Planetary transmission fault diagnosis method and system based on time domain characteristics and frequency domain characteristics of frequency response function, medium and equipment

Abstract

The invention relates to the field of planetary transmission fault diagnosis, and discloses a planetary transmission fault diagnosis method and system based on time domain characteristics and frequency domain characteristics of a frequency response function, a medium and equipment, and the method comprises the steps: carrying out the angle resampling of collected time domain signals of a normal state and a fault state of a planetary transmission; selecting a frequency band by adopting a fast spectral kurtosis algorithm, decomposing and reconstructing the signal after the frequency band is selected, and constructing a frequency response function by using the reconstructed signal and white noise; extracting a time domain characteristic parameter in the reconstructed signal, extracting a frequency domain characteristic parameter in a frequency response function, and establishing a characteristic vector according to the time domain characteristic parameter and the frequency domain characteristic parameter; inputting the characteristic vector into a BYOL model, and training the BYOL model; and after characteristic vectors are extracted from vibration signals collected in real time in the industrial process, the vibration signals are input to the trained BYOL model for fault diagnosis. Effective feature information can be extracted, and the fault diagnosis accuracy of the planetary transmission is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of planetary transmission fault diagnosis, and in particular to a planetary transmission fault diagnosis method, system, medium and equipment based on time domain characteristics and frequency domain characteristics of frequency response functions. Background Art

[0002] Vibration test signals from planetary transmissions exhibit significant complexity, time-varying properties, and modulation. This is due to the combined effects of their inherently complex motion, dynamic nature, and variable environmental excitations. These complex and variable vibration signals not only include the characteristic frequencies of individual components, but also the vibration effects of the power input and load equipment. They are also subject to interference from background noise and exhibit a wide variety of waveform variations, ranging from simple harmonics and impulse characteristics to multi-component modulation and even time-varying modulation. To effectively extract fault signatures from planetary transmissions, signal noise reduction is essential. However, effective feature extraction based on the characteristics of vibration signals for accurate fault diagnosis remains a hot research topic in this field. Summary of the Invention

[0003] In response to the above problems, the purpose of the present invention is to provide a planetary transmission fault diagnosis method, system, medium and equipment based on time domain characteristics and frequency domain characteristics of frequency response function, which can extract effective feature information and improve the accuracy of planetary transmission fault diagnosis.

[0004] To achieve the above objectives, in a first aspect, the present invention adopts a technical solution: a planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response functions, comprising: performing angle resampling on the time domain signals of the planetary transmission in normal and faulty states, selecting a frequency band using a fast spectral kurtosis algorithm, decomposing and reconstructing the signal after the selected frequency band, and constructing a frequency response function by combining the reconstructed signal and white noise; extracting time domain feature parameters from the reconstructed signal, extracting frequency domain feature parameters from the frequency response function, and establishing a feature vector from the time domain feature parameters and the frequency domain feature parameters; inputting the feature vector into a BYOL model to train the BYOL model; extracting feature vectors from vibration signals collected in real time during industrial processes, and inputting the extracted feature vectors into the trained BYOL model for fault diagnosis.

[0005] Furthermore, the collected time domain signals of the planetary transmission in the normal state and the fault state are subjected to angle resampling, including:

[0006] Fitting the rotational speed signal in the collected time domain signal and converting it into an angular velocity signal; wherein the time domain signal includes the rotational speed signal and the time domain vibration acceleration signal;

[0007] The angular velocity signal is integrated to obtain a time-angle function, and the time-angle function is resampled. 1024 points of data are collected per rotation, and a new time-angle function is obtained by interpolation.

[0008] Linear interpolation is performed on the new time-angle function and the time-domain vibration acceleration signal to obtain the angle-domain signal.

[0009] Furthermore, the reconstructed signal and white noise are combined to construct a frequency response function, including:

[0010] The noise signal w(t) is added to the original time domain signal X(t) to obtain a new time domain signal, and the new time domain signal is subjected to EMD decomposition to obtain multiple IMF components;

[0011] Repeat the above process, add white noise of the same intensity but different sequences to the time domain signal each time, and perform EMD decomposition on the time domain signal with added white noise;

[0012] Using the white noise spectrum with a mean of zero, the average of all decomposed IMFs is calculated to obtain the final IMF component to obtain the reconstructed signal;

[0013] After correlating the reconstructed signal with the white noise and performing Fourier transform, the frequency response function is obtained.

[0014] Furthermore, the frequency response function is:

[0015]

[0016] Where g(θ) is the frequency response function and w(θ) is the amplitude of 0.2A. mp Gaussian white noise, A mp =RMS(y(θ)); y(θ) is the reconstructed signal; y(θ)*w(θ) is the cross-correlation function of the reconstructed signal and white noise, and w(θ)*w(θ) is the autocorrelation function of the white noise; FFT is the fast Fourier transform.

[0017] Furthermore, the time domain characteristic parameters in the reconstructed signal are extracted, and the frequency domain characteristic parameters in the frequency response function are extracted, and a characteristic vector is established by the time domain characteristic parameters and the frequency domain characteristic parameters, including:

[0018] Extract the time domain characteristic parameters of the reconstructed signal, including kurtosis K and waveform factor w i , pulse factor C i , peak factor Imp, margin factor M i ;

[0019] Extract the frequency domain characteristic parameters of the frequency response function, including the center of gravity frequency FC, mean square frequency MSF, root mean square frequency RMSF, frequency variance VF and frequency standard deviation RVF;

[0020] Establish the feature vector Z: Z=[K,w i ,C i ,Imp,M i ,FC,MSF,RMSF,VF,RVF] T .

[0021] Furthermore, the feature vector is input into the BYOL model and the BYOL model is trained, including: dividing the feature vector into a test set and a training set, training the BYOL model with the data in the training set, understanding the characteristics of different operating states under the feature vector, completing the classification and correct identification of the data, and testing the accuracy of the model based on the test set.

[0022] Furthermore, the BYOL model is built on the momentum encoder by introducing a multi-layer perceptron for prediction;

[0023] The BYOL model consists of two parts: an online network and a target network. The online network consists of an encoder, a projector, and a predictor; the target network consists of an encoder and a projector.

[0024] In a second aspect, the technical solution adopted by the present invention is: a planetary transmission fault diagnosis system based on time domain characteristics and frequency domain characteristics of frequency response functions, which includes: a function construction module, which performs angle resampling on the time domain signals of the planetary transmission in normal and faulty states, selects the frequency band using a fast spectral kurtosis algorithm, decomposes and reconstructs the signal after the selected frequency band, and constructs a frequency response function by combining the reconstructed signal and white noise; a feature vector establishment module, which extracts time domain feature parameters from the reconstructed signal and frequency domain feature parameters from the frequency response function, and establishes a feature vector from the time domain feature parameters and the frequency domain feature parameters; a model training module, which inputs the feature vector into a BYOL model to train the BYOL model; and a fault diagnosis module, which extracts feature vectors from vibration signals collected in real time in industrial processes, and inputs the extracted feature vectors into the trained BYOL model for fault diagnosis.

[0025] In a third aspect, the technical solution adopted by the present invention is: a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by a computing device, the computing device executes any one of the above methods.

[0026] In a fourth aspect, the technical solution adopted by the present invention is: a computing device, comprising: one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the above methods.

[0027] The present invention has the following advantages due to the adoption of the above technical solution:

[0028] 1. The present invention greatly reduces the difference in characteristic parameters caused by rotation speed through angle resampling, and quickly and effectively extracts effective components in the signal through fast spectral kurtosis algorithm and EEMD algorithm.

[0029] 2. The frequency response function transformation adopted by the present invention can effectively clarify parameters such as the characteristic frequency of the components causing vibration in the system, and can reduce the impact of noise on the effective signal.

[0030] 3. The BYOL model adopted in the present invention belongs to contrastive learning, which can efficiently realize the classification of different operating states of planetary transmissions, is independent of negative samples and has strong adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Flowchart of a planetary transmission fault diagnosis method based on frequency response function and time domain characteristics in an embodiment of the present invention;

[0032] Figure 2 This is a flow chart of a method for constructing a frequency response function according to an embodiment of the present invention;

[0033] Figure 3 2 is a structural diagram of the BYOL model in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] Researchers in the field of planetary transmission fault diagnosis have devoted considerable effort to developing effective planetary transmission fault diagnosis techniques. However, because planetary transmission vibration signals are often coupled with vibration signals from other components in the transmission system, effective feature extraction and fault diagnosis based on these signals remain a hot research topic. Planetary transmissions are a crucial component of mechanical transmission systems. The monitored signal components encompass vibration information from the entire transmission system. By constructing a response transfer function, the characteristic parameters of the corresponding components in the vibration signal can be identified, effectively providing more accurate and detailed operational status information. Therefore, the present invention proposes a planetary transmission fault diagnosis method, system, medium, and device based on time-domain and frequency-domain characteristics of frequency response functions. Using a pre-built planetary transmission work platform, time-domain signals of normal and faulty planetary transmissions are collected. The time-domain signals are angularly resampled based on the rotational speed, then frequency bands are selected using fast spectral kurtosis and reconstructed using EEMD decomposition. A frequency response function is constructed by combining the reconstructed signal with white noise. Time-domain characteristic signals are then extracted from the reconstructed signal, and frequency-domain characteristic parameters are extracted from the constructed frequency response function. The data are then divided into training and test sets. A BYOL model is trained based on the training set and tested using the test set, with the performance indicator being the fault diagnosis accuracy. After feature extraction, the real-time collected planetary transmission signals are input into the trained BYOL model for diagnosis. The present invention extracts effective time-domain and frequency-domain characteristic information, demonstrating advantages such as high diagnostic efficiency and rich fault information in diagnosis and prediction.

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] In one embodiment of the present invention, a planetary transmission fault diagnosis method based on time domain features and frequency domain features of frequency response functions is provided. The fault features of the planetary transmission are extracted by extracting time domain feature parameters from the reconstructed signal and constructing feature vectors from frequency domain feature parameters extracted based on the method of constructing frequency response functions. Specifically, in this embodiment, if Figure 1 As shown, the method includes the following steps:

[0038] Step 1: After angle resampling of the collected time domain signals of the planetary transmission in normal and faulty states, a fast spectral kurtosis algorithm is used to select the frequency band, the signals after the selected frequency band are decomposed and reconstructed, and a frequency response function is constructed by combining the reconstructed signals with white noise;

[0039] Step 2: extracting time domain characteristic parameters from the reconstructed signal and frequency domain characteristic parameters from the frequency response function, and establishing a characteristic vector from the time domain characteristic parameters and the frequency domain characteristic parameters;

[0040] Step 3: Input the feature vector into the BYOL model to train the BYOL model;

[0041] Step 4: After extracting the feature vector from the vibration signal collected in real time during the industrial process, the feature vector is input into the trained BYOL model for fault diagnosis.

[0042] In the above step 1, a vibration acceleration sensor is used to collect vibration signals of the planetary gearbox, and at least 600 signals with a length of more than 1 second are collected for each fault.

[0043] In the above step 1, the time domain signals of the planetary transmission in the normal state and the fault state are collected and angle resampled. In this embodiment, the time domain signals are angle resampled, and the number of points after each resampling is 1024. Specifically, the following steps are included:

[0044] Step 101: Fit the speed signal n(t) in the collected time domain signal to obtain the angular velocity signal ω(t); wherein the time domain signal includes the speed signal n(t) and the time domain vibration acceleration signal A(t), t=0, 1, 2, ..., M-1, and the sampling frequency is f.

[0045]

[0046] Step 102: Integrate the angular velocity signal ω(t) to obtain a time-angle function θ(t), and resample the time-angle function θ(t). 1024 points of data are collected per rotation, and a new time-angle function θ′(t) is obtained by interpolation.

[0047] in:

[0048]

[0049] Step 103: Perform linear interpolation on the new time-angle function θ′(t) and the time-domain vibration acceleration signal A(t) to obtain an angle-domain signal A(θ).

[0050] In this embodiment, a fast spectral kurtosis algorithm is used to select the signal frequency band. Specifically, the fast spectral kurtosis algorithm is used to select a parameter combination that can most effectively distinguish signal characteristics: a filter length of 16 and a normalized cutoff frequency of the filter of 0.4.

[0051] In the above step 1, the reconstructed signal and white noise are combined to construct a frequency response function, and every 2048 points are selected as a sample. Specifically, Figure 2 As shown, the following steps are included:

[0052] Step 111: Add the noise signal w(t) to the original time domain signal X(t) to obtain a new time domain signal X'(t), and perform EMD decomposition on the new time domain signal X'(t) to obtain multiple IMF components;

[0053] X`(t)=X(t)+w(t)

[0054]

[0055] Among them, r n (t) is the residual component after decomposition; c j (t) is the jth IMF component, and n represents the total number of IMF components.

[0056] Step 112: Repeat the above step 111, adding white noise of the same intensity but different sequences to the time domain signal each time, and performing EMD decomposition on the time domain signal with the white noise added;

[0057] The process is as follows:

[0058]

[0059] Among them, c ij (t) represents the jth IMF obtained by adding white noise to the decomposition for the i-th time; w in (t) is the remaining component after further decomposition.

[0060] Step 113: Using the white noise spectrum with a mean of zero, calculate the mean of all decomposed IMFs to obtain the final IMF component c n (t), to obtain the reconstructed signal;

[0061]

[0062] Step 114: Correlation processing and Fourier transform are performed on the reconstructed signal and white noise to obtain a frequency response function g(θ).

[0063] The frequency response function is:

[0064]

[0065] Where g(θ) is the frequency response function and w(θ) is the amplitude of 0.2A. mp Gaussian white noise, A mp =RMS(y(θ)); y(θ) is the reconstructed signal; y(θ)*w(θ) is the cross-correlation function of the reconstructed signal and white noise, and w(θ)*w(θ) is the autocorrelation function of the white noise; FFT is the fast Fourier transform.

[0066] In the above step 2, extracting the time domain characteristic parameters from the reconstructed signal and the frequency domain characteristic parameters from the frequency response function, and establishing a characteristic vector from the time domain characteristic parameters and the frequency domain characteristic parameters includes the following steps:

[0067] Step 201: Extract the time domain characteristic parameters of the reconstructed signal y(θ), including the kurtosis K, the waveform factor w i , pulse factor C i , peak factor Imp, margin factor M i ;

[0068] Specifically:

[0069] Kurtosis:

[0070] Form Factor:

[0071] Pulse Factor:

[0072] Crest Factor:

[0073] Margin Factor:

[0074] Among them, x i represents the vibration signal sequence, i = 1 to N; N represents the number of sampling points; RMS represents the root mean square value of the signal; μ represents the average value of the signal; X r Indicates the root square amplitude of the signal.

[0075] Step 202: extracting frequency domain characteristic parameters of the frequency response function, including center of gravity frequency FC, mean square frequency MSF, root mean square frequency RMSF, frequency variance VF, and frequency standard deviation RVF;

[0076] Specifically:

[0077] Center of gravity frequency:

[0078] Mean square frequency:

[0079] RMS frequency:

[0080] Frequency variance: VF = MSF - FC 2

[0081] Frequency standard deviation:

[0082] Where S(f) represents the spectrum of the signal.

[0083] Step 203: Create a feature vector Z: Z = [K, w i ,C i ,Imp,M i ,FC,MSF,RMSF,VF,RVF] T .

[0084] In step 3 above, the feature vector Z is input into the BYOL model (the BYOL model structure is as follows Figure 3 The BYOL model is trained, including dividing the feature vector Z into a test set and a training set, training the BYOL model with the data in the training set, understanding the characteristics of different operating states under the feature vectors, completing data classification and correct identification, and testing the accuracy of the model based on the test set.

[0085] In this embodiment, Figure 3 As shown in the figure, x represents the input sample of a given dataset, v represents the data enhancement method of the online network, v′ represents the data enhancement method of the target network, and y θ , y′ ξ They represent the characteristics of the input training samples learned by the corresponding network, q θ represents the predictor of the network, g θ Represents the projector of the network, Z θ Represents the output of the online network, Z′ ξ The BYOL model is built on the basis of the momentum encoder (Mo Co) and introduces a multi-layer perceptron for prediction. The BYOL model consists of two parts: the online network and the target network. The online network consists of the encoder f θ , projector θ and predictor q θ The target network is similar to the online network structure, except that it lacks the predictor. The target network consists of the encoder f ε and projectorg εconstitute.

[0086] Based on the above steps, any signal from any planetary gearbox is collected, features are extracted, and then input into the BYOL model for diagnosis. Based on the output, the planetary gearbox is determined to be faulty, indicating the corresponding fault type. The collected data is then added to a database, accumulating a large amount of data in actual operation to improve the accuracy of the model.

[0087] In one embodiment of the present invention, a planetary transmission fault diagnosis system based on time domain characteristics and frequency domain characteristics of a frequency response function is provided, comprising:

[0088] The function construction module performs angle resampling on the collected time domain signals of the planetary transmission in normal and faulty states, selects the frequency band using a fast spectral kurtosis algorithm, decomposes and reconstructs the signals after the selected frequency band, and constructs a frequency response function by combining the reconstructed signals with white noise.

[0089] A feature vector establishment module extracts time domain feature parameters from the reconstructed signal and frequency domain feature parameters from the frequency response function, and establishes a feature vector based on the time domain feature parameters and the frequency domain feature parameters;

[0090] The model training module inputs the feature vector into the BYOL model to train the BYOL model;

[0091] The fault diagnosis module extracts feature vectors from the vibration signals collected in real time during the industrial process and inputs them into the trained BYOL model for fault diagnosis.

[0092] In the above embodiment, the angle resampling of the collected time domain signals of the planetary transmission in the normal state and the fault state includes:

[0093] Fitting the rotational speed signal in the collected time domain signal and converting it into an angular velocity signal; wherein the time domain signal includes the rotational speed signal and the time domain vibration acceleration signal;

[0094] The angular velocity signal is integrated to obtain a time-angle function, and the time-angle function is resampled. 1024 points of data are collected per rotation, and a new time-angle function is obtained by interpolation.

[0095] Linear interpolation is performed on the new time-angle function and the time-domain vibration acceleration signal to obtain the angle-domain signal.

[0096] In the above embodiment, constructing a frequency response function by combining the reconstructed signal and white noise includes:

[0097] The noise signal w(t) is added to the original time domain signal X(t) to obtain a new time domain signal, and the new time domain signal is subjected to EMD decomposition to obtain multiple IMF components;

[0098] Repeat the above process, add white noise of the same intensity but different sequences to the time domain signal each time, and perform EMD decomposition on the time domain signal with added white noise;

[0099] Using the white noise spectrum with a mean of zero, the average of all decomposed IMFs is calculated to obtain the final IMF component to obtain the reconstructed signal;

[0100] After correlating the reconstructed signal with the white noise and performing Fourier transform, the frequency response function is obtained.

[0101] In this embodiment, the frequency response function is:

[0102]

[0103] Where g(θ) is the frequency response function and w(θ) is the amplitude of 0.2A. mp Gaussian white noise, A mp =RMS(y(θ)); y(θ) is the reconstructed signal; y(θ)*w(θ) is the cross-correlation function of the reconstructed signal and white noise, and w(θ)*w(θ) is the autocorrelation function of the white noise; FFT is the fast Fourier transform.

[0104] In the above embodiment, extracting the time domain characteristic parameters from the reconstructed signal and extracting the frequency domain characteristic parameters from the frequency response function, and establishing a characteristic vector from the time domain characteristic parameters and the frequency domain characteristic parameters includes:

[0105] Extract the time domain characteristic parameters of the reconstructed signal, including kurtosis K and waveform factor w i , pulse factor C i , peak factor Imp, margin factor M i ;

[0106] Extract the frequency domain characteristic parameters of the frequency response function, including the center of gravity frequency FC, mean square frequency MSF, root mean square frequency RMSF, frequency variance VF and frequency standard deviation RVF;

[0107] Establish the feature vector Z: Z=[K,w i ,C i ,Imp,M i ,FC,MSF,RMSF,VF,RVF] T .

[0108] In the above embodiment, the feature vector is input into the BYOL model, and the BYOL model is trained, including: dividing the feature vector into a test set and a training set, training the BYOL model with the data in the training set, understanding the characteristics of different operating states under the feature vector, completing data classification and correct recognition, and testing the accuracy of the model based on the test set.

[0109] In the above embodiment, the BYOL model is built on the basis of the momentum encoder and a multi-layer perceptron is introduced to perform predictions;

[0110] The BYOL model consists of two parts: an online network and a target network. The online network consists of an encoder, a projector, and a predictor; the target network consists of an encoder and a projector.

[0111] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0112] A computing device provided in one embodiment of the present invention may be a terminal and may include: a processor, a communications interface, a memory, a display screen, and an input device. The processor, communications interface, and memory communicate with each other via a communications bus. The processor is configured to provide computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a computer program. When executed by the processor, the computer program implements the methods described in the aforementioned embodiments. The internal memory provides an environment for the operating system and computer program in the non-volatile storage medium to run. The communications interface is configured to communicate with an external terminal via wired or wireless communication, where wireless communication may be achieved via Wi-Fi, a network management service provider, NFC (near-field communication), or other technologies. The display screen may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen layer covering the display screen, or may be buttons, a trackball, or a touchpad provided on the computing device housing, or may be an external keyboard, touchpad, or mouse. The processor may invoke logic instructions stored in the memory.

[0113] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0114] In one embodiment of the present invention, a computer program product is provided, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments.

[0115] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided. The non-transitory computer-readable storage medium stores server instructions. The computer instructions enable a computer to execute the methods provided in the above embodiments.

[0116] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effects are similar to those of the above method embodiment, and will not be repeated here.

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

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

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

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function, characterized in that: include: After angle resampling of the collected time domain signals of the planetary transmission in normal and faulty states, a fast spectral kurtosis algorithm is used to select the frequency band. The signals after the selected frequency band are decomposed and reconstructed, and a frequency response function is constructed by combining the reconstructed signals with white noise. Extracting time domain characteristic parameters from the reconstructed signal and extracting frequency domain characteristic parameters from the frequency response function, and establishing a characteristic vector from the time domain characteristic parameters and the frequency domain characteristic parameters; Input the feature vector into the BYOL model to train the BYOL model; After extracting feature vectors from the vibration signals collected in real time during industrial processes, they are input into the trained BYOL model for fault diagnosis.

2. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function according to claim 1, characterized in that: The collected time domain signals of the planetary transmission in normal state and fault state are subjected to angle resampling, including: Fitting the rotational speed signal in the collected time domain signal and converting it into an angular velocity signal; wherein the time domain signal includes the rotational speed signal and the time domain vibration acceleration signal; The angular velocity signal is integrated to obtain a time-angle function, and the time-angle function is resampled. 1024 points of data are collected per rotation, and a new time-angle function is obtained by interpolation. Linear interpolation is performed on the new time-angle function and the time-domain vibration acceleration signal to obtain the angle-domain signal.

3. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function according to claim 1, characterized in that: The reconstructed signal and white noise are combined to construct a frequency response function, including: The noise signal w(t) is added to the original time domain signal X(t) to obtain a new time domain signal, and the new time domain signal is subjected to EMD decomposition to obtain multiple IMF components; Repeat the above process, add white noise of the same intensity but different sequences to the time domain signal each time, and perform EMD decomposition on the time domain signal with added white noise; Using the white noise spectrum with a mean of zero, the average of all decomposed IMFs is calculated to obtain the final IMF component to obtain the reconstructed signal; After correlating the reconstructed signal with the white noise and performing Fourier transform, the frequency response function is obtained.

4. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function as claimed in claim 3, characterized in that: The frequency response function is: Where g(θ) is the frequency response function and w(θ) is the amplitude of 0.2A. mp Gaussian white noise, A mp =RMS(y(θ)); y(θ) is the reconstructed signal; y(θ)*w(θ) is the cross-correlation function of the reconstructed signal and white noise, and w(θ)*w(θ) is the autocorrelation function of the white noise; FFT is the fast Fourier transform.

5. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function according to claim 1, characterized in that: Extract the time domain characteristic parameters from the reconstructed signal and the frequency domain characteristic parameters from the frequency response function, and establish a characteristic vector based on the time domain characteristic parameters and the frequency domain characteristic parameters, including: Extract the time domain characteristic parameters of the reconstructed signal, including kurtosis K and waveform factor w i , pulse factor C i , peak factor Imp, margin factor M i ; Extract the frequency domain characteristic parameters of the frequency response function, including the center of gravity frequency FC, mean square frequency MSF, root mean square frequency RMSF, frequency variance VF and frequency standard deviation RVF; Establish the feature vector Z: Z=[K,w i ,C i ,Imp,M i ,FC,MSF,RMSF,VF,RVF] T .

6. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function as claimed in claim 1, characterized in that: Input the feature vector into the BYOL model and train the BYOL model. This includes dividing the feature vector into a test set and a training set, training the BYOL model with the data in the training set, understanding the characteristics of different operating states under the feature vector, classifying and correctly identifying the data, and testing the model's accuracy based on the test set.

7. The planetary transmission fault diagnosis method based on time domain characteristics and frequency domain characteristics of frequency response function as claimed in claim 6, characterized in that: The BYOL model is built on the momentum encoder by introducing a multi-layer perceptron for prediction; The BYOL model consists of two parts: an online network and a target network. The online network consists of an encoder, a projector, and a predictor; the target network consists of an encoder and a projector.

8. A planetary transmission fault diagnosis system based on time domain characteristics and frequency domain characteristics of frequency response function, characterized in that: include: The function construction module performs angle resampling on the collected time domain signals of the planetary transmission in normal and faulty states, selects the frequency band using a fast spectral kurtosis algorithm, decomposes and reconstructs the signals after the selected frequency band, and constructs a frequency response function by combining the reconstructed signals with white noise. A feature vector establishment module extracts time domain feature parameters from the reconstructed signal and frequency domain feature parameters from the frequency response function, and establishes a feature vector based on the time domain feature parameters and the frequency domain feature parameters; The model training module inputs the feature vector into the BYOL model to train the BYOL model; The fault diagnosis module extracts feature vectors from the vibration signals collected in real time during the industrial process and inputs them into the trained BYOL model for fault diagnosis.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any one of the methods of claims 1 to 7 .

10. A computing device, characterized in that include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any one of the methods according to claims 1 to 7.

Images