Nondestructive testing method and system for internal cracks of prebaked raw anode block

Abstract

The invention discloses a nondestructive testing method and system for internal cracks of a prebaked raw anode block, and mainly relates to the technical field of industrial nondestructive testing. Sound vibration is generated through force hammer excitation, a block sound vibration response signal is collected, a frequency spectrum is obtained through band-pass filtering and fast Fourier transform, a plurality of main peak frequencies and amplitudes are extracted as initial features, and dimensionality reduction is carried out through principal component analysis. And adjusting parameters of a support vector regression machine by adopting a particle swarm optimization algorithm, and establishing a mapping relationship between the dimension reduction characteristics and the crack area and position height, so as to realize quantitative prediction of the crack parameters. The method has the beneficial effects that the detection speed is high, the precision is high, the positioning and quantification integrated analysis of cracks can be realized, the method is suitable for the automatic quality detection of the prebaked anode green block, and the qualified rate of final products is improved while the comprehensive energy consumption of production is reduced.

Description

Technical Field

[0001] This invention relates to the field of industrial non-destructive testing technology, specifically a method and system for non-destructive testing of internal cracks in prebaked anode green blocks based on acoustic vibration signals and artificial intelligence. Background Technology

[0002] Prebaked anodes are the core material in aluminum electrolysis production, and their quality directly affects production efficiency and energy consumption. During manufacturing, cracks can easily form inside the green anode due to uneven process parameters. These defects can significantly reduce anode performance, leading to problems such as chipping and breakage during electrolysis, increasing energy consumption, and even causing accidents. Therefore, effective detection of internal cracks during the green anode stage is crucial. Currently, the industry mainly uses the manual hammering and listening method. This method relies on the experience of inspectors, cannot quantitatively assess defects, has poor reliability, and no longer meets modern quality control requirements. Non-destructive testing technologies such as X-rays and ultrasound are difficult to promote in industrial settings due to expensive equipment, low testing efficiency, or poor applicability.

[0003] Acoustic vibration analysis, as an emerging technology, offers advantages such as low cost, high efficiency, and non-contact operation, identifying internal defects by analyzing the vibration response of components. However, its application in prebaked anodized blocks faces significant challenges: the influence of cracks on vibration characteristics is complex and nonlinear, making it difficult for traditional methods to extract effective features from complex spectra; there is a lack of clear linear relationship between crack parameters and acoustic features; and obtaining a large number of labeled samples is challenging in industrial practice. Existing research has attempted to introduce machine learning algorithms, but most methods suffer from limitations such as single feature sets, unoptimized parameters, and only qualitative identification capabilities, failing to simultaneously achieve precise crack localization and quantitative analysis, severely restricting the industrial application of this technology.

[0004] Therefore, there is an urgent need for an innovative method that can combine acoustic vibration mechanism with artificial intelligence technology to achieve rapid, accurate and automated detection of internal cracks in prebaked anode blocks. Summary of the Invention

[0005] The purpose of this invention is to provide a non-destructive testing method and system for internal cracks in prebaked anode green blocks based on acoustic vibration signals and artificial intelligence. It enables rapid and accurate identification and screening of defective products with internal cracks, while reducing overall production energy consumption and improving the pass rate and overall quality level of the final product.

[0006] To achieve the above objectives, the present invention employs the following technical solution: On the one hand, the present invention provides a non-destructive testing method for internal cracks in prebaked anode green blocks, comprising the following steps: S1: Signal acquisition step, applying instantaneous excitation to several preset measurement points on the surface of the prebaked anode green block, and acquiring the acoustic vibration response signal generated by the prebaked anode green block; S2: Signal preprocessing and frequency domain analysis steps, preprocessing the acquired acoustic and vibration response signals, including denoising and truncating the effective segment, and then performing frequency domain transformation to obtain the spectrum; S3: Feature extraction and dimensionality reduction step: Extract initial features from the spectrum, the initial features include multiple peak frequencies and corresponding amplitudes, and then perform dimensionality reduction processing on the initial features to obtain the dimensionality-reduced feature vector; S4: Artificial intelligence crack analysis model construction and training steps, using support vector regression machine as the regression prediction model, optimizing the parameters of the support vector regression machine through particle swarm optimization algorithm, and training with a sample dataset of known crack parameters to establish the mapping relationship between the dimensionality-reduced feature vector and the crack parameters, the crack parameters including crack area ratio and crack location height; S5: Crack parameter prediction step. For the prebaked anode block to be detected, repeat S1 to S3 to obtain its dimensionality-reduced feature vector, input it into the trained artificial intelligence crack analysis model, and output the predicted values ​​of the crack area ratio and crack location height of the prebaked anode block.

[0007] Preferably, in the signal acquisition step, the instantaneous excitation is achieved by striking with a force hammer, which is equipped with a force sensor to synchronously record the excitation force signal; the acoustic vibration response signal is acquired by a non-contact sound pressure sensor or accelerometer, the sound pressure sensor being fixed 50mm from the side of the prebaked anode block; the sampling frequency is set to 10kHz; the prebaked anode block is placed on a soft rubber pad to simulate free boundary conditions; the preset measurement points are 5 fixed positions, evenly distributed on the upper surface of the prebaked anode block.

[0008] Preferably, in the signal preprocessing and frequency domain analysis steps, the preprocessing includes bandpass filtering of the acoustic vibration response signal with a cutoff frequency of 10Hz to 4kHz to remove low-frequency vibrations and high-frequency noise; the frequency domain analysis uses fast Fourier transform and applies Hanning window function to reduce spectral leakage to obtain an amplitude spectrum of 0Hz to 5kHz; the effective segment truncation length is 1 second.

[0009] Preferably, in the feature extraction and dimensionality reduction steps, the initial feature extraction includes automatically identifying the preceding features from the spectrum. The maximum peak value is used to extract its corresponding frequency and amplitude values, which together constitute a... An initial feature vector of dimension, where It is a positive integer.

[0010] Preferably, the dimensionality reduction process employs principal component analysis, specifically including: The dataset composed of the initial feature vectors is standardized to have zero mean and unit variance. Perform principal component analysis and select the k principal components whose cumulative contribution rate exceeds a preset threshold as the dimensionality-reduced feature vectors.

[0011] Preferably, in the artificial intelligence crack analysis model construction and training steps, the support vector regression machine uses a radial basis function as its kernel function, and its prediction function is: ; in, and It is a Lagrange multiplier. It is the radial basis function kernel, defined as: ; in, These are kernel parameters. It is the input feature vector. They are support vectors. It is a bias term; The particle swarm optimization algorithm is used to optimize the penalty coefficient of the support vector regression machine. and kernel parameters ,in The search range is [0.1, 100]. The search range is [0.001, 10]. The parameters of the particle swarm optimization algorithm are set as follows: particle swarm size is 20, and maximum number of iterations is 100. The optimization objective is to minimize the root mean square error of 5-fold cross-validation, which is defined as: ; in, It is the number of validation samples. These are the actual crack parameter values. This is a predicted value.

[0012] Preferably, training is performed using a sample dataset with known crack parameters, the sample dataset including qualified blocks and preset crack blocks, wherein the crack parameters are accurately obtained through dissection and measurement; Use 70% of the sample dataset as the training set and 30% as the test set; After the model is trained, its performance is validated using a test set, and the evaluation metrics include the coefficient of determination. and root mean square error ; For predicting the crack area ratio, it is required that... Greater than 0.85 and Less than 0.05; For predicting the location and height of cracks, the following requirements are required. Greater than 0.85 and Less than 0.03; The predicted value of a qualified block should meet the following requirements: the crack area ratio is less than 0.1% and the crack height is less than 2mm.

[0013] On the other hand, the present invention also provides a non-destructive testing system for internal cracks in prebaked anode green blocks, used to implement the above-described non-destructive testing method for internal cracks in prebaked anode green blocks, comprising: The signal transmission link module is used for the electrical connection between the excitation module and the signal acquisition module. The acquired signal is output to the artificial intelligence analysis module via the signal processing module. The excitation module is used to apply instantaneous excitation to the prebaked anode green block; The signal acquisition module is used to acquire the acoustic and vibration response signals generated by the prebaked anode green block; The signal processing module is used to preprocess, frequency domain transform, and extract and reduce features of the acoustic vibration response signal; The artificial intelligence analysis module has a built-in trained support vector regression model, which is used to predict crack parameters based on the extracted features. The parameter optimization module integrates a particle swarm optimization algorithm to optimize the parameters of the support vector regression machine model. The results output module is used to display and record the location and quantitative results of the crack.

[0014] Preferably, the excitation module includes a force hammer equipped with a force sensor and can be handheld or fixed by a robotic arm; the signal acquisition module includes a non-contact sound pressure sensor or accelerometer, and a multi-channel dynamic signal analyzer with a sampling frequency set to 10kHz; the prebaked anode block is placed on a soft rubber pad; the signal processing module and the artificial intelligence analysis module run on an industrial computer configured with an Intel i7 CPU or higher, 16GB or more of memory, and a Python 3.8 software platform integrating Scikit-learn and LibSVM libraries.

[0015] Preferably, the signal processing module is configured to: perform bandpass filtering on the signal with a cutoff frequency of 10Hz to 4kHz; perform fast Fourier transform using a Hanning window function to obtain the amplitude spectrum; extract the frequencies and amplitudes corresponding to the top 20 maximum peaks from the spectrum as initial feature vectors; and use principal component analysis to reduce the dimensionality of the initial feature vectors, with a cumulative contribution rate exceeding 95%. In the artificial intelligence analysis module, the support vector regression machine model uses a radial basis function kernel, and its parameters are optimized using a particle swarm optimization algorithm. The optimized parameters include a penalty coefficient. and kernel parameters , The search range is [0.1, 100]. The search range is [0.001, 10]. The result output module automatically determines whether a result is qualified or unqualified based on a preset threshold, and stores and displays the results.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Moving the quality control checkpoint forward: By accurately screening internal cracks at the raw block stage, defective products are effectively prevented from entering the roasting process; the roasting process is extremely energy-intensive, and removing waste products in advance can directly save a lot of natural gas or electricity consumption, while improving the effective capacity of the kiln and the yield of good products, thus achieving the green production goal of "improving quality and efficiency, saving energy and reducing consumption" from the source. 2. This method overcomes the limitations of traditional manual listening methods, which can only make qualitative judgments, and achieves integrated and precise detection of both the location (location) and size (quantification) of internal cracks. This provides unprecedentedly refined data support for process improvement, enabling quality control to move from vague experience-based judgments to precise data-driven approaches. Compared to manual listening methods, this method improves detection accuracy by approximately 40%. 3. The entire testing process is completed automatically by equipment and algorithms, which minimizes the uncertainty and subjectivity of human factors. The test results are objective, reliable and traceable, which is conducive to establishing unified and strict quality standards and laying the foundation for big data analysis of production quality. 4. The acoustic vibration detection method itself is fast and non-contact. Combined with an optimized artificial intelligence model, it can complete the detection and result analysis of a sample within 30 seconds, making it very suitable for online or batch rapid sampling inspection in production sites. The equipment investment and maintenance costs of this method are far lower than those of large equipment such as X-rays, making it extremely cost-effective and worthy of widespread application. 5. The "acoustic and vibration signal-artificial intelligence" analysis framework constructed in this invention has strong versatility and scalability. By changing the training data, this technical approach can be further applied to detect other types of internal defects in prebaked anodes, such as voids and porosity, and even extended to the non-destructive testing of other carbon products or ceramic materials, showing broad application prospects. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0018] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined in this application.

[0019] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0020] Example: like Figure 1 As shown in the figure, the specific implementation steps of the non-destructive testing method for internal cracks in prebaked anode green blocks provided in this embodiment are as follows: (1) Signal acquisition: Forty prebaked anode green block samples were prepared, including 15 qualified blocks and 25 samples with different cracks pre-formed using a water-drilling method. The samples were placed on soft rubber pads to simulate free boundary conditions. A handheld hammer equipped with a force sensor was used to apply vertical impact excitation at five pre-marked, evenly distributed measuring points on the upper surface of the green block. During each impact, a high-sensitivity acoustic pressure sensor fixed 50 mm from the side of the green block was used to collect the acoustic vibration response signal. The sampling frequency was set to 10 kHz, and the recording time was 1 second.

[0021] (2) Signal preprocessing and frequency domain analysis: The acquired raw signal was bandpass filtered from 10Hz to 4kHz to remove low-frequency jitter and high-frequency noise. After applying a Hanning window function to the preprocessed time-domain signal, a Fast Fourier Transform was performed to convert it to the frequency domain, obtaining the amplitude spectrum from 0Hz to 5kHz.

[0022] (3) Feature extraction and dimensionality reduction: The frequency and amplitude values ​​corresponding to the top 20 largest peaks are automatically identified and extracted from the spectrum of each sample to form a 40-dimensional initial feature vector. The dataset composed of the initial feature vectors of all samples is standardized (zero mean, unit variance), and then principal component analysis is performed to select the top k principal components (k=6 in this example) with a cumulative contribution rate of more than 95% as the dimensionality-reduced feature vector.

[0023] (4) Artificial intelligence model construction and training: Support Vector Regression (SVR) was used as the regression prediction model, with radial basis functions (RBF) as the kernel function. Particle swarm optimization (PSO) was employed to optimize the penalty coefficient *c* and kernel parameter *γ* of the SVR, where the search range for *c* was [0.1, 100] and for *γ* was [0.001, 10]. The particle swarm size was set to 20, and the maximum number of iterations was 100. The optimization objective was to minimize the root mean square error (RMSE) of the 5-fold cross-validation.

[0024] 70% (28 samples) of the total samples were used as the training set. The dimensionality-reduced feature vectors were used as inputs, and the corresponding crack area ratio and crack location height (obtained precisely through dissection measurement) were used as outputs to train two independent SVR models.

[0025] (5) Model validation and prediction: The remaining 30% of the samples (12) were used as the test set for performance validation. The performance was validated to be accurate in predicting the crack area percentage. The value was 0.87, and the RMSE was 0.043; the predicted crack location height was... The accuracy was 0.86 and the RMSE was 0.028. The predicted values ​​of qualified blocks were all less than the threshold (area ratio <0.1%, height <2mm), indicating that the model has high accuracy and low false alarm rate.

[0026] For the raw block to be detected, repeat steps (1)-(3) to obtain its feature vector, and input it into the trained model to obtain the predicted value of the crack parameters.

[0027] like Figure 2 As shown, this embodiment also provides a non-destructive testing system for internal cracks in prebaked anode green blocks, including: Excitation module: A handheld hammer equipped with a force sensor is used, which can be fixed by a robotic arm to achieve automated striking.

[0028] Signal acquisition module: includes a high-sensitivity sound pressure sensor and a multi-channel dynamic signal analyzer, with a sampling frequency set to 10kHz.

[0029] Signal processing module: running on an industrial computer (CPU Intel i7, memory 16GB), configured with Python 3.8 platform, integrating Scikit-learn library, responsible for signal filtering, FFT transformation, feature extraction and PCA dimensionality reduction.

[0030] Artificial intelligence analysis module: Built-in PSO-optimized SVR model, using radial basis function kernel, parameters and The optimization ranges are [0.1, 100] and [0.001, 10], respectively.

[0031] Parameter optimization module: integrates particle swarm optimization algorithm, population size 20, maximum number of iterations 100.

[0032] Results output module: Displays crack location and quantitative results, and automatically determines whether the result is qualified or not based on preset thresholds.

[0033] Through the coordinated operation of its various modules, this system enables rapid, accurate, and automated detection of internal cracks in prebaked anode green blocks, providing a reliable quality control method for industrial production.

[0034] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A non-destructive testing method for internal cracks in prebaked anode green blocks, characterized in that, Includes the following steps: S1: Signal acquisition step, applying instantaneous excitation to several preset measurement points on the surface of the prebaked anode green block, and acquiring the acoustic vibration response signal generated by the prebaked anode green block; S2: Signal preprocessing and frequency domain analysis steps, preprocessing the acquired acoustic and vibration response signals, including denoising and truncating the effective segment, and then performing frequency domain transformation to obtain the spectrum; S3: Feature extraction and dimensionality reduction step: Extract initial features from the spectrum, the initial features include multiple peak frequencies and corresponding amplitudes, and then perform dimensionality reduction processing on the initial features to obtain the dimensionality-reduced feature vector; S4: Artificial intelligence crack analysis model construction and training steps, using support vector regression machine as the regression prediction model, optimizing the parameters of the support vector regression machine through particle swarm optimization algorithm, and training with a sample dataset of known crack parameters to establish the mapping relationship between the dimensionality-reduced feature vector and the crack parameters, the crack parameters including crack area ratio and crack location height; S5: Crack parameter prediction step. For the prebaked anode block to be detected, repeat S1 to S3 to obtain its dimensionality-reduced feature vector, input it into the trained artificial intelligence crack analysis model, and output the predicted values ​​of the crack area ratio and crack location height of the prebaked anode block.

2. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 1, characterized in that, In the signal acquisition step, the instantaneous excitation is achieved by striking with a force hammer, which is equipped with a force sensor to synchronously record the excitation force signal; the acoustic vibration response signal is acquired by a non-contact sound pressure sensor or accelerometer, with the sound pressure sensor fixed 50mm from the side of the prebaked anode block; the sampling frequency is set to 10kHz; the prebaked anode block is placed on a soft rubber pad to simulate free boundary conditions; there are 5 preset measurement points at fixed positions, evenly distributed on the upper surface of the prebaked anode block.

3. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 1, characterized in that, In the signal preprocessing and frequency domain analysis steps, the preprocessing includes bandpass filtering of the acoustic vibration response signal with a cutoff frequency of 10Hz to 4kHz to remove low-frequency vibrations and high-frequency noise; the frequency domain analysis uses fast Fourier transform and applies Hanning window function to reduce spectral leakage to obtain an amplitude spectrum of 0Hz to 5kHz; the effective segment truncation length is 1 second.

4. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 1, characterized in that, In the feature extraction and dimensionality reduction steps, the initial feature extraction includes automatically identifying the preceding features from the spectrum. The maximum peak value is used to extract its corresponding frequency and amplitude values, which together constitute a... An initial feature vector of dimension, where It is a positive integer.

5. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 4, characterized in that, Dimensionality reduction was performed using principal component analysis, specifically including: The dataset composed of the initial feature vectors is standardized to have zero mean and unit variance. Perform principal component analysis and select the k principal components whose cumulative contribution rate exceeds a preset threshold as the dimensionality-reduced feature vectors.

6. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 1, characterized in that, In the construction and training steps of the artificial intelligence crack analysis model, the support vector regression machine uses the radial basis function as the kernel function, and its prediction function is: ； in, and It is a Lagrange multiplier. It is the radial basis function kernel, defined as: ； in, These are kernel parameters. It is the input feature vector. They are support vectors. It is a bias term; The particle swarm optimization algorithm is used to optimize the penalty coefficient of the support vector regression machine. and kernel parameters ,in The search range is [0.1, 100]. The search range is [0.001, 10]. The parameters of the particle swarm optimization algorithm are set as follows: particle swarm size is 20, and maximum number of iterations is 100. The optimization objective is to minimize the root mean square error of 5-fold cross-validation, which is defined as: ； in, It is the number of validation samples. These are the actual crack parameter values. This is a predicted value.

7. The non-destructive testing method for internal cracks in prebaked anode green blocks according to claim 6, characterized in that, Training is performed using a sample dataset with known crack parameters, which includes qualified blocks and pre-set cracked blocks, wherein the crack parameters are precisely obtained through dissection and measurement. Use 70% of the sample dataset as the training set and 30% as the test set; After the model is trained, its performance is validated using a test set, and the evaluation metrics include the coefficient of determination. and root mean square error ; For predicting the crack area ratio, it is required that... Greater than 0.85 and Less than 0.05; For predicting the location and height of cracks, the following requirements are required. Greater than 0.85 and Less than 0.03; The predicted value of a qualified block should meet the following requirements: the crack area ratio is less than 0.1% and the crack height is less than 2mm.

8. A non-destructive testing system for internal cracks in prebaked anode green blocks, used to implement the non-destructive testing method for internal cracks in prebaked anode green blocks as described in any one of claims 1-7, characterized in that, include: The signal transmission link module is used for the electrical connection between the excitation module and the signal acquisition module. The acquired signal is output to the artificial intelligence analysis module via the signal processing module. The excitation module is used to apply instantaneous excitation to the prebaked anode green block; The signal acquisition module is used to acquire the acoustic and vibration response signals generated by the prebaked anode green block; The signal processing module is used to preprocess, frequency domain transform, and extract and reduce features of the acoustic vibration response signal; The artificial intelligence analysis module has a built-in trained support vector regression model, which is used to predict crack parameters based on the extracted features. The parameter optimization module integrates a particle swarm optimization algorithm to optimize the parameters of the support vector regression machine model. The results output module is used to display and record the location and quantitative results of the crack.

9. The non-destructive testing system for internal cracks in prebaked anode green blocks according to claim 8, characterized in that, The excitation module includes a force hammer equipped with a force sensor and can be handheld or fixed by a robotic arm; the signal acquisition module includes a non-contact sound pressure sensor or accelerometer, and a multi-channel dynamic signal analyzer with a sampling frequency set to 10kHz; the prebaked anode block is placed on a soft rubber pad; the signal processing module and the artificial intelligence analysis module run on an industrial computer configured with an Intel i7 CPU or higher, 16GB or more of memory, and a Python 3.8 software platform integrating Scikit-learn and LibSVM libraries.

10. The non-destructive testing system for internal cracks in prebaked anode green blocks according to claim 8, characterized in that, The signal processing module is configured to: perform bandpass filtering on the signal with a cutoff frequency of 10Hz to 4kHz; perform fast Fourier transform using the Hanning window function to obtain the amplitude spectrum; extract the frequencies and amplitudes corresponding to the top 20 maximum peaks from the spectrum as initial feature vectors; and use principal component analysis to reduce the dimensionality of the initial feature vectors, with a cumulative contribution rate exceeding 95%. In the artificial intelligence analysis module, the support vector regression machine model uses a radial basis function kernel, and its parameters are optimized using a particle swarm optimization algorithm. The optimized parameters include a penalty coefficient. and kernel parameters , The search range is [0.1, 100]. The search range is [0.001, 10]. The result output module automatically determines whether a result is qualified or unqualified based on a preset threshold, and stores and displays the results.

Images