Percussion and PSO-SVM-Based Damage Detection for Refractory Materials

Refractory materials are basic materials widely used in industrial furnaces and thermal equipment. Their microstructure is similar to that of many heterogeneous high-performance materials used in micro/nanodevices. The presence of damage can reduce the mechanical properties and service life of refractory materials and even cause serious safety accidents. In this paper, a novel percussion and particle swarm optimization-support vector machine (PSO-SVM)-based method is proposed to detect damage in refractory materials. An impact is applied to the material and the generated sound is recorded. The percussion-induced sound signals are fed into a mel filter bank to generate time–frequency representations in the form of mel spectrograms. Then, two image descriptors—the local binary pattern (LBP) and histogram of oriented gradient (HOG)—are used to extract the texture information of the mel spectrogram. Finally, combining both HOG and LBP features, the fused features are input to the PSO-SVM algorithm to realize damage detection in refractory materials. The results demonstrated that the proposed method could identify five different degrees of damage of refractory materials, with an accuracy rate greater than 97%. Therefore, the percussion and PSO-SVM-based method proposed in this paper has high potential for field applications in damage detection in refractory material, and also has the potential to be extended to research on damage detection methods for other materials used in micro/nanodevices.


Introduction
Refractory materials are composite materials resistant to thermal shock and chemical erosion [1], and their heterogeneity and microstructural complexity have commonalities with related materials [2][3][4] used in micro/nano-devices. Multiple forms of damage in refractory material such as cracks and holes can occur in complex production environments. These damages can reduce the performance of the material and seriously jeopardize the normal operation of devices. Therefore, damage detection for refractory materials after manufacture is vital to ensure their quality and, thus, the stable operation of industrial production.
Currently, several common non-destructive testing (NDT) methods have been attempted to detect damage of certain structures, including the acoustic emission (AE) method [5,6], the ultrasonic method [7,8], the radiography method [9], etc. AE is the phenomenon of transient elastic waves generated by the rapid release of local energy within a material (or structure). Liu et al. [10]. proposed a damage detection method of refractory materials using principle component analysis and the Gaussian mixture model to reduce the dimensions of the relevant parameters of the AE signal and describe the overall properties of material is used to identify the degree of damage to the refractory materials. The rest of this paper is organized as follows. Section 2 provides the methodologies and related principles of the proposed method. Section 3 describes the experimental device and experimental procedure. Section 4 discusses the identification results of the method and provides a comparative analysis with other strategies. Section 5 is the conclusion of the paper.

Methodologies
The schematic diagram of the proposed percussion method is shown in Figure 1. The method involves three main steps: percussion signal acquisition, feature extraction and damage classification and recognition based on PSO-SVM. Firstly, percussion signals are collected for five different degrees of damage to refractory materials. Then, in order to reveal the variation pattern of the percussion sound signals, the mel spectrogram was used to reflect a large amount of feature information of the signals. In addition, the HOG feature and LBP feature were adopted to capture the texture information of the mel spectrogram variations. Finally, the combination of LBP and HOG was entered into the PSO-SVM classifier, and the output of the PSO-SVM represented the degree of damage to the refractory material.
In this study, an easy-to-implement and efficient novel percussion method is us detect damage to refractory materials. During detection, the percussion-induced s signals are first transformed into mel spectrograms, which can depict the singular different signals. LBP and HOG methods are used to extract the unique textural fea of the mel spectrogram to further uncover the hidden damage-related inform Thereafter, the feature vector of the signal is obtained by the fusion of HOG and features. Finally, the PSO-SVM classifier is used to identify the degree of damage refractory materials. The rest of this paper is organized as follows. Section 2 provid methodologies and related principles of the proposed method. Section 3 describe experimental device and experimental procedure. Section 4 discusses the identific results of the method and provides a comparative analysis with other strategies. Se 5 is the conclusion of the paper.

Methodologies
The schematic diagram of the proposed percussion method is shown in Figure 1 method involves three main steps: percussion signal acquisition, feature extraction damage classification and recognition based on PSO-SVM. Firstly, percussion signa collected for five different degrees of damage to refractory materials. Then, in ord reveal the variation pattern of the percussion sound signals, the mel spectrogram used to reflect a large amount of feature information of the signals. In addition, the feature and LBP feature were adopted to capture the texture information of th spectrogram variations. Finally, the combination of LBP and HOG was entered int PSO-SVM classifier, and the output of the PSO-SVM represented the degree of dama the refractory material.

Mel Spectrogram
The mel spectrogram [36] is a combination of the mel scale and a spectrogram acquisition process is simple; the input signal is preprocessed (framing, windowing Fourier transform (FFT) is applied and the signal is passed through a mel filter bank 2.1.2. Histogram of Oriented Gradient HOG is an excellent local feature descriptor proposed by Dalal and Triggs at C [37,38] for obtaining image feature vectors. The specific steps are as follows: (1) Grayscale the input images; (2) Perform color space normalization of grayscale images by the Gamma corre method  The mel spectrogram [36] is a combination of the mel scale and a spectrogram. The acquisition process is simple; the input signal is preprocessed (framing, windowing), fast Fourier transform (FFT) is applied and the signal is passed through a mel filter bank.

Histogram of Oriented Gradient
HOG is an excellent local feature descriptor proposed by Dalal and Triggs at CVPR [37,38] for obtaining image feature vectors. The specific steps are as follows: (1) Grayscale the input images; (2) Perform color space normalization of grayscale images by the Gamma correction method (3) Calculate the gradient (direction and intensity distribution) of each pixel to further obtain the image contour information.
where G x (x,y), G y (x,y) and I(x,y) denote horizontal gradient, vertical gradient and pixel value, respectively. In addition, the direct and intensity distributions are as follows: (4) Divide the image into small cells; then count the gradient histograms of each cell and obtain the statistics of gradient distribution in different directions; (5) Several cells form a block. HOG features of all blocks of the image are obtained through a sliding block.

Local Binary Patterns
LBP is a simple and efficient method for texture description [39] that is able to obtain a binary pattern of differences between a central pixel and neighboring pixels. Specifically, the pixels of the image are labeled by thresholding the 3 × 3 neighborhood of each pixel with the center value and using the result as a binary number. The central pixel value is calculated as follows: where i c is the grayscale of the central pixel, the grayscale of the neighbourhood in the pixel matrix is i p and S is the step function, as shown below:

Support Vector Machine
Support vector machine (SVM) is one of the best algorithms in machine learning and is commonly used to solve classification and regression problems [40][41][42]. The data points for each sample can be expressed as {x i ,y i }, where i = 1, 2, 3, . . . , N, x i ∈ R n , y i ∈ {+1,−1}. The expression of the hyperplane in the linearly divisible case can be written as: where f (x) is the separating hyperplane, the parameter w is the weight and b is the bias. When encountering a linearly indistinguishable problem, the above equation needs to be extended by adding the slack variable ξ i and the penalty c factor to obtain the model for solving the nonlinear problem, as follows: This equation is then transformed into a pairwise problem using the Lagrange multiplier method to obtain the decision function as follows: where α i is only one variable included in the Lagrangian function; k(x i ,y i ) = ϕ(x i ) T ϕ(x j ) is the kernel function introduced to address the difficulty of direct calculation due to the high dimensionality of the features.

Particle Swarm Optimization Algorithm
The PSO algorithm [43] is inspired by the predatory behavior of a flock of birds searching for food randomly and is used to solve optimization problems. PSO is initialized as a flock of random particles; then, the optimal solution is found by iteration. At each iteration, the particles update their velocity and position towards the individual extremes and the global extremes. The update formula is as follows: (11) where w is the inertia weight, k is the number of current iterations, c 1 and c 2 are learning factors and r 1 and r 2 are uniform random numbers in the range of (0, 1).

The PSO-SVM Method
The RBF kernel function is used in SVM to solve the nonlinear classification problem; classification accuracy is mainly influenced by the penalty factor C and the kernel parameter g. The PSO algorithm is adopted to search for the ideal parameters C and g of SVM to achieve the accurate and reliable detection of damage in refractory materials. The specific process of PSO-SVM is as follows: (1) The specific implementation steps of the PSO-SVM algorithm are as follows. Set the relevant parameters: the population size is 20, the inertia factor is 0.6, the acceleration constants are 1.5 and 1.7, and the number of iterations is 50; (2) Prepare the train and test data sets. The training sets adopt the five-fold crossvalidation method and the classification accuracy is set to the particle fitness value; (3) The penalty parameter C and kernel parameter g are [-4, 4]; in addition, the velocity and location of each particle are randomly initialized; (4) Calculate the adaptability value of each particle and calibrate it; (5) Update particle velocity and position; (6) Determine if the termination condition is reached; if so, stop the update; otherwise, return to step (5); (7) When the number of iterations reaches the initial setting value, the optimal parameters C and g are obtained.
The PSO-SVM model construction process is shown in Figure 2.
where w is the inertia weight, k is the number of current iterations, c1 and c2 a factors and r1 and r2 are uniform random numbers in the range of (0, 1).

The PSO-SVM Method
The RBF kernel function is used in SVM to solve the nonlinear classificatio classification accuracy is mainly influenced by the penalty factor C and the ker eter g. The PSO algorithm is adopted to search for the ideal parameters C and g achieve the accurate and reliable detection of damage in refractory materials. T process of PSO-SVM is as follows: (1) The specific implementation steps of the PSO-SVM algorithm are as follo relevant parameters: the population size is 20, the inertia factor is 0.6, the a constants are 1.5 and 1.7, and the number of iterations is 50; (2) Prepare the train and test data sets. The training sets adopt the fivevalidation method and the classification accuracy is set to the particle fitn (3) The penalty parameter C and kernel parameter g are [-4, 4]; in addition, t and location of each particle are randomly initialized; (4) Calculate the adaptability value of each particle and calibrate it; (5) Update particle velocity and position; (6) Determine if the termination condition is reached; if so, stop the update; return to step (5); (7) When the number of iterations reaches the initial setting value, the optim ters C and g are obtained.
The PSO-SVM model construction process is shown in Figure 2.

Experimental Setup and Procedure
The samples and devices used for the experiments are given in Figure  alumina refractory material was manually impacted using an impact hamm percussion-induced sound was captured by a microphone (B&K Type 4966-H sound signals were collected by a data acquisition system (a NI cDAQ-9174 c

Experimental Setup and Procedure
The samples and devices used for the experiments are given in Figure 3. The high alumina refractory material was manually impacted using an impact hammer, and the percussion-induced sound was captured by a microphone (B&K Type 4966-H-041). The sound signals were collected by a data acquisition system (a NI cDAQ-9174 chassis with a NI-9232C data acquisition module) and stored on a laptop. The microphone was placed at a fixed position approximately 5 cm from the material and the sound signal was acquired at a sampling rate of 100 kHz. a NI-9232C data acquisition module) and stored on a laptop. The microphone was placed at a fixed position approximately 5 cm from the material and the sound signal was acquired at a sampling rate of 100 kHz. The size of the refractory material was 230 × 114 × 65 mm. During the experiment, a saw was used to cut slits of different depth on the refractory material to simulate five different degrees of damage (denoted as D1, D2, D3, D4, D5) to the refractory material. The specimen was percussed 100 times for each degree. The locations of the impact point and the simulated damage are shown in Figure 4. The degree of damage in the refractory materials in the impact test is given in Table 1.

Experimental Results and Analysis
In the experiment, all percussive sound signals were pre-processed by normalization, and the length of each signal was 0.1 s (sample points = 0.1 × 100,000 = 10,000). Figure 5 depicts the sound signal samples for each of the five damage degrees. It can be seen that the trend of the signals is similar in the time domain. The size of the refractory material was 230 × 114 × 65 mm. During the experiment, a saw was used to cut slits of different depth on the refractory material to simulate five different degrees of damage (denoted as D1, D2, D3, D4, D5) to the refractory material. The specimen was percussed 100 times for each degree. The locations of the impact point and the simulated damage are shown in Figure 4. The degree of damage in the refractory materials in the impact test is given in Table 1.
a NI-9232C data acquisition module) and stored on a laptop. The microphone was placed at a fixed position approximately 5 cm from the material and the sound signal was acquired at a sampling rate of 100 kHz. The size of the refractory material was 230 × 114 × 65 mm. During the experiment, a saw was used to cut slits of different depth on the refractory material to simulate five different degrees of damage (denoted as D1, D2, D3, D4, D5) to the refractory material. The specimen was percussed 100 times for each degree. The locations of the impact point and the simulated damage are shown in Figure 4. The degree of damage in the refractory materials in the impact test is given in Table 1.

Experimental Results and Analysis
In the experiment, all percussive sound signals were pre-processed by normalization, and the length of each signal was 0.1 s (sample points = 0.1 × 100,000 = 10,000). Figure 5 depicts the sound signal samples for each of the five damage degrees. It can be seen that the trend of the signals is similar in the time domain.

Experimental Results and Analysis
In the experiment, all percussive sound signals were pre-processed by normalization, and the length of each signal was 0.1 s (sample points = 0.1 × 100,000 = 10,000). Figure 5 depicts the sound signal samples for each of the five damage degrees. It can be seen that the trend of the signals is similar in the time domain. Time−frequency analysis was used to convert the signals into mel spectrograms. In mel spectrogram representation, the Hamming window is considered, and the parameters window length and overlap length [44] are set to 2048 sample points and 1024 sample points, respectively. Figure 6 shows the extracted mel spectrogram features. From the figure, it can be seen that the mel spectrogram depicts the energy variations of differen frequency bands over time, and that the mel spectrogram representation of sound signals has texture. This texture can be used to differentiate the percussion-induced acoustic signals with different degrees of damage. Next, two powerful texture descriptors−LBP and HOG−were considered to extrac features from the mel spectrogram. Since the generation of HOG feature depends on cel size, block size and the number of bins [37], these three parameters are discussed in this paper, as shown in Figure 7. In general, the size of the cell has a greater influence on texture information encoding. As illustrated in Figure 7, when the cell size was 16 × 16, the highest recognition accuracy was obtained, followed by 8 × 8. The cell size of 32 × 32 gen erated the worst performance. In addition, in contrast to the other three options, setting the block size and bins to 2 × 2 and 9 yielded better performance. Therefore, in this work the parameters of the HOG algorithm were set as shown in Table 2. On the other hand Time−frequency analysis was used to convert the signals into mel spectrograms. In mel spectrogram representation, the Hamming window is considered, and the parameters window length and overlap length [44] are set to 2048 sample points and 1024 sample points, respectively. Figure 6 shows the extracted mel spectrogram features. From the figure, it can be seen that the mel spectrogram depicts the energy variations of different frequency bands over time, and that the mel spectrogram representation of sound signals has texture. This texture can be used to differentiate the percussion-induced acoustic signals with different degrees of damage. Time−frequency analysis was used to convert the signals into mel spectrograms. In mel spectrogram representation, the Hamming window is considered, and the parameters window length and overlap length [44] are set to 2048 sample points and 1024 sample points, respectively. Figure 6 shows the extracted mel spectrogram features. From the figure, it can be seen that the mel spectrogram depicts the energy variations of different frequency bands over time, and that the mel spectrogram representation of sound signals has texture. This texture can be used to differentiate the percussion-induced acoustic signals with different degrees of damage. Next, two powerful texture descriptors−LBP and HOG−were considered to extract features from the mel spectrogram. Since the generation of HOG feature depends on cell size, block size and the number of bins [37], these three parameters are discussed in this paper, as shown in Figure 7. In general, the size of the cell has a greater influence on texture information encoding. As illustrated in Figure 7, when the cell size was 16 × 16, the highest recognition accuracy was obtained, followed by 8 × 8. The cell size of 32 × 32 generated the worst performance. In addition, in contrast to the other three options, setting the block size and bins to 2 × 2 and 9 yielded better performance. Therefore, in this work, the parameters of the HOG algorithm were set as shown in Table 2. On the other hand, different sampling radii R and numbers of sampling points P result in different image Next, two powerful texture descriptors−LBP and HOG−were considered to extract features from the mel spectrogram. Since the generation of HOG feature depends on cell size, block size and the number of bins [37], these three parameters are discussed in this paper, as shown in Figure 7. In general, the size of the cell has a greater influence on texture information encoding. As illustrated in Figure 7, when the cell size was 16 × 16, the highest recognition accuracy was obtained, followed by 8 × 8. The cell size of 32 × 32 generated the worst performance. In addition, in contrast to the other three options, setting the block size and bins to 2 × 2 and 9 yielded better performance. Therefore, in this work, the parameters of the HOG algorithm were set as shown in Table 2. On the other hand, different sampling radii R and numbers of sampling points P result in different image texture extraction capabilities for LBP features [39]. Six (P, R) pair values were chosen; the accuracy of the LBP features with different parameters is shown in Figure 8. As R becomes larger and the number of P increases, the texture description capability of LBP decreases. It is obvious that the best feature extraction is achieved when R = 1, P = 8. The LBP and HOG features emphasize the different texture information of the mel spectrogram. As the features are complementary, fusion was applied to concatenate the LBP and HOG feature vectors into enhanced vectors (LBP&HOG).
Micromachines 2023, 13, x FOR PEER REVIEW 8 of 15 texture extraction capabilities for LBP features [39]. Six (P, R) pair values were chosen; the accuracy of the LBP features with different parameters is shown in Figure 8. As R becomes larger and the number of P increases, the texture description capability of LBP decreases. It is obvious that the best feature extraction is achieved when R = 1, P = 8. The LBP and HOG features emphasize the different texture information of the mel spectrogram. As the features are complementary, fusion was applied to concatenate the LBP and HOG feature vectors into enhanced vectors (LBP&HOG).   After that, the entire dataset was divided into a training set and a test set. The overall speed and accuracy of the model is closely related to the reasonable partitioning of the data set. Table 3 presents the accuracy and time of different training-to-test set ratios. As can be seen from the table, the best accuracy and fastest speed were achieved when the training-to-test set ratio was 7:3. The total data size for the five damage degrees was 500. Therefore, for each damage degree, 70% of the data from the data sets were randomly taken as the training set, and the other 30% of the data were used as the testing set.   [39]. Six (P, R) pair values were chosen; the accuracy of the LBP features with different parameters is shown in Figure 8. As R becomes larger and the number of P increases, the texture description capability of LBP decreases. It is obvious that the best feature extraction is achieved when R = 1, P = 8. The LBP and HOG features emphasize the different texture information of the mel spectrogram. As the features are complementary, fusion was applied to concatenate the LBP and HOG feature vectors into enhanced vectors (LBP&HOG).   After that, the entire dataset was divided into a training set and a test set. The overall speed and accuracy of the model is closely related to the reasonable partitioning of the data set. Table 3 presents the accuracy and time of different training-to-test set ratios. As can be seen from the table, the best accuracy and fastest speed were achieved when the training-to-test set ratio was 7:3. The total data size for the five damage degrees was 500. Therefore, for each damage degree, 70% of the data from the data sets were randomly taken as the training set, and the other 30% of the data were used as the testing set. After that, the entire dataset was divided into a training set and a test set. The overall speed and accuracy of the model is closely related to the reasonable partitioning of the data set. Table 3 presents the accuracy and time of different training-to-test set ratios. As can be seen from the table, the best accuracy and fastest speed were achieved when the training-to-test set ratio was 7:3. The total data size for the five damage degrees was 500. Therefore, for each damage degree, 70% of the data from the data sets were randomly taken as the training set, and the other 30% of the data were used as the testing set. The features obtained in the previous step were used as the input of PSO-SVM and three damage detection models (HOG&LBP + PSO-SVM, HOG + PSO-SVM, LBP + PSO-SVM) were trained. Table 4 shows the optimal parameter values obtained by PSO searching in the solution space. The trained models were used to perform recognition on the test samples. The details and visualization of the predicted and real classes are shown in Figure 9. In Figure 9, the horizontal coordinate indicates the data set, and the vertical coordinate indicates the degree of damage in the refractory material. The ordinates of 1, 2, 3, 4 and 5, respectively denote D1~D5, which are also shown in Table 1. From the figure, it can be seen that the accuracy of HOG features is 89.33%, the accuracy of LBP features is 82.67% and the accuracy of LBP&HOG features is 98.67%. Meanwhile, Figure 9c shows that only four cases in the test set are classified into incorrect classes. It is obvious that the fused features (LBP&HOG) as input outperformed the typical single features in terms of classification performance. The quality of the output of the PSO-SVM classifier was evaluated by the performance parameters [45] precision, recall, F1-score and error rate in the classification task, as shown in Table 5. The quality of the output of the PSO-SVM classifier was evaluated by the performance parameters [45] precision, recall, F1-score and error rate in the classification task, as shown in Table 5. It is observed from Table 5 that the output of the PSO-SVM model had a precision of 0.94-1, a recall of 0.93-1, an F1-score of 0.95-1 and an error rate of 0-0.05. The results demonstrate that the PSO-SVM method yields excellent detection of damage severity in refractory materials.
To assess the effectiveness and superiority of the method proposed in this paper, some recognition results of well-known classifiers were used to the data. These included k-nearest neighbor (KNN), random forest (RF) and convolutional neural network (CNN). The results after repeated experiments are shown in Figure 10. The mean accuracy and implementation time for 10 experiments were calculated, as shown in Table 6. In the KNN classifier, several experiments were performed with various K values. The best performance was obtained with K = 12; however, the recognition was poor at only 91.06%. For the RF classifier, the number of trees was set to 100 and a recognition rate of 91.87% was obtained. Comparing the recognition results of the four classifiers, the proposed PSO-SVM classification is the best; it had the highest average accuracy, achieving a maximum recognition accuracy of 98.67%. The performance of CNN was second, with a maximum accuracy of 96.13% and an average accuracy second only to PSO-SVM. However, with CNN, due to the presence of a large number of convolutional operations, the number of calculation operations for trainable parameters increases significantly, which leads to a longer implementation time the worst performance in terms of implementation speed. As a result, it is proven that the method proposed in this paper obtains the best classification accuracy and saves valuable computational resources with its relatively low time required to complete the recognition task.
Micromachines 2023, 13, x FOR PEER REVIEW 11 of 15 accuracy of 96.13% and an average accuracy second only to PSO-SVM. However, with CNN, due to the presence of a large number of convolutional operations, the number of calculation operations for trainable parameters increases significantly, which leads to a longer implementation time the worst performance in terms of implementation speed. As a result, it is proven that the method proposed in this paper obtains the best classification accuracy and saves valuable computational resources with its relatively low time required to complete the recognition task.    To further investigate the effect of damage distribution at different locations on the performance of the proposed method, extended experiments were conducted. The shape of the refractory material is symmetrical. With the left side as the baseline, slits with a depth of 5 mm were fabricated at 46 mm, 92 mm, 138 mm and 184 mm (denoted as L1, L2, L3, and L4) from each of the four specimens to simulate damage at different locations. The distribution of damage at different locations is shown in Figure 11.  To further investigate the effect of damage distribution at different locations on the performance of the proposed method, extended experiments were conducted. The shape of the refractory material is symmetrical. With the left side as the baseline, slits with a depth of 5 mm were fabricated at 46 mm, 92 mm, 138 mm and 184 mm (denoted as L1, L2, L3, and L4) from each of the four specimens to simulate damage at different locations. The distribution of damage at different locations is shown in Figure 11. The percussion method was used to obtain percussion sound signals for the damage at different locations. Similarly, mel spectrogram and HOG&LBP were used to process each sound signal to acquire the feature vector. The PSO-SVM algorithm completed the The percussion method was used to obtain percussion sound signals for the damage at different locations. Similarly, mel spectrogram and HOG&LBP were used to process each sound signal to acquire the feature vector. The PSO-SVM algorithm completed the classification detection. All parameters were set in accordance with the best settings obtained from the analysis. A confusion matrix was used to visualize the recognition results of different damage locations, as shown in Figure 12.
Micromachines 2023, 13, x FOR PEER REVIEW 12 of 15 classification detection. All parameters were set in accordance with the best settings obtained from the analysis. A confusion matrix was used to visualize the recognition results of different damage locations, as shown in Figure 12. In Figure 12, the horizontal and vertical coordinates represent the true and predict labels of different damage locations, respectively. As shown in Figure 12, only a small number of samples were incorrectly identified (e.g., one L2 sample was classified as L1 and two L3 samples were classified as L4), while the rest of the samples were classified to the correct categories. The overall identification accuracy was 97.5%. From the results, it can be seen that the method has good generalization ability and can achieve effective recognition of damage with different location distributions. Therefore, the method proposed in this paper can accurately extract the key features of damage and realize the In Figure 12, the horizontal and vertical coordinates represent the true and predict labels of different damage locations, respectively. As shown in Figure 12, only a small number of samples were incorrectly identified (e.g., one L2 sample was classified as L1 and two L3 samples were classified as L4), while the rest of the samples were classified to the correct categories. The overall identification accuracy was 97.5%. From the results, it can be seen that the method has good generalization ability and can achieve effective recognition of damage with different location distributions. Therefore, the method proposed in this paper can accurately extract the key features of damage and realize the accurate detection.
In addition, the performance results of the proposed method were compared with other newly published methods for solving the damage detection problem. As can be seen in Table 7, the proposed method in this paper yields better accuracy scores than multiple data processing methods and classification models.

Conclusions
Based on percussion and PSO-SVM, a novel method for damage detection and identification in refractory materials is proposed in this paper. Sound signals generated by manually controlled percussions are converted to mel spectrograms, and LBP features and HOG features are extracted from mel spectrograms. Then, the two features are fused and input into a PSO-SVM for training to realize damage detection in refractory materials. The experimental results verified the effectiveness of the proposed method. It is worth noting that converting percussion-induced acoustic signals into mel spectrograms and using the fused HOG&LBP method achieved better damage detection in refractory materials than typical single LBP and HOG texture features. Furthermore, the recognition accuracy of the PSO-SVM was in the range of 96-98.67%, with a more stable classification performance and higher classification accuracy than the other three classification strategies. Overall, the damage detection method for refractory materials proposed in this paper is convenient and reliable, with potential for field application. In future work, the combination of robotics and machine learning can be further implemented. Specifically, a robotic arm will carry a tapping device and a microphone for automatic tapping and signal acquisition to produce a system for automatic detection of refractory material damage. In addition, environmental noise is not considered in the paper; further investigation will be conducted on this issue.  Data Availability Statement: Due to the nature of the research, the data of this study are not shared publicly and are only available upon reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.