A device and method for detecting internal defects of magnetic tiles based on acoustic emission signals

Abstract

The present invention discloses a device and method for detecting internal defects of magnetic tiles based on acoustic emission signals, and relates to the field of internal defect detection of magnetic tiles. The present invention constructs a detection device / method that can be directly adapted to a magnetic tile production line, preprocesses the collected collision acoustic vibration signal and converts it into a digital signal, and then decomposes it into three component signals by an improved clustering symplectic geometric modal decomposition method, selects the optimal component signal as a feature signal by using the minimum decomposition evaluation index, and uses an IITN network to classify the input feature signal to obtain the feature signal classification result, that is, the magnetic tile defect detection result, thereby forming a set of fast and effective magnetic tile defect detection solutions that can be directly used in magnetic tile production lines.

Description

Technical Field

[0001] The present invention relates to the field of magnetic tile internal defect detection, and in particular to a magnetic tile internal defect detection device and method based on acoustic emission signals. Background Art

[0002] Magnetic tiles are a key component widely used in electromagnetic equipment such as motors and generators. They are usually made of permanent magnetic materials and can provide a stable magnetic field. The magnetic properties, mechanical strength and internal structural integrity of the magnetic tiles directly affect the efficiency, life and reliability of the motor. However, the manufacturing process of magnetic tiles is highly complex. Factors such as mold pressure control during pressing, temperature management during the sintering stage, and various other uncertainties can all cause defects in the magnetic tiles. Among the defects of magnetic tiles, internal defects account for more than 60%, and the defects are of uncertain location, size, and shape, which undoubtedly brings great challenges and difficulties to the detection of internal defects.

[0003] The detection of internal defects in the current magnetic tile production process still relies heavily on manual inspection, where inspectors analyze the sound produced when the magnetic tile hits the metal surface to assess the presence of internal defects. This traditional method relies heavily on the auditory perception and subjective judgment of inspectors, resulting in reduced operational efficiency, vague inspection standards, and inconsistent accuracy. The accuracy and speed of inspection are difficult to meet inspection requirements. In response to the problems existing in manual inspection of internal defects in magnetic tiles, a fast and low-cost acoustic inspection method is selected as a detection method to match the fast and low-cost requirements of magnetic tile quality inspection. However, the signal form used for internal defect detection of magnetic tiles is single, and the inherent nonlinearity and non-stationarity of the signal are not conducive to feature extraction and identification. Therefore, it is of great significance to develop a fast, efficient and novel device and method for detecting internal defects in magnetic tiles. Summary of the Invention

[0004] In response to the above-mentioned deficiencies in the prior art, the present invention provides a device and method for detecting internal defects of magnetic tiles based on acoustic emission signals, which solves the problems of low detection efficiency and large influence of human factors in magnetic tile internal defect detection in magnetic tile production lines.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] Provided is a magnetic tile internal defect detection device based on acoustic emission signals, which includes a defect detection component, an excitation inclined plane and an acoustic emission sensor; the acoustic emission sensor is arranged on the excitation inclined plane;

[0007] Acoustic emission sensor, used to capture the impact acoustic vibration signal generated when the magnetic tile falls on the exciting inclined plane;

[0008] The defect detection component is used to receive the collision acoustic vibration signal captured by the acoustic emission sensor and detect the internal defects of the magnetic tile based on the collision acoustic vibration signal.

[0009] The beneficial effects of this device are as follows: the present invention detects magnetic tile defects through the collision acoustic vibration signal generated when the magnetic tile falls on the exciting inclined plane, which is convenient for application on the production line of magnetic tiles, thereby realizing defect detection of magnetic tiles on the production line.

[0010] Furthermore, the defect detection component includes a signal amplifier, a signal acquisition card, a signal decomposition module and a defect recognition module, wherein:

[0011] The signal amplifier is used to amplify and filter the collision sound vibration signal output by the acoustic emission sensor to obtain a pre-processed analog electrical signal;

[0012] A signal acquisition card is used to convert the pre-processed analog electrical signal into a digital signal;

[0013] The signal decomposition module deploys an improved clustered symplectic geometric mode decomposition method to decompose the digital signal into three component signals and select the component signal corresponding to the minimum decomposition evaluation index as its output feature signal;

[0014] The defect recognition module is deployed with a pre-trained IITN network, which is used to use the feature signal output by the signal decomposition module as the input of the pre-trained IITN network. The input feature signal is classified through the IITN network to obtain the feature signal classification result, that is, the defect detection result of the magnetic tile is obtained.

[0015] The beneficial effects of adopting the above-mentioned further scheme are as follows: after the collected collision acoustic vibration signal is preprocessed and converted into a digital signal, it is decomposed into three component signals by the improved clustering symplectic geometric modal decomposition method, and the optimal (with instantaneous impact component) component signal is selected as the feature signal through the minimum decomposition evaluation index. The input feature signal is classified using the IITN network to obtain the feature signal classification result, that is, the defect detection result of the magnetic tile is obtained, thereby forming a set of fast and effective magnetic tile defect detection scheme.

[0016] Furthermore, the IITN network includes a first Inception module, a second Inception module, and a third Inception module connected in sequence, and a global average pooling layer, a flattening layer, and a fully connected layer connected in sequence; the input of the first Inception module is processed by the CBAM attention mechanism and then residually connected with the output of the second Inception module, and the residual connection result is the input of the third Inception module; the output of the third Inception module is the input of the global average pooling layer; the input of the first Inception module is the input of the IITN network; and the output of the fully connected layer is the output of the IITN network.

[0017] The beneficial effects of adopting the above further scheme are: this IITN network removes the repeated redundant modules in the InceptionTime network architecture, and uses three Inception modules and residual connections with CBAM attention mechanism to form a single InceptionTime module. Combined with the global average pooling layer, flattening layer and fully connected layer, it can efficiently and quickly classify the characteristic signals corresponding to the magnetic tiles and determine whether the magnetic tiles have internal defects.

[0018] Furthermore, a first bracket and a second bracket are respectively provided on both sides of the upper end of the excitation slope; a photoelectric transmitter is provided on the first bracket, and a photoelectric receiver is provided on the second bracket; the connection line between the photoelectric transmitter and the photoelectric receiver intersects with the falling path of the magnetic tile, so that the magnetic tile prevents the photoelectric receiver from receiving the signal from the photoelectric transmitter before falling onto the excitation slope, and the time point when the photoelectric receiver does not receive the signal from the photoelectric transmitter is used as the starting point for signal collection of the acoustic emission sensor.

[0019] Furthermore, a first conveyor belt is provided above the excitation slope and next to the photoelectric receiver, and a second conveyor belt is provided at the lower end of the excitation slope; a third conveyor belt is provided on one side of the second conveyor belt, and a pusher is provided on the other side of the second conveyor belt, facing the third conveyor belt;

[0020] The first conveyor belt is used to transport the magnetic tile to the top of the exciting inclined plane, so that the magnetic tile passes through the connection line between the photoelectric transmitter and the photoelectric receiver and falls onto the exciting inclined plane;

[0021] The second conveyor belt is used to transport the fallen magnetic tiles that are detected to be defect-free toward the qualified magnetic tile discharge port;

[0022] A pusher, used for pushing the magnetic tile that has fallen and is detected as defective from the second conveyor belt to the third conveyor belt;

[0023] The third conveyor belt is used to transport the magnetic tiles detected as having defects toward a waste recycling location.

[0024] Furthermore, the magnetic tile is transported on the first conveyor belt in a standing posture, with the convex surface of the magnetic tile facing the transport direction; the falling height of the magnetic tile is 20 mm.

[0025] The beneficial effects of adopting this further solution are as follows: The present invention uses a photoelectric transmitter and a photoelectric receiver to generate a measurement start signal, utilizes a first conveyor belt to drop the magnetic tile onto an excitation ramp, and is then transported by a second conveyor belt when the tile reaches the bottom of the excitation ramp. During transportation, the pusher is activated based on the defect detection results to sort the magnetic tiles, and the tiles are then transported to different stages based on the classification results. The entire device can be directly deployed in a magnetic tile production line, completing defect detection and automatic sorting and transportation of the tiles without human intervention, greatly improving production efficiency.

[0026] A method for detecting internal defects of a magnetic tile based on acoustic emission signals is provided, which comprises the following steps:

[0027] The magnetic tile is dropped onto the vibration inclined surface, and the acoustic emission sensor is used to capture the collision acoustic vibration signal generated when the magnetic tile drops onto the vibration inclined surface;

[0028] The defect detection component receives the collision acoustic vibration signal captured by the acoustic emission sensor, and detects the internal defects of the magnetic tile based on the collision acoustic vibration signal.

[0029] Furthermore, the specific method of receiving the collision acoustic vibration signal captured by the acoustic emission sensor through the defect detection component and detecting the internal defects of the magnetic tile according to the collision acoustic vibration signal includes the following steps:

[0030] A signal amplifier is used to amplify and filter the collision sound vibration signal output by the acoustic emission sensor to obtain a pre-processed analog electrical signal;

[0031] A signal acquisition card is used to convert the pre-processed analog electrical signal into a digital signal;

[0032] A signal decomposition module with an improved clustered symplectic geometric mode decomposition method is used to decompose the digital signal into three component signals, and the component signal corresponding to the minimum decomposition evaluation index is selected as the output feature signal;

[0033] A defect recognition module deployed with a pre-trained IITN network is used to take the characteristic signal output by the signal decomposition module as the input of the pre-trained IITN network. The input characteristic signal is classified through the IITN network to obtain the characteristic signal classification result, that is, the defect detection result of the magnetic tile is obtained.

[0034] Furthermore, the specific method of decomposing the digital signal into three component signals includes the following steps:

[0035] Apply the time series topology equivalence method to construct a one-dimensional time series and form a trajectory matrix :

[0036]

[0037] in The embedding dimension representing the topological equivalence of time series; represents the delay time of the time series topological equivalence method, , Indicates the data length of the digital signal; express In the delay The value after time; Represents the trajectory matrix Middle The element in the first column of the row;

[0038] Trajectory Matrix Perform autocorrelation analysis to obtain the covariance matrix , and then we get the matrix ;

[0039] Get Matrix The eigenvalues ​​and eigenvectors of , using the feature vector and trajectory matrix Construct coefficient matrix , and then get the initial single-component reconstruction matrix ;in ;

[0040] Adjust the elements in each initial single-component reconstruction matrix to obtain the adjusted initial single-component reconstruction matrix: Middle m No.1 in the industry d -1 elements in the column, if ,make ; For the initial single-component reconstruction matrix Other elements in ;in Reconstruct the matrix for the initial single component Middle a Rank b Elements of the column; Reconstruct the matrix for the initial single component Middle b Rank a Elements of the column;

[0041] The first The adjusted initial single-component reconstruction matrix Convert to a one-dimensional time series , and then get a one-dimensional time series ;in Middle k The expression for each element is:

[0042]

[0043] in Reconstruct the matrix for the adjusted initial single component Middle Rank Elements of the column; , min means taking the minimum value; , max means taking the maximum value;

[0044] Agglomerative hierarchical clustering algorithm for one-dimensional time series The signals are reorganized into three components through clustering.

[0045] Furthermore, the calculation expression of the minimum decomposition evaluation index is:

[0046]

[0047] in Represents a single component signal With digital signal The minimum decomposition evaluation index of ; Indicates time; Indicates the data length of a digital signal, that is, the number of data points in a single component signal.

[0048] The beneficial effects of this method are as follows: after the collected collision acoustic vibration signal is preprocessed and converted into a digital signal, it is decomposed into three component signals by the improved clustering symplectic geometric modal decomposition method, and the optimal component signal (with instantaneous impact component) is selected as the feature signal through the minimum decomposition evaluation index. The input feature signal is classified using the IITN network to obtain the feature signal classification result, that is, the defect detection result of the magnetic tile is obtained, thereby forming a fast and effective magnetic tile defect detection solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the structure of the device;

[0050] Figure 2 The waveforms of the acoustic emission signals of qualified and defective magnetic tiles in time domain and frequency domain are shown; (a) is the time domain diagram of the acoustic emission signal of qualified magnetic tiles, (b) is the frequency domain diagram of the acoustic emission signal of qualified magnetic tiles, (c) is the time domain diagram of the acoustic emission signal of defective magnetic tiles, and (d) is the frequency domain diagram of the acoustic emission signal of defective magnetic tiles;

[0051] Figure 3is the decomposition effect of clustering symplectic geometric components on a random sample;

[0052] Figure 4 This is a schematic diagram of the IITN network structure;

[0053] Figure 5 Schematic diagram comparing the loss function convergence curve and accuracy convergence curve of the original InceptionTime network and the IITN network; (a) is the loss function convergence curve of the original InceptionTime network; (b) is the loss function convergence curve of the IITN network; (c) is the accuracy convergence curve of the original InceptionTime network; (d) is the accuracy convergence curve of the IITN network.

[0054] Among them: 1. Excitation slope; 2. First bracket; 3. Second bracket; 4. First conveyor belt; 5. Second conveyor belt; 6. Third conveyor belt; 7. Pusher; 8. Magnetic tile; 9. Acoustic emission sensor; 10. Photoelectric receiver. DETAILED DESCRIPTION

[0055] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0056] like Figure 1 As shown, the magnetic tile internal defect detection device based on acoustic emission signals includes a defect detection component, an excitation inclined plane 1 and an acoustic emission sensor 9; the acoustic emission sensor 9 is arranged on the excitation inclined plane 1;

[0057] The acoustic emission sensor 9 is used to capture the impact acoustic vibration signal generated when the magnetic tile 8 falls on the exciting slope 1;

[0058] The defect detection component is used to receive the collision acoustic vibration signal captured by the acoustic emission sensor 9 and detect the internal defects of the magnetic tile according to the collision acoustic vibration signal.

[0059] The time domain and frequency domain waveforms of the acoustic emission signals of a random qualified and defective magnetic tile are shown in the following figure: Figure 2 As shown in the figure, it can be seen that there is a certain similarity between the time domain and frequency domain waveforms of qualified and defective samples, and there is no obvious standard to measure the difference between the two. If the collision acoustic vibration signal is directly used as the basis for defect detection, it may lead to low defect detection accuracy. Therefore, an effective signal decomposition method is needed to extract detailed feature information.

[0060] In this embodiment, the defect detection component includes a signal amplifier, a signal acquisition card, a signal decomposition module, and a defect recognition module, wherein:

[0061] A signal amplifier is used to amplify and filter the collision acoustic vibration signal output by the acoustic emission sensor 9 to obtain a pre-processed analog electrical signal;

[0062] A signal acquisition card is used to convert the pre-processed analog electrical signal into a digital signal at a sampling frequency of 120kHz;

[0063] The signal decomposition module uses an improved clustered symplectic geometry mode decomposition (SGMD) method to decompose the digital signal into three component signals and select the component signal corresponding to the minimum evaluation index of decomposition (EIDC) as its output feature signal. The EIDC calculates the error between the signal before and after decomposition. The lower the EIDC value, the higher the similarity between the component and the signal.

[0064] The defect recognition module is deployed with a pre-trained IITN network, which is used to use the characteristic signal output by the signal decomposition module as the input of the pre-trained IITN network, classify the input characteristic signal through the IITN network, and obtain the characteristic signal classification result, that is, obtain the defect detection result of the magnetic tile 8.

[0065] Figure 3 The time domain and frequency domain waveforms of a random qualified magnetic tile original signal and three component signals are shown. It can be seen that among the three resulting component signals decomposed by the improved clustered symplectic geometric mode decomposition, component signal 1 represents the instantaneous impact frequency, component signal 2 and component signal 3 represent meaningless low-frequency and high-frequency components, among which component signal 1 is the most contributing characteristic component (i.e., characteristic signal). Component signal 1 is selected as the input of the IITN network through EIDC.

[0066] In this embodiment, if Figure 4As shown in the figure, the IITN network includes the first Inception module, the second Inception module and the third Inception module connected in sequence, and the global average pooling layer, the flattening layer and the fully connected layer connected in sequence; the input of the first Inception module is processed by the CBAM attention mechanism and then residually connected with the output of the second Inception module, and the residual connection result is the input of the third Inception module; the output of the third Inception module is the input of the global average pooling layer; the input of the first Inception module is the input of the IITN network; and the output of the fully connected layer is the output of the IITN network.

[0067] Unlike conventional Inception networks that use 3×3 convolutions, the existing InceptionTime network captures features at different time scales by running multi-scale convolution kernels in parallel with 10, 20, and 40 time steps. Conventional Inception networks rely on pooling layers to compress the time dimension, while the existing InceptionTime network stacks six Inception modules to gradually expand the receptive field. Furthermore, the existing InceptionTime network enforces the use of residual connections, directly adding the Inception module output to the input, effectively alleviating the vanishing gradient problem.

[0068] The IITN network improves upon the existing InceptionTime network. Because signal decomposition methods already extract characteristic components, processing the six layers of Inception modules in the existing InceptionTime network not only results in excessive resource consumption but also in long processing times. Therefore, the redundant InceptionTime modules (three Inception modules and their residual connections) are removed from the existing InceptionTime network. This architectural simplification strategy reduces the depth and breadth of feature extraction, which in turn affects the network's ability to capture complex features. To address this, this embodiment incorporates a CBAM attention mechanism within the residual connections. By combining channel-wise and spatial-attention mechanisms, the CBAM attention mechanism enhances the precision of feature representation at different scales and spatial locations, thereby improving network recognition accuracy and reducing data processing time.

[0069] Figure 5 The loss function convergence curves and accuracy convergence curves of the original InceptionTime network and the IITN network are shown. It can be seen that the original InceptionTime network has a relatively slow convergence speed in loss function and accuracy due to its 6-layer Inception module, while the IITN network has a simple architecture and a relatively fast convergence speed. The addition of the CBAM attention mechanism also improves the model accuracy.

[0070] In the Inception module, for the dimension The time series first passes through a bottleneck layer that executes with a step length of 1 Convolution reduces the input dimension to , three filters of size 10, 20, and 40 are applied to the time series at the same time to extract features of different scales and introduce a parallel MaxPooling layer to maintain model robustness. The output of each independent parallel convolution and maximum pooling is connected to form the output of the Inception module. After the three Inception modules, the data is classified through global average pooling and a fully connected layer.

[0071] In this embodiment, a first bracket 2 and a second bracket 3 are respectively provided on both sides of the upper end of the excitation slope 1; a photoelectric transmitter is provided on the first bracket 2, and a photoelectric receiver 10 is provided on the second bracket 3; the connection line between the photoelectric transmitter and the photoelectric receiver 10 intersects with the falling path of the magnetic tile 8, so that the magnetic tile 8 prevents the photoelectric receiver 10 from receiving the signal from the photoelectric transmitter before falling onto the excitation slope 1, and the time point when the photoelectric receiver 10 does not receive the signal from the photoelectric transmitter is used as the signal collection starting point of the acoustic emission sensor 9.

[0072] A first conveyor belt 4 is provided above the excitation slope 1 and next to the photoelectric receiver 10, and a second conveyor belt 5 is provided at the lower end of the excitation slope 1; a third conveyor belt 6 is provided on one side of the second conveyor belt 5, and a pusher 7 is provided on the other side of the second conveyor belt 5 and directly opposite to the third conveyor belt 6;

[0073] The first conveyor belt 4 is used to transport the magnetic tile 8 to the top of the exciting inclined plane 1, so that the magnetic tile 8 passes through the line connecting the photoelectric transmitter and the photoelectric receiver 10 and falls onto the exciting inclined plane 1;

[0074] The second conveyor belt 5 is used to transport the fallen magnetic tiles 8 that are detected to be defect-free toward the qualified magnetic tile discharge port;

[0075] A pusher 7 is used to push the fallen magnetic tile 8 detected as defective from the second conveyor belt 5 to the third conveyor belt 6;

[0076] The third conveyor belt 6 is used to transport the magnetic tiles 8 that are detected to have defects toward the waste recycling area.

[0077] The magnetic tile 8 is transported on the first conveyor belt 4 in a standing posture, with the convex surface of the magnetic tile 8 facing the conveying direction; the falling height of the magnetic tile 8 is 20 mm. This setting can not only make the magnetic tile 8 generate an effective signal, but also prevent the magnetic tile 8 from breaking.

[0078] Corresponding to the device, this embodiment also provides a method for detecting internal defects of a magnetic tile based on acoustic emission signals, which includes the following steps:

[0079] S1, making the magnetic tile 8 fall onto the exciting inclined plane 1, and capturing the collision acoustic vibration signal generated when the magnetic tile 8 falls onto the exciting inclined plane 1 by the acoustic emission sensor 9;

[0080] S2. Receive the collision acoustic vibration signal captured by the acoustic emission sensor 9 through the defect detection component, and detect the internal defects of the magnetic tile according to the collision acoustic vibration signal.

[0081] In step S2, the specific method of receiving the collision acoustic vibration signal captured by the acoustic emission sensor 9 through the defect detection component and detecting the internal defects of the magnetic tile according to the collision acoustic vibration signal includes the following steps:

[0082] S2-1, using a signal amplifier to amplify and filter the collision sound vibration signal output by the acoustic emission sensor 9 to obtain a pre-processed analog electrical signal;

[0083] S2-2, using a signal acquisition card to convert the pre-processed analog electrical signal into a digital signal;

[0084] S2-3, using a signal decomposition module that deploys an improved clustered symplectic geometric mode decomposition method to decompose the digital signal into three component signals, and selecting the component signal corresponding to the minimum decomposition evaluation index as its output feature signal;

[0085] S2-4. Use the defect recognition module deployed with the pre-trained IITN network to take the characteristic signal output by the signal decomposition module as the input of the pre-trained IITN network, classify the input characteristic signal through the IITN network, and obtain the characteristic signal classification result, that is, obtain the defect detection result of the magnetic tile 8.

[0086] In this embodiment, the specific method of decomposing the digital signal into three component signals in step S2-3 includes the following steps:

[0087] S2-3-1. Apply the time series topology equivalence method to digital signals to construct a one-dimensional time series and form a trajectory matrix :

[0088]

[0089] in The embedding dimension representing the topological equivalence of time series; represents the delay time of the time series topological equivalence method, , Indicates the data length of the digital signal; express In the delay The value after time; Represents the trajectory matrix Middle The element in the first column of the row;

[0090] S2-3-2. Trajectory Matrix Perform autocorrelation analysis to obtain the covariance matrix , and then we get the matrix ;

[0091] S2-3-3. Get the matrix The eigenvalues ​​and eigenvectors of , using the feature vector and trajectory matrix Construct coefficient matrix , and then get the initial single-component reconstruction matrix ;in ;

[0092] S2-3-4. Adjust the elements in each initial single-component reconstruction matrix to obtain the adjusted initial single-component reconstruction matrix: Middle m No.1 in the industry d -1 elements in the column, if ,make ; For the initial single-component reconstruction matrix Other elements in ;in Reconstruct the matrix for the initial single component Middle a Rank b Elements of the column; Reconstruct the matrix for the initial single component Middle b Rank a Elements of the column;

[0093] S2-3-5, the The adjusted initial single-component reconstruction matrix Convert to a one-dimensional time series , and then get a one-dimensional time series ;in Middle k The expression for each element is:

[0094]

[0095] in Reconstruct the matrix for the adjusted initial single component Middle Rank Elements of the column; , min means taking the minimum value; , max means taking the maximum value;

[0096] S2-3-6. Clustering of one-dimensional time series using agglomerative hierarchical clustering algorithm The signals are reorganized into three components through clustering.

[0097] The calculation expression of the minimum decomposition evaluation index in step S2-3 is:

[0098]

[0099] in Represents a single component signal With digital signal The minimum decomposition evaluation index of ; Indicates time; Indicates the data length of a digital signal, that is, the number of data points in a single component signal.

[0100] In step S2-3-6, the initial single component set Each sample in is considered as an independent cluster, forming Initial clusters. Using the Euclidean distance metric, the distances between clusters are calculated to establish their relative spatial relationships. and clusters The inter-cluster distance between is mathematically expressed as follows:

[0101]

[0102] Merge the two clusters with the shortest inter-cluster distance, thereby reducing the total number of clusters in the current iteration. Update the distance matrix by recalculating the distances between the newly formed cluster and all remaining clusters. When , the iterative process will terminate, otherwise, it will repeatedly calculate the distance between clusters, merge clusters and update the distance matrix. , That is, through agglomerative hierarchical clustering The collection is reconstructed. The value is 3, resulting in 3 clustered symplectic geometric components, that is, three component signals.

[0103] In one embodiment of the present invention, the experimental platform is Windows 10, an NVIDIA GeForce GTX 3060Ti GPU, and CUDA 11.3. The number of magnetic tile samples is 2,000, with an equal number of qualified and defective samples. The sampling frequency is 120 kHz, the number of sampling points is 20,000, and the size information is shown in Table 1.

[0104] Table 1: Basic information of magnetic tile samples

[0105]

[0106] The feature signals are divided into training samples and test samples. The training samples are input into the InceptionTime pruning model (IITN network) with the CBAM attention mechanism for training, and the trained IITN network is tested with the test samples. The training loss is 0.0029 and the validation loss is 0.0458. The accuracy of the present invention in the test set reaches 99.17%, the precision rate is 99.65%, and the recall rate is 98.62%, which meets the enterprise's requirement for a qualified magnetic tile recognition rate of no less than 95%.

[0107] Table 2 shows a performance comparison of the original InceptionTime network, the InceptionTime with three-layer Inception modules (3IM-ITN), and the IITN network proposed in this invention, including parameters such as parameter count, accuracy, precision, and recall. The original InceptionTime network achieved an accuracy of 98.83%. To reduce computational cost, the accuracy dropped to 98.17% after reducing the number of Inception modules to three. To enhance the feature extraction capabilities of the three-layer Inception module network, this invention incorporates the CBAM mechanism. CBAM combines channel-attention and spatial-attention mechanisms to adaptively adjust the importance of each channel based on global information, capturing detailed information at key locations in the feature map. This improves accuracy to 99.17%, demonstrating the effectiveness of combining the three-layer Inception module with the CBAM attention mechanism in enhancing network robustness and overall classification performance. The results also outperform those of the original InceptionTime network.

[0108] Table 2: Comparison of three network configurations

[0109]

[0110] To sum up, the present invention constructs a detection device / method that can be directly adapted to the magnetic tile production line. After preprocessing the collected collision acoustic vibration signal and converting it into a digital signal, it is decomposed into three component signals by the improved clustering symplectic geometric modal decomposition method, and the optimal component signal is selected as the feature signal through the minimum decomposition evaluation index. The input feature signal is classified using the IITN network to obtain the feature signal classification result, that is, the magnetic tile defect detection result, thereby forming a fast and effective magnetic tile defect detection solution that can be directly used in the magnetic tile production line.

Claims

1. A magnetic tile internal defect detection device based on acoustic emission signals, characterized in that: It comprises a defect detection component, an excitation inclined plane (1) and an acoustic emission sensor (9); the acoustic emission sensor (9) is arranged on the excitation inclined plane (1); An acoustic emission sensor (9) is used to capture the collision acoustic vibration signal generated when the magnetic tile (8) falls on the exciting inclined plane (1); A defect detection component is used to receive the collision acoustic vibration signal captured by the acoustic emission sensor (9) and detect internal defects of the magnetic tile based on the collision acoustic vibration signal; The defect detection component includes a signal amplifier, a signal acquisition card, a signal decomposition module, and a defect recognition module, among which: A signal amplifier, used to amplify and filter the collision acoustic vibration signal output by the acoustic emission sensor (9) to obtain a pre-processed analog electrical signal; A signal acquisition card is used to convert the pre-processed analog electrical signal into a digital signal; The signal decomposition module deploys an improved clustered symplectic geometric mode decomposition method to decompose the digital signal into three component signals and select the component signal corresponding to the minimum decomposition evaluation index as its output feature signal; The defect recognition module is deployed with a pre-trained IITN network, which is used to take the characteristic signal output by the signal decomposition module as the input of the pre-trained IITN network, classify the input characteristic signal through the IITN network, and obtain the characteristic signal classification result, that is, obtain the defect detection result of the magnetic tile (8); The IITN network includes the first Inception module, the second Inception module and the third Inception module connected in sequence, and the global average pooling layer, the flattening layer and the fully connected layer connected in sequence; the input of the first Inception module is processed by the CBAM attention mechanism and then residually connected with the output of the second Inception module, and the residual connection result is the input of the third Inception module; the output of the third Inception module is the input of the global average pooling layer; the input of the first Inception module is the input of the IITN network; and the output of the fully connected layer is the output of the IITN network.

2. The device for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 1, characterized in that: A first bracket (2) and a second bracket (3) are respectively provided on both sides of the upper end of the excitation inclined plane (1); a photoelectric transmitter is provided on the first bracket (2), and a photoelectric receiver (10) is provided on the second bracket (3); a line connecting the photoelectric transmitter and the photoelectric receiver (10) intersects with a falling path of the magnetic tile (8), so that the magnetic tile (8) blocks the photoelectric receiver (10) from receiving a signal from the photoelectric transmitter before falling onto the excitation inclined plane (1), and a time point at which the photoelectric receiver (10) does not receive a signal from the photoelectric transmitter is used as a signal collection starting point of the acoustic emission sensor (9).

3. The device for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 2, characterized in that: A first conveyor belt (4) is provided above the excitation slope (1) and beside the photoelectric receiver (10), and a second conveyor belt (5) is provided at the lower end of the excitation slope (1); a third conveyor belt (6) is provided on one side of the second conveyor belt (5), and a pusher (7) is provided on the other side of the second conveyor belt (5) and is directly opposite to the third conveyor belt (6); A first conveyor belt (4) is used to transport the magnetic tile (8) to the top of the exciting inclined plane (1), so that the magnetic tile (8) passes through the connection line between the photoelectric transmitter and the photoelectric receiver (10) and falls onto the exciting inclined plane (1); A second conveyor belt (5) is used to transport the fallen magnetic tiles (8) that are detected as defect-free toward a qualified magnetic tile discharge port; A pusher (7) is used to push the magnetic tile (8) that has fallen and is detected as defective from the second conveyor belt (5) to the third conveyor belt (6); The third conveyor belt (6) is used to transport the magnetic tiles (8) detected as having defects toward a waste recycling location.

4. The device for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 3, characterized in that: The magnetic tile (8) is conveyed on the first conveyor belt (4) in a standing posture, with the convex surface of the magnetic tile (8) facing the conveying direction; the falling height of the magnetic tile (8) is 20 mm.

5. A method for detecting internal defects of a magnetic tile based on acoustic emission signals, based on the device for detecting internal defects of a magnetic tile based on acoustic emission signals according to any one of claims 1 to 4, characterized in that: The following steps are involved: The magnetic tile (8) is caused to fall onto the exciting inclined plane (1), and the collision acoustic vibration signal generated when the magnetic tile (8) falls onto the exciting inclined plane (1) is captured by the acoustic emission sensor (9); The collision acoustic vibration signal captured by the acoustic emission sensor (9) is received by the defect detection component, and the internal defects of the magnetic tile are detected based on the collision acoustic vibration signal.

6. The method for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 5, characterized in that: The specific method of receiving the collision acoustic vibration signal captured by the acoustic emission sensor (9) through the defect detection component and detecting the internal defects of the magnetic tile according to the collision acoustic vibration signal includes the following steps: A signal amplifier is used to amplify and filter the collision acoustic vibration signal output by the acoustic emission sensor (9) to obtain a pre-processed analog electrical signal; A signal acquisition card is used to convert the pre-processed analog electrical signal into a digital signal; A signal decomposition module with an improved clustered symplectic geometric mode decomposition method is used to decompose the digital signal into three component signals, and the component signal corresponding to the minimum decomposition evaluation index is selected as the output feature signal; The defect recognition module deployed with the pre-trained IITN network uses the characteristic signal output by the signal decomposition module as the input of the pre-trained IITN network, and the input characteristic signal is classified through the IITN network to obtain the characteristic signal classification result, that is, the defect detection result of the magnetic tile (8) is obtained.

7. The method for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 6, characterized in that: The specific method of decomposing a digital signal into three component signals includes the following steps: Apply the time series topology equivalence method to construct a one-dimensional time series and form a trajectory matrix : in The embedding dimension representing the topological equivalence of time series; represents the delay time of the time series topological equivalence method, , Indicates the data length of the digital signal; express In the delay The value after time; Represents the trajectory matrix Middle The element in the first column of the row; Trajectory Matrix Perform autocorrelation analysis to obtain the covariance matrix , and then we get the matrix ; Get Matrix The eigenvalues ​​and eigenvectors of , using the feature vector and trajectory matrix Construct coefficient matrix , and then get the initial single-component reconstruction matrix ;in ; Adjust the elements in each initial single-component reconstruction matrix to obtain the adjusted initial single-component reconstruction matrix: Middle m No.1 in the industry d -1 elements in the column, if ,make ; For the initial single-component reconstruction matrix Other elements in ;in Reconstruct the matrix for the initial single component Middle a Rank b Elements of the column; Reconstruct the matrix for the initial single component Middle b Rank a Elements of the column; The first The adjusted initial single-component reconstruction matrix Convert to a one-dimensional time series , and then get a one-dimensional time series ;in Middle k The expression for each element is: in Reconstruct the matrix for the adjusted initial single component Middle Rank Elements of the column; , min means taking the minimum value; , max means taking the maximum value; Agglomerative hierarchical clustering algorithm for one-dimensional time series The signals are reorganized into three components through clustering.

8. The method for detecting internal defects of magnetic tiles based on acoustic emission signals according to claim 6, characterized in that: The calculation expression of the minimum decomposition evaluation index is: in Represents a single component signal With digital signal The minimum decomposition evaluation index of ; Indicates time; Indicates the data length of a digital signal, that is, the number of data points in a single component signal.

Images