A method and system for diagnosing faults in gears and bearings in a transmission gearbox

Abstract

The present application relates to a kind of gear and bearing fault diagnosis method and system in transmission gearbox, method includes: obtaining the fault data set of bearing in transmission gearbox and the composite fault data set of gear, and data processing is carried out;After training, the bearing fault diagnosis model and gear fault diagnosis model are obtained;The running data of the fault bearing to be diagnosed in the preset transmission gearbox is obtained, and the fault mode of the fault bearing to be diagnosed is determined using bearing fault diagnosis model, the running data of the fault gear to be diagnosed in the preset transmission gearbox is obtained, and the fault mode of the fault gear to be diagnosed is determined using gear fault diagnosis model.Classification is carried out by predicting the gear and bearing composite fault of transmission gearbox, different fault data set is analyzed and handled using specific method, so that diagnosis target is completely realized in fault diagnosis, and the accuracy of fault diagnosis is high.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault diagnosis of gears and bearings in a transmission gear box, and in particular to a fault diagnosis method and system for gears and bearings in a transmission gear box. BACKGROUND

[0002] The accessory gear system is an important structure of a very important mechanical component in a gas turbine, which plays a key role in changing speed and transmitting torque. The gears and rolling bearings in the transmission gear box are the most important components of the gear box. In actual engineering, the gears and bearings in the transmission gear box often have compound faults. Intelligent diagnosis of such faults has certain practical engineering significance. Compound faults usually refer to the simultaneous occurrence of two or more faults on a mechanical device. Different faults may occur simultaneously, or after a fault occurs in a component, the fault is not handled in time, causing other components to also fail. In complex systems, the occurrence of faults is random and concurrent, and as the size of the system continues to expand, the situation between multiple faults detected by the system is also more complex. Because the working structure of the transmission gear box is complex, its signals are generally multi-component modulation non-stationary signals coupled with different fault characteristics, and traditional signal analysis and processing methods often cannot completely achieve the diagnosis target in fault diagnosis. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a fault diagnosis method and system for gears and bearings in a transmission gear box to solve the problems of the prior art.

[0004] The technical scheme of the fault diagnosis method for gears and bearings in a transmission gear box of the present application is as follows:

[0005] Obtain the fault data set of the bearings and the compound fault data set of the gears in the transmission gear box, and perform data processing;

[0006] Based on the fault data set after data processing, train the first deep learning model to obtain a bearing fault diagnosis model, and based on the compound fault data set after data processing, train the second deep learning model to obtain a gear fault diagnosis model;

[0007] Obtain the running data of a fault bearing to be diagnosed in a preset transmission gear box, and use the bearing fault diagnosis model to determine the fault mode of the fault bearing to be diagnosed, obtain the running data of a fault gear to be diagnosed in the preset transmission gear box, and use the gear fault diagnosis model to determine the fault mode of the fault gear to be diagnosed.

[0008] The technical scheme of the fault diagnosis system for gears and bearings in a transmission gear box of the present application is as follows:

[0009] The method comprises a data processing module, a training module and a fault diagnosis module.

[0010] The data processing module is configured to acquire a fault data set of a bearing and a compound fault data set of a gear in a transmission gear box, and perform data processing.

[0011] The training module is configured to train a first deep learning model based on the processed fault data set to obtain a bearing fault diagnosis model, and train a second deep learning model based on the processed compound fault data set to obtain a gear fault diagnosis model.

[0012] The fault diagnosis module is configured to acquire running data of a bearing to be diagnosed in a preset transmission gear box, and determine a fault mode of the bearing to be diagnosed by using the bearing fault diagnosis model, and acquire running data of a gear to be diagnosed in the preset transmission gear box, and determine a fault mode of the gear to be diagnosed by using the gear fault diagnosis model.

[0013] The technical scheme of the electronic device of the present application is as follows:

[0014] The method comprises a memory, a processor and a program stored in the memory and running on the processor, and the processor executes the program to realize the steps of the fault diagnosis method of the gear and the bearing in the transmission gear box.

[0015] The beneficial effects of the present application are as follows:

[0016] By predicting and classifying the compound faults of the gear and the bearing in the transmission gear box, different fault data sets are analyzed and processed by using specific methods, so that the diagnosis target is completely realized in fault diagnosis, and the accuracy of fault diagnosis is high. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the fault diagnosis method of the gear and the bearing in the transmission gear box is shown in the embodiment of the present application.

[0018] Figure 2 The normal data waveform chart in the embodiment of the present application is shown in the embodiment of the present application.

[0019] Figure 3 The drive end rolling body fault waveform chart in the embodiment of the present application is shown in the embodiment of the present application.

[0020] Figure 4 The drive end inner ring fault waveform chart in the embodiment of the present application is shown in the embodiment of the present application.

[0021] Figure 5 The drive end outer ring fault waveform chart in the embodiment of the present application is shown in the embodiment of the present application.

[0022] Figure 6 This is a waveform diagram of a rolling element failure at the fan end according to an embodiment of the present invention;

[0023] Figure 7 This is a waveform diagram of a fan end inner ring fault in an embodiment of the present invention;

[0024] Figure 8 This is a waveform diagram of a fan end outer ring fault in an embodiment of the present invention;

[0025] Figure 9 This is a normal data waveform diagram in an embodiment of the present invention;

[0026] Figure 10 This is a waveform diagram of bearing rolling element defect data in an embodiment of the present invention;

[0027] Figure 11 This is a waveform diagram of gear tooth breakage defect data in an embodiment of the present invention;

[0028] Figure 12 This is a waveform diagram of gear tooth breakage + bearing outer ring defect data in an embodiment of the present invention;

[0029] Figure 13 This is a waveform diagram of gear tooth breakage + bearing inner ring defect data in an embodiment of the present invention;

[0030] Figure 14 This is a waveform diagram of gear tooth missing defect data in an embodiment of the present invention;

[0031] Figure 15 This is a waveform diagram of the gear tooth missing + bearing outer ring defect data in an embodiment of the present invention;

[0032] Figure 16 This is a waveform diagram of the gear tooth missing + bearing inner ring defect data in an embodiment of the present invention;

[0033] Figure 17 is the loss / accuracy of the predicted fault classification in the embodiment of the present invention;

[0034] Figure 18 This is the confusion matrix obtained by the prediction network in the embodiment of the present invention.

[0035] Figure 19 The figure is a schematic structural diagram of a fault diagnosis system for gears and bearings in a transmission gearbox according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] like Figure 1 As shown, a fault diagnosis method for gears and bearings in a transmission gearbox according to an embodiment of the present invention includes the following steps:

[0037] S1, acquire a bearing fault dataset and a gear compound fault dataset in a transmission gearbox, and perform data processing;

[0038] The bearing fault dataset includes multiple fault data acquired at a first preset sampling frequency.

[0039] S10, from the bearing fault dataset, acquire a vibration signal corresponding to each fault mode of the bearing, define the vibration signal as a first time sequence, obtain a first frequency domain signal corresponding to each first time sequence by using FFT transformation, add a one-hot label to each first frequency domain signal, and obtain the data-processed fault dataset.

[0040] The first preset sampling frequency is 12 kHz or 48 kHz, or can be set according to actual conditions.

[0041] The multiple fault data in the bearing fault dataset include bearing rolling element defect data, gear tooth breakage defect data, gear tooth breakage-outer ring defect data, gear tooth breakage-inner ring defect data, gear tooth missing defect data, gear tooth missing-outer ring defect data, and gear tooth missing-inner ring defect data.

[0042] Each first time sequence includes a preset number of data points, such as 500 or 1000 data points, the data-processed fault dataset includes all first frequency domain signals added with one-hot labels, the first deep learning model is trained based on all first frequency domain signals added with one-hot labels, and a bearing fault diagnosis model is obtained.

[0043] The process of adding a one-hot label to each first frequency domain signal is specifically as follows:

[0044] The digital label of the bearing rolling element defect data is set to 0, the digital label of the gear tooth breakage defect data is set to 1, the digital label of the gear tooth breakage-outer ring defect data is set to 2, the digital label of the gear tooth breakage-inner ring defect data is set to 3, the digital label of the gear tooth missing defect data is set to 4, the digital label of the gear tooth missing-outer ring defect data is set to 5, and the digital label of the gear tooth missing-inner ring defect data is set to 6. The one-hot label of the vibration signal corresponding to each fault mode of the bearing is obtained according to the digital labels of the multiple fault data in the bearing fault dataset, and the one-hot label of the vibration signal corresponding to each fault mode of the bearing is the one-hot label of the corresponding first frequency domain signal.

[0045] The composite fault data set of the gear includes multi-class fault data obtained at a second preset sampling frequency, and in S1, a process of data processing on the composite fault data set of the gear includes:

[0046] In S11, vibration signals corresponding to each fault mode of the gear are obtained from the composite fault data set of the gear, the vibration signals are defined as second time series, second frequency domain signals corresponding to each second time series are obtained by using FFT transformation, one-hot labels are added to each second frequency domain signal, and a data-processed composite fault data set is obtained.

[0047] The second preset sampling frequency is 12 kHz or 48 kHz. The second preset sampling frequency can also be set according to actual conditions.

[0048] The multi-class fault data in the composite fault data set of the gear includes:

[0049] normal bearing data, bearing rolling body defect data, gear tooth breakage defect data, gear tooth breakage-outer bearing ring defect data, gear tooth breakage-inner bearing ring defect data, gear tooth missing defect data, gear tooth missing-outer bearing ring defect data, and gear tooth missing-inner bearing ring defect data.

[0050] Each first time series includes a preset number of data points, such as 500 or 1000 data points, the data-processed composite fault data set includes all second frequency domain signals added with one-hot labels, a second deep learning model is trained based on all second frequency domain signals added with one-hot labels, and a gear fault diagnosis model is obtained.

[0051] The process of adding one-hot labels to each second frequency domain signal is specifically as follows:

[0052] The digital label of the normal bearing data is set to 0, the digital label of the bearing rolling body defect data is set to 1, the digital label of the gear tooth breakage defect data is set to 2, the digital label of the gear tooth breakage-outer bearing ring defect data is set to 3, the digital label of the gear tooth breakage-inner bearing ring defect data is set to 4, the digital label of the gear tooth missing defect data is set to 5, the digital label of the gear tooth missing-outer bearing ring defect data is set to 6, the digital label of the gear tooth missing-inner bearing ring defect data is set to 7, one-hot labels of the vibration signals corresponding to each fault mode of the gear are obtained according to the digital labels of the multi-class fault data in the composite fault data set of the gear, and the one-hot labels of the vibration signals corresponding to each fault mode of the gear are one-hot labels of the corresponding second frequency domain signals.

[0053] S2, training the first deep learning model based on the processed fault data set to obtain a bearing fault diagnosis model, and training the second deep learning model based on the processed composite fault data set to obtain a gear fault diagnosis model;

[0054] Wherein, all the first frequency domain signals added with one-hot labels are shuffled in order, 90% of the data is used for training, and 10% of the data is verified to obtain the bearing fault diagnosis model.

[0055] Wherein, all the second frequency domain signals added with one-hot labels can be shuffled in order, 90% of the data can be used for training, and 10% of the data can be verified to obtain the gear fault diagnosis model.

[0056] Wherein, the input data of the bearing fault diagnosis model is the first frequency domain signal, the output data of the bearing fault diagnosis model is the fault mode of the bearing, the input data of the gear fault diagnosis model is the second frequency domain signal, and the output data of the gear fault diagnosis model is the fault mode of the gear.

[0057] S3, obtaining the running data of the bearing to be diagnosed in the preset transmission gearbox, and determining the fault mode of the bearing to be diagnosed by using the bearing fault diagnosis model, obtaining the running data of the gear to be diagnosed in the preset transmission gearbox, and determining the fault mode of the gear to be diagnosed by using the gear fault diagnosis model.

[0058] Wherein, the running data of the bearing to be diagnosed in the preset transmission gearbox includes: one-dimensional waveform, i.e. vibration signal, collected by vibration acceleration sensor under rolling bearing fault state, i.e. one-dimensional time series data, the data is subjected to FFT transformation to obtain the frequency domain signal corresponding to the data, the frequency domain signal is input into the bearing fault diagnosis model to determine the fault mode of the bearing to be diagnosed.

[0059] Wherein, the running data of the gear to be diagnosed in the preset transmission gearbox includes: one-dimensional waveform, i.e. vibration signal, collected by vibration acceleration sensor under gear fault state, i.e. one-dimensional time series data, the data is subjected to FFT transformation to obtain the frequency domain signal corresponding to the data, the frequency domain signal is input into the gear fault diagnosis model to determine the fault mode of the gear to be diagnosed.

[0060] Wherein, the fault mode of the bearing to be diagnosed includes: bearing rolling element defect, bearing outer ring defect, bearing inner ring defect.

[0061] Wherein, the fault mode of the gear to be diagnosed includes: gear tooth breakage, gear tooth defect.

[0062] Optionally, in the above technical solution, the first deep learning model and the second deep learning model are both self-attention MLP networks, the self-attention MLP network comprises an input layer, a self-attention layer, a flatten layer, a first full connection layer, a second full connection layer and an output layer arranged in sequence, wherein the number of parameters of the self-attention layer is 3, the number of parameters of the first full connection layer is 364*364, and the number of parameters of the second full connection layer is 375*95. The output shape and the number of parameters of the self-attention MLP network are shown in Table 1.

[0063] Table 1:

[0064] In another embodiment, the first deep learning model and the second deep learning model can be neural networks.

[0065] In the following, another embodiment of the fault diagnosis method for gears and bearings in a transmission gearbox is described, which comprises:

[0066] S1, obtaining a bearing fault data set and a gear composite fault data set, and performing data processing;

[0067] S2, training the first deep learning model based on the data-processed fault data set to obtain a bearing fault diagnosis model, and training the second deep learning model based on the data-processed composite fault data set to obtain a gear composite fault diagnosis model;

[0068] S3, obtaining running data of a fault-bearing to be diagnosed, and using the bearing fault diagnosis model to determine the fault mode of the fault-bearing to be diagnosed, and obtaining running data of a composite fault gear to be diagnosed, and using the gear composite fault diagnosis model to determine the fault mode of the composite fault gear to be diagnosed.

[0069] Two data sets are used in this embodiment, which are a bearing data set (CWRU bearing data set) and a gearbox composite fault data set made in the laboratory.

[0070] Specifically, the processing of the CWRU bearing data set is provided in this embodiment, which is as follows:

[0071] The running data of the fault-bearing to be diagnosed in the CWRU bearing data set comprises one-dimensional waveform data, i.e. one-dimensional time series data, collected by a vibration acceleration sensor under a rolling bearing fault state.

[0072] The fault mode of the fault-bearing to be diagnosed comprises bearing rolling element defects, bearing outer ring defects and bearing inner ring defects.

[0073] Specifically, the data in the CWRU bearing dataset includes normal baseline data, 12 kHz drive end fault data, 48 kHz drive end fault data, and 12 kHz fan end fault data; the fan end bearing is of model SKF6203, and the drive end bearing is of model SKF6205; the bearing geometric parameters are shown in Table 2, where α is 0 because the drive end and fan end bearings are both ball bearings.

[0074] Table 2:

[0075] Rolling element diameter b / inch Pitch diameter D / inch Number of rolling elements z Contact angle a Drive end 0.3126 1.537 9 0 Fan end 0.2656 1.122 8 0

[0076] The data in the CWRU dataset is collected at four different rotating speeds, and the rotating speeds and the fault characteristic frequencies of the drive end bearing and the fan end bearing at different rotating speeds are shown in Table 3, where the fault frequency refers to a multiple of the rotating frequency (Hz).

[0077] Table 3:

[0078]

[0079] As can be seen from Table 3, each fault frequency at different rotating speeds is similar, so the data at different rotating speeds can be classified as one type for fault recognition.

[0080] Optionally, in the above technical solution, the CWRU bearing dataset includes multiple types of fault data obtained according to a first preset sampling frequency, and in S1, the process of data processing on the bearing fault dataset includes:

[0081] S10, generating a first time sequence corresponding to each type of fault data in the CWRU bearing dataset, using FFT transformation to obtain a first frequency domain signal corresponding to each first time sequence, adding a one-hot label to each first frequency domain signal to obtain a data-processed fault dataset.

[0082] The first preset sampling frequency is 12 kHz.

[0083] The multiple types of fault data in the bearing fault dataset are divided into seven types, namely normal, fan end inner ring fault, fan end outer ring fault, fan end rolling element fault, drive end inner ring fault, drive end outer ring fault, and drive end rolling element fault.

[0084] The number of fault data in each type of fault data in the bearing fault dataset is set to 500 to form a first time sequence, the data-processed fault dataset includes all first frequency domain signals with one-hot labels, the first deep learning model is trained based on all first frequency domain signals with one-hot labels to obtain a bearing fault diagnosis model, and specifically:

[0085] The first time sequence corresponding to each type of fault data in the bearing fault data set is transformed by FFT to obtain a first frequency domain signal corresponding to each first time sequence, specifically:

[0086] 1) The time sequence with a length of N=2m is represented as x(n), n=1, 2, …N, and the discrete Fourier transform (DFT) of x(n) is:

[0087]

[0088] wherein k=0, 1, …, N-1,

[0089] 2) x(n) is decomposed into the sum of two sequences of even and odd numbers, that is:

[0090] x(n)=x1(n)+x2(n) Equation (2-2)

[0091] 3) According to equation (2-2), the lengths of x1(n) and x2(n) are both N / 2, assuming that x1(n) is an even sequence and x2(n) is an odd sequence, then:

[0092]

[0093] Therefore:

[0094]

[0095] Since then:

[0096]

[0097] wherein X1(k) and X2(k) are N / 2-point DFTs of x1(n) and x2(n) respectively. Since X1(k) and X2(k) both have a period of N / 2, and Therefore, X(k) can also be represented as:

[0098]

[0099]

[0100] wherein the typical frequency domain fault graphs of the seven types of faults are as shown in Figures 2-8 , and the title of each graph is DE_TIME, representing the data obtained by the drive end acceleration sensor; Figures 2-8 The collection of the above is the frequency domain waveform graph of the typical fault state in the CWRU bearing data set.

[0101] wherein the process of adding a one-hot label to each second frequency domain signal is specifically:

[0102] The normal data is labeled as 0, the drive end rolling body fault is labeled as 1, the drive end inner ring fault is labeled as 2, the drive end outer ring fault is labeled as 3, the fan end rolling body fault is labeled as 4, the fan end inner ring fault is labeled as 5, and the fan end outer ring fault is labeled as 6, and then a one-hot label is generated.

[0103] The first frequency domain signal to which the one-hot label is added can be shuffled in sequence, 90% of the data can be used for training, and 10% of the data can be verified to obtain a CWRU bearing data set fault diagnosis model.

[0104] Specifically, the processing of the composite fault data set in the embodiment is as follows:

[0105] The operation data of the fault gear to be diagnosed in the transmission gearbox includes one-dimensional waveform data, i.e., one-dimensional time series data, collected by a vibration acceleration sensor in a gear fault state.

[0106] The fault modes of the fault bearing to be diagnosed include bearing rolling body defects, bearing outer ring defects, and bearing inner ring defects.

[0107] The fault modes of the fault gearbox to be diagnosed include gear tooth breakage and gear tooth defect.

[0108] Specifically, the composite fault data set includes normal bearing data, bearing rolling body defect data, bearing inner ring defect data, bearing outer ring defect data, gear tooth breakage data, gear tooth breakage + bearing outer ring defect data, gear tooth breakage + bearing inner ring defect data, gear tooth defect data, gear tooth defect + bearing outer ring defect data, and gear tooth defect + bearing inner ring defect data. The model of the fault bearing is ER-16K, and the number of fault gears is 29 (Z1=29). The data is collected at five different speeds, i.e., 1200 rpm, 1500 rpm, 1800 rpm, 2100 rpm, and 2400 rpm, and the corresponding frequencies are 20 Hz, 25 Hz, 30 Hz, 35 Hz, and 40 Hz, respectively. The data file contains data of two acceleration sensors, and the embodiment of the application uses the data of each Channel_1_Data in the file with a speed of 1200 rmp.

[0109] Alternatively, in the above technical solution, the composite fault data set of the gearbox includes multi-class fault data obtained according to a second preset sampling frequency, and the process of data processing of the composite fault data set of the gear in S1 includes:

[0110] S11. Generate a second time series corresponding to each type of fault data in the gear composite fault dataset, perform an FFT transform, obtain a second frequency domain signal corresponding to each second time series, and add a one-hot label to each second frequency domain signal to obtain a processed composite fault dataset. The data acquisition frequency is 16384 Hz.

[0111] Among them, the multi-class fault data in the gear composite fault dataset includes:

[0112] Normal bearing data, bearing defect data, gear tooth breakage data, gear tooth breakage-bearing outer ring defect data, gear tooth breakage-bearing inner ring defect data, gear tooth missing defect data, gear tooth missing-bearing outer ring defect data, and gear tooth missing-bearing inner ring defect data.

[0113] Among them, the set number of each type of fault data in the composite fault data set of the gear is 1,000 fault data to form a second time series, and the data is processed using FFT transformation. The FFT transformation process is shown in S10. The processed composite fault data set includes all second frequency domain signals with one-hot labels added. The second deep learning model is trained based on all second frequency domain signals with one-hot labels added to obtain a gear fault diagnosis model. Specifically: all second frequency domain signals with one-hot labels added are shuffled in order, 90% of the data are used for training, and 10% of the data are used for verification to obtain a gear fault diagnosis model.

[0114] Among them, the process of adding a one-hot label to each second frequency domain signal is specifically as follows: setting the digital label of normal bearing data to 0, setting the digital label of bearing defect data to 1, setting the digital label of gear tooth breakage data to 2, setting the digital label of gear tooth breakage-bearing outer ring defect data to 3, setting the digital label of gear tooth breakage-bearing inner ring defect data to 4, setting the digital label of gear tooth missing defect data to 5, setting the digital label of gear tooth missing-bearing outer ring defect data to 6, and setting the digital label of gear tooth missing-bearing inner ring defect data to 7, and then generating corresponding one-hot labels.

[0115] Specifically, the typical frequency domain fault diagrams of eight types of faults are as follows: Figures 9-16 As shown, the title of each figure is Channel_1_Data, which represents the data obtained by the Channel_1_Data channel acceleration sensor. Figures 9-16 The collection is the frequency domain waveform diagram of typical fault states in the composite fault data set.

[0116] Optionally, in the above technical solution, the first deep learning model and the second deep learning model are both self-attention MLP networks, and the output shape and the number of parameters of the self-attention MLP network are as shown in Table 1, which is not described herein.

[0117] The "loss / accuracy curve" of the gearbox compound fault classification is as shown in Figure 17 The confusion matrix obtained by testing the self-attention MLP network is as shown in Figure 18 The results of the test set are output to a two-dimensional plane. The final accuracy of the network prediction is 0.97.

[0118] Table 4 shows that more than 10 data of the gear tooth missing + bearing inner ring defect and the gear tooth missing + bearing inner ring defect are identified into the category of the other, and the F1 score of the two types of faults is also low; one disadvantage of the network is that the classification boundary of each type of fault is not clear. In addition, another disadvantage of the network is long response time, and Table 4 shows that:

[0119] Table 4:

[0120] In the above embodiments, although the steps are numbered S1, S2, etc., this is only a specific embodiment given by the present application, and those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation, which is within the protection scope of the present application. It can be understood that in some embodiments, some or all of the above embodiments can be included.

[0121] As shown in Figure 19 The fault diagnosis system 200 of the gear and bearing in the transmission gearbox according to an embodiment of the present application includes a data processing module 210, a training module 220, and a fault diagnosis module 230.

[0122] The data processing module 210 is configured to acquire a fault data set of the bearing and a compound fault data set of the gear in the transmission gearbox, and perform data processing.

[0123] The training module 220 is configured to train the first deep learning model based on the processed fault data set to obtain a bearing fault diagnosis model, and train the second deep learning model based on the processed compound fault data set to obtain a gear fault diagnosis model.

[0124] The fault diagnosis module 230 is configured to acquire running data of a to-be-diagnosed fault bearing in a preset transmission gearbox, and determine a fault mode of the to-be-diagnosed fault bearing by using the bearing fault diagnosis model, and acquire running data of a to-be-diagnosed fault gear in the preset transmission gearbox, and determine a fault mode of the to-be-diagnosed fault gear by using the gear fault diagnosis model.

[0125] Optionally, in the above technical solution, the fault data set of the bearing includes multiple types of fault data obtained at a first preset sampling frequency, and the process of data processing of the fault data set of the bearing by the data processing module 210 includes:

[0126] From the fault data set of the bearing, the vibration signal corresponding to each fault mode of the bearing is obtained, which is defined as a first time sequence, and the first frequency domain signal corresponding to each first time sequence is obtained by using FFT transformation. A one-hot label is added to each first frequency domain signal to obtain the processed fault data set.

[0127] Optionally, in the above technical solution, the multiple types of fault data in the fault data set of the bearing include: bearing rolling body defect data, gear tooth breakage defect data, gear tooth breakage- bearing outer ring defect data, gear tooth breakage- bearing inner ring defect data, gear tooth missing defect data, gear tooth missing- bearing outer ring defect data, and gear tooth missing- bearing inner ring defect data.

[0128] Optionally, in the above technical solution, the first preset sampling frequency is 12 kHz or 48 kHz.

[0129] Optionally, in the above technical solution, the composite fault data set of the gear includes multiple types of fault data obtained at a second preset sampling frequency, and the process of data processing of the composite fault data set of the gear by the data processing module 210 includes:

[0130] From the composite fault data of the gear, the vibration signal corresponding to each fault mode of the gear is obtained, which is defined as a second time sequence, and the second frequency domain signal corresponding to each second time sequence is obtained by using FFT transformation. A one-hot label is added to each second frequency domain signal to obtain the processed composite fault data set.

[0131] Optionally, in the above technical solution, the multiple types of fault data in the composite fault data set of the gear include:

[0132] Normal bearing data, bearing rolling body defect data, gear tooth breakage defect data, gear tooth breakage- bearing outer ring defect data, gear tooth breakage- bearing inner ring defect data, gear tooth missing defect data, gear tooth missing- bearing outer ring defect data, and gear tooth missing- bearing inner ring defect data.

[0133] Optionally, in the above technical solution, the second preset sampling frequency is 12 kHz or 48 kHz.

[0134] Optionally, in the technical solution above, the first deep learning model and the second deep learning model are both self-attention MLP networks, the self-attention MLP network comprises an input layer, a self-attention layer, a flatten layer, a first full connection layer, a second full connection layer and an output layer arranged in sequence, wherein the number of parameters of the self-attention layer is 3, the number of parameters of the first full connection layer is 364*364, and the number of parameters of the second full connection layer is 375*95.

[0135] The steps of implementing the corresponding functions of each parameter and each unit module in the fault diagnosis system 200 of the gear and bearing in the transmission gear box according to the embodiments of the fault diagnosis method of the gear and bearing in the transmission gear box are described above, and will not be repeated here.

[0136] The electronic device according to the embodiments of the present application comprises a memory, a processor and a program stored in the memory and running on the processor, and the processor implements the steps of the fault diagnosis method of the gear and bearing in the transmission gear box according to any one of the embodiments when executing the program.

[0137] Correspondingly, the program is computer software or mobile phone APP, and the parameters and steps in the electronic device according to the present application are described above in the embodiments of the fault diagnosis method of the gear and bearing in the transmission gear box, and will not be repeated here.

[0138] Those skilled in the art know that the present application can be implemented as a system, a method or a computer program product.

[0139] Therefore, the present disclosure can be embodied in the form of a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present application can also be embodied in the form of a computer program product in one or more computer readable media, which contains computer readable program code.

[0140] Any combination of one or more computer readable medium can be utilized. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0141] Although the embodiments of the present application have been shown and described above, it should be understood by those skilled in the art that the above embodiments are exemplary, and cannot be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements, and variations to the above embodiments within the scope of the present application.

Claims

1. A method of fault diagnosis of gears and bearings in a transmission gearbox, c h a r a c t e r i s e d in that, The method comprises the following steps: obtaining a fault data set of a bearing in a transmission gear box and a composite fault data set of a gear, and performing data processing; training a first deep learning model based on the processed fault data set to obtain a bearing fault diagnosis model, and training a second deep learning model based on the processed composite fault data set to obtain a gear fault diagnosis model; obtaining running data of a bearing to be diagnosed in a preset transmission gear box, and determining a fault mode of the bearing to be diagnosed by using the bearing fault diagnosis model, and obtaining running data of a gear to be diagnosed in the preset transmission gear box, and determining a fault mode of the gear to be diagnosed by using the gear fault diagnosis model; The fault data set of the bearing includes multiple types of fault data obtained at a first preset sampling frequency, and the data processing process of the fault data set of the bearing comprises: from the fault data set of the bearing, obtain the vibration signal corresponding to each fault mode of the bearing, define the vibration signal as a first time sequence, obtain the first frequency domain signal corresponding to each first time sequence by using FFT transformation, add a one-hot label to each first frequency domain signal to obtain the processed fault data set; The composite fault data set of the gear includes multiple types of fault data obtained at a second preset sampling frequency, and the data processing process of the composite fault data set of the gear comprises: from the composite fault data of the gear, obtain the vibration signal corresponding to each fault mode of the gear, define the vibration signal as a second time sequence, obtain the second frequency domain signal corresponding to each second time sequence by using FFT transformation, add a one-hot label to each second frequency domain signal to obtain the processed composite fault data set; The first deep learning model and the second deep learning model are both self-attention MLP networks, and the self-attention MLP network comprises an input layer, a self-attention layer, a flatten layer, a first full connection layer, a second full connection layer and an output layer arranged in sequence, wherein the number of parameters of the self-attention layer is 3, the number of parameters of the first full connection layer is 364*364, and the number of parameters of the second full connection layer is 375*95.

2. A method of diagnosing a fault of a gear and a bearing in a transmission gear case according to claim 1, characterized by, The multiple types of fault data in the fault data set of the bearing include: bearing rolling element defect data, gear tooth breakage defect data, gear tooth breakage- bearing outer ring defect data, gear tooth breakage- bearing inner ring defect data, gear tooth missing defect data, gear tooth missing- bearing outer ring defect data, and gear tooth missing- bearing inner ring defect data.

3. A method of diagnosing a fault in gears and bearings in a transmission gearbox according to claim 1, characterized in that, The first preset sampling frequency is 12kHz or 48kHz.

4. A method of diagnosing a fault in gears and bearings in a transmission gearbox according to claim 1, characterized in that, The multiple types of fault data in the composite fault data set of the gear include: normal bearing data, bearing rolling element defect data, gear tooth breakage defect data, gear tooth breakage- bearing outer ring defect data, gear tooth breakage- bearing inner ring defect data, gear tooth missing defect data, gear tooth missing- bearing outer ring defect data, and gear tooth missing- bearing inner ring defect data.

5. A method of fault diagnosis of gears and bearings in a transmission gearbox according to claim 4, characterized in that, The second preset sampling frequency is 12 kHz or 48 kHz.

6. A fault diagnostic system for gears and bearings in a transmission gearbox, characterized in that The data processing module, the training module, and the fault diagnosis module are included. The data processing module is configured to acquire a fault data set of a bearing and a compound fault data set of a gear in a transmission gear box, and perform data processing. The training module is configured to train a first deep learning model based on the processed fault data set of the bearing to obtain a bearing fault diagnosis model, and train a second deep learning model based on the processed compound fault data set to obtain a gear fault diagnosis model. The fault diagnosis module is configured to acquire running data of a bearing to be diagnosed in a preset transmission gear box, and determine a fault mode of the bearing to be diagnosed by using the bearing fault diagnosis model, and acquire running data of a gear to be diagnosed in the preset transmission gear box, and determine a fault mode of the gear to be diagnosed by using the gear fault diagnosis model. The fault data set of the bearing includes multi-class fault data acquired at a first preset sampling frequency, and the process of performing data processing on the fault data set of the bearing includes: From the fault data set of the bearing, a vibration signal corresponding to each fault mode of the bearing is acquired, the vibration signal is defined as a first time sequence, FFT transformation is used to obtain a first frequency domain signal corresponding to each first time sequence, a one-hot label is added to each first frequency domain signal to obtain the processed fault data set. The compound fault data set of the gear includes multi-class fault data acquired at a second preset sampling frequency, and the process of performing data processing on the compound fault data set of the gear includes: From the compound fault data set of the gear, a vibration signal corresponding to each fault mode of the gear is acquired, the vibration signal is defined as a second time sequence, FFT transformation is used to obtain a second frequency domain signal corresponding to each second time sequence, a one-hot label is added to each second frequency domain signal to obtain the processed compound fault data set. The first deep learning model and the second deep learning model are both self-attention MLP networks, and the self-attention MLP network includes an input layer, a self-attention layer, a flatten layer, a first full connection layer, a second full connection layer, and an output layer arranged in sequence, wherein the number of parameters of the self-attention layer is 3, the number of parameters of the first full connection layer is 364*364, and the number of parameters of the second full connection layer is 375*95.

7. An electronic device comprising a memory, a processor, and a program stored on the memory and running on the processor, characterized in that, The processor executes the program to implement the steps of the fault diagnosis method of the gear and the bearing in the transmission gear box according to any one of claims 1 to 5.

Images