Rolling bearing fault diagnosis method and system based on deep learning

Abstract

The invention particularly relates to a rolling bearing fault diagnosis method based on deep learning. The method comprises the following steps: inputting rolling bearing inspection data; the rolling bearing actual operation vibration signal in the recorded rolling bearing inspection data is automatically converted into a rolling bearing actual operation vibration signal time-frequency diagram; and the rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis according to the actual operation vibration signal time-frequency diagram of the rolling bearing. The invention further relates to a rolling bearing fault diagnosis system based on deep learning. Based on a rolling bearing fault mechanism and simulation fault data, rolling bearing fault intelligent diagnosis is carried out.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis of rotating machinery faults, and in particular to a rolling bearing fault diagnosis method and system based on deep learning. Background Art

[0002] Rolling bearings are widely used in various types of rotating machinery, and the state of rolling bearings is related to the safe and stable operation of the equipment. In the nuclear power field, which pays more attention to safety and reliability, rolling bearings are usually replaced before they are damaged. In actual operation, there is a lack of fault samples, and it is impossible to directly carry out intelligent diagnosis of bearing faults based on deep learning neural network technology. It can only rely on expert experience to carry out spectrum analysis and use characteristic frequencies to identify bearing faults. This method is inefficient and its accuracy is also heavily dependent on expert experience. Summary of the invention

[0003] Based on this, it is necessary to provide a rolling bearing fault diagnosis method and system based on deep learning to address the problem of lack of fault samples of rolling bearings in the nuclear power field and inability to use intelligent technologies such as deep learning to carry out intelligent diagnosis of bearing faults. Based on the rolling bearing fault mechanism and simulated fault data, intelligent diagnosis of rolling bearing faults is carried out.

[0004] In order to solve the above problems, the present invention provides a rolling bearing fault diagnosis method based on deep learning, comprising the following steps:

[0005] Enter rolling bearing inspection data;

[0006] Automatically convert the actual running vibration signal of the rolling bearing in the input rolling bearing inspection data into a time-frequency diagram of the actual running vibration signal of the rolling bearing;

[0007] The rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis according to the time-frequency diagram of the actual operation vibration signal of the rolling bearing.

[0008] In the present invention, a rolling bearing fault diagnosis model based on deep learning is established through the following steps:

[0009] Obtain the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation and the time-frequency diagram of the vibration signal of the rolling bearing in historical normal operation;

[0010] Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal are fused to generate the time-frequency diagram of the rolling bearing fault operation vibration signal;

[0011] The time-frequency diagram of the rolling bearing fault operation vibration signal is used as the rolling bearing fault sample to train the deep learning neural network, and a rolling bearing fault diagnosis model based on deep learning is established.

[0012] In the present invention, obtaining the time-frequency diagram of the vibration signal of the rolling bearing during normal operation includes the following steps:

[0013] Automatically collect historical inspection data of rolling bearings;

[0014] Filter out the historical normal operation vibration signal of the rolling bearing from the historical inspection data of the rolling bearing;

[0015] Automatically convert the rolling bearing's historical normal operation vibration signal into a time-frequency diagram of the rolling bearing's historical normal operation vibration signal.

[0016] In the present invention, the historical normal operation vibration signal of the rolling bearing is screened out from the historical inspection data of the rolling bearing, which includes the following steps:

[0017] Expert experience is used to mark the historical normal operation vibration signals of rolling bearings on the historical inspection data of rolling bearings, and the historical normal operation vibration signals of rolling bearings are screened out from the historical inspection data of rolling bearings.

[0018] In the present invention, obtaining the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation includes the following steps:

[0019] Generate rolling bearing fault operation vibration simulation signal;

[0020] Automatically convert the rolling bearing fault operation vibration simulation signal into the rolling bearing fault operation vibration simulation signal time-frequency diagram.

[0021] In the present invention, generating a rolling bearing fault operation vibration simulation signal comprises the following steps:

[0022] Set rolling bearing models, rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings;

[0023] According to the set rolling bearing model, a rolling bearing three-dimensional model is automatically generated; according to the set rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings, a rolling bearing fault operation vibration simulation signal is automatically generated.

[0024] In the present invention, a transfer learning algorithm is used to fuse the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal to generate a time-frequency diagram of the rolling bearing fault operation vibration signal, including the following steps:

[0025] Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing's historical normal operation vibration signal are fused, and the background noise of the time-frequency diagram of the rolling bearing's historical normal operation vibration signal is transferred to the time-frequency diagram of the rolling bearing fault operation vibration simulation signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal close to the actual operating state.

[0026] In the present invention, the actual operation vibration signal of the rolling bearing, the historical normal operation vibration signal of the rolling bearing and the vibration simulation signal of the rolling bearing fault operation are all rolling bearing vibration signals, and the conversion of the rolling bearing vibration signal time-frequency diagram includes the following steps:

[0027] Perform at least M layers of wavelet packet decomposition on the rolling bearing vibration signal, and decompose the rolling bearing vibration signal with a sampling rate of 2f and a length of N into components of M frequency bands;

[0028] Calculate the energy of the signal in each frequency band;

[0029] The energy of different frequency bands is sorted, and the components with energy proportion exceeding γ are retained. The changes of these components over time are assembled into a time-frequency diagram to obtain the time-frequency diagram of the rolling bearing vibration signal after noise reduction.

[0030] In order to solve the above problems, the present invention also provides a rolling bearing fault diagnosis system based on deep learning, comprising:

[0031] The rolling bearing inspection data management and analysis module is equipped with a rolling bearing fault diagnosis model based on deep learning, which is used to input rolling bearing inspection data. The rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis based on the corresponding rolling bearing actual operation vibration signal time-frequency diagram; automatically collects rolling bearing historical inspection data, and screens out the rolling bearing historical normal operation vibration signal from the rolling bearing historical inspection data; manually confirms and automatically marks whether the diagnosis result of the rolling bearing actual operation vibration signal time-frequency diagram corresponding to the rolling bearing inspection data is accurate, and stores the rolling bearing actual operation vibration signal time-frequency diagram corresponding to the rolling bearing inspection data and its diagnosis result as a rolling bearing test sample;

[0032] The rolling bearing vibration signal time-frequency diagram conversion module is used to automatically convert the rolling bearing actual operation vibration signal in the recorded rolling bearing inspection data into the rolling bearing actual operation vibration signal time-frequency diagram; automatically convert the rolling bearing historical normal operation vibration signal into the rolling bearing historical normal operation vibration signal time-frequency diagram; automatically convert the rolling bearing fault operation vibration simulation signal into the rolling bearing fault operation vibration simulation signal time-frequency diagram;

[0033] Rolling bearing three-dimensional dynamics simulation program, used to generate rolling bearing fault operation vibration simulation signal;

[0034] The rolling bearing fault diagnosis model management module is equipped with a deep learning neural network, which is used to train the deep learning neural network by using the time-frequency diagram of the rolling bearing fault operation vibration signal as the rolling bearing fault sample, and establish a rolling bearing fault diagnosis model based on deep learning; it automatically optimizes the learning-based rolling bearing fault diagnosis model regularly according to the rolling bearing test samples;

[0035] The rolling bearing fault sample generation module is equipped with a transfer learning algorithm, which is used to use the transfer learning algorithm to fuse the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal that is close to the actual operating state.

[0036] Beneficial technical effects of the present invention:

[0037] The rolling bearing fault diagnosis method and system based on deep learning of the present invention have a unified rolling bearing structure and a clear fault mechanism, and fully utilize the rolling bearing fault mechanism simulation method to generate a rolling bearing fault operation vibration simulation signal; the rolling bearing fault operation vibration simulation signal and the rolling bearing historical normal operation vibration signal obtained through the rolling bearing historical inspection are converted into a video image; the rolling bearing fault operation vibration simulation signal time-frequency graph and the rolling bearing historical normal operation vibration signal time-frequency graph are merged by a transfer learning algorithm to generate a rolling bearing fault operation vibration signal time-frequency graph close to the actual operation state; the rolling bearing fault operation vibration signal time-frequency graph is used as a rolling bearing fault sample to train a deep learning neural network, and a rolling bearing fault diagnosis model based on deep learning is established, which realizes the rapid diagnosis of rolling bearing faults during the rolling bearing inspection process, solves the problem that the development and training of the intelligent fault diagnosis model cannot be carried out due to the lack of rolling bearing fault samples, effectively reduces the dependence on rolling bearing fault samples, and expands the source of rolling bearing fault samples; rolling bearing test samples are continuously accumulated during the rolling bearing inspection process, and the learning-based rolling bearing fault diagnosis model is regularly optimized according to the rolling bearing test samples, thereby improving the accuracy of the learning-based rolling bearing fault diagnosis model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of an embodiment of a rolling bearing fault diagnosis method based on deep learning of the present invention;

[0039] Figure 2 This is a schematic structural view of an embodiment of the rolling bearing fault diagnosis system based on deep learning of the present invention. DETAILED DESCRIPTION

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used in the specification of the application are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0041] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0042] The technical solution of the present invention is clearly and completely described below in conjunction with the accompanying drawings and specific embodiments.

[0043] refer to Figure 1 , a flow chart of an embodiment of a rolling bearing fault diagnosis method based on deep learning is shown, and the rolling bearing fault diagnosis method based on deep learning comprises the following steps:

[0044] Enter rolling bearing inspection data;

[0045] Automatically convert the actual running vibration signal of the rolling bearing in the input rolling bearing inspection data into a time-frequency diagram of the actual running vibration signal of the rolling bearing;

[0046] The rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis according to the time-frequency diagram of the actual operation vibration signal of the rolling bearing.

[0047] In this embodiment, the rolling bearing fault diagnosis model based on deep learning is established through the following steps:

[0048] Obtain the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation and the time-frequency diagram of the vibration signal of the rolling bearing in historical normal operation;

[0049] Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal are fused to generate the time-frequency diagram of the rolling bearing fault operation vibration signal;

[0050] The time-frequency diagram of the rolling bearing fault operation vibration signal is used as the rolling bearing fault sample to train the deep learning neural network, and a rolling bearing fault diagnosis model based on deep learning is established.

[0051] In this embodiment, obtaining the time-frequency diagram of the rolling bearing's historical normal operation vibration signal includes the following steps:

[0052] Automatically collect historical inspection data of rolling bearings;

[0053] Filter out the historical normal operation vibration signal of the rolling bearing from the historical inspection data of the rolling bearing;

[0054] Automatically convert the rolling bearing's historical normal operation vibration signal into a time-frequency diagram of the rolling bearing's historical normal operation vibration signal.

[0055] In this embodiment, filtering out the historical normal operation vibration signal of the rolling bearing from the historical inspection data of the rolling bearing includes the following steps:

[0056] Expert experience is used to mark the historical normal operation vibration signals of rolling bearings on the historical inspection data of rolling bearings, and the historical normal operation vibration signals of rolling bearings are screened out from the historical inspection data of rolling bearings.

[0057] In this embodiment, obtaining the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation includes the following steps:

[0058] Generate rolling bearing fault operation vibration simulation signal;

[0059] Automatically convert the rolling bearing fault operation vibration simulation signal into the rolling bearing fault operation vibration simulation signal time-frequency diagram.

[0060] In this embodiment, generating a rolling bearing fault operation vibration simulation signal includes the following steps:

[0061] Set rolling bearing models, rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings;

[0062] Automatically generate a rolling bearing three-dimensional model according to the set rolling bearing model; automatically generate a rolling bearing fault operation vibration simulation signal according to the set rolling bearing operating conditions and typical rolling bearing faults of different types and degrees;

[0063] Rolling bearing operating conditions include speed and load.

[0064] In this embodiment, a transfer learning algorithm is used to fuse the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal to generate a time-frequency diagram of the rolling bearing fault operation vibration signal, including the following steps:

[0065] Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing's historical normal operation vibration signal are fused, and the background noise of the time-frequency diagram of the rolling bearing's historical normal operation vibration signal is transferred to the time-frequency diagram of the rolling bearing fault operation vibration simulation signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal close to the actual operating state.

[0066] In this embodiment, the transfer learning algorithm is the CycleGAN transfer learning algorithm.

[0067] In this embodiment, the rolling bearing inspection data is manually entered in order, and the rolling bearing fault diagnosis model based on deep learning is used to perform rolling bearing fault diagnosis according to the corresponding rolling bearing actual operation vibration signal time-frequency diagram;

[0068] Manually confirm whether the diagnosis result of the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data is accurate, automatically mark whether the diagnosis result of the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data is accurate, and store the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data and its diagnosis result as a rolling bearing test sample;

[0069] Automatically optimize the rolling bearing fault diagnosis model based on learning based on rolling bearing test samples regularly.

[0070] In this embodiment, the actual operation vibration signal of the rolling bearing, the historical normal operation vibration signal of the rolling bearing and the fault operation vibration simulation signal of the rolling bearing are all rolling bearing vibration signals, and the conversion of the rolling bearing vibration signal time-frequency diagram includes the following steps:

[0071] Perform at least M layers of wavelet packet decomposition on the rolling bearing vibration signal, and decompose the rolling bearing vibration signal with a sampling rate of 2f and a length of N into components of M frequency bands;

[0072] Calculate the energy of the signal in each frequency band;

[0073] The energy of different frequency bands is sorted, and the components with energy proportion exceeding γ are retained. The changes of these components over time are assembled into a time-frequency diagram to obtain the time-frequency diagram of the rolling bearing vibration signal after noise reduction.

[0074] In this embodiment, taking M as 3 as an example, the signals in the frequency band include [0, f / 8), [f / 8, f / 4), [f / 4, 3f / 8), [3f / 8, f / 2), [f / 2, 5f / 8), [5f / 8, 3f / 4), [3f / 4, 7f / 8) and [7f / 8, f], and the length of the signal in the frequency band is N / 8; the energy of the signal in the frequency band is calculated according to the following formula:

[0075]

[0076] Where E represents the energy of the signal within the frequency band; x i Indicates the amplitude of the signal within the frequency band.

[0077] In this embodiment, γ is set to 0.2.

[0078] refer to Figure 2 As an implementation of the above method, the present invention provides an embodiment of a rolling bearing fault diagnosis system based on deep learning. The system embodiment corresponds to the embodiment of the above rolling bearing fault diagnosis method based on deep learning. The system can be specifically applied to various electronic devices.

[0079] The rolling bearing fault diagnosis system based on deep learning described in this embodiment includes:

[0080] The rolling bearing inspection data management and analysis module is equipped with a rolling bearing fault diagnosis model based on deep learning, which is used to manually enter the rolling bearing inspection data in order; the rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis based on the corresponding rolling bearing actual operation vibration signal time-frequency diagram;

[0081] The rolling bearing vibration signal time-frequency diagram conversion module is used to automatically convert the rolling bearing actual operation vibration signal in the input rolling bearing inspection data into the rolling bearing actual operation vibration signal time-frequency diagram.

[0082] In this embodiment, the rolling bearing fault diagnosis system based on deep learning also includes: a rolling bearing fault diagnosis model management module, a rolling bearing fault sample generation module and a rolling bearing three-dimensional dynamics simulation program;

[0083] The rolling bearing inspection data management and analysis module is also used to automatically collect the historical inspection data of rolling bearings; use expert experience to mark the historical normal operation vibration signal of rolling bearings on the historical inspection data of rolling bearings, and filter out the historical normal operation vibration signal of rolling bearings from the historical inspection data of rolling bearings;

[0084] The rolling bearing vibration signal time-frequency diagram conversion module is also used to automatically convert the rolling bearing's historical normal operation vibration signal into the rolling bearing's historical normal operation vibration signal time-frequency diagram; automatically convert the rolling bearing's fault operation vibration simulation signal into the rolling bearing's fault operation vibration simulation signal time-frequency diagram;

[0085] The rolling bearing three-dimensional dynamics simulation program is used to set the rolling bearing model, rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings; according to the set rolling bearing model, the rolling bearing three-dimensional model is automatically generated; according to the set rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings, the rolling bearing failure operation vibration simulation signal is automatically generated;

[0086] The rolling bearing fault diagnosis model management module is provided with a deep learning neural network, which is used to train the deep learning neural network by using the time-frequency diagram of the rolling bearing fault operation vibration signal as a rolling bearing fault sample, and establish a rolling bearing fault diagnosis model based on deep learning;

[0087] The rolling bearing fault sample generation module is equipped with a transfer learning algorithm, which is used to use the transfer learning algorithm to fuse the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing's historical normal operation vibration signal, and to migrate the background noise of the time-frequency diagram of the rolling bearing's historical normal operation vibration signal to the time-frequency diagram of the rolling bearing fault operation vibration simulation signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal that is close to the actual operating state.

[0088] In this embodiment, the rolling bearing inspection data management and analysis module is also used to manually confirm whether the diagnosis result of the time-frequency diagram of the rolling bearing actual operation vibration signal corresponding to the rolling bearing inspection data is accurate, automatically mark whether the diagnosis result of the time-frequency diagram of the rolling bearing actual operation vibration signal corresponding to the rolling bearing inspection data is accurate, and store the time-frequency diagram of the rolling bearing actual operation vibration signal corresponding to the rolling bearing inspection data and its diagnosis result as a rolling bearing test sample;

[0089] The rolling bearing fault diagnosis model management module is also used to automatically optimize the learning-based rolling bearing fault diagnosis model regularly according to the rolling bearing test samples.

[0090] In this embodiment, the actual operation vibration signal of the rolling bearing, the historical normal operation vibration signal of the rolling bearing, and the fault operation vibration simulation signal of the rolling bearing are all rolling bearing vibration signals; the rolling bearing vibration signal time-frequency diagram conversion module converts the rolling bearing vibration signal time-frequency diagram, including the following steps:

[0091] Perform at least M layers of wavelet packet decomposition on the rolling bearing vibration signal, and decompose the rolling bearing vibration signal with a sampling rate of 2f and a length of N into components of M frequency bands;

[0092] Calculate the energy of the signal in each frequency band;

[0093] The energy of different frequency bands is sorted, and the components with energy proportion exceeding γ are retained. The changes of these components over time are assembled into a time-frequency diagram to obtain the time-frequency diagram of the rolling bearing vibration signal after noise reduction.

[0094] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A rolling bearing fault diagnosis method based on deep learning, characterized in that: The following steps are involved: Enter rolling bearing inspection data; Automatically convert the actual running vibration signal of the rolling bearing in the input rolling bearing inspection data into a time-frequency diagram of the actual running vibration signal of the rolling bearing; The rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis according to the time-frequency diagram of the actual operation vibration signal of the rolling bearing.

2. The rolling bearing fault diagnosis method based on deep learning according to claim 1 is characterized in that: The rolling bearing fault diagnosis model based on deep learning is established through the following steps: Obtain the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation and the time-frequency diagram of the vibration signal of the rolling bearing in historical normal operation; Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal are fused to generate the time-frequency diagram of the rolling bearing fault operation vibration signal; The time-frequency diagram of the rolling bearing fault operation vibration signal is used as the rolling bearing fault sample to train the deep learning neural network, and a rolling bearing fault diagnosis model based on deep learning is established.

3. The rolling bearing fault diagnosis method based on deep learning according to claim 2 is characterized in that: Obtaining the time-frequency diagram of the rolling bearing's historical normal operation vibration signal includes the following steps: Automatically collect historical inspection data of rolling bearings; Filter out the historical normal operation vibration signal of the rolling bearing from the historical inspection data of the rolling bearing; Automatically convert the rolling bearing's historical normal operation vibration signal into a time-frequency diagram of the rolling bearing's historical normal operation vibration signal.

4. The rolling bearing fault diagnosis method based on deep learning according to claim 3 is characterized in that: Screening out the historical normal operation vibration signal of the rolling bearing from the historical inspection data of the rolling bearing includes the following steps: Expert experience is used to mark the historical normal operation vibration signals of rolling bearings on the historical inspection data of rolling bearings, and the historical normal operation vibration signals of rolling bearings are screened out from the historical inspection data of rolling bearings.

5. The rolling bearing fault diagnosis method based on deep learning according to claim 2 is characterized in that: Obtaining the time-frequency diagram of the vibration simulation signal of the rolling bearing fault operation includes the following steps: Generate rolling bearing fault operation vibration simulation signal; Automatically convert the rolling bearing fault operation vibration simulation signal into the rolling bearing fault operation vibration simulation signal time-frequency diagram.

6. The rolling bearing fault diagnosis method based on deep learning according to claim 5 is characterized in that: Generating a rolling bearing fault operation vibration simulation signal includes the following steps: Set rolling bearing models, rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings; According to the set rolling bearing model, a rolling bearing three-dimensional model is automatically generated; according to the set rolling bearing operating conditions and typical failures of different types and degrees of rolling bearings, a rolling bearing fault operation vibration simulation signal is automatically generated.

7. The rolling bearing fault diagnosis method based on deep learning according to claim 2 is characterized in that: Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal are fused to generate the time-frequency diagram of the rolling bearing fault operation vibration signal, including the following steps: Using the transfer learning algorithm, the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing's historical normal operation vibration signal are fused, and the background noise of the time-frequency diagram of the rolling bearing's historical normal operation vibration signal is transferred to the time-frequency diagram of the rolling bearing fault operation vibration simulation signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal close to the actual operating state.

8. The rolling bearing fault diagnosis method based on deep learning according to claim 2 is characterized in that: The actual running vibration signal of the rolling bearing, the historical normal running vibration signal of the rolling bearing and the vibration simulation signal of the rolling bearing fault running are all rolling bearing vibration signals. The conversion of the time-frequency diagram of the rolling bearing vibration signal includes the following steps: Perform at least M layers of wavelet packet decomposition on the rolling bearing vibration signal, and decompose the rolling bearing vibration signal with a sampling rate of 2f and a length of N into components of M frequency bands; Calculate the energy of the signal in each frequency band; The energy of different frequency bands is sorted, and the components with energy proportion exceeding γ are retained. The changes of these components over time are assembled into a time-frequency diagram to obtain the time-frequency diagram of the rolling bearing vibration signal after noise reduction.

9. The rolling bearing fault diagnosis method based on deep learning according to claim 1, characterized in that: Manually input rolling bearing inspection data one by one, and use the rolling bearing fault diagnosis model based on deep learning to perform rolling bearing fault diagnosis according to the corresponding rolling bearing actual operation vibration signal time-frequency diagram; Manually confirm whether the diagnosis result of the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data is accurate, automatically mark whether the diagnosis result of the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data is accurate, and store the time-frequency diagram of the actual operation vibration signal of the rolling bearing corresponding to the rolling bearing inspection data and its diagnosis result as a rolling bearing test sample; Automatically optimize the rolling bearing fault diagnosis model based on learning based on rolling bearing test samples regularly.

10. A rolling bearing fault diagnosis system based on deep learning, characterized in that: include: The rolling bearing inspection data management and analysis module is equipped with a rolling bearing fault diagnosis model based on deep learning, which is used to input rolling bearing inspection data. The rolling bearing fault diagnosis model based on deep learning performs rolling bearing fault diagnosis based on the corresponding rolling bearing actual operation vibration signal time-frequency diagram; automatically collects rolling bearing historical inspection data, and screens out the rolling bearing historical normal operation vibration signal from the rolling bearing historical inspection data; manually confirms and automatically marks whether the diagnosis result of the rolling bearing actual operation vibration signal time-frequency diagram corresponding to the rolling bearing inspection data is accurate, and stores the rolling bearing actual operation vibration signal time-frequency diagram corresponding to the rolling bearing inspection data and its diagnosis result as a rolling bearing test sample; The rolling bearing vibration signal time-frequency diagram conversion module is used to automatically convert the rolling bearing actual operation vibration signal in the recorded rolling bearing inspection data into the rolling bearing actual operation vibration signal time-frequency diagram; automatically convert the rolling bearing historical normal operation vibration signal into the rolling bearing historical normal operation vibration signal time-frequency diagram; automatically convert the rolling bearing fault operation vibration simulation signal into the rolling bearing fault operation vibration simulation signal time-frequency diagram; Rolling bearing three-dimensional dynamics simulation program, used to generate rolling bearing fault operation vibration simulation signal; The rolling bearing fault diagnosis model management module is equipped with a deep learning neural network, which is used to train the deep learning neural network by using the time-frequency diagram of the rolling bearing fault operation vibration signal as the rolling bearing fault sample, and establish a rolling bearing fault diagnosis model based on deep learning; it automatically optimizes the learning-based rolling bearing fault diagnosis model regularly according to the rolling bearing test samples; The rolling bearing fault sample generation module is equipped with a transfer learning algorithm, which is used to use the transfer learning algorithm to fuse the time-frequency diagram of the rolling bearing fault operation vibration simulation signal and the time-frequency diagram of the rolling bearing historical normal operation vibration signal, so as to generate a time-frequency diagram of the rolling bearing fault operation vibration signal that is close to the actual operating state.

Images