Roller bearing abnormal vibration fault diagnosis method and system based on improved SVM

Abstract

The invention provides a fault diagnosis method and system for abnormal vibration of a roller bearing based on an improved SVM (Support Vector Machine). The method comprises the following steps: acquiring a vibration temperature signal in a target area; performing signal preprocessing on the vibration temperature signal to obtain a de-noised vibration temperature signal; feature extraction is carried out based on the de-noised vibration temperature signal to obtain a de-noised vibration temperature signal fault feature vector; inputting the de-noised vibration temperature signal fault feature vector into a particle swarm support vector so as to distinguish different types of bearing faults; and integrating different types of bearing faults, displaying the bearing faults through a webpage end, and pushing the bearing faults to a user to guide equipment repair and maintenance. According to the method, diagnosis of 3-5 abnormal vibration fault reasons of the roller (bearing) can be realized, the diagnosis accuracy based on a historical test data set reaches 70%, and the method is expected to improve the maintenance work efficiency, shorten the maintenance time and avoid economic loss caused by serious faults.

Description

Technical Field

[0001] The invention belongs to the technical field of electronic communications and relates to a fault diagnosis method, in particular to a fault diagnosis method and system for abnormal vibration of a roller bearing based on an improved SVM. Background Art

[0002] Belt conveyor roller (bearing) is a very important component of belt conveyor, and its performance directly affects the operation effect of the entire conveyor.

[0003] The roller is the main component of the belt conveyor for transmitting power, and is also a part prone to failure. The roller has several key functions. On the one hand, the roller transmits power to the conveyor belt through rotation to make the conveyor belt run; on the other hand, as a supporting component, it can effectively share the load of the conveyor belt to prevent the conveyor belt from breaking or being damaged due to excessive force. The roller bearing, as the core component of the roller, affects the normal operation of the roller.

[0004] Therefore, it is necessary to pay attention to the maintenance of the roller bearings during use, and promptly check and repair roller bearing failures to ensure the normal operation of the conveyor.

[0005] At present, vibration sensors are installed on the roller bearings to monitor the operating status of the roller bearings, but there is a lack of judgment on the cause of the abnormal vibration of the roller bearings. It is necessary to study an intelligent diagnosis model for abnormal vibration of the roller (bearing), which can perform fault diagnosis based on the real-time detection data of the sensor and the historical fault inspection and repair information.

[0006] In order to avoid the above-mentioned effects caused by abnormal vibration of the roller (bearing) of the belt conveyor, an effective fault diagnosis method for abnormal vibration of the roller (bearing) of the belt conveyor is urgently needed. At present, there are mainly the following problems:

[0007] First, it is inconvenient to conduct manual inspection when the inspection environment is bad, and manual inspection is likely to cause casualties;

[0008] Secondly, there are obvious defects such as high inspection costs and low efficiency, which seriously affect production and transportation work.

[0009] Therefore, in the existing belt conveyor roller (bearing) abnormal vibration fault diagnosis technology, the roller bearing fault is discovered through manual inspection, which leads to high inspection costs, long working time and low results, and thus the maintenance work efficiency cannot be completed efficiently, resulting in long maintenance time, and even more serious faults and greater economic losses due to inadequate fault diagnosis. Summary of the invention

[0010] In view of the shortcomings of the prior art mentioned above, the purpose of the present application is to provide a fault diagnosis method and system for abnormal vibration of roller bearings based on improved SVM, which is used to solve the problem that in the process of implementing the fault diagnosis technology of abnormal vibration of rollers (bearings) of belt conveyors in the prior art, the inspection cost is high, the working time is long and the effect is low due to the discovery of roller bearing faults through manual detection, so that the maintenance work cannot be completed efficiently, resulting in long maintenance time, and even more serious faults and greater economic losses due to inadequate fault diagnosis.

[0011] To achieve the above-mentioned purpose and other related purposes, in the first aspect, the present application provides a fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM, comprising the following steps: obtaining a vibration temperature signal within a target area; performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal; performing feature extraction based on the denoised vibration temperature signal to obtain a denoised vibration temperature signal fault feature vector; inputting the denoised vibration temperature signal fault feature vector into a particle swarm support vector to distinguish different types of bearing faults; integrating different types of bearing faults, displaying them through a web page, and pushing them to users to guide equipment repair and maintenance.

[0012] In an implementation of the first aspect, obtaining the vibration temperature signal in the target area includes the following steps: installing a temperature vibration probe in the target area; the temperature vibration probe includes a temperature vibration composite sensor; and collecting the vibration temperature signals of each setting position in real time based on the temperature vibration probe.

[0013] In an implementation of the first aspect, signal preprocessing is performed on the vibration temperature signal to obtain a denoised vibration temperature signal, including the following steps: performing wavelet decomposition based on the vibration temperature signal to obtain wavelet coefficients of the vibration temperature signal; denoising the wavelet coefficients of the vibration temperature signal based on threshold function processing; and reconstructing the denoised vibration temperature signal based on an inverse wavelet transform to obtain a denoised vibration temperature signal.

[0014] In an implementation manner of the first aspect, a calculation formula of the threshold function is:

[0015]

[0016] Among them, σ j Expressed as standard deviation; w j It is represented by the wavelet coefficient; λ is represented by the threshold value calculated according to the threshold rule; median() is represented by the median function.

[0017] In an implementation of the first aspect, feature extraction is performed based on the denoised vibration-temperature signal to obtain a denoised vibration-temperature signal fault feature vector, including the following steps: constructing a denoised vibration-temperature signal probability density model based on the denoised vibration-temperature signal; performing feature extraction on the denoised vibration-temperature signal probability density model to obtain a denoised vibration-temperature signal feature vector within a sampling period; constructing a signal histogram based on the denoised vibration-temperature signal feature vector; performing probability density fitting and parameter estimation based on the signal histogram to obtain a denoised vibration-temperature signal fault feature vector.

[0018] In an implementation of the first aspect, the probability density fitting curve formula is:

[0019]

[0020] The Gaussian mixture distribution expression is:

[0021] g(xμ k ,σ k )

[0022] Where k represents the number of components; w k Expressed as a mixing weight, 0≤w k ≤1; g(xμ k ,σ k ) is the distribution of the kth Gaussian component.

[0023] In an implementation of the first aspect, the denoised vibration and temperature signal fault feature vector is input into a particle swarm support vector to distinguish different types of bearing faults, including the following steps: constructing a multi-fault classification model based on the denoised vibration and temperature signal fault feature vector; the multi-fault classification model includes a directed acyclic classifier composed of several binary classification SVMs; based on the multi-fault classification model, determining a penalty factor and a kernel function parameter; optimizing the SVM based on a particle swarm optimization algorithm to distinguish different types of bearing faults.

[0024] In an implementation of the first aspect, optimizing SVM based on a particle swarm optimization algorithm to distinguish different types of bearing faults includes the following steps: defining a number of particles in space; initializing a population and setting parameters for the particles; calculating the fitness value of the particles according to a fitness function; calculating a single optimal and a global optimal particle, and updating the particle position and particle velocity; judging the direction and distance of particle flight based on the particle velocity, and judging whether it meets preset conditions; when the preset conditions are met, outputting the optimal parameters of SVM; constructing a PSO-SVM model based on the optimal parameters of SVM to complete the optimization process of SVM.

[0025] In the second aspect, the present application provides a fault diagnosis system for abnormal vibration of a roller bearing based on an improved SVM, comprising: an acquisition module for acquiring a vibration temperature signal within a target area; a data processing module for performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal; a feature extraction module for performing feature extraction based on the denoised vibration temperature signal to obtain a denoised vibration temperature signal fault feature vector; a fault diagnosis module for inputting the denoised vibration temperature signal fault feature vector into a particle swarm support vector to distinguish different types of bearing faults; and a display module for integrating different types of bearing faults, displaying them through a web page, and pushing them to users to guide equipment repair and maintenance.

[0026] In the last aspect, the present application provides a fault diagnosis device for abnormal vibration of a roller bearing based on an improved SVM, comprising: a processor and a memory. The memory is used to store a computer program; the processor is connected to the memory and is used to execute the computer program stored in the memory, so that the fault diagnosis device for abnormal vibration of a roller bearing based on the improved SVM executes the fault diagnosis method for abnormal vibration of a roller bearing based on the improved SVM.

[0027] As described above, the fault diagnosis method and system for abnormal vibration of a roller bearing based on improved SVM of the present invention have the following beneficial effects:

[0028] The fault diagnosis of abnormal vibration of roller bearings based on improved SVM in this application can realize the diagnosis of 3-5 causes of abnormal vibration of rollers (bearings), and the diagnosis accuracy rate based on historical test data sets reaches 70%. The application of this method improves the maintenance efficiency of rollers (bearings) of belt conveyors, shortens the maintenance time, and avoids economic losses caused by serious faults. At the same time, the fault diagnosis method of abnormal vibration of roller bearings based on improved SVM provided by this application can realize timely detection and elimination of faults by combining vibration sensor data, avoid serious impact caused by further development of faults, avoid large expenses and operation time caused by manual inspections, and reduce equipment maintenance operation costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic flow chart of a fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM in one embodiment of the present invention.

[0030] Figure 2 It is a schematic flow chart of S11 in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0031] Figure 3AIt is a schematic flow chart of S12 in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0032] Figure 3B Shown is a schematic diagram of the wavelet transform process in the fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0033] Figure 4A It is a schematic flow chart of S13 in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0034] Figure 4B It shows a schematic diagram of signal probability density fitting in the fault diagnosis method of abnormal vibration of roller bearing based on improved SVM of the present invention.

[0035] Figure 5A It is a schematic flow chart of S14 in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0036] Figure 5B It is a schematic diagram of a directed acyclic classifier in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0037] Figure 5C Shown is a schematic diagram of the SVM process using the PSO algorithm to optimize the SVM in the fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0038] Figure 6 The display platform shows the integrated effect of the fault diagnosis algorithm in the fault diagnosis method for abnormal vibration of the roller bearing based on the improved SVM of the present invention.

[0039] Figure 7 It is a schematic diagram of the principle structure of a fault diagnosis system for abnormal vibration of a roller bearing based on improved SVM in one embodiment of the present invention.

[0040] Figure 8 It is a schematic diagram showing the principle structure of a fault diagnosis device for abnormal vibration of a roller bearing based on an improved SVM in one embodiment of the present invention.

[0041] Component number description

[0042] 71 Get Module

[0043] 72 Data processing module

[0044] 73 Feature extraction module

[0045] 74 Fault diagnosis module

[0046] 75 Display Module

[0047] 81 Processor

[0048] 82 Memory

[0049] Steps S11 to S15 DETAILED DESCRIPTION

[0050] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0051] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0052] The fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM provided in the embodiments of the present application will be described in detail below in conjunction with the drawings in the embodiments of the present application.

[0053] See also Figure 1 , which is a flow chart of a fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM in one embodiment of the present invention. Figure 1 As shown, this embodiment provides a fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM.

[0054] The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM specifically comprises the following steps:

[0055] S11, obtaining a vibration temperature signal in a target area.

[0056] See also Figure 2 , which is a flow chart of S11 in the fault diagnosis method of abnormal vibration of a roller bearing based on improved SVM of the present invention. The specific steps are as follows:

[0057] Belt conveyor is a friction-driven machine that transports materials in a continuous manner. It is mainly composed of a frame, conveyor belt, rollers, drums, tensioning devices, transmission devices and other parts. It can form a material conveying process from the initial feeding point to the final unloading point on a certain conveying line. It can convey both bulk materials and pieced items. In addition to pure material conveying, it can also cooperate with the requirements of the process in the production process of various industrial enterprises to form a rhythmic flow operation transportation line.

[0058] The belt conveyor is mainly composed of two end rollers and a closed conveyor belt tightly sleeved thereon. The roller that drives the conveyor belt to rotate is called the drive roller (transmission roller); the other roller that only changes the direction of movement of the conveyor belt is called the redirection roller. The drive roller is driven by an electric motor through a reducer, and the conveyor belt is dragged by the friction between the drive roller and the conveyor belt. The drive roller is generally installed at the discharge end to increase traction and facilitate dragging. The material is fed from the feeding end, falls on the rotating conveyor belt, and is driven by the friction of the conveyor belt to be transported to the discharge end for discharge.

[0059] The roller in the belt conveyor is the main component for transmitting power. It is mainly driven by the motor to drive the conveyor belt through the friction between the belts. The main forms of the belt conveyor not being able to operate normally due to defects in the design and processing of the roller include: belt deviation, belt surface slippage, vibration, noise and other problems. Among them, the vibration is caused by the vibration of the rollers and roller groups on the belt conveyor when it is running, which leads to fatigue damage of the structure, loosening of the equipment, failure, noise, etc., which will have a huge impact on the smooth operation, running resistance and safety of the whole machine.

[0060] S111, installing a temperature vibration probe in the target area; the temperature vibration probe includes a temperature vibration composite sensor.

[0061] In this embodiment, several nodes are set and installed at the positions on both sides of the roller of the belt conveyor in the target area, and a corresponding temperature vibration probe is configured on each node to realize the fault diagnosis function of abnormal vibration of the roller (bearing) of the belt conveyor. The temperature vibration probe here adopts a temperature vibration composite sensor. The node includes the location information of the temperature vibration probe installation, etc.

[0062] Specifically, several temperature-vibration composite sensors are installed upward at the vertical diameter position of the roller bearings on both sides of the belt conveyor roller in the target area. For example, two temperature-vibration composite sensors are installed at the monitoring position of the driving side roller (i.e., the transmission roller) of the belt conveyor; at the same time, 10 temperature-vibration composite sensors are installed at the monitoring position of the non-driving side roller (i.e., the redirecting roller) of the belt conveyor.

[0063] S112, collecting vibration temperature signals of each setting position in real time based on the temperature vibration probe.

[0064] In this embodiment, the vibration temperature signal data of the roller bearing during operation is collected in real time by the temperature-vibration composite sensors arranged at each monitoring position. The vibration temperature signal data includes: temperature characteristic data and vibration characteristic data. Among them, the temperature characteristic data includes: temperature, surface temperature, transmitter circuit operating temperature, etc.; the vibration characteristic data includes: vibration frequency, phase, vibration displacement, vibration measurement direction, vibration displacement, vibration speed, acceleration, etc.

[0065] S12, performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal. Figure 3A and Figure 3B , respectively showing a flow chart of S12 in the fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM of the present invention and a flow chart of wavelet transform in the fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM of the present invention. Figure 3A and Figure 3B As shown, the S12 comprises the following steps:

[0066] Since the noise signal of the roller bearing of the belt conveyor contains a lot of background noise, the characteristic information may be submerged by the noise, resulting in inaccurate fault diagnosis results. Therefore, it is necessary to perform noise reduction on the roller bearing vibration signal.

[0067] S121, performing wavelet decomposition based on the vibration temperature signal to obtain vibration temperature signal wavelet coefficients.

[0068] In this embodiment, the collected vibration temperature signal containing noise is decomposed by wavelet transform to obtain wavelet coefficients of several layers.

[0069] Specifically, the real-time vibration temperature signal data collected in the above step is filtered and denoised to improve signal quality.

[0070] Select the appropriate wavelet basis function and use the wavelet transform algorithm to decompose the real-time vibration temperature signal. Different wavelet bases have different local performances in the time domain and frequency domain, which makes the wavelet transform have different abilities to characterize the local characteristics of the signal in the time domain and frequency domain, so it is particularly important to select the appropriate wavelet basis.

[0071] Then select an appropriate decomposition level, perform multi-layer wavelet decomposition on the preprocessed vibration temperature signal, and obtain wavelet coefficients of different scales in several layers. In this stage, the order of the wavelet basis needs to be determined. Generally speaking, the higher the order, the stronger the ability to represent the local characteristics of the signal, but the amount of calculation will also increase accordingly. When the order is higher than 5, increasing the order of the wavelet basis has little effect on improving the ability of the wavelet basis to represent the local characteristics of the speech signal.

[0072] It should be noted that the wavelet basis function can be implemented using wavelet basis functions of types such as Haar, Daubechies wavelet basis function, Symlets wavelet basis function, etc.

[0073] S122, denoising the wavelet coefficients of the vibration temperature signal based on threshold function processing.

[0074] In this embodiment, the wavelet coefficients of the vibration temperature signal at the threshold function are used to remove the components related to the noise.

[0075] Specifically, according to the wavelet threshold denoising principle, a suitable threshold function is selected to perform threshold processing on the wavelet coefficients. By comparing the wavelet coefficients processed by the threshold function with the original wavelet coefficients, the components related to noise are found, such as the wavelet coefficients and other data that are less than the threshold; and the components related to noise are removed. That is, when the wavelet coefficients processed by the threshold function are greater than the original wavelet coefficients, they are removed; when the wavelet coefficients processed by the threshold function are less than the original wavelet coefficients, the wavelet coefficients are retained.

[0076] The calculation formula of the threshold function is:

[0077]

[0078] Among them, σ j Expressed as standard deviation; w j It is represented as wavelet coefficient; λ is the threshold value calculated according to the threshold rule; median() is the median function. The threshold function calculation formula can satisfy the requirement of taking a higher threshold value at a small scale to avoid constant deviation of the global threshold.

[0079] It should be noted that the threshold function may be processed using, for example, a soft threshold function, a hard threshold function, etc.

[0080] S123, reconstructing the denoised vibration temperature signal based on inverse wavelet transform to obtain a denoised vibration temperature signal.

[0081] In this embodiment, the wavelet coefficients obtained in the above step are analyzed and visualized, and then the wavelet reconstruction technology is used to reconstruct the wavelet coefficients into a denoised vibration temperature signal for subsequent processing.

[0082] Specifically, the obtained wavelet coefficients are counted and analyzed to extract useful feature information, such as: analyzing the detail coefficients, the feature information of the signal at different scales can be obtained; analyzing the approximate coefficients, the feature information of the low-frequency part of the vibration temperature signal can be obtained. Then, the wavelet coefficients can be visualized using tools (such as MATLAB). Then, the wavelet inverse transform is used for wavelet reconstruction, that is, the denoised wavelet coefficients are inversely transformed to obtain the low-frequency part and high-frequency part of the denoised vibration temperature signal; then the low-frequency part and the high-frequency part are superimposed to obtain the denoised vibration temperature signal; then, the denoised vibration temperature signal is post-processed, such as smoothing, normalization, etc., to obtain a more accurate result; then, the reconstructed vibration temperature signal is feature extracted, such as time domain analysis, frequency domain analysis, etc., to finally obtain an accurate denoised vibration temperature signal.

[0083] It should be noted that the accuracy of the inverse wavelet transform is closely related to the quality of the reconstructed signal, so it is necessary to select appropriate wavelet basis functions and decomposition levels to ensure the accuracy of the reconstructed signal. At the same time, in order to obtain better reconstruction results, the signal needs to be properly pre-processed and post-processed.

[0084] S13, extracting features based on the denoised vibration and temperature signal to obtain a denoised vibration and temperature signal fault feature vector. Figure 4A , which is a flow chart of S13 in the fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM of the present invention. Figure 4A As shown, the S13 comprises the following steps:

[0085] S131, constructing a denoised vibration-temperature signal probability density model based on the denoised vibration-temperature signal.

[0086] Since the fault signal of the belt conveyor has the characteristics of strong impulse and poor symmetry, the probability density function of the signal can be modeled using the Gaussian mixture distribution model, which can obtain the characteristic quantity within a sampling period and serve as the input of abnormality analysis.

[0087] In this embodiment, modeling is performed based on the accurate denoised vibration temperature signal obtained in the above steps, and the Gaussian mixture distribution model is used for the modeling.

[0088] The Gaussian Mixture Model (GMM) is a probability density function model used to describe the probability distribution of data. A Gaussian mixture model is a probability model composed of a mixture of multiple Gaussian distributions and is used to describe the probability density function of data. This process usually involves selecting appropriate model parameters (e.g., the number of Gaussian distributions, and the mean and covariance of each Gaussian distribution) and estimating the parameters using maximum likelihood estimation or other optimization techniques.

[0089] The Gaussian mixture model uses the Gaussian probability density function (normal distribution curve) to accurately quantify things and decomposes one thing into several models based on the Gaussian probability density function (normal distribution curve).

[0090] Specifically, according to the obtained accurate denoised vibration temperature signal, based on the characteristics of the signal related to the probability distribution, such as mean, variance, covariance and other data. According to the extracted data features, the parameters of the Gaussian mixture model are determined, such as the number of Gaussian distributions, the mean, variance and weight of each Gaussian distribution; these parameters can be determined by maximum likelihood estimation or iterative optimization algorithm. These determined parameters are input into the Gaussian mixture model, and a probability density function is generated according to the input features to obtain a denoised vibration temperature signal probability density model. Then, the trained denoised vibration temperature signal probability density model can be evaluated by cross-validation, holdout validation or Bayesian information criterion to judge the fitting results of the model and the prediction ability of the new data; if necessary, the model parameters can be further adjusted or the number of Gaussian distributions can be increased or decreased to retrain the model to achieve the purpose of optimizing the model.

[0091] S132, performing feature extraction on the denoised vibration-temperature signal probability density model to obtain a denoised vibration-temperature signal feature vector within a sampling period.

[0092] In this embodiment, a feature vector is extracted from the probability density model of the denoised vibration temperature signal. The feature vector includes: various parameters of the model, such as: mean, variance, weight, skewness, vibration amplitude, phase, frequency domain, time domain, time-frequency domain and other feature data.

[0093] S133: construct a signal histogram based on the denoised vibration temperature signal feature vector.

[0094] In this embodiment, a signal histogram is constructed based on the denoised vibration temperature signal feature vector provided in the above steps. Then, by observing the shape and distribution of the histogram, a deeper understanding of the operating status of the belt conveyor can be obtained. For example, if the histogram of the vibration frequency shows that it is mainly concentrated on a certain frequency, it may indicate that the equipment is vibrating at an abnormal frequency.

[0095] It should be noted that when constructing a histogram, it is necessary to determine the appropriate binning range. For vibration signals, binning can be done by frequency or amplitude; for temperature signals, binning can be done by temperature range. The number of bins and width can be determined based on actual needs and signal characteristics.

[0096] S134, performing probability density fitting and parameter estimation based on the signal histogram to obtain a denoised vibration temperature signal fault feature vector. Figure 4B , which is a schematic diagram of signal probability density fitting in the fault diagnosis method of abnormal vibration of roller bearing based on improved SVM of the present invention.

[0097] In this embodiment, a histogram is created for the vibration signal of each sample after preprocessing. A suitable probability density function is selected to fit the histogram data. The probability density function is used to estimate the parameters of the function by maximum likelihood estimation or least squares method. One or more eigenvectors are obtained by calculating the parameters of the probability density function. These eigenvectors can reflect whether the vibration signal is an abnormal fault, so that the denoised vibration temperature signal fault eigenvector can be obtained. The samples here include: standard samples and test samples. Among them, the standard sample is a theoretical value, and the test sample is a historical data set of belt conveyor roller bearing fault signals.

[0098] When a new vibration temperature signal is collected and a new histogram is generated, the possible condition or behavior of the equipment can be predicted by looking up the trained model.

[0099] The Gaussian mixture distribution formula is:

[0100]

[0101] Where k represents the number of components; w k Expressed as a mixing weight, 0≤w k ≤1; g(xμ k ,σ k ) is the distribution of the kth Gaussian component.

[0102] Finally, the maximum expectation algorithm is used to estimate the fault feature vector of the denoised vibration and temperature signal, and the signal characteristics of the fault feature vector of the denoised vibration and temperature signal are further simplified.

[0103] S14, inputting the denoised vibration temperature signal fault feature vector into the particle swarm support vector to distinguish different types of bearing faults. Figure 5A , which is a flow chart of S14 in the fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM of the present invention. Figure 5A As shown, the S14 comprises the following steps:

[0104] The fault features extracted in the data preprocessing stage are input as feature vectors into the support vector machine optimized by particle swarm optimization, which can be used to distinguish different types of bearing faults.

[0105] S141, construct a multi-fault classification model based on the denoised vibration temperature signal fault feature vector. The multi-fault classification model includes a directed acyclic classifier composed of several binary classification SVMs. Figure 5B , which is a schematic diagram of a directed acyclic classifier in the fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM of the present invention.

[0106] SVM (Support Vector Machine) is a generalized linear classifier that performs binary classification of data in a supervised learning manner. The core idea of ​​SVM classification is the optimal classification hyperplane. When using SVM to classify a given sample, first find a hyperplane that meets the classification requirements, try to keep the coordinate points in the sample away from the plane, and ensure that the areas on both sides of the classification hyperplane are large enough.

[0107] In this embodiment, the features of each denoised vibration and temperature signal fault sample are combined into a feature vector. The denoised vibration and temperature signal fault feature vectors extracted in the above step are respectively input into multiple SVM classifiers, and each SVM classifier is trained using known denoised vibration and temperature signal fault samples. The outputs of multiple SVM classifiers are then selected and combined to form a more complex directed acyclic classifier to obtain a multi-fault classification model, so as to implement the classification process through different strategies. Next, the multi-fault classification model is evaluated using test samples, and indicators such as classification accuracy and recall rate are calculated; then the model is optimized based on the evaluation results, such as: adjusting SVM parameters, adding features, optimizing integrated learning strategies, etc. Then, the results of the multi-fault classification are analyzed, and the reasons for the classification results are explained based on the output of the multi-fault classification model and the actual fault conditions.

[0108] Specifically, a directed acyclic classifier is formed based on multiple SVM classifiers to construct a multi-fault classification model, wherein the directed acyclic classifier can be obtained through the fault mode relationship between the components in the fault map.

[0109] For example: SVM14 represents four types of faults from Class 1 to Class 4, which are considered as abnormal vibration faults of the reducer as a whole. First, input the denoised vibration temperature signal fault feature vector into the SVM14 classifier to determine whether the denoised vibration temperature signal fault feature vector is not Class 1 or Class 4; if it is not Class 1, go left and ask the SVM24 classifier; however, if the denoised vibration temperature signal fault feature vector is not Class 4, go right to the SVM13 classifier. Then, if the SVM13 classifier determines that the denoised vibration temperature signal fault feature vector is not Class 1, then go left to SVM23. At this time, if the denoised vibration temperature signal fault feature vector is determined to be not Class 2, then continue to go left until Class 3. At this time, the denoised vibration temperature signal fault feature vector can be determined to be Class 3 fault.

[0110] S142: Determine a penalty factor and a kernel function parameter based on the multi-fault classification model.

[0111] In this embodiment, a Gaussian radial kernel function is selected as the kernel function of the multi-classifier in the multi-fault classification model, and the penalty factor and kernel function parameters are determined. Maximizing the interval between classification hyperplanes is essentially an improvement and optimization of the generalization ability of SVM. By maximizing the interval between classification hyperplanes, the structural risk of SVM can be minimized, while the generalization ability of the algorithm is enhanced, meeting the core idea of ​​SVM.

[0112] Specifically, the Gaussian radial kernel function is used as the kernel function. The penalty factor and kernel function parameters are important parameters of the SVM classifier. First, the multi-fault classification model is initialized; the optimal penalty factor and kernel function parameters are found through grid search, the parameter grid is defined, and the performance of each parameter combination is evaluated using cross validation; then, according to the results of the grid search, the parameters of the multi-fault classification model are adjusted; the multi-fault classification model is trained, and the model performance is evaluated and optimized.

[0113] S143, optimizes SVM based on particle swarm optimization algorithm to distinguish different types of bearing faults. Figure 5C , which is a schematic diagram of the SVM process optimized by the PSO algorithm in the fault diagnosis method of abnormal vibration of the roller bearing based on the improved SVM of the present invention.

[0114] In this embodiment, PSO (Particle Swarm Optimization) is used to optimize SVM. That is, a group of particles is defined in space, and the particle state can be described by initial velocity, position and fitness value. The particle can remember the best position found in the design space, and its fitness value is calculated according to the fitness function. The direction and distance of the particle flight are determined based on the particle speed, and the particle can fly and search in the optimal solution space. In the process of continuous iterative optimization, the particle updates its velocity and position by comparing fitness and individual and global extreme values.

[0115] Specifically, a number of particles (i.e., fault vector feature points) are defined in the sample space; the population of the particles is initialized and the parameters are set; the fitness value of the particles is calculated according to the fitness function; the single optimal and global optimal particles are calculated, and the particle position and particle velocity are updated; the direction and distance of the particle flight are determined based on the particle velocity, and whether the preset conditions are met; when the preset conditions are met, the optimal parameters of the SVM are output; the PSO-SVM model is constructed based on the optimal parameters of the SVM to complete the optimization process of the SVM.

[0116] For example: historical vibration temperature signals and real-time vibration temperature signals are used as training data sets and test data sets.

[0117] First, define the size of the particle swarm (i.e., the number of particles), and initialize the initial position and initial velocity of each particle, as well as the individual best position and global best position of each particle. Calculate the fitness value of each particle according to the fitness function, such as the minimum value of the function. If the fitness value of a particle's current position is better than the fitness value of its individual best position, update the individual best position to the current position; further, if the fitness value of the current particle is better than the fitness value of the global best position, update the global best position to the position of the current particle.

[0118] Then, the speed and position of each particle are updated according to the speed update formula and the position update formula. The speed update formula involved is as follows:

[0119] v[i]=wv[i]+c1rand()(pBest[i]-pos[i])+c2rand()*(gBest-pos[i])

[0120] Similarly, the position update formula is as follows:

[0121] pos[i]=pos[i]+v[i]

[0122] Where v[i] is the velocity of particle i, pos[i] is the position of particle i, w is the inertia weight, c1 and c2 are acceleration constants, rand() is a random function, pBest[i] is the individual best position of particle i, and gBest is the global best position.

[0123] Next, determine whether the termination condition is met, that is, whether the preset maximum number of iterations is reached. When the termination condition is met, the iteration is stopped, and the global optimal position and the fitness value corresponding to the global optimal position are output; otherwise, the particle fitness is returned to be calculated to enter the next round of optimization process. In the process of continuous iterative optimization, the particle updates its speed and position by comparing the fitness and individual and global extreme values. That is, this process is trained by combining the standard samples of the public data set with the existing data samples. Since the existing data cannot be fitted with the standard samples, the alarm probability and the solution are set, but the judgment here is not completely accurate. Therefore, it is necessary to judge whether the fault feature vector is pushed and the corresponding solution when it appears again. If it is a pushed fault, the weight of the corresponding fault is increased. If it is not a pushed fault, it depends on whether it is an existing fault. If it is an existing fault, the weight of the fault needs to be increased; if it is not an existing fault, the fault content needs to be added.

[0124] After the termination condition is met, the updated particle position and velocity are output to the multi-fault classification model and optimized to output the global optimal position of the particle (i.e., the optimal SVM parameter) and the corresponding fitness value. Finally, the global optimal position of the particle and the corresponding fitness value are input to the multi-fault classification model to complete the optimization process of the multi-fault classification model.

[0125] For example, take the fault structure tree constructed by the abnormal vibration of the belt conveyor reducer as an example. Set the abnormal vibration of the reducer to SVM14; SVM24 indicates the abnormal transmission mechanism, and SVM13 indicates the abnormal installation state. The SVM classifier determines that the fault M at this time is not Class 1, which means that the fault belongs to the abnormal transmission mechanism problem; then, enter SVM24 for the next step of judgment. It can be seen from the judgment that the fault M is not Class 4, so it is judged that the fault does not belong to the abnormal state of the low-speed bearing. It can be seen that the fault M may be an abnormal adjustment shaft or an abnormal intermediate shaft; then judge in sequence according to this method, and get the fault M as an adjustment shaft fracture fault. At this time, it belongs to Class 2 fault. Then, select the Gaussian radial kernel function as the multi-fault classification model to determine the penalty factor and kernel function parameters of the fault. Then optimize the multi-fault classification model according to the particle swarm optimization algorithm to make the fault judgment more accurate.

[0126] It is important to note that PSO and SVM are two widely used optimization techniques and machine learning models, but they are usually used for different purposes. PSO is used to optimize continuous or discrete global optimization problems, while SVM is a supervised learning model for classification and regression. This application uses PSO to optimize the parameters of SVM to find the best hyperparameter combination.

[0127] S15 integrates different types of bearing faults, displays them on the web page, and pushes them to users to guide equipment repair and maintenance. Figure 6 , showing the integrated effect of the fault diagnosis algorithm in the fault diagnosis method of abnormal vibration of the roller bearing based on improved SVM of the present invention on the display platform display diagram.

[0128] In this embodiment, the fault diagnosis results of the abnormal vibration of the roller bearing in the above steps are integrated on the display platform on the web page. The fault occurrence time, fault data, fault structure tree, possible fault causes, and related maintenance suggestions of the abnormal vibration of the roller bearing can be pushed to the user. The user can guide the maintenance work of the belt conveyor roller bearing abnormality according to the received information.

[0129] The fault diagnosis method for abnormal vibration of roller bearings based on improved SVM involved in this application is being developed as a part of the research on the promotion and application of intelligent operation and maintenance systems for belt conveyors, and will soon be installed and applied in the field.

[0130] The fault diagnosis method for abnormal vibration of roller bearings based on improved SVM provided in the present application realizes timely detection and elimination of faults by combining vibration sensor data, avoids serious impact caused by further development of faults, avoids large expenses and operation time incurred by manual inspections, improves the work efficiency of roller (bearing) maintenance of belt conveyors, and reduces equipment maintenance operation costs.

[0131] The fault diagnosis method for abnormal vibration of roller bearings based on improved SVM provided in this application can diagnose 3-5 causes of abnormal vibration of rollers (bearings), and the diagnostic accuracy rate based on historical test data sets reaches 70%. The application of this method improves maintenance work efficiency, shortens maintenance time, and avoids economic losses caused by serious faults.

[0132] The protection scope of the fault diagnosis method for abnormal vibration of roller bearings based on improved SVM described in the embodiment of the present application is not limited to the execution order of the steps listed in this embodiment. All solutions implemented by adding, reducing or replacing steps in the prior art based on the principles of the present application are included in the protection scope of the present application.

[0133] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the following is achieved: Figure 1 The fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM.

[0134] At any possible level of technical detail combination, the present application may be a system, method and / or computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present application.

[0135] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, (but not limited to) an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0136] The computer-readable program described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device through a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives a computer-readable program instruction from the network, and forwards the computer-readable program instruction for storage in the computer-readable storage medium in each computing / processing device. The computer program instruction for performing the operation of the present application can be an assembly instruction, an instruction set architecture (ISA) instruction, a machine instruction, a machine-related instruction, a microcode, a firmware instruction, a state setting data, an integrated circuit configuration data, or a source code or object code written in any combination of one or more programming languages, wherein the programming language includes an object-oriented programming language, such as Smalltalk, C++, etc., and a procedural programming language, such as "C" language or similar programming language. Computer readable program instructions can be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In certain embodiments, by using the state information of computer readable program instructions to personalize electronic circuits, such as programmable logic circuits, field programmable gate arrays (FPGAs) or programmable logic arrays (PLAs), the electronic circuits can execute computer readable program instructions, thereby realizing various aspects of the present application.

[0137] The embodiment of the present application also provides a fault diagnosis system for abnormal vibration of a roller bearing based on an improved SVM. The fault diagnosis system for abnormal vibration of a roller bearing based on an improved SVM can implement the fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM described in the present application. However, the implementation device of the fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM described in the present application includes but is not limited to the structure of the fault diagnosis system for abnormal vibration of a roller bearing based on an improved SVM listed in the present embodiment. All structural deformations and replacements of the prior art made according to the principles of the present application are included in the protection scope of the present application.

[0138] The fault diagnosis system for abnormal vibration of a roller bearing based on improved SVM provided in this embodiment will be described in detail below with reference to the drawings.

[0139] This embodiment provides a fault diagnosis system for abnormal vibration of a drum bearing based on an improved SVM, comprising:

[0140] See also Figure 7 , which is a schematic diagram of the principle structure of a fault diagnosis system for abnormal vibration of a roller bearing based on improved SVM in one embodiment of the present invention. Figure 7 As shown, the fault diagnosis system for abnormal vibration of a roller bearing based on improved SVM includes: an acquisition module 71 , a data processing module 72 , a feature extraction module 73 , a fault diagnosis module 74 and a display module 75 .

[0141] The acquisition module 71 is used to acquire the vibration temperature signal in the target area.

[0142] A temperature vibration probe is installed in the target area; the temperature vibration probe includes a temperature vibration composite sensor.

[0143] In this embodiment, several nodes are set and installed at the positions on both sides of the roller of the belt conveyor in the target area, and a corresponding temperature vibration probe is configured on each node to realize the fault diagnosis function of abnormal vibration of the roller (bearing) of the belt conveyor. The temperature vibration probe here adopts a temperature vibration composite sensor. The node includes the location information of the temperature vibration probe installation, etc.

[0144] Specifically, a plurality of temperature-vibration composite sensors are respectively installed upward at the vertical diameter positions of the roller bearings on both sides of the roller of the belt conveyor in the target area.

[0145] The vibration temperature signals of each setting position are collected in real time based on the temperature vibration probe.

[0146] In this embodiment, the vibration temperature signal data of the roller bearing during operation is collected in real time by the temperature-vibration composite sensors arranged at each monitoring position. The vibration temperature signal data includes: temperature characteristic data and vibration characteristic data. Among them, the temperature characteristic data includes: temperature, surface temperature, transmitter circuit operating temperature, etc.; the vibration characteristic data includes: vibration frequency, phase, vibration displacement, vibration measurement direction, vibration displacement, vibration speed, acceleration, etc.

[0147] The data processing module 72 is connected to the acquisition module 71 and is used to perform signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal.

[0148] Since the noise signal of the roller bearing of the belt conveyor contains a lot of background noise, the characteristic information may be submerged by the noise, resulting in inaccurate fault diagnosis results. Therefore, it is necessary to perform noise reduction on the roller bearing vibration signal.

[0149] The vibration temperature signal is decomposed by wavelet to obtain the vibration temperature signal wavelet coefficients. The vibration temperature signal wavelet coefficients are denoised based on threshold function processing. The denoised vibration temperature signal is reconstructed based on inverse wavelet transform to obtain the denoised vibration temperature signal.

[0150] In this embodiment, the collected vibration temperature signal containing noise is decomposed by wavelet transform to obtain wavelet coefficients of several layers.

[0151] Specifically, the collected real-time vibration temperature signal data is filtered and denoised to improve signal quality.

[0152] Select the appropriate wavelet basis function and use the wavelet transform algorithm to decompose the real-time vibration temperature signal. Different wavelet bases have different local performances in the time domain and frequency domain, which makes the wavelet transform have different abilities to characterize the local characteristics of the signal in the time domain and frequency domain, so it is particularly important to select the appropriate wavelet basis.

[0153] Then select the appropriate decomposition level, perform multi-layer wavelet decomposition on the pre-processed vibration temperature signal, and obtain wavelet coefficients of different scales in several layers. In this stage, the order of the wavelet basis needs to be determined.

[0154] In this embodiment, the wavelet coefficients of the vibration temperature signal at the threshold function are used to remove the components related to the noise.

[0155] Specifically, according to the wavelet threshold denoising principle, a suitable threshold function is selected to perform threshold processing on the wavelet coefficients. By comparing the wavelet coefficients processed by the threshold function with the original wavelet coefficients, the components related to the noise are found and removed. That is, when the wavelet coefficients processed by the threshold function are greater than the original wavelet coefficients, they are removed; when the wavelet coefficients processed by the threshold function are less than the original wavelet coefficients, the wavelet coefficients are retained.

[0156] In this embodiment, analysis is performed based on wavelet coefficients and visualization is performed on them. Then, wavelet reconstruction technology is used to reconstruct the wavelet coefficients into denoised vibration temperature signals for subsequent processing.

[0157] Specifically, the obtained wavelet coefficients are counted and analyzed to extract useful feature information, such as: by analyzing the detail coefficients, the feature information of the signal at different scales can be obtained; by analyzing the approximate coefficients, the feature information of the low-frequency part of the vibration temperature signal can be obtained. Then, the wavelet coefficients can be visualized using tools. Then, the wavelet inverse transform is used for wavelet reconstruction, that is, the denoised wavelet coefficients are inversely transformed to obtain the low-frequency and high-frequency parts of the denoised vibration temperature signal; then the low-frequency and high-frequency parts are superimposed to obtain the denoised vibration temperature signal; then, the denoised vibration temperature signal is post-processed, such as smoothing, normalization, etc., to obtain more accurate results; then, the reconstructed vibration temperature signal is feature extracted, such as time domain analysis, frequency domain analysis, etc., to finally obtain an accurate denoised vibration temperature signal.

[0158] The feature extraction module 73 is used to perform feature extraction based on the denoised vibration-temperature signal to obtain a denoised vibration-temperature signal fault feature vector.

[0159] A denoised vibration-temperature signal probability density model is constructed based on the denoised vibration-temperature signal.

[0160] Since the fault signal of the belt conveyor has the characteristics of strong impulse and poor symmetry, the probability density function of the signal can be modeled using the Gaussian mixture distribution model, which can obtain the characteristic quantity within a sampling period and serve as the input of abnormality analysis.

[0161] In this embodiment, modeling is performed based on the accurate denoised vibration temperature signal obtained in the above steps, and the Gaussian mixture distribution model is used for the modeling.

[0162] Specifically, according to the obtained accurate denoised vibration temperature signal, based on the characteristics of the signal related to the probability distribution. According to the extracted data features, the parameters of the Gaussian mixture model are determined; these parameters can be determined by maximum likelihood estimation or iterative optimization algorithm. These determined parameters are input into the Gaussian mixture model, and a probability density function is generated according to the input features to obtain a probability density model of the denoised vibration temperature signal. Then, the trained probability density model of the denoised vibration temperature signal can be evaluated by methods such as cross-validation, hold-out validation or Bayesian information criterion to judge the fitting results of the model and its predictive ability for new data; if necessary, the model parameters can be further adjusted or the number of Gaussian distributions can be increased or decreased to retrain the model to achieve the purpose of optimizing the model.

[0163] Feature extraction is performed on the probability density model of the denoised vibration and temperature signal to obtain a feature vector of the denoised vibration and temperature signal within a sampling period.

[0164] In this embodiment, a feature vector is extracted from the probability density model of the denoised vibration temperature signal. The feature vector includes: various parameters of the model, such as: mean, variance, weight, skewness, vibration amplitude, phase, frequency domain, time domain, time-frequency domain and other feature data.

[0165] A signal histogram is constructed based on the denoised vibration temperature signal feature vector.

[0166] In this embodiment, a signal histogram is constructed based on the denoised vibration temperature signal feature vector provided in the above steps, and then a deeper understanding of the operating state of the belt conveyor is obtained by observing the shape and distribution of the histogram.

[0167] Probability density fitting and parameter estimation are performed based on the signal histogram to obtain a denoised vibration temperature signal fault feature vector.

[0168] In this embodiment, a histogram is created for the vibration signal of each sample after preprocessing. A suitable probability density function is selected to fit the histogram data. The probability density function is used to estimate the parameters of the function by maximum likelihood estimation or least squares method. One or more eigenvectors are obtained by calculating the parameters of the probability density function. These eigenvectors can reflect whether the vibration signal is an abnormal fault, so that the denoised vibration temperature signal fault eigenvector can be obtained. The samples here include: standard samples and test samples. Among them, the standard sample is a theoretical value, and the test sample is a historical data set of belt conveyor roller bearing fault signals.

[0169] When a new vibration temperature signal is collected and a new histogram is generated, the possible condition or behavior of the equipment can be predicted by looking up the trained model.

[0170] Finally, the maximum expectation algorithm is used to estimate the fault feature vector of the denoised vibration and temperature signal, and the signal characteristics of the fault feature vector of the denoised vibration and temperature signal are further simplified.

[0171] The fault diagnosis module 74 is used to input the denoised vibration temperature signal fault feature vector into the particle swarm support vector to distinguish different types of bearing faults. The fault features extracted in the data preprocessing step are input as feature vectors into the particle swarm optimized support vector machine to distinguish different types of bearing faults.

[0172] Firstly, a multi-fault classification model is constructed based on the denoised vibration temperature signal fault feature vector. The multi-fault classification model includes a directed acyclic classifier composed of several binary classification SVMs.

[0173] In this embodiment, the features of each denoised vibration and temperature signal fault sample are combined into a feature vector. The extracted denoised vibration and temperature signal fault feature vectors are respectively input into multiple SVM classifiers, and each SVM classifier is trained using known denoised vibration and temperature signal fault samples. The outputs of multiple SVM classifiers are then selected and combined to form a more complex directed acyclic classifier to obtain a multi-fault classification model, so as to implement the classification process through different strategies. Next, the multi-fault classification model is evaluated using test samples, and indicators such as classification accuracy and recall rate are calculated; then the model is optimized based on the evaluation results. Then, the results of the multi-fault classification are analyzed, and the reasons for the classification results are explained based on the output of the multi-fault classification model and the actual fault conditions.

[0174] Specifically, a directed acyclic classifier is formed based on multiple SVM classifiers to construct a multi-fault classification model, wherein the directed acyclic classifier can be obtained through the fault mode relationship between the components in the fault map.

[0175] Then, based on the multi-fault classification model, a penalty factor and a kernel function parameter are determined.

[0176] In this embodiment, a Gaussian radial kernel function is selected as the kernel function of the multi-classifier in the multi-fault classification model, and the penalty factor and kernel function parameters are determined. Maximizing the interval between classification hyperplanes is essentially an improvement and optimization of the generalization ability of SVM. By maximizing the interval between classification hyperplanes, the structural risk of SVM can be minimized, while the generalization ability of the algorithm is enhanced, meeting the core idea of ​​SVM.

[0177] Specifically, the Gaussian radial kernel function is used as the kernel function. The penalty factor and kernel function parameters are important parameters of the SVM classifier. First, the multi-fault classification model is initialized; the optimal penalty factor and kernel function parameters are found through grid search, the parameter grid is defined, and the performance of each parameter combination is evaluated using cross validation; then, according to the results of the grid search, the parameters of the multi-fault classification model are adjusted; the multi-fault classification model is trained, and the model performance is evaluated and optimized.

[0178] Then, the SVM is optimized based on the particle swarm optimization algorithm to distinguish different types of bearing faults.

[0179] In this embodiment, PSO is used to optimize SVM. A number of particles (i.e., fault vector feature points) are defined in the sample space; the population of the particles is initialized and the parameters are set; the fitness value of the particles is calculated according to the fitness function; the single optimal and global optimal particles are calculated, and the particle position and particle speed are updated; the direction and distance of the particle flight are determined based on the particle speed, and whether it meets the preset conditions; when the preset conditions are met, the optimal parameters of SVM are output; the PSO-SVM model is constructed based on the optimal parameters of SVM to complete the optimization process of SVM.

[0180] Specifically, the size of the particle swarm (i.e., the number of particles) is first defined, and the initial position and initial velocity of each particle, as well as the individual best position and global best position of each particle are initialized. The fitness value of each particle is calculated according to the fitness function, such as the minimum value of the function. If the fitness value of a particle's current position is better than the fitness value of its individual best position, the individual best position is updated to the current position; further, if the fitness value of the current particle is better than the fitness value of the global best position, the global best position is updated to the position of the current particle.

[0181] Then, the speed and position of each particle are updated according to the speed update formula and the position update formula.

[0182] Next, determine whether the termination condition is met, that is, whether the preset maximum number of iterations is reached. When the termination condition is met, the iteration is stopped, and the global optimal position and the fitness value corresponding to the global optimal position are output; otherwise, the particle fitness is returned to be calculated to enter the next round of optimization process. In the process of continuous iterative optimization, the particle updates its speed and position by comparing the fitness and individual and global extreme values. That is, this process is trained by combining the standard samples of the public data set with the existing data samples. Since the existing data cannot be fitted with the standard samples, the alarm probability and the solution are set, but the judgment here is not completely accurate. Therefore, it is necessary to judge whether the fault feature vector is pushed and the corresponding solution when it appears again. If it is a pushed fault, the weight of the corresponding fault is increased. If it is not a pushed fault, it depends on whether it is an existing fault. If it is an existing fault, the weight of the fault needs to be increased; if it is not an existing fault, the fault content needs to be added.

[0183] After the termination condition is met, the updated particle position and velocity are output to the multi-fault classification model and optimized to output the global optimal position of the particle (i.e., the optimal SVM parameter) and the corresponding fitness value. Finally, the global optimal position of the particle and the corresponding fitness value are input to the multi-fault classification model to complete the optimization process of the multi-fault classification model.

[0184] The display module 75 is used to integrate different types of bearing faults, display them through a web page, and push them to users to guide equipment repair and maintenance.

[0185] In this embodiment, the fault diagnosis results of the abnormal vibration of the roller bearing in the above steps are integrated on the display platform on the web page. The fault occurrence time, fault data, fault structure tree, possible fault causes, and related maintenance suggestions of the abnormal vibration of the roller bearing can be pushed to the user. The user can guide the maintenance work of the belt conveyor roller bearing abnormality according to the received information.

[0186] The fault diagnosis model of abnormal vibration of roller bearing based on improved SVM is built. The fault diagnosis system of abnormal vibration of roller bearing based on improved SVM can diagnose the causes of abnormal vibration of 3-5 rollers (bearings), and the diagnosis accuracy rate based on historical test data sets is 70%. At the same time, it improves the efficiency of maintenance work, shortens the maintenance time, avoids the economic losses caused by serious faults, realizes timely detection and elimination of faults, avoids the serious impact caused by further development of faults, avoids the large expenses and operation time caused by manual inspections, and reduces the cost of equipment maintenance operations.

[0187] It should be noted that it should be understood that the division of the various modules of the above system is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the x module can be a separately established processing element, or it can be integrated in a certain chip of the above system for implementation. In addition, it can also be stored in the memory of the above system in the form of program code, and called and executed by a certain processing element of the above system. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0188] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital singnal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0189] See also Figure 8 , which is a schematic diagram showing the principle structure of a fault diagnosis device for abnormal vibration of a roller bearing based on an improved SVM in one embodiment of the present invention. Figure 8 As shown, this embodiment provides a fault diagnosis device for abnormal vibration of a roller bearing based on an improved SVM, and the fault diagnosis device for abnormal vibration of a roller bearing based on an improved SVM comprises: a processor 81 and a memory 82; the memory 82 is used to store computer programs; the processor 81 is connected to the memory 82, and is used to execute the computer program stored in the memory 82, so that the fault diagnosis device for abnormal vibration of a roller bearing based on the improved SVM performs the various steps of the fault diagnosis method for abnormal vibration of a roller bearing based on the improved SVM as described above.

[0190] Preferably, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0191] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0192] In summary, the fault diagnosis method and system for abnormal vibration of a drum bearing based on improved SVM provided in the present application have the following beneficial effects:

[0193] The present application provides a fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM, which can diagnose the causes of 3-5 types of abnormal vibration faults of the roller (bearing), and the diagnostic accuracy rate based on the historical test data set is 70%. The application of this method improves the efficiency of maintenance work, shortens the maintenance time, and avoids the economic losses caused by serious faults. At the same time, the fault diagnosis method for abnormal vibration of a roller bearing based on an improved SVM provided by the present application can realize the timely discovery and elimination of faults by combining vibration sensor data, avoid the serious impact caused by further development of the fault, avoid the large expenses and operation time generated by manual inspection, and reduce the cost of equipment maintenance operations.

[0194] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM, characterized in that: The following steps are involved: Acquire vibration temperature signals in the target area; Performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal; Perform feature extraction based on the denoised vibration and temperature signal to obtain a denoised vibration and temperature signal fault feature vector; Inputting the denoised vibration temperature signal fault feature vector into a particle swarm support vector to distinguish different types of bearing faults; Different types of bearing failures are integrated, displayed through the web page, and pushed to users to guide equipment repair and maintenance.

2. The fault diagnosis method for abnormal vibration of a roller bearing based on improved SVM according to claim 1 is characterized in that: Acquiring the vibration temperature signal in the target area includes the following steps: Installing a temperature vibration probe in the target area; the temperature vibration probe includes a temperature vibration composite sensor; The vibration temperature signals of each setting position are collected in real time based on the temperature vibration probe.

3. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 1 is characterized in that: Performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal comprises the following steps: Perform wavelet decomposition based on the vibration temperature signal to obtain the vibration temperature signal wavelet coefficient; De-noising the wavelet coefficients of the vibration temperature signal based on threshold function processing; The denoised vibration temperature signal is reconstructed based on inverse wavelet transform to obtain a denoised vibration temperature signal.

4. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 3 is characterized in that: The calculation formula of the threshold function is: Among them, σ j Expressed as standard deviation; w j It is represented by the wavelet coefficient; λ is represented by the threshold value calculated according to the threshold rule; median() is represented by the median function.

5. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 1 is characterized in that: Performing feature extraction based on the denoised vibration and temperature signal to obtain a denoised vibration and temperature signal fault feature vector comprises the following steps: Constructing a denoised vibration temperature signal probability density model based on the denoised vibration temperature signal; Performing feature extraction on the probability density model of the denoised vibration and temperature signal to obtain a feature vector of the denoised vibration and temperature signal within a sampling period; Constructing a signal histogram based on the denoised vibration temperature signal feature vector; Probability density fitting and parameter estimation are performed based on the signal histogram to obtain a denoised vibration temperature signal fault feature vector.

6. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 5 is characterized in that: The probability density fitting curve formula is: The Gaussian mixture distribution expression is: g(xμ k ,σ k ) Where k represents the number of components; w k Expressed as a mixing weight, 0≤w k ≤1; g(xμ k ,σ k ) is the distribution of the kth Gaussian component.

7. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 1 is characterized in that: Inputting the denoised vibration temperature signal fault feature vector into the particle swarm support vector to distinguish different types of bearing faults includes the following steps: Constructing a multi-fault classification model based on the denoised vibration temperature signal fault feature vector; the multi-fault classification model includes a directed acyclic classifier composed of several binary classification SVMs; Based on the multi-fault classification model, determining a penalty factor and a kernel function parameter; The SVM is optimized based on the particle swarm optimization algorithm to distinguish different types of bearing faults.

8. The fault diagnosis method for abnormal vibration of a drum bearing based on improved SVM according to claim 7 is characterized in that: Optimizing SVM based on particle swarm optimization algorithm to distinguish different types of bearing faults includes the following steps: Define a number of particles in space; Initializing the population and setting parameters of the particles; Calculating the fitness value of the particle according to the fitness function; Calculate the single optimal and global optimal particles, and update the particle positions and particle velocities; Determine the direction and distance of particle flight based on the particle velocity, and determine whether it meets preset conditions; When the preset conditions are met, the optimal parameters of SVM are output; A PSO-SVM model is constructed based on the optimal parameters of the SVM to complete the optimization process of the SVM.

9. A fault diagnosis system for abnormal vibration of a drum bearing based on improved SVM, characterized in that: include: An acquisition module, used for acquiring vibration temperature signals in a target area; A data processing module, used for performing signal preprocessing on the vibration temperature signal to obtain a denoised vibration temperature signal; A feature extraction module, used for performing feature extraction based on the denoised vibration and temperature signal to obtain a denoised vibration and temperature signal fault feature vector; A fault diagnosis module, used for inputting the denoised vibration temperature signal fault feature vector into a particle swarm support vector to distinguish different types of bearing faults; The display module is used to integrate different types of bearing faults, display them through the web page, and push them to users to guide equipment repair and maintenance.

10. A fault diagnosis device for abnormal vibration of a drum bearing based on improved SVM, characterized in that: include: Processor and memory; The memory is used to store computer programs; The processor is connected to the memory and is used to execute the computer program stored in the memory so that the fault diagnosis device for abnormal vibration of the roller bearing based on the improved SVM executes the fault diagnosis method for abnormal vibration of the roller bearing based on the improved SVM as described in any one of claims 1 to 8.

Images