A method and apparatus for diagnosing faults in rolling bearings

Abstract

This invention provides a method and apparatus for fault diagnosis of rolling bearings. The method involves acquiring the vibration acceleration signal of the rolling bearing and then obtaining its time-frequency distribution, time-domain waveform recursion graph, and spectral recursion graph. A combined tensor is then formed based on these graphs to construct a fault diagnosis model for the rolling bearing. Finally, test sample signals are input into the trained fault diagnosis model to obtain the fault diagnosis result. This invention combines the recursion graph analysis method with time-domain and frequency-domain signals, improving the diagnostic accuracy and generalization performance of the fault diagnosis model. Furthermore, through training and testing the fault diagnosis model, the diagnostic accuracy of the fault model under complex environments can be further improved.

Description

Technical Field

[0001] This invention relates to the field of bearing fault detection technology, and specifically to a method and apparatus for diagnosing faults in rolling bearings. Background Technology

[0002] In modern industry, rotating machinery is becoming increasingly complex and sophisticated, making potential failures difficult to monitor and predict. Therefore, monitoring and diagnosing the condition of rolling bearings holds significant strategic importance for both national defense and modern industry. Currently, deep learning methods are flourishing in various fields such as image recognition and speech processing, achieving remarkable results in numerous applications, including mechanical equipment condition monitoring.

[0003] Vibration signals from most equipment are non-stationary and nonlinear. Increasingly, nonlinear information processing techniques are being used in fault diagnosis. Among these, recursive graph analysis algorithms can extract dynamic information about system operation from the time series of signals, directly describing the mainstream information on key signals, and are suitable for expressing vibration signal characteristics. However, current analysis of recursive graphs mainly relies on recursive quantitative analysis, which is largely dependent on human experience. Although some works use recursive graphs as input and employ graphical image analysis for fault analysis, the methods are relatively simplistic and the signal analysis is not comprehensive enough.

[0004] Therefore, how to integrate multi-type information of rolling bearings and extract the fault characteristics of rolling bearings to achieve accurate rolling bearing fault diagnosis has become an urgent problem for technicians in the field. Summary of the Invention

[0005] In view of this, it is necessary to provide a fault diagnosis method and device for rolling bearings to solve the problems of single fault signals and low accuracy of diagnosis results in existing rolling bearings.

[0006] To address the aforementioned technical problems, this invention provides a method for diagnosing faults in rolling bearings, comprising:

[0007] The vibration acceleration signal of the rolling bearing is acquired, and the vibration acceleration signal is divided into training sample signal and test sample signal according to a preset rule;

[0008] Calculate the time-frequency distribution of the vibration acceleration signal;

[0009] The time-domain waveform of the vibration acceleration signal is reconstructed in phase space to obtain a recursive graph of the time-domain waveform of the vibration acceleration signal.

[0010] The time-domain waveform of the vibration acceleration signal is converted into a spectrum, and the spectrum is reconstructed in phase space to obtain the spectrum recursion diagram of the vibration acceleration signal.

[0011] A fault diagnosis model for rolling bearings is constructed by forming a combined tensor based on time-frequency distribution, time-domain waveform recursion graph, and spectrum recursion graph.

[0012] The fault diagnosis model is trained based on the training sample signals, and the test sample signals are input into the trained fault diagnosis model to obtain a fully trained fault diagnosis model. The fault diagnosis result of the rolling bearing is obtained based on the fully trained fault diagnosis model.

[0013] In one possible implementation, acquiring the vibration acceleration signal of the rolling bearing includes:

[0014] The vibration acceleration signals of the rolling bearing under different operating conditions are obtained based on the acceleration sensor, and the vibration acceleration signals are labeled according to different operating conditions;

[0015] The labeled vibration acceleration signals are divided into training sample signals and test sample signals according to a preset ratio;

[0016] The operating states of rolling bearings include normal operation, inner ring failure, outer ring failure, and rolling element failure.

[0017] In one possible implementation, after acquiring the vibration acceleration signal of the rolling bearing, the process further includes:

[0018] Gaussian white noise in the vibration acceleration signal is filtered out using wavelet threshold denoising method to obtain the denoised vibration acceleration signal.

[0019] In one possible implementation, calculating the time-frequency distribution of the vibration acceleration signal includes:

[0020] Wavelet transform of vibration acceleration signal based on wavelet transform function;

[0021] The wavelet transform of the vibration acceleration signal is converted into a two-dimensional time-frequency matrix using bilinear interpolation, thus obtaining the time-frequency distribution of the vibration acceleration signal.

[0022] In one possible implementation, the time-domain waveform of the vibration acceleration signal is reconstructed in phase space to obtain a recursive time-domain waveform of the vibration acceleration signal, including:

[0023] Reconstruction of pseudo-phase space based on vibration acceleration signal of rolling bearing and preset parameters;

[0024] Calculate the distances between target points on the pseudo-phase space trajectory;

[0025] Based on the target rules, a square graph based on a two-dimensional dot matrix is ​​constructed to obtain the time-domain waveform recursive graph of the vibration acceleration signal.

[0026] In one possible implementation, the preset parameters include the target dimension and a delay constant. The reconstruction of the pseudo-phase space based on the vibration acceleration signal of the rolling bearing and the preset parameters includes:

[0027] Based on the discrete-time sequence of the vibration acceleration signal, the target dimension, and the delay constant, a pseudo-space reconstruction of the vibration acceleration signal is performed to obtain the state variables of the vibration acceleration signal.

[0028] The expression for the discrete-time sequence of the vibration acceleration signal is: ;

[0029] in, Represents vibration acceleration signals at different times;

[0030] The state variable expression for the vibration acceleration signal is:

[0031] ;

[0032] in, This can be expressed as a state variable expression, where N represents the dimension and m represents the target dimension. It is represented as a delay constant.

[0033] In one possible implementation, the formula for calculating the distance between target points on the pseudo-phase space trajectory is:

[0034] ;

[0035] in, Represents point i on the pseudo-directional spatial trajectory to J o'clock The distance.

[0036] In one possible implementation, the target rule expression is:

[0037] ;

[0038] Where r represents the preset threshold.

[0039] In one possible implementation, the fault diagnosis model for rolling bearings uses the cross-entropy function as the loss function to optimize the interlayer connectivity weights.

[0040] To address the above problems, the present invention also provides a fault diagnosis device for rolling bearings, comprising:

[0041] The signal acquisition module is used to acquire the vibration acceleration signal of the rolling bearing and divide the vibration acceleration signal into training sample signal and test sample signal according to a preset rule.

[0042] The time-frequency distribution processing module is used to calculate the time-frequency distribution of the vibration acceleration signal;

[0043] The time-domain processing module is used to perform phase space reconstruction processing on the time-domain waveform of the vibration acceleration signal to obtain the recursive graph of the time-domain waveform of the vibration acceleration signal.

[0044] The spectrum processing module is used to convert the time-domain waveform of the vibration acceleration signal into a spectrum, and to perform phase space reconstruction processing on the spectrum to obtain the spectrum recursion diagram of the vibration acceleration signal.

[0045] The model building module is used to construct a fault diagnosis model for rolling bearings by forming a combined tensor based on time-frequency distribution, time-domain waveform recursion graph and spectral recursion graph.

[0046] The execution module is used to train the fault diagnosis model based on the training sample signals, input the test sample signals into the trained fault diagnosis model to obtain a fully trained fault diagnosis model, and obtain the fault diagnosis result of the rolling bearing based on the fully trained fault diagnosis model.

[0047] The beneficial effect of using the above embodiments is that after acquiring the vibration acceleration signal of the rolling bearing, the...

[0048] The vibration acceleration signal is divided into training sample signals and test sample signals according to a preset rule, and the time-frequency distribution, time-domain waveform recursion graph, and spectral recursion graph of the vibration acceleration signal are obtained respectively. Then, a combined tensor is formed based on the time-frequency distribution, time-domain waveform recursion graph, and spectral recursion graph to construct a fault diagnosis model for rolling bearings. Finally, the fault diagnosis model is trained based on the training sample signals, and the test sample signals are input into the trained fault diagnosis model to obtain the fault diagnosis results of rolling bearings. This invention combines the recursion graph analysis method with time-domain and frequency-domain signals, which can improve the diagnostic accuracy and generalization performance of the fault diagnosis model; moreover, through training and testing the fault diagnosis model, the diagnostic accuracy of the fault model under complex environments can be further improved. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0050] Figure 1 A schematic flowchart of an embodiment of the fault diagnosis method for rolling bearings provided by the present invention;

[0051] Figure 2 For the present invention Figure 1A schematic diagram of a scenario of an embodiment of S105;

[0052] Figure 3 A schematic flowchart of a sub-embodiment of the fault diagnosis device for rolling bearings provided by the present invention;

[0053] Figure 4 This is a flowchart illustrating another sub-implementation of the fault diagnosis device for rolling bearings provided by the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor systems and / or microcontroller systems.

[0056] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0057] This invention provides a method and apparatus for diagnosing faults in rolling bearings, which will be described below.

[0058] Figure 1 This is a schematic flowchart of an embodiment of the rolling bearing fault diagnosis method provided by the present invention.

[0059] Reference Figure 1 This invention provides a method for diagnosing faults in rolling bearings, comprising:

[0060] S101. Obtain the vibration acceleration signal of the rolling bearing, and divide the vibration acceleration signal into training sample signal and test sample signal according to a preset rule;

[0061] S102. Calculate the time-frequency distribution of the vibration acceleration signal;

[0062] S103. Perform phase space reconstruction processing on the time-domain waveform of the vibration acceleration signal to obtain the recursive graph of the time-domain waveform of the vibration acceleration signal.

[0063] S104. Convert the time-domain waveform of the vibration acceleration signal into a spectrum, and perform phase space reconstruction processing on the spectrum to obtain the spectrum recursion diagram of the vibration acceleration signal.

[0064] S105. Based on the time-frequency distribution, time-domain waveform recursion graph and spectrum recursion graph, a combined tensor is formed to construct a fault diagnosis model for rolling bearings;

[0065] S106. The fault diagnosis model is trained based on the training sample signal. The test sample signal is input into the trained fault diagnosis model to obtain a fully trained fault diagnosis model. The fault diagnosis result of the rolling bearing is obtained based on the fully trained fault diagnosis model.

[0066] In embodiments of this invention, after acquiring the vibration acceleration signal of a rolling bearing, the vibration acceleration signal is divided into training sample signals and test sample signals according to a preset rule, and the time-frequency distribution, time-domain waveform recursion graph, and spectral recursion graph of the vibration acceleration signal are obtained respectively. Then, a combined tensor is formed based on the time-frequency distribution, time-domain waveform recursion graph, and spectral recursion graph to construct a fault diagnosis model for the rolling bearing. Finally, the fault diagnosis model is trained based on the training sample signals, and the test sample signals are input into the trained fault diagnosis model to obtain a fully trained fault diagnosis model. The fault diagnosis result of the rolling bearing is obtained based on the fully trained fault diagnosis model. This invention, by combining the recursion graph analysis method with time-domain and frequency-domain signals, can improve the diagnostic accuracy and generalization performance of the fault diagnosis model; moreover, through training and testing the fault diagnosis model, the diagnostic accuracy of the fault model under complex environments can be further improved.

[0067] The embodiments of the present invention will be described in detail below.

[0068] In one embodiment, obtaining the vibration acceleration signal of the rolling bearing in step S101 includes:

[0069] The vibration acceleration signals of the rolling bearing under different operating conditions are obtained based on the acceleration sensor, and the vibration acceleration signals are labeled according to different operating conditions;

[0070] The labeled vibration acceleration signals are divided into training sample signals and test sample signals according to a preset ratio;

[0071] The operating states of rolling bearings include normal operation, inner ring failure, outer ring failure, and rolling element failure.

[0072] Understandably, by acquiring vibration acceleration signals of rolling bearings under different operating conditions through accelerometers, and labeling the acceleration signals according to the different operating conditions of the bearings, vibration acceleration signal samples of rolling bearings can be obtained, thereby providing data support for the next step of data processing and model training and testing.

[0073] Furthermore, the rolling bearing vibration acceleration signal The sampling frequency should be higher than the theoretical impact frequency of the rolling bearing during operation. The sampling duration of the rolling bearing vibration acceleration signal should be no less than two vibration cycles of the rolling bearing to ensure the effectiveness of the rolling bearing vibration acceleration signal. The number of sampling points for the rolling bearing vibration acceleration signal is set to n. For example, in a sampling scenario, the sampling frequency fs of the vibration acceleration signal is set to 25600, the sampling duration of the vibration acceleration signal is no less than two vibration cycles of the rolling bearing, and the number of sampling points is 2560.

[0074] In one embodiment, after acquiring the vibration acceleration signal of the rolling bearing, the method further includes:

[0075] Gaussian white noise in the vibration acceleration signal is filtered out using wavelet threshold denoising to obtain the denoised vibration acceleration signal. This avoids the influence of environmental noise on the vibration acceleration signal of the roller bearing.

[0076] In one embodiment, calculating the time-frequency distribution of the vibration acceleration signal includes:

[0077] Wavelet transform of vibration acceleration signal based on wavelet transform function;

[0078] The wavelet transform of the vibration acceleration signal is converted into a two-dimensional time-frequency matrix of dimension [N, N] using bilinear interpolation, thus obtaining the time-frequency distribution of the vibration acceleration signal.

[0079] Furthermore, the wavelet transform method (Continuous Wavelet Transform, CWT) in the above embodiments uses complex wavelet functions and sets the basic wavelet function. As the basic wavelet of continuous wavelets, the basic wavelet function The expression is:

[0080] ;

[0081] in, Represented as bandwidth, Let represent the wavelet center frequency, t represent the time variable, and j represent the imaginary number.

[0082] The fundamental wavelet function of the continuous wavelet transform is obtained by extending and transforming the basic wavelet of the continuous wavelet transform. The expression is:

[0083] ;

[0084] Where a represents the expansion variable and b represents the translation variable.

[0085] In one embodiment, a composite Morlet wavelet with a frequency band and a center frequency of 3 is selected as the mother wavelet, the scale sequence length is set to 256, a is set to 2, and b is set to 5.

[0086] CWT can partially pinpoint time, and the size of the time window can be flexibly changed. It is suitable for unstable signals with large frequency fluctuations and is an adaptive time-frequency analysis method that can perform multi-resolution analysis. In the above embodiment, the implementation of CWT is based on the composite Morlet wavelet function, and the obtained rolling bearing acceleration signal time-frequency distribution can maintain high time-frequency concentration.

[0087] In one embodiment, the vibration acceleration signal is processed based on wavelet function transformation. The CWT transformation formula is:

[0088] ;

[0089] in, Let represent the conjugate function of the fundamental function, t represent the time variable, a represent the expansion variable, and b represent the translation variable, which are continuous variables in the formula.

[0090] Wavelet transform of vibration acceleration signal using bilinear interpolation Transform into a two-dimensional time-frequency matrix of dimension [N, N]. The time-frequency distribution of the vibration acceleration signal is obtained.

[0091] In one embodiment, the time-domain waveform of the vibration acceleration signal is reconstructed in phase space to obtain a recursive time-domain waveform of the vibration acceleration signal, including:

[0092] Reconstruction of pseudo-phase space based on vibration acceleration signal of rolling bearing and preset parameters;

[0093] Calculate the distances between target points on the pseudo-phase space trajectory;

[0094] Based on the target rules, a square graph based on a two-dimensional dot matrix is ​​constructed to obtain the time-domain waveform recursive graph of the vibration acceleration signal.

[0095] Furthermore, the preset parameters include the target dimension m and the delay constant. The reconstruction of the pseudo-phase space based on the vibration acceleration signal of the rolling bearing and preset parameters includes:

[0096] Based on the discrete-time sequence of vibration acceleration signal, target dimension m, and time delay constant Pseudo-space reconstruction of the vibration acceleration signal is performed to obtain the state variables of the vibration acceleration signal.

[0097] The expression for the discrete-time sequence of the vibration acceleration signal is: ;

[0098] in, Represents vibration acceleration signals at different times;

[0099] The state variable expression for the vibration acceleration signal is:

[0100] ;

[0101] in, This can be expressed as a state variable expression, where N represents the dimension and m represents the target dimension. It is represented as a delay constant.

[0102] The formula for calculating the distance between target points on the pseudo-phase space trajectory is:

[0103] ;

[0104] in, Represents point i on the pseudo-directional spatial trajectory to J o'clock The distance.

[0105] Furthermore, the target rule expression is:

[0106] ;

[0107] Where r represents the preset threshold.

[0108] The time-domain recurrence relation matrix of the vibration acceleration signal of the rolling bearing is as follows: .

[0109] In one embodiment, the time-domain waveform of the vibration acceleration signal is converted into a spectrum, and phase space reconstruction processing is performed on the spectrum to obtain a recursive spectrum diagram of the vibration acceleration signal, including:

[0110] Discrete-time series of vibration acceleration signals It is converted into a spectral sequence using a Fast Fourier Transform. ;

[0111] in, Represents the frequency spectrum of vibration acceleration at different times;

[0112] Based on the spectral sequence of rolling bearings, the target dimension m, and the delay constant Reconstruct the pseudo-phase space;

[0113] Calculate the distances between target points on the pseudo-phase space trajectory;

[0114] Based on the target rules, a square graph based on a two-dimensional dot matrix is ​​constructed to obtain the time-domain waveform recursive graph of the vibration acceleration signal.

[0115] The state variable expression for the spectral sequence is:

[0116] ;

[0117] in, This can be expressed as a state variable expression, where N represents the dimension and m represents the target dimension. It is represented as a delay constant.

[0118] The formula for calculating the distance between target points on the pseudo-phase space trajectory is:

[0119] ;

[0120] in, Represents point i on the pseudo-directional spatial trajectory to J o'clock The distance.

[0121] Furthermore, the target rule expression is:

[0122] ;

[0123] Where r represents the preset threshold.

[0124] The recurrence relation matrix of the vibration acceleration signal of the rolling bearing is as follows: .

[0125] Figure 2 For the present invention Figure 1 A schematic diagram of a scenario of one embodiment of S105.

[0126] Reference Figure 2 In this embodiment, a combined tensor is formed based on the time-frequency distribution, the time-domain waveform recursion graph, and the spectral recursion graph to construct a fault diagnosis model for the rolling bearing. This fault diagnosis model is a two-dimensional convolutional neural network model. In one scenario, the structure of the convolutional neural network is described as follows:

[0127] enter( The sequence is as follows: (parameters correspond to LBP spectral tensor) → 2D convolutional layer (kernel size: 5×5, channels: 16, stride: 1) → batch normalization layer (features: 16, eps: 10⁻⁵) → ReLU activation → 2D max pooling layer (kernel size: 2×2) → 2D convolutional layer (kernel size: 3×3, channels: 32, stride: 1) → batch normalization layer (features: 16, eps: 10⁻⁵) → ReLU activation → 2D max pooling layer (kernel size: 2×2) → 2D convolutional layer (kernel size: 3×3, channels: 64, stride: 1) → batch normalization layer (features: 16, eps: 10⁻⁵) → ReLU activation → adaptive average pooling layer (kernel size: 1×1) → fully connected layer (size: 64×1×1, output dimension: 128, no bias) → fully connected layer (size: 128×1×1, output dimension: 4, no bias).

[0128] Furthermore, the fault diagnosis model for rolling bearings uses the cross-entropy function as the loss function to optimize the inter-layer connection weights, with a learning step size of 128, a learning rate of 0.01, and 500 training epochs.

[0129] In one embodiment, 20% of the vibration acceleration signals of rolling bearings are randomly selected as test sample signals during the training of the fault diagnosis model. The RMSProp (Root Mean Square Prop) algorithm is applied to adaptively search for the optimal solution of the objective function. The network model is built in the form of multi-scale feature analysis, thereby obtaining a comprehensive analysis of the encoding of LBP spectrum information on three different convolutional layers, which can achieve better fault diagnosis results.

[0130] Figure 3 This is a flowchart illustrating a sub-implementation of the fault diagnosis device for rolling bearings provided by the present invention.

[0131] Reference Figure 3 The present invention also provides a fault diagnosis device for rolling bearings, comprising:

[0132] The signal acquisition module 301 is used to acquire the vibration acceleration signal of the rolling bearing and divide the vibration acceleration signal into training sample signal and test sample signal according to a preset rule.

[0133] The time-frequency distribution processing module 302 is used to calculate the time-frequency distribution of the vibration acceleration signal;

[0134] The time-domain processing module 303 is used to perform phase space reconstruction processing on the time-domain waveform of the vibration acceleration signal to obtain a recursive graph of the time-domain waveform of the vibration acceleration signal.

[0135] The spectrum processing module 304 is used to convert the time-domain waveform of the vibration acceleration signal into a spectrum, and to perform phase space reconstruction processing on the spectrum to obtain the spectrum recursion diagram of the vibration acceleration signal.

[0136] The model building module 305 is used to construct a fault diagnosis model for rolling bearings by forming a combined tensor based on time-frequency distribution, time-domain waveform recursion graph and spectrum recursion graph.

[0137] The execution module 306 is used to train the fault diagnosis model based on the training sample signal, input the test sample signal into the trained fault diagnosis model to obtain a fully trained fault diagnosis model, and obtain the fault diagnosis result of the rolling bearing based on the fully trained fault diagnosis model.

[0138] The beneficial effects of the above embodiments are as follows: After the signal acquisition module 301 acquires the vibration acceleration signal of the rolling bearing, it divides the vibration acceleration signal into training sample signals and test sample signals according to preset rules. The time-frequency distribution processing module 302, the time-domain processing module 303, and the spectrum processing module 304 respectively obtain the time-frequency distribution, time-domain waveform recursion graph, and spectrum recursion graph of the vibration acceleration signal. The model construction module 305 is used to form a combined tensor based on the time-frequency distribution, time-domain waveform recursion graph, and spectrum recursion graph to construct a fault diagnosis model for the rolling bearing. The execution module 306 trains the fault diagnosis model based on the training sample signals, inputs the test sample signals into the trained fault diagnosis model to obtain a fully trained fault diagnosis model, and obtains the fault diagnosis result of the rolling bearing based on the fully trained fault diagnosis model. This invention combines the recursion graph analysis method with time-domain and frequency-domain signals, which can improve the diagnostic accuracy and generalization performance of the fault diagnosis model. Moreover, through training and testing the fault diagnosis model, the diagnostic accuracy of the fault model under complex environments can be further improved.

[0139] like Figure 4 As shown, the present invention also provides a fault diagnosis device 400 for rolling bearings. The fault diagnosis device 400 for rolling bearings includes a processor 401, a memory 402, and a display 403. Figure 4 Only some components of the rolling bearing fault diagnosis device 400 are shown; however, it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0140] In some embodiments, memory 402 may be an internal storage unit of the rolling bearing fault diagnosis device 400, such as a hard disk or memory of the rolling bearing fault diagnosis device 400. In other embodiments, memory 402 may also be an external storage device of the rolling bearing fault diagnosis device 400, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the rolling bearing fault diagnosis device 400.

[0141] Furthermore, the memory 402 may include both internal storage units of the rolling bearing fault diagnosis device 400 and external storage devices. The memory 402 is used to store application software and various types of data of the rolling bearing fault diagnosis device 400.

[0142] In some embodiments, processor 401 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 402 or process data, such as the rolling bearing fault diagnosis method of the present invention.

[0143] In some embodiments, display 403 may be an LED display, a liquid crystal display, a touch-screen liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 403 is used to display information from the rolling bearing fault diagnosis device 400 and to display a visual user interface. Components 401-403 of the rolling bearing fault diagnosis device 400 communicate with each other via a system bus.

[0144] In some embodiments of the present invention, when the processor 401 executes the rolling bearing fault diagnosis method in the memory 402, the following steps can be implemented:

[0145] The vibration acceleration signal of the rolling bearing is acquired, and the vibration acceleration signal is divided into training sample signal and test sample signal according to a preset rule;

[0146] Calculate the time-frequency distribution of the vibration acceleration signal;

[0147] The time-domain waveform of the vibration acceleration signal is reconstructed in phase space to obtain a recursive graph of the time-domain waveform of the vibration acceleration signal.

[0148] The time-domain waveform of the vibration acceleration signal is converted into a spectrum, and the spectrum is reconstructed in phase space to obtain the spectrum recursion diagram of the vibration acceleration signal.

[0149] A fault diagnosis model for rolling bearings is constructed by forming a combined tensor based on time-frequency distribution, time-domain waveform recursion graph, and spectrum recursion graph.

[0150] The fault diagnosis model is trained based on the training sample signals, and the test sample signals are input into the trained fault diagnosis model to obtain a fully trained fault diagnosis model. The fault diagnosis results of the rolling bearing are obtained based on the fully trained fault diagnosis model.

[0151] It should be understood that when the processor 401 executes the fault diagnosis program for the rolling bearing in the memory 402, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0152] Accordingly, this application also provides a computer-readable storage medium for storing a computer-readable program or instruction. When the program or instruction is executed by a processor, it can implement the steps or functions of the rolling bearing fault diagnosis method provided in the above-described method embodiments.

[0153] The above embodiments provide a method for diagnosing rolling bearing faults that can realize the technical solutions described in the embodiments of the rolling bearing fault diagnosis device. The specific implementation principles of each module or unit can be found in the corresponding content of the embodiments of the rolling bearing fault diagnosis device, which will not be repeated here.

[0154] The fault diagnosis device for rolling bearings provided by the present invention has been described in detail above. Specific examples have been used to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation and application scope based on the idea of ​​the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for diagnosing faults in rolling bearings, characterized in that, include: The vibration acceleration signal of the rolling bearing is acquired, and the vibration acceleration signal is divided into training sample signal and test sample signal according to a preset rule; Calculate the time-frequency distribution of the vibration acceleration signal; The time-domain waveform of the vibration acceleration signal is reconstructed in phase space to obtain a recursive time-domain waveform of the vibration acceleration signal. The time-domain waveform of the vibration acceleration signal is converted into a spectrum, and the spectrum is reconstructed in phase space to obtain the spectrum recursion diagram of the vibration acceleration signal. Based on the time-frequency distribution, the time-domain waveform recursion graph, and the spectrum recursion graph, a combined tensor is formed to construct the fault diagnosis model of the rolling bearing; The fault diagnosis model is trained based on the training sample signals, and the test sample signals are input into the trained fault diagnosis model to obtain a fully trained fault diagnosis model. The fault diagnosis result of the rolling bearing is obtained based on the fully trained fault diagnosis model.

2. The fault diagnosis method for rolling bearings according to claim 1, characterized in that, The acquisition of the vibration acceleration signal of the rolling bearing includes: Vibration acceleration signals of rolling bearings under different operating conditions are obtained based on acceleration sensors, and the vibration acceleration signals are labeled according to different operating conditions; The labeled vibration acceleration signals are divided into training sample signals and test sample signals according to a preset ratio; The operating states of the rolling bearing include normal operation, inner ring failure, outer ring failure, and rolling element failure.

3. The fault diagnosis method for rolling bearings according to claim 2, characterized in that, After acquiring the vibration acceleration signal of the rolling bearing, the method further includes: The Gaussian white noise in the vibration acceleration signal is filtered out using the wavelet threshold denoising method to obtain the denoised vibration acceleration signal.

4. The fault diagnosis method for rolling bearings according to claim 1, characterized in that, The calculation of the time-frequency distribution of the vibration acceleration signal includes: The wavelet transform of the vibration acceleration signal is obtained based on the wavelet transform function; The vibration acceleration signal is transformed into a two-dimensional time-frequency matrix using bilinear interpolation to obtain the time-frequency distribution of the vibration acceleration signal.

5. The fault diagnosis method for rolling bearings according to claim 1, characterized in that, The phase space reconstruction processing of the time-domain waveform of the vibration acceleration signal to obtain a recursive time-domain waveform of the vibration acceleration signal includes: The pseudo-phase space is reconstructed based on the vibration acceleration signal of the rolling bearing and preset parameters; Calculate the distance between target points on the pseudo-phase space trajectory; Based on the target rules, a square graph based on a two-dimensional dot matrix is ​​constructed to obtain the time-domain waveform recursive graph of the vibration acceleration signal.

6. The fault diagnosis method for rolling bearings according to claim 5, characterized in that, The preset parameters include the target dimension and the delay constant; The reconstruction of the pseudo-phase space based on the vibration acceleration signal of the rolling bearing and preset parameters includes: Based on the discrete-time sequence of the vibration acceleration signal, the target dimension, and the delay constant, a pseudo-space reconstruction of the vibration acceleration signal is performed to obtain the state variables of the vibration acceleration signal. The expression for the discrete-time sequence of the vibration acceleration signal is: {x i ,i=1,2,…,N; Where, x i Represents vibration acceleration signals at different times; The state variable expression for the vibration acceleration signal is: Y i ={x i ,x i+τ ,…,x i+(m-1)τ },i=1,2,…,N-(m-1)τ; Among them, Y i It is expressed as a state variable expression, where N represents the dimension, m represents the target dimension, and τ represents the delay constant.

7. The fault diagnosis method for rolling bearings according to claim 6, characterized in that, The formula for calculating the distance between target points on the pseudo-phase spatial trajectory is as follows: Where dist(,j;m) represents point Y on the pseudo-directional spatial trajectory. i to point Y j The distance.

8. The fault diagnosis method for rolling bearings according to claim 7, characterized in that, The target rule expression is: Where r represents the preset threshold.

9. The fault diagnosis method for rolling bearings according to claim 1, characterized in that, The fault diagnosis model for the rolling bearing uses the cross-entropy function as the loss function to optimize the inter-layer connection weights.

10. A fault diagnosis device for rolling bearings, characterized in that, include: The signal acquisition module is used to acquire the vibration acceleration signal of the rolling bearing and divide the vibration acceleration signal into training sample signal and test sample signal according to a preset rule. The time-frequency distribution processing module is used to calculate the time-frequency distribution of the vibration acceleration signal; The time-domain processing module is used to perform phase space reconstruction processing on the time-domain waveform of the vibration acceleration signal to obtain a recursive graph of the time-domain waveform of the vibration acceleration signal. The spectrum processing module is used to convert the time-domain waveform of the vibration acceleration signal into a spectrum, and to perform phase space reconstruction processing on the spectrum to obtain the spectrum recursion diagram of the vibration acceleration signal. The model building module is used to form a combined tensor based on the time-frequency distribution, the time-domain waveform recursion graph, and the spectrum recursion graph to construct a fault diagnosis model for the rolling bearing; The execution module is used to train the fault diagnosis model based on the training sample signal, input the test sample signal into the trained fault diagnosis model to obtain a fully trained fault diagnosis model, and obtain the fault diagnosis result of the rolling bearing based on the fully trained fault diagnosis model.

Images