Method and device for monitoring loosening of multiple bolts based on oversampling of a gaussian mixture model

Abstract

The application discloses a method and related device for over-sampling multi-bolt loosening monitoring based on a Gaussian mixture model, which comprises the following steps: extracting Lamb wave signals of different bolt loosening conditions, extracting characteristic values, performing characteristic reduction and normalization; dividing each minority class into different Gaussian component clusters and calculating the number of new samples to be generated for each cluster; applying an SMOTE over-sampling method to each cluster; after generating a new sample, a criterion based on KL distance is proposed to judge whether the new sample is an optimal sample; new samples are continuously generated through the SMOTE until the number of new samples is satisfied, and a KNN model is established to obtain a bolt loosening monitoring result. The obtained bolt loosening classification result is close to the actual bolt loosening condition, the classification result is accurate, and the problem that all bolt loosening states are not considered in traditional multi-bolt loosening monitoring is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of multiple bolt loosening monitoring, and in particular to an oversampling multiple bolt loosening monitoring method based on a Gaussian mixture model and a related device. Background Art

[0002] Bolted connections are the most common type of connection in high-speed train structures. Due to dynamic loads, improper installation, and severe exposure to the working environment, structural integrity can naturally loosen, causing serious safety accidents. Existing bolt loosening monitoring methods generally only consider inter-class balance, such as the imbalance in the amount of data between the fully tightened state and the loosened state. Because repeated tightening of bolts can cause certain structural changes and create uncertainty, data on some bolt loosening states is very limited. Therefore, traditional bolt loosening methods cannot take into account all bolt loosening states, and bolt loosening monitoring based on inter-class imbalance is not comprehensive. Summary of the Invention

[0003] The purpose of the present invention is to provide an oversampling multi-bolt loosening monitoring method and related devices based on a Gaussian mixture model to solve the problem that traditional multi-bolt loosening monitoring methods only consider inter-class balance but not intra-class balance.

[0004] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0005] The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model includes:

[0006] Extract Lamb wave signals of different bolt loosening conditions, extract eigenvalues ​​and perform feature reduction and normalization;

[0007] Divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster;

[0008] Apply the SMOTE oversampling method to each cluster;

[0009] After generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample;

[0010] New samples are continuously generated through SMOTE until the number of new samples is met. Then, the KNN model is established to obtain the bolt loosening monitoring results.

[0011] Furthermore, twenty eigenvalues ​​were extracted and feature reduction was performed through principal component analysis, reduced to three eigenvalues, and all features were normalized.

[0012] Furthermore, the twenty eigenvalues ​​are mean, peak-to-peak value, rectified mean, variance, standard deviation, kurtosis, skewness, root mean square, form factor, crest factor, impulse factor, margin factor, power spectrum entropy, energy entropy, approximate entropy, sample entropy, fuzzy entropy, permutation entropy, envelope entropy and correlation coefficient.

[0013] Furthermore, each minority class is divided into different Gaussian component clusters and the number of new samples required to be generated for each cluster is calculated, including:

[0014] Each minority class is divided into k clusters using the Gaussian mixture clustering method. The specific process is as follows:

[0015]

[0016] Where λ i is the weight of cluster i, is the probability density distribution of cluster i, which is expressed as:

[0017]

[0018] In the formula, d is the dimension of the feature, is the sample point of cluster i, u i and ∑ i are the mean and variance of cluster i respectively;

[0019] The number of new samples required to generate each cluster is calculated as follows:

[0020] N i(new) =W i ×NN i

[0021] Where N i(new) is the number of samples that need to be newly generated, N is the number of majority class samples, N i is the original number of samples in the ith cluster, and W is the weight of the ith cluster, which is calculated as follows:

[0022]

[0023] Where V i is the volume of the i-th cluster, which is the volume of a sphere with the mean of the cluster as the center and the distance from the farthest point to the mean as the radius.

[0024] Furthermore, after generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample, including:

[0025] After generating a new sample, construct two criteria D ra and D er , to determine whether this new sample is the optimal sample, the calculation process is as follows:

[0026]

[0027]

[0028] in Represents the sample distribution composed of a new sample and the original sample cluster, Represents the sample distribution of the original sample cluster, p(U h ) represents the distribution of majority class samples, and KL(,) represents the KL distance between two distributions;

[0029] Determine whether the new sample is the optimal sample by judging whether the following formula is satisfied:

[0030] D ra <0.5

[0031] D er >0

[0032] If it is the best sample, the sample is retained; otherwise, the sample is discarded.

[0033] Furthermore, after establishing the KNN model, the bolt loosening monitoring results were obtained, including:

[0034] The parameters of the KNN model are set to the three adjacent points. After the model is established, the multi-bolt loosening monitoring results are given.

[0035] Furthermore, M lead zirconate titanate piezoelectric ceramic sensors are provided at the target structure for collecting signals, where M is a positive integer.

[0036] Furthermore, the oversampling multi-bolt loosening monitoring system based on the Gaussian mixture model includes:

[0037] Signal acquisition module, used to extract Lamb wave signals of different bolt loosening conditions, extract eigenvalues ​​and perform feature reduction and normalization;

[0038] The classification module is used to divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster;

[0039] The judgment module is used to apply the SMOTE oversampling method to each cluster; after generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample;

[0040] The output module is used to continuously generate new samples through SMOTE until the number of new samples is met, and then the KNN model is established to obtain the bolt loosening monitoring results.

[0041] Furthermore, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the oversampling multi-bolt loosening monitoring method based on the Gaussian mixture model are implemented.

[0042] Furthermore, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the over-sampling multi-bolt loosening monitoring method based on a Gaussian mixture model are implemented.

[0043] Compared with the prior art, the present invention has the following technical effects:

[0044] The present invention provides a multi-bolt loosening monitoring method based on an oversampling method of a Gaussian mixture model. Using a multi-bolt loosening monitoring experiment, signals under different loosening conditions are obtained. Feature reduction is performed after extracting eigenvalues. Minority class samples are clustered and the number of new samples required to generate each cluster is calculated. New sample points are generated using the original SMOTE method. Based on the criteria proposed in this article, whether the new sample is the optimal sample is determined. After the number of new samples generated meets the standard, a bolt loosening classification model is constructed using the support vector method to determine the loosening status of multiple bolts. In this way, the obtained bolt loosening classification results are close to the actual bolt loosening conditions, the classification results are accurate, and the problem of not considering the loosening status of all bolts in traditional multi-bolt loosening monitoring is overcome.

[0045] The multi-bolt loosening monitoring system provided by the present invention using an oversampling method based on a Gaussian mixture model can accurately sense the structural strain changes caused by bolt loosening through lead zirconate titanate piezoelectric ceramic sensors arranged at key positions of the structure. By extracting the eigenvalues ​​of the Lamb waves emitted by the lead zirconate titanate piezoelectric ceramic sensors, a quantitative crack monitoring model can be established to achieve real-time monitoring of crack damage. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 Flowchart of a method for monitoring multiple bolt loosening based on an oversampling method of a Gaussian mixture model according to a preferred embodiment of the present invention;

[0047] Figure 2 A schematic diagram of the dimensions of a simulation specimen according to a preferred embodiment of the present invention;

[0048] Figure 3 A sample diagram of a balanced data set according to a preferred embodiment of the present invention;

[0049] Figure 4 This is a confusion matrix diagram in the experiment of the preferred embodiment of the present invention;

[0050] Figure 5 4 is a ROC curve diagram of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following is a clear and complete description of the technical solutions of the present invention. It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0052] Unless otherwise defined, the technical or scientific terms used in the present invention shall have the usual meanings understood by persons of ordinary skill in the field to which the present invention belongs. The words "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "a" do not indicate a quantity limitation, but rather indicate the existence of at least one. Words such as "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0053] It should be noted that due to the complex operating environment of high-speed rail, bolts often become loose. Single bolt loosening events are relatively rare, and multiple bolt loosening events are even rarer. Such loosening events can easily cause structural damage and threaten driving safety. Therefore, monitoring the loosening of each bolt on a high-speed rail is essential. Based on this, the present application provides a multi-bolt loosening monitoring method based on an oversampling method using a Gaussian mixture model.

[0054] See Figure 1 The present invention provides a method for monitoring the looseness of multiple bolts using an oversampling method based on a Gaussian mixture model, including:

[0055] Lamb wave signals of different bolt loosening conditions were extracted through the structural health monitoring system, twenty eigenvalues ​​were extracted and feature reduction and normalization were performed through principal component analysis;

[0056] Divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster;

[0057] The SMOTE oversampling method is used for each cluster;

[0058] After generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample;

[0059] New samples are continuously generated through SMOTE until the number of new samples is met. Then, the KNN model is established to obtain the bolt loosening monitoring results.

[0060] In this embodiment, Lamb wave signals of different bolt loosening conditions are extracted through the structural health monitoring system. Specifically, the loosening conditions include bolt tightening status, single bolt loosening status, and multiple bolt loosening status signals. Twenty eigenvalues ​​are extracted and feature reduction and normalization are performed through principal component analysis. The eigenvalues ​​include: mean, peak-to-peak value, rectified mean, variance, standard deviation, kurtosis, skewness, root mean square, form factor, crest factor, impulse factor, margin factor, power spectrum entropy, energy entropy, approximate entropy, sample entropy, fuzzy entropy, permutation entropy, envelope entropy, and correlation coefficient. The expression is as follows:

[0061]

[0062]

[0063] The twenty eigenvalues ​​are reduced to three eigenvalues ​​through principal component analysis. Then, the three reduced features are normalized to -1 to 1, where the fully tightened state signal is classified into the majority class, and the single bolt loosening state signal and the multiple bolt loosening state signal are classified into the minority class.

[0064] Among them, the difference in the number of samples between the majority class and the minority class is inter-class imbalance, and the difference in the number / density of samples in each cluster in the minority class is intra-class imbalance.

[0065] In this embodiment, each minority class is divided into multiple Gaussian component clusters and the number of new samples required to be generated for each cluster is calculated, including:

[0066] Each minority class is divided into k Gaussian component clusters by the Gaussian mixture clustering method. The calculation process is as follows:

[0067]

[0068] Where λ i is the weight of cluster i, is the probability density distribution of cluster i, which is expressed as:

[0069]

[0070] In the formula, d is the dimension of the feature, is the sample point of cluster i, u i and ∑ i are the mean and variance of the i-th cluster respectively. Bayesian principle and EM algorithm are used to calculate each cluster. Each Gaussian component is determined by the calculated mean and covariance, which represents the sample distribution in each cluster. After clustering by GMM, each minority class is defined as

[0071] The number of samples in each Gaussian component is different, and the number of samples in a few Gaussian component clusters is very small, which leads to intra-class imbalance. Therefore, we need to get the number of new samples that need to be generated for each cluster. The calculation process is as follows:

[0072] N i(new) =W i ×NN i

[0073] Where N i(new) is the number of samples that need to be newly generated, N is the number of majority class samples, N i is the original number of samples in the ith cluster, and W is the weight of the ith cluster, which is calculated as follows:

[0074]

[0075] Where V i is the volume of the i-th cluster, which is the volume of a sphere with the mean of the cluster as the center and the distance from the farthest point to the mean as the radius.

[0076] In this embodiment, the sample generation method of the original SMOTE method is used to generate a new sample for each cluster in each minority class. The calculation formula is as follows:

[0077] x new =x+Rand(0,1)×(x n -x)

[0078] Where x is the original sample in the cluster. Rand(0,1) is a random number between 0 and 1. n represents the nth adjacent sample, x new Represents a new sample.

[0079] In this embodiment, after a new sample is generated, a criterion based on KL distance is proposed to determine whether the new sample is the best sample, including:

[0080] After generating a new sample, construct two criteria D ra and D er , to determine whether this new sample is the optimal sample, the calculation process is as follows:

[0081]

[0082]

[0083] in Represents the sample distribution composed of a new sample and the original sample cluster, Represents the sample distribution of the original sample cluster, p(Uh ) represents the distribution of the majority class samples, KL(, ) represents the KL distance between two distributions.

[0084] The calculated result is brought into the following formula to observe whether the condition is met to determine whether the new sample is the optimal sample:

[0085] D ra <0.5

[0086] D er >0

[0087] If it is the optimal sample, the sample is retained, otherwise the sample is discarded.

[0088] In this embodiment, after retaining or discarding a sample, the SMOTE operation is repeated until the number of generated samples reaches the number calculated above. The generated dataset includes one majority class and multiple minority classes, wherein the number of samples of the majority class is the same as that of each minority class. The number / density of samples of different clusters in each minority class is the same. Then a KNN model is generated through the dataset, which can be used to diagnose the bolt loosening state.

[0089] In summary, the application establishes a multi-bolt loosening condition classification model by extracting multiple feature values. Specifically, when a multi-bolt loosening condition classification model is established using the KNN method, part of the sample data is selected as training samples, and the multi-bolt loosening condition classification model is trained. The training accuracy depends on the selection of the number of neighboring samples N. The cross-validation method can be used to select N iteratively to obtain the optimal model parameters. Finally, the remaining test samples can be used to classify the multi-bolt loosening condition. In this way, the diagnosis efficiency and accuracy can be improved.

[0090] Next, the method steps of the application will be further illustrated by a certain test example.

[0091] In this example, a SHM structure health monitoring system is used to obtain Lamb wave signals under different bolt loosening conditions. The mechanical property parameters of the connected component materials are shown in Table 1, and the sensor parameters are shown in Table 2.

[0092] Table 1 Mechanical property parameters of connected component materials

[0093]

[0094] Table 2 Sensor parameters

[0095]

[0096] Please refer to Figure 2 , Figure 2 is a schematic diagram of the connected component plate, sensor position and bolt position, etc. Figure 2 In the figure, B1 represents bolt No. 1; B2 represents bolt No. 2; B3 represents bolt No. 3; PZT1 represents sensor No. 1; PZT2 represents sensor No. 2; PZT3 represents sensor No. 3; and PZT4 represents sensor No. 4. 1 represents the path formed by PZT1-PZT2; 2 represents the path formed by PZT2-PZT1; 3 represents the path formed by PZT3-PZT4; 4 represents the path formed by PZT4-PZT3; 5 represents the path formed by PZT3-PZT2; 6 represents the path formed by PZT2-PZT3; 7 represents the path formed by PZT1-PZT4; and 8 represents the path formed by PZT4-PZT1. Three 11 mm diameter holes were drilled in the center of the plate, and three M10 bolts and nuts were installed. The three bolts have eight different loosening conditions, as shown in Table 3. In each test process, the bolt is first loosened to the specified torque, then tightened and loosened again. This test design not only simulates bolt loosening, but also covers the impact of repeated tightening on bolt monitoring, which is more in line with engineering practice. In this embodiment, a 5-cycle sine wave of 145kHz, 150kHz, and 155kHz modulated by a Hanning window is selected as the experimental excitation signal. After each loosening, three frequencies of Lamb waves are used for excitation and acquisition, and the excitation signals are shown in the figure. Each case is repeated 25 times of loosening and tightening, with a total of 600 (25×8×3) Lamb wave signals, a total of 4800 signals. Among them, 480 groups for each case are training sets, and the remaining 120 groups are test sets.

[0097] Table 3 Bolt looseness (T: 60Nm; L: 50-55Nm)

[0098]

[0099] Twenty eigenvalues ​​were extracted from 4800 Lmab wave signals, resulting in a 4800 × 20 feature matrix. Feature reduction was then performed on this matrix, yielding 4800 feature samples, each with three features. These three features were then normalized to a range of -1 to 1. Some single-bolt loosening samples and multiple-bolt loosening samples were randomly deleted to create an unbalanced dataset. The proposed oversampling method was then used to create a balanced dataset, as shown in Table 4.

[0100] Table 4 Number of bolt samples

[0101]

[0102]

[0103] See Figure 3 , Figure 3This figure shows how the samples change after oversampling. It can be seen that the number of samples has become the same. The circles represent the original samples, and the triangles represent the new samples generated using the above method.

[0104] A KNN classification model is constructed based on the balanced dataset to establish a bolt loosening classification model to achieve the classification of bolt loosening. 120 sets of data are used as training sets for each case, i.e. a total of 960 sets of training sets. The training results can be found in Figure 4 , Figure 4 The classification accuracy of the balanced dataset after sample generation is shown. Figure 5 The ROC curve for the balanced dataset is shown. The AUC is defined as the area between the ROC curve and the x-axis. The larger the AUC, the better the training effect.

[0105] according to Figure 4 and Figure 5 It can be determined that the predicted bolt loosening classification is very close to the actual bolt loosening situation. It can be seen that the classification result of this application is relatively accurate.

[0106] The present application provides a multi-bolt loosening monitoring system based on an oversampling method of a Gaussian mixture model, comprising M lead zirconate titanate piezoelectric ceramic sensors arranged at a target structure of a bolt-connected part, where M is a positive integer.

[0107] The above-mentioned multi-bolt loosening monitoring system based on the oversampling method of the Gaussian mixture model can accurately sense the structural strain changes caused by bolt loosening through the lead zirconate titanate piezoelectric ceramic sensors arranged at key positions of the structure. The bolt loosening classification model established through the Lamb wave signal of the bolt loosening can realize real-time monitoring of the bolt loosening situation.

[0108] In another embodiment of the present invention, an oversampling multi-bolt loosening monitoring system based on a Gaussian mixture model is provided, which can be used to implement the above-mentioned oversampling multi-bolt loosening monitoring method based on a Gaussian mixture model. Specifically, the system includes:

[0109] The signal acquisition module is used to extract Lamb wave signals of different bolt loosening conditions, extract eigenvalues, and perform feature reduction and normalization;

[0110] The classification module is used to divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster;

[0111] The judgment module is used to apply the SMOTE oversampling method to each cluster; after generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample;

[0112] The output module is used for continuously generating new samples by SMOTE until the number of new samples meets the requirement, and then a KNN model is established to obtain the bolt loosening monitoring result.

[0113] The division of the modules in the embodiments of the present application is illustrative, and is only a logical function division. In actual implementation, another division manner can be used. In addition, each function module in each embodiment of the present application can be integrated in one processor, or can be a separate physical existence, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module.

[0114] In another embodiment of the present application, a computer device is provided, which includes a processor and a memory. The memory is used to store a computer program, and the computer program includes program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor is the computing core and control core of the terminal, and is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method flow or a corresponding function. The processor in the embodiments of the present application can be used for the operation of the oversampling multi-bolt loosening monitoring method based on the Gaussian mixture model.

[0115] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the oversampling multi-bolt loosening monitoring method based on the Gaussian mixture model in the above embodiment.

[0116] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. An oversampling multi-bolt loosening monitoring method based on Gaussian mixture model is characterized by: include: Extract Lamb wave signals of different bolt loosening conditions, extract eigenvalues ​​and perform feature reduction and normalization; Divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster; Apply the SMOTE oversampling method to each cluster; After generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample; New samples are continuously generated through SMOTE until the number of new samples is met. Then, the KNN model is established to obtain the bolt loosening monitoring results. After generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample, including: After generating a new sample, construct two criteria D ra and D er , to determine whether this new sample is the optimal sample, the calculation process is as follows: in Represents the sample distribution composed of a new sample and the original sample cluster, Represents the sample distribution of the original sample cluster, p(U h ) represents the distribution of majority class samples, and KL(,) represents the KL distance between two distributions; By judging whether the following formula is satisfied, we can determine whether the new sample is the optimal sample: D ra ＜0.5 D er ＞0 If it is the best sample, the sample is retained; otherwise, the sample is discarded.

2. The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model according to claim 1 is characterized in that: Twenty eigenvalues ​​were extracted and reduced to three eigenvalues ​​through principal component analysis, and all features were normalized.

3. The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model according to claim 2 is characterized in that: The twenty eigenvalues ​​are mean, peak-to-peak value, rectified mean, variance, standard deviation, kurtosis, skewness, root mean square, form factor, crest factor, impulse factor, margin factor, power spectrum entropy, energy entropy, approximate entropy, sample entropy, fuzzy entropy, permutation entropy, envelope entropy and correlation coefficient.

4. The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model according to claim 1 is characterized in that: Divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster, including: Each minority class is divided into k clusters using the Gaussian mixture clustering method. The specific process is as follows: Where λ i is the weight of cluster i, is the probability density distribution of cluster i, which is expressed as: In the formula, d is the dimension of the feature, is the sample point of cluster i, u i and Σ i are the mean and variance of cluster i respectively; The number of new samples required to generate each cluster is calculated as follows: N i(new) =W i ×N-N i Where N i(new) is the number of samples that need to be newly generated, N is the number of majority class samples, N i is the original number of samples in the ith cluster, and W is the weight of the ith cluster, which is calculated as follows: Where V i is the volume of the i-th cluster, which is the volume of a sphere with the mean of the cluster as the center and the distance from the farthest point to the mean as the radius.

5. The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model according to claim 1 is characterized in that: After establishing the KNN model, the bolt loosening monitoring results were obtained, including: The parameters of the KNN model are set to the three adjacent points. After the model is established, the multi-bolt loosening monitoring results are given.

6. The oversampling multi-bolt loosening monitoring method based on Gaussian mixture model according to claim 1 is characterized in that: M lead zirconate titanate piezoelectric ceramic sensors are arranged at the target structure to collect signals, where M is a positive integer.

7. Oversampling multi-bolt loosening monitoring system based on Gaussian mixture model, characterized by: include: The signal acquisition module is used to extract Lamb wave signals of different bolt loosening conditions, extract eigenvalues, and perform feature reduction and normalization; The classification module is used to divide each minority class into different Gaussian component clusters and calculate the number of new samples that need to be generated for each cluster; The judgment module is used to apply the SMOTE oversampling method to each cluster; after generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample; The output module is used to continuously generate new samples through SMOTE until the number of new samples is met, and then the KNN model is established to obtain the bolt loosening monitoring results; After generating a new sample, a criterion based on KL distance is proposed to determine whether the new sample is the best sample, including: After generating a new sample, construct two criteria D ra and D er , to determine whether this new sample is the optimal sample, the calculation process is as follows: in Represents the sample distribution composed of a new sample and the original sample cluster, Represents the sample distribution of the original sample cluster, p(U h ) represents the distribution of majority class samples, and KL(,) represents the KL distance between two distributions; By judging whether the following formula is satisfied, we can determine whether the new sample is the optimal sample: D ra ＜0.5 D er ＞0 If it is the best sample, the sample is retained; otherwise, the sample is discarded.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the oversampling multi-bolt loosening monitoring method based on the Gaussian mixture model as described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the over-sampling multi-bolt loosening monitoring method based on a Gaussian mixture model as claimed in any one of claims 1 to 6 are implemented.

Images