vibration signal of the good magnetic tile, the modal frequency points are concentrated near three frequency points. Therefore, the two-mode decomposition of the good magnetic tile has three situations through the free combination of the three frequency points. For the three situations, the mean of the eigenvectors is calculated as the cluster center (C i(j) , i=1,2,3,j=1,2,3), and then set the decision threshold θi(i=1,2,3).

8. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 7 is characterized in that: The specific judgment process is as follows: Calculate the eigenvectors and three cluster centers C of the acoustic vibration signal of the magnetic tile to be tested respectively i The distance Ri, i = 1, 2, 3; the good product judgment is carried out through the following formula to obtain the dark crack defect information of the magnetic tile: Ri<θi i=1,2,3 (7) If the formula (6) is satisfied, it is judged as a good product, otherwise it is judged as a defective product, that is, the magnetic tile contains a hidden crack defect.

9. The method for detecting hidden cracks in magnetic tiles based on acoustic vibration signal decomposition according to claim 8, characterized in that: The distance Ri is obtained by the following formula: In the formula, F j (j=1,…,10) is the eigenvector of the magnetic tile to be tested, and α is the weight coefficient.

Images