A rolling bearing fault diagnosis method and system based on deep learning

Abstract

The present invention discloses a rolling bearing fault diagnosis method and system based on deep learning, which relates to the field of bearing testing technology. The method includes training a deep learning model to obtain a trained learning model; determining multiple test operating condition information of the rolling bearing, and testing the rolling bearing respectively according to the multiple test operating condition information; inputting a vibration signal into the trained learning model after preprocessing and feature extraction, outputting the bearing fault category through the learning model, and obtaining historical fault diagnosis data, the historical fault diagnosis data including historical test operating condition information, vibration signals corresponding to the test operating condition information, and corresponding fault categories; and training the deep learning model based on the historical fault diagnosis data to obtain a trained learning model. The present invention sets multiple test operating condition information to obtain corresponding vibration signals, and can more comprehensively analyze the rolling bearing fault, avoiding misjudgment of the rolling bearing due to improper use.

Description

Technical Field

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

[0002] With the advancement of industrial automation and intelligentization, industrial equipment is moving toward high speed, heavy load, and high precision, placing higher demands on the performance and reliability of rolling bearings. A rolling bearing failure can cause equipment downtime, leading to production interruptions, economic losses, and safety hazards. Research into rolling bearing failure mechanisms continues to deepen, clarifying the characteristic manifestations of different fault types and severity in vibration signals. Bearing fault diagnosis requires obtaining bearing vibration signals. However, incomplete testing conditions can affect the deviation of rolling bearing vibration signals, making it difficult to improve the accuracy of fault diagnosis. Summary of the Invention

[0003] The purpose of the present invention is to provide a rolling bearing fault diagnosis method and system based on deep learning to address the shortcomings of the background technology.

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

[0005] Train the deep learning model to obtain a trained learning model;

[0006] Determine multiple test working condition information of the rolling bearing, test the rolling bearing respectively according to the multiple test working condition information, and obtain vibration signals;

[0007] After preprocessing and feature extraction, the vibration signal is input into the trained learning model, and the fault category of the bearing is output through the learning model.

[0008] In a preferred embodiment, the step of training the deep learning model to obtain a trained learning model includes:

[0009] Acquiring historical fault diagnosis data, wherein the historical fault diagnosis data includes historical test condition information, vibration signals corresponding to the test condition information, and corresponding fault categories;

[0010] The deep learning model is trained based on historical fault diagnosis data to obtain a trained learning model.

[0011] In a preferred embodiment, the training of the deep learning model based on historical fault diagnosis data to obtain a trained learning model includes:

[0012] Divide historical fault diagnosis data into training set and test set;

[0013] Use the training set to train the deep learning model, use the test set to evaluate the trained model, and output the deep learning model that meets the preset accuracy as the trained learning model.

[0014] In a preferred embodiment, the step of determining multiple test condition information of the rolling bearing, testing the rolling bearing respectively according to the multiple test condition information, and obtaining the vibration signal includes:

[0015] Determining multiple test operating condition information of the rolling bearing, wherein the test operating condition information includes operating conditions of different speeds and operating environments;

[0016] Corresponding test models are constructed according to multiple test conditions, and a test star disk is formulated according to the test model. The rolling bearing is tested according to the test star disk to obtain corresponding vibration information.

[0017] In a preferred embodiment, the steps of constructing corresponding test models according to the plurality of test condition information, formulating a test star chart according to the test models, testing the rolling bearing according to the test star chart, and obtaining corresponding vibration information include:

[0018] Constructing corresponding test models corresponding to the plurality of test working condition information, wherein the test model is a three-dimensional working model of the rolling bearing, and marking the test working condition information in the three-dimensional working model;

[0019] A test layout is formulated, wherein the test layout includes multiple test groups, control points of test condition information are set in the test groups, and corresponding control points in the multiple test groups are connected to obtain a test star disk.

[0020] In a preferred embodiment, the step of connecting corresponding control points in a plurality of test clusters to obtain a test star disk includes:

[0021] Arrange multiple test models, and set data planes corresponding to the multiple arranged test models;

[0022] On the data plane, control intervals corresponding to the number of test condition information types are set. The control intervals include speed control intervals and operating environment control intervals. Both speed control intervals and operating environment control intervals cover each test model. The control intervals are divided into test groups corresponding to each test model.

[0023] Mark the control parameters in the speed control range and the operating environment control range respectively and configure the corresponding control points;

[0024] According to the arrangement of the test models, the control points of the same test condition information type in multiple test models are connected in sequence, the connection lines between the control points are bound to the corresponding control intervals and the control parameters of the corresponding positions are recorded as the test star disk.

[0025] In a preferred embodiment, the step of inputting the vibration signal into a trained learning model after preprocessing and feature extraction, and outputting the bearing fault category through the learning model includes:

[0026] De-noising and standardizing the vibration information to obtain a pre-processed vibration signal, and extracting features from the vibration signal to obtain feature data;

[0027] The feature data is input into the trained learning model, and the fault category of the bearing is output through the learning model.

[0028] The present invention also provides a rolling bearing fault diagnosis system based on deep learning, comprising:

[0029] The training module is used to train the deep learning model to obtain a trained learning model;

[0030] An information acquisition module is connected to the training model and is used to determine multiple test working condition information of the rolling bearing, and test the rolling bearing according to the multiple test working condition information to obtain a vibration signal;

[0031] The analysis module is connected to the information acquisition module and is used to input the vibration signal into the trained learning model after preprocessing and feature extraction, and output the fault category of the bearing through the learning model.

[0032] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0033] The present invention sets multiple test condition information to obtain corresponding vibration signals, which can perform fault analysis on the rolling bearing more comprehensively and avoid misjudgment of the rolling bearing due to improper use. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0035] Figure 1 Flow chart of the method of the present invention.

[0036] Figure 2 This is a system block diagram of the present invention. DETAILED DESCRIPTION

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in 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. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0038] Example 1, please refer to Figure 1 As shown, the rolling bearing fault diagnosis method based on deep learning described in this embodiment includes the following steps:

[0039] S1. Train the deep learning model to obtain a trained learning model;

[0040] S2. Determine multiple test condition information of the rolling bearing, and test the rolling bearing according to the multiple test condition information to obtain a vibration signal;

[0041] S3. After preprocessing and feature extraction, the vibration signal is input into the trained learning model, and the fault category of the bearing is output through the learning model.

[0042] In one embodiment, the step S1 of training the deep learning model to obtain a trained learning model includes:

[0043] S11. Acquire historical fault diagnosis data, wherein the historical fault diagnosis data includes historical test condition information, vibration signals corresponding to the test condition information, and corresponding fault categories;

[0044] S12. Train the deep learning model based on historical fault diagnosis data to obtain a trained learning model.

[0045] In one embodiment, the step of training the deep learning model according to the historical fault diagnosis data to obtain the trained learning model S12 includes:

[0046] S121, dividing the historical fault diagnosis data into a training set and a test set;

[0047] S122. Use the training set to train the deep learning model, use the test set to evaluate the trained model, and output the deep learning model that meets the preset accuracy as the trained learning model.

[0048] As described in steps S11 and S12 above, historical fault diagnosis data is obtained. This historical fault diagnosis data includes historical test condition information, the vibration signal corresponding to the test condition information, and the corresponding fault category. The vibration signal here is time-domain data and includes amplitude information: indicating the intensity of the vibration, such as peak value, mean value, and effective value. The peak value reflects the maximum amplitude during the vibration process and can be used to determine whether a sudden impact fault has occurred. The mean value is the average amplitude of the vibration signal over a period of time, reflecting the overall signal level. The effective value comprehensively considers the signal's amplitude and time distribution, providing a good representation of the vibration energy. Waveform characteristics: These include the shape and period of the vibration signal. During normal operation, the vibration waveform of a rolling bearing exhibits a certain regularity. When a fault occurs, the waveform will be distorted, such as spikes, jagged patterns, or periodic variations. These variations can help determine the type and severity of the fault. Time series: Records the changes in the vibration signal over time. By analyzing time series, it is possible to understand the dynamic process of vibration and capture instantaneous vibration changes, such as the instantaneous impact response when a fault occurs. Frequency domain data includes frequency components: By applying spectral analysis methods such as Fourier transform to the time domain vibration signal, the frequency composition of the signal is obtained. Faults in different parts of a rolling bearing will produce characteristic frequencies in specific frequency bands. For example, rolling element faults, inner race faults, and outer race faults all have corresponding characteristic frequencies. Identifying these characteristic frequencies can determine the location and type of the fault. The amplitude spectrum: This represents the amplitude corresponding to different frequency components and visually illustrates the distribution of vibration energy at each frequency. When a fault occurs, the amplitude of certain frequency components increases significantly. By observing changes in the amplitude spectrum, the development trend of the fault can be determined. Phase information: In the frequency domain, phase represents the relative time relationship between different frequency components. Although phase information is not as commonly used as amplitude information and frequency components in rolling bearing fault diagnosis, it can provide additional diagnostic clues in some complex fault diagnosis methods, such as those based on modal analysis. Time-frequency domain data includes time-frequency distribution: Because rolling bearing vibration signals are often non-stationary, their frequency components vary over time. Time-frequency analysis methods such as wavelet transform and short-time Fourier transform can jointly analyze signals in the time-frequency domain to produce a time-frequency distribution image. This image simultaneously displays both the time and frequency information of the signal, more comprehensively reflecting the characteristics of the vibration signal and being very effective in identifying time-varying fault characteristics. Time-frequency characteristic parameters: Certain characteristic parameters can be extracted from the time-frequency distribution, such as the center frequency, bandwidth, and entropy of the time-frequency energy distribution. These parameters can be used as feature vectors for rolling bearing fault diagnosis, with high diagnostic accuracy and robustness. Historical test condition information includes operating conditions at different speeds (rolling bearings) and operating environments (temperature and humidity of the rolling bearing operating environment).When the fault category is normal operation, the rolling bearing's components work well together, operating smoothly. The vibration signal is relatively regular and has a low amplitude. For example, in a normally operating motor, the rolling bearing supports the rotation of the motor shaft. The inner and outer rings, rolling elements, and cage of the bearing all move in unison, without abnormal friction, collision, or looseness. The resulting vibration signal is primarily caused by factors such as slight imbalance within the bearing itself and air turbulence caused by rotation. The vibration amplitude typically fluctuates within a relatively small range. Different types of fault conditions: Wear failure: Wear in rolling bearings is typically caused by the gradual wear of the surface material of the bearing components over long-term use, resulting in increased clearance and surface roughness. For example, in some long-term industrial equipment, such as conveyor belt motors in mines, the heavy loads borne by the bearings over time cause friction between the rolling elements and the inner and outer ring raceways, leading to wear grooves on the raceway surfaces. This changes the bearing's vibration signal, with the vibration amplitude gradually increasing with increasing wear. The frequency components of the signal also shift, potentially revealing characteristic frequencies associated with wear. Fatigue failure: Fatigue failure is a common type of failure in rolling element bearings. It is primarily caused by the formation and gradual propagation of fatigue cracks within the bearing material under long-term alternating loads. For example, in the main shaft bearings of wind turbines, due to wind instability, the bearings are subjected to complex alternating loads. After prolonged operation, fatigue cracks may develop in the inner or outer rings of the bearings. When cracks appear, the vibration signal exhibits a distinct impact signature, with characteristic frequencies associated with crack propagation appearing on the spectrum. The vibration amplitude gradually increases as the crack propagates. Lubrication failure: Lubrication is crucial for the proper operation of rolling element bearings. Inadequate lubrication or deterioration of the lubricant can increase friction between bearing components, leading to overheating and abnormal vibration. For example, in the crankshaft bearings of automobile engines, insufficient lubricant or deterioration due to long-term unreplacement can prevent the formation of a good oil film between the rolling elements and raceways of the bearing, resulting in direct metal-to-metal contact and increased friction. Consequently, the vibration signal exhibits increased high-frequency components, unstable amplitude, and may also be accompanied by elevated temperatures.

[0049] Fault states of varying severity: Early-stage faults: Taking minor wear on a rolling bearing as an example, in the early stages, the wear is relatively mild, perhaps manifested only by slight changes in the raceway surface roughness or a small number of tiny wear particles. At this stage, the vibration signal changes less noticeably, with a slight increase in amplitude and the emergence of characteristic frequencies. However, the signal-to-noise ratio is relatively low, requiring sophisticated signal processing methods to identify signs of the fault. For example, wavelet analysis or spectrum analysis of the vibration signal can reveal weak low-frequency characteristic frequencies associated with bearing wear. Mid-stage faults: When the fault progresses to the mid-stage, such as when wear worsens, wear grooves deepen on the raceway surface, or fatigue cracks extend to a certain length, the vibration signal changes more significantly, with a significant increase in amplitude, more prominent characteristic frequencies, and a higher signal-to-noise ratio. At this stage, conventional vibration monitoring methods, such as spectrum analysis and time-domain analysis, can readily identify fault characteristics. For example, characteristic frequency peaks associated with wear or fatigue can be clearly seen on the spectrum graph, and the amplitude of these peaks increases as the fault progresses. Late failure: In the late failure stage, the damage to the bearing is already very serious, such as rolling element rupture, severe wear of the inner or outer ring, or even fracture. At this time, the vibration signal will change dramatically, the amplitude will increase significantly, the signal will become very unstable, and a variety of complex frequency components may appear. The operating state of the equipment will also be seriously affected, and there may be obvious abnormal noise, increased vibration, or even equipment shutdown. For example, when the rolling element is broken, a strong impact pulse will appear in the vibration signal. The signal mutation can be clearly seen in the time domain, and in the frequency domain it is manifested as a wide-band energy distribution, with various frequency components mixed together. The historical fault diagnosis data is divided into a training set and a test set, and a machine learning model is constructed. The machine learning model is trained to obtain an initial machine learning model. The initial machine learning model is tested (evaluated) using the test set, and an initial machine learning model that meets the preset accuracy is output. The machine learning model is a logistic regression model; the historical fault diagnosis data is grouped and grouped to obtain a group number. The calculation formula for the prediction accuracy is ,in, is the group number of historical fault diagnosis data, For the Group prediction accuracy, For the The prediction number corresponding to the historical fault diagnosis data of the group, For the The actual serial numbers corresponding to the historical fault diagnosis data of the group can efficiently and accurately train the deep learning model, which is then convenient for using the deep learning model.

[0050] In one embodiment, the step S2 of determining multiple test condition information of the rolling bearing, testing the rolling bearing according to the multiple test condition information, and obtaining the vibration signal includes:

[0051] S21. Determine multiple test operating condition information of the rolling bearing, wherein the test operating condition information includes operating conditions of different speeds and operating environments;

[0052] S22, constructing corresponding test models according to the multiple test condition information, formulating a test star plate according to the test model, and testing the rolling bearing according to the test star plate to obtain corresponding vibration information;

[0053] In one embodiment, the step S22 of constructing corresponding test models according to the plurality of test condition information, formulating a test star chart according to the test models, testing the rolling bearing according to the test star chart, and obtaining corresponding vibration information includes:

[0054] S221, constructing corresponding test models for each of the plurality of test condition information, wherein the test model is a three-dimensional working model of the rolling bearing, and marking the test condition information in the three-dimensional working model;

[0055] S222: Develop a test layout, wherein the test layout includes multiple test groups, set control points of test condition information in the test groups, and connect corresponding control points in the multiple test groups to obtain a test star chart;

[0056] In one embodiment, the step S222 of connecting corresponding control points in a plurality of test clusters to obtain a test star disk includes:

[0057] S2221. Arrange multiple test models, and set data planes corresponding to the multiple arranged test models;

[0058] S2222. Control intervals corresponding to the number of test condition information types are set on the data plane. The control intervals include a speed control interval and an operating environment control interval. Both the speed control interval and the operating environment control interval cover each test model. The control intervals are divided into test groups corresponding to each test model.

[0059] S2223. Mark the control parameters in the speed control interval and the operating environment control interval respectively and configure corresponding control points;

[0060] S2224. Connect the control points of the same test condition information type in multiple test models in sequence according to the arrangement of the test models, bind the lines between the control points to the corresponding control intervals and record the control parameters of the corresponding positions as a test star disk.

[0061] As described in the above steps S1-S3, when diagnosing rolling bearing faults, in order to ensure the comprehensiveness of the collected data and avoid the influence of the test working condition information that leads to misjudgment of the fault, multiple test variables are set here. The working conditions of different speeds and operating environments are used as test working condition information, that is, test variables. A test working condition information has a certain speed and working condition of the operating environment. Multiple test working condition information is set here for collecting data, because the failure reaction of the bearing caused by different working conditions may be a problem of the test variables during use, rather than a problem of the rolling bearing itself. According to multiple test condition information, corresponding test models are constructed respectively, and test star disks are formulated according to the test models. The rolling bearings are tested according to the test star disks to obtain corresponding vibration information. The rolling bearings can be tested efficiently and continuously, thereby improving the test efficiency and continuity. Specifically, corresponding test models are constructed respectively for multiple test condition information, wherein the test model is a three-dimensional working model of the rolling bearing, and the test condition information is marked in the three-dimensional working model. First, the three-dimensional working model of the rolling bearing is set. The three-dimensional working model here is the three-dimensional model of the rolling bearing and the working environment model. For example, the rolling bearing is set inside the motor and can be constructed in the corresponding use scenario. It exists as a carrier for test information collection and control. After obtaining multiple test models, in order to better perform continuous testing and collection on multiple test models, it is necessary to configure a data plane for the corresponding test models. The data plane here is a cloud server. Multiple test models are set in the cloud server. The processing space in the cloud server here exists as a data plane. The data plane is a data space. The test model can be stored inside the data space. The control parameters are marked in the corresponding data space. The control parameters are marked in sequence from small to large. Control Parameters are bound to the data space through information points. A corresponding information point is set at a position corresponding to a control parameter mark in the data space. The information point is used to store the corresponding control parameter (whether it is a speed control parameter or an environment control parameter). The information point is a virtual machine used to store and bind control parameters. A control point parameter information can be directly provided later to complete the control of the corresponding controller. A control interval corresponding to the number of test condition information types is set on the data surface. The control interval includes a speed control interval and an environment control interval. The speed control interval and the environment control interval both cover each test model. The control interval is divided into a test group corresponding to each test model. The test group represents the control interval of the corresponding test model. The control parameters are marked and corresponding control points are configured in the speed control interval and the environment control interval respectively. The control point is used to correspond to the control parameter. The control point in the speed control interval is directly connected to the controller that controls the speed, and the control point in the environment control interval is directly connected to the controller that controls the environment parameter. This can improve the control responsiveness. The control points of the same test condition information type in multiple test models are connected in sequence according to the arrangement of the test models.Bind the connection between the control points and the corresponding control interval and record the control parameters of the corresponding position. The control point is an information port and can be moved on the data surface and connected to different information points. Specifically: multiple test models are sorted in order, and the arrangement rule is that the test parameters of the test model are sorted from low to high. This can make it more efficient to test multiple test models in sequence. In multiple test models, the control points of multiple test models in the speed control interval are connected in sequence; the control points of multiple test models in the use environment control interval are connected in sequence, wherein the connection between the control points in the control interval will exist on the data surface, and a test star disk is obtained. In subsequent use, when multiple test models are switched for testing, it is necessary to use the parameters on the data surface corresponding to the connection. Switching makes the switching between multiple test models more orderly and stable, and can automatically test multiple test models in sequence. Through the connection between the control point and the corresponding controller, the rolling bearing can be tested more efficiently according to the test model, ensuring the smoothness of the test, thereby improving the test efficiency, and better analyzing the fault category through the subsequent learning model. After that, the controller responsible for the speed and the controller responsible for the environmental parameters are controlled according to the test star disk to obtain the vibration information of the rolling bearing. Setting multiple test working condition information can obtain the corresponding vibration signal, which can conduct a more comprehensive fault analysis of the rolling bearing, avoid misjudgment of the rolling bearing due to improper use, and use the speed and environmental parameters as variables to comprehensively analyze the fault conditions of the rolling bearing under multiple working conditions.

[0062] In one embodiment, the step S3 of inputting the vibration signal into a trained learning model after preprocessing and feature extraction, and outputting the bearing fault category through the learning model includes:

[0063] S31, performing denoising and standardization processing on the vibration information to obtain a pre-processed vibration signal, and extracting features from the vibration signal to obtain feature data;

[0064] S32. Input the feature data into the trained learning model, and output the bearing fault category through the learning model;

[0065] As described in the above steps S31 and S32, the collected vibration signal is denoised, for example, by using methods such as wavelet denoising to remove noise interference in the signal and improve the quality of the signal. Data normalization or standardization is performed, and the data is mapped to a certain range to facilitate the training and convergence of the neural network. The vibration information is denoised and standardized to obtain a vibration signal after preprocessing. The features in the vibration signal are extracted to obtain feature data, where the feature data includes time domain data, frequency domain data and time-frequency domain data. The feature information is then used as input and analyzed through a trained learning model to obtain a fault category. The environment and speed of the rolling bearing are used as variables, so that the rolling bearing fault can be analyzed in various situations with good analysis results.

[0066] Example 2, please refer to Figure 2 As shown, the rolling bearing fault diagnosis system based on deep learning described in this embodiment includes:

[0067] The training module is used to train the deep learning model to obtain a trained learning model;

[0068] An information acquisition module is connected to the training model and is used to determine multiple test working condition information of the rolling bearing, and test the rolling bearing according to the multiple test working condition information to obtain a vibration signal;

[0069] The analysis module is connected to the information acquisition module and is used to input the vibration signal into the trained learning model after preprocessing and feature extraction, and output the fault category of the bearing through the learning model.

[0070] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A rolling bearing fault diagnosis method based on deep learning, characterized in that: The following steps are involved: Train the deep learning model to obtain a trained learning model; Determine multiple test working condition information of the rolling bearing, test the rolling bearing respectively according to the multiple test working condition information, and obtain vibration signals; The step of determining multiple test condition information of the rolling bearing, testing the rolling bearing respectively according to the multiple test condition information, and obtaining the vibration signal includes: Determining multiple test operating condition information of the rolling bearing, wherein the test operating condition information includes operating conditions of different speeds and operating environments; Build corresponding test models according to multiple test conditions, formulate test star charts based on the test models, test rolling bearings according to the test star charts, and obtain corresponding vibration information; The steps of constructing corresponding test models according to the plurality of test condition information, formulating a test star plate according to the test models, testing the rolling bearing according to the test star plate, and obtaining corresponding vibration information include: Constructing corresponding test models corresponding to the plurality of test working condition information, wherein the test model is a three-dimensional working model of the rolling bearing, and marking the test working condition information in the three-dimensional working model; Formulate a test layout, wherein the test layout includes multiple test groups, set control points of test working condition information in the test groups, and connect corresponding control points in the multiple test groups to obtain a test star chart; The step of connecting corresponding control points in a plurality of test clusters to obtain a test star chart includes: Arrange multiple test models, and set data planes corresponding to the multiple arranged test models; On the data plane, control intervals corresponding to the number of test condition information types are set. The control intervals include speed control intervals and operating environment control intervals. Both speed control intervals and operating environment control intervals cover each test model. The control intervals are divided into test groups corresponding to each test model. Mark the control parameters in the speed control range and the operating environment control range respectively and configure corresponding control points. In the data plane, a corresponding information point is set at the position corresponding to a control parameter mark. The information point is used to store the corresponding control parameter. The control point can be moved on the data plane and connected to different information points. According to the arrangement of the test models, the control points of the same test condition information type in multiple test models are connected in sequence, the connecting lines between the control points are bound to the corresponding control intervals, and the control parameters of the corresponding positions are recorded as the test star chart; After preprocessing and feature extraction, the vibration signal is input into the trained learning model, and the fault category of the bearing is output through the learning model.

2. The rolling bearing fault diagnosis method based on deep learning according to claim 1, characterized in that: The step of training the deep learning model to obtain a trained learning model includes: Acquiring historical fault diagnosis data, wherein the historical fault diagnosis data includes historical test condition information, vibration signals corresponding to the test condition information, and corresponding fault categories; The deep learning model is trained based on historical fault diagnosis data to obtain a trained learning model.

3. The rolling bearing fault diagnosis method based on deep learning according to claim 2, characterized in that: The deep learning model is trained according to the historical fault diagnosis data to obtain a trained learning model, including: Divide historical fault diagnosis data into training set and test set; Use the training set to train the deep learning model, use the test set to evaluate the trained model, and output the deep learning model that meets the preset accuracy as the trained learning model.

4. The rolling bearing fault diagnosis method based on deep learning according to claim 1, characterized in that: The step of inputting the vibration signal into a trained learning model after preprocessing and feature extraction, and outputting the bearing fault category through the learning model includes: De-noising and standardizing the vibration information to obtain a pre-processed vibration signal, and extracting features from the vibration signal to obtain feature data; The feature data is input into the trained learning model, and the fault category of the bearing is output through the learning model.

5. A rolling bearing fault diagnosis system based on deep learning, used to implement the rolling bearing fault diagnosis method based on deep learning according to any one of claims 1 to 4, characterized in that: include: The training module is used to train the deep learning model to obtain a trained learning model; An information acquisition module is connected to the training model and is used to determine multiple test working condition information of the rolling bearing, and test the rolling bearing according to the multiple test working condition information to obtain a vibration signal; The analysis module is connected to the information acquisition module and is used to input the vibration signal into the trained learning model after preprocessing and feature extraction, and output the fault category of the bearing through the learning model.

Images