Construction method, system and equipment of sliding bearing fault diagnosis model and medium

Abstract

The invention belongs to the technical field of part detection, and relates to a sliding bearing fault diagnosis model construction method and system, equipment and a medium. The method comprises the following steps: acquiring a sound signal of a bearing; preprocessing the sound signal of the bearing to obtain a training data set; adopting a feature selection method to obtain bearing fault feature data from the training data set; and performing model training on the training data set and the bearing fault feature data by adopting an online transfer learning method to obtain a sliding bearing fault diagnosis model. And after the sliding bearing fault diagnosis model is obtained, the sound signal of the bearing is input into the sliding bearing fault diagnosis model for sliding bearing fault diagnosis. The bearing does not need to be in direct contact, damage to the bearing is avoided, and meanwhile the maintenance cost and risk are reduced. In addition, compared with an existing fault diagnosis method, the machine learning model can process complex data features more effectively, and the accuracy of fault diagnosis is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of parts detection, and relates to a method, system, equipment and medium for constructing a sliding bearing fault diagnosis model. Background Art

[0002] In modern industrial processes, early fault detection is crucial, which can not only avoid potential safety hazards, but also reduce maintenance costs and machine downtime. Sliding bearing fault detection mainly focuses on the condition of the bearing itself, as well as the friction, wear and lubrication between the bearing and other mechanical parts, such as the friction, wear and lubrication between the bearing and the shaft, housing, etc.

[0003] Contact detection methods usually require the monitoring element (such as a sensor, probe, etc.) to be in physical contact with the monitored object (such as the bearing surface) to obtain the required monitoring data. In this process, the bearing surface may be damaged due to contact force, friction, wear and other factors.

[0004] As a non-contact detection method, acoustic emission detection technology can achieve non-destructive measurement by detecting the longitudinal waves, transverse waves and surface waves generated by bearing faults, so it has a wide range of applications. The source data obtained by acoustic emission detection technology can be processed by deep learning methods and applied to other types of sliding bearing fault diagnosis, thereby achieving a wider range of fault identification.

[0005] However, existing acoustic signal bearing fault diagnosis technologies based on non-traditional feature extraction and deep learning methods have problems such as poor accuracy.

[0006] In summary, the existing sliding bearing fault diagnosis method has the problem of low accuracy, and the contact detection method has the problem of easily damaging the sliding bearing. Summary of the invention

[0007] In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a method, system, device and medium for constructing a sliding bearing fault diagnosis model. The present invention can realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] In a first aspect, the present invention provides a method for constructing a sliding bearing fault diagnosis model, comprising the following steps:

[0010] Get the sound signal of the bearing;

[0011] Preprocess the sound signal of the bearing to obtain a training data set;

[0012] The feature selection method is used to obtain bearing fault feature data from the training data set;

[0013] The online transfer learning method is used to train the model on the training data set and bearing fault feature data to obtain the sliding bearing fault diagnosis model.

[0014] In a second aspect, the present invention provides a sliding bearing fault diagnosis method, comprising the following steps:

[0015] Get the sound signal of the bearing;

[0016] The sound signal of the bearing is input into the sliding bearing fault diagnosis model to perform sliding bearing fault diagnosis.

[0017] In a third aspect, the present invention provides a system for constructing a sliding bearing fault diagnosis model, comprising:

[0018] Data acquisition module: used to obtain the sound signal of the bearing;

[0019] Preprocessing module: used to preprocess the sound signal of the bearing to obtain a training data set;

[0020] Feature extraction module: used to obtain bearing fault feature data from the training data set;

[0021] Model generation module: used to use online transfer learning method to perform model training on training data set and bearing fault feature data to obtain sliding bearing fault diagnosis model.

[0022] In a fourth aspect, the present invention provides an electronic device, comprising: a processor; a memory for storing computer program instructions; and steps for implementing a method for constructing a sliding bearing fault diagnosis model when executing the computer program.

[0023] In a fifth aspect, the present invention provides a storage medium storing computer program instructions. When the computer program instructions are loaded and run by a processor, the processor executes a method for constructing a sliding bearing fault diagnosis model.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. The present invention obtains the sound signal of the bearing; pre-processes the sound signal of the bearing to obtain a training data set, which is conducive to improving the prediction accuracy of the generated model. The bearing fault feature data is obtained from the training data set using a feature selection method; the training data set and the bearing fault feature data are trained using an online transfer learning method to obtain a sliding bearing fault diagnosis model. The present invention does not require direct contact with the bearing, thus avoiding damage to the bearing and reducing maintenance costs and risks. In addition, compared with existing fault diagnosis methods, machine learning models can more effectively process complex data features and improve the accuracy of fault diagnosis.

[0026] 2. The system of the present invention includes a data acquisition module, a preprocessing module, a feature extraction module and a model generation module. The data acquisition module is used to acquire the sound signal of the bearing; the preprocessing module is used to preprocess the sound signal of the bearing to obtain a training data set; the feature extraction module is used to acquire the bearing fault feature data from the training data set; the model generation module is used to perform model training on the training data set and the bearing fault feature data using an online transfer learning method to obtain a sliding bearing fault diagnosis model. The various modules of the system of the present invention cooperate with each other to realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy.

[0027] 3. The device and medium of the present invention can also realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a method flow chart of an embodiment of the present invention;

[0029] Figure 2 A fault diagnosis result diagram of an embodiment of the present invention;

[0030] Figure 3 The figure is a flow chart of training and testing of the fault diagnosis model according to an embodiment of the present invention.

[0031] Figure 4 is a system module diagram of an embodiment of the present invention;

[0032] Figure 5 The present invention is a flow chart of the method for constructing a sliding bearing fault diagnosis model.

[0033] Figure 6 It is a flow chart of the sliding bearing fault diagnosis method of the present invention;

[0034] Figure 7 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0036] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0037] Before introducing the embodiments of the present invention, the relevant terms involved in the embodiments of the present invention are first explained as follows:

[0038] Nyquist-Shannon Theorem: Nyquist theorem, also known as Nyquist's criterion, states that in an ideal low-pass channel (no noise and limited bandwidth), in order to avoid inter-symbol interference, the limit symbol transmission rate is 2W baud, where W is the bandwidth of the ideal low-pass channel. Shannon's theorem gives the limit data transmission rate of a channel with limited bandwidth and Gaussian white noise interference = Wlog2(1+S / N) (in b / s). When transmitting at this rate, no error can be generated.

[0039] Refined decision tree algorithm: A decision tree is a tree structure in which each internal node represents a judgment on an attribute, each branch represents the output of a judgment result, and finally each leaf node represents a classification result.

[0040] Online transfer learning: This is a machine learning method that uses knowledge learned from one task to improve performance on another related task. Traditional machine learning methods usually require that the distribution of training data is the same or very similar to that of test data, while online transfer learning allows knowledge to be shared between different domains or tasks, thus enabling wider applications. The core idea is to help the learning of the target domain by leveraging knowledge from the source domain.

[0041] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0042] See also Figure 5 The present invention discloses a method for constructing a sliding bearing fault diagnosis model, comprising the following steps:

[0043] S1. Obtain the sound signal of the bearing;

[0044] Preferably, the sound signal of the bearing includes: a sound signal of a healthy bearing, a sound signal of a bearing wear, a sound signal of a bearing crack, a sound signal of poor lubrication, and a sound signal of the presence of impurities and particles.

[0045] S2. Preprocess the sound signal of the bearing to obtain a training data set, as follows:

[0046] Normalizing the sound signal of the bearing to obtain a normalized sound signal;

[0047] The environmental noise and mechanical interference in the normalized sound signal are removed to obtain the bearing fault sound signal, as follows:

[0048] Get the minimum and maximum values ​​in the sound signal of the bearing;

[0049] Normalization is performed by subtracting the minimum value from each signal value in the sound signal of the bearing and dividing the resultant value by the difference between the maximum value and the minimum value to obtain a normalized sound signal;

[0050] The bearing fault sound signal is converted from the time domain to the frequency domain through fast Fourier transform to obtain the training data set.

[0051] S3. Use feature selection method to obtain bearing fault feature data from the training data set, as follows:

[0052] Extract fault-related features from the time domain signal of the training data set and enhance the fault-related features as follows:

[0053] Generate C functions using MATLAB Coder and use the C functions to extract fault-related features from the time domain signals of the training dataset;

[0054] A three-layer neural network autoencoder consisting of an input layer, a hidden layer, and an output layer is introduced to enhance and reconstruct fault-related features;

[0055] The enhanced fault-related features are sorted using a variance analysis-based method to obtain bearing fault feature data.

[0056] S4. Use the online transfer learning method to train the model on the training data set and bearing fault feature data to obtain the sliding bearing fault diagnosis model.

[0057] See also Figure 5 In another feasible embodiment of the present invention, the following is adaptively modified according to the situation. It includes the following steps: obtaining the sound signal of the bearing; preprocessing the sound signal of the bearing to obtain a training data set, which is conducive to improving the prediction accuracy of the generated model. The bearing fault feature data is obtained from the training data set by a feature selection method; the training data set and the bearing fault feature data are trained by an online transfer learning method to obtain a sliding bearing fault diagnosis model. The present invention does not require direct contact with the bearing, avoids damage to the bearing, and reduces maintenance costs and risks. On the other hand, compared with existing fault diagnosis methods, machine learning models can more effectively process complex data features and improve the accuracy of fault diagnosis. That is, the present invention can realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy.

[0058] Based on the above method, the present invention also discloses a sliding bearing fault diagnosis method, see Figure 6 , including the following steps:

[0059] S1. Obtain the sound signal of the bearing;

[0060] S2. Input the sound signal of the bearing into the sliding bearing fault diagnosis model to perform sliding bearing fault diagnosis.

[0061] The present invention can realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy.

[0062] Embodiment 1:

[0063] This embodiment discloses a sliding bearing fault diagnosis method, comprising the following steps:

[0064] S1. Install the acoustic emission sensor: Fix the sensor on the sliding bearing and connect it to the data acquisition system.

[0065] S2. Acoustic signal acquisition: According to the Nyquist-Shannon theorem, a data acquisition card is used to record the acoustic signal of the bearing in a wide frequency range and store it in an SD card.

[0066] S3. Signal preprocessing: Use digital filters to remove irrelevant signals such as environmental noise and mechanical interference, and retain the sound signal related to the bearing fault. Then, convert the time domain signal into a frequency domain signal for spectrum analysis and feature extraction.

[0067] S4. Feature extraction and selection: According to the characteristic frequency of bearing fault, features are extracted and enhanced from the time domain signal, and the features are sorted using a variance analysis-based method to achieve feature visualization.

[0068] S5. Model training and fault diagnosis: The online transfer learning method is used to train the model on the existing data, and the model is used to diagnose the faults of various types of sliding bearings. The model performance is evaluated through cross-validation.

[0069] Embodiment 2:

[0070] This embodiment discloses a sliding bearing fault diagnosis method, see Figure 1 and Figure 5 , use acoustic emission sensors to capture the sound signals of the bearing when it is running, and analyze them in the time domain and frequency domain to extract the required features. The importance of the features is ranked by one-way ANOVA. Then, the selected features are used as source data for online transfer learning to train the model, and the fault diagnosis of other different types of sliding bearings is performed through the source data. Finally, the model is deployed on the edge device for sliding bearing fault detection. The specific steps include:

[0071] S1. The acoustic emission sensor is installed on the outside of the sliding bearing, with one side close to the surface of the workpiece and the other side connected to the preamplifier through a low-noise cable. The output of the preamplifier is connected to the acoustic emission data acquisition card through a coaxial cable. The acoustic emission sensor and the workpiece can be fixed mechanically, adhesively or magnetically. In addition, the acoustic emission sensor can also be other smart devices that can record and save sound to receive acoustic emission signals from different locations. According to the Nyquist-Shannon theorem, the acoustic emission data acquisition card records the sound signal of the bearing at a wide range of sampling frequencies and stores it in a secure digital (SD) card.

[0072] S2. Based on the sound signals collected by the acoustic emission sensor, they are preprocessed to generate a training data set; normalization is achieved by subtracting the minimum value from each signal value and then dividing it by the difference between the maximum and minimum values ​​to eliminate the impact of data dimension and scale. Digital filters are used to remove irrelevant signal components such as environmental noise and mechanical interference to retain the sound signals generated by bearing faults. The sound signal is then converted from the time domain to the frequency domain through fast Fourier transform (FFT) for better spectrum analysis and feature extraction.

[0073] S3. For the characteristic frequencies generated by bearing faults, these features are extracted and enhanced from the time domain and frequency domain signal analysis to complete the state indicator identification. After all the features are extracted, they are sorted by importance through the feature selection method based on variance analysis to realize feature visualization and obtain the bearing fault feature data.

[0074] S4. The training dataset and bearing fault feature data were input into the online transfer learning model, which adopted a refined decision tree algorithm and optimized the performance of the decision tree model through pruning, feature selection, and threshold optimization. The performance of the classification model was evaluated using a confusion matrix, and the overall performance of the model was evaluated through cross-validation.

[0075] Furthermore, the model in the source domain is migrated to different types of sliding bearing fault diagnosis, experimentally verified in the target domain, and the model is tuned according to the verification results, and finally the model is updated and improved as needed.

[0076] The present invention uses an acoustic emission sensor to capture the sound signal of the bearing during operation, extracts the bearing fault characteristics through time domain and frequency domain analysis, and then combines the fine decision tree machine learning technology to realize sliding bearing fault detection through training models, thereby significantly improving the efficiency and accuracy of fault diagnosis. On the one hand, compared with the traditional contact monitoring method, the present invention does not need to directly contact the bearing, avoids damage to the bearing, and reduces maintenance costs and risks. On the other hand, compared with traditional statistical methods or rule-based fault diagnosis methods, machine learning models can more effectively process complex data features and improve the accuracy of fault diagnosis. In the model training process, the present invention adopts an effective feature selection method, that is, using single-factor variance analysis to enhance and optimize the performance of the model, and using the empirical cumulative distribution function to improve the stability of the model. In addition, by performing transfer learning in different types of sliding bearing target fields, the model of the source field is applied to a wider range of bearing types, and the model can be trained based on a single source data, thereby realizing the fault diagnosis of various sliding bearings. The model can also be optimized in real time according to the test samples to provide more accurate fault diagnosis results.

[0077] Embodiment three:

[0078] This embodiment discloses a method for diagnosing faults of a turbine pump sliding bearing. Figure 1 , the specific steps are as follows:

[0079] Step 1: Installation of acoustic emission sensor. Select the acoustic emission sensor model 1045D, which has an operating frequency range of 100-1300kHz and is fixed in the axial position of the end cover of the hinge sliding bearing by magnetic adsorption. The preamplifier uses a PXPA5 low-noise amplifier with a gain of 40dB; the acoustic emission data acquisition card model is PCI-2, which is used to collect sensor signals. The sensor and preamplifier are connected by a low-noise cable, and the preamplifier and data acquisition card are connected by a coaxial cable. Optionally, a digital MEMS microphone can be used instead of the acoustic emission sensor for sound collection, beamforming, source positioning and echo cancellation.

[0080] Step 2: Acoustic signal acquisition. According to the Nyquist-Shannon theorem, the data acquisition card uses a cyclostationary signal acquisition method to continuously record the acoustic signals of the bearings at a wide range of sampling frequencies and store them in a secure digital (SD) card. Each record lasts 100 seconds. In order to diagnose different types of sliding bearing faults, the workpiece sound signals are divided into five categories: healthy, bearing wear, bearing cracks, poor lubrication, and the presence of impurities and particles. All known faulty and normal bearings are tested, and the sampling frequency is set to 40kHz to cover the entire audible range, so as to obtain known faulty and normal bearing signals.

[0081] Step 3: Data preprocessing. The collected signal is de-noised and filtered to improve the signal-to-noise ratio. The digital bandpass filter is used as an anti-aliasing filter to remove the DC offset at the cutoff frequency of 20Hz and 20kHz, filter out irrelevant signals such as environmental noise and mechanical interference, and retain the sound signal generated by the bearing fault. 4000 sampling data are selected from the sound signal for time domain analysis; considering the frequency resolution and spectrum leakage, the digital signal is fast Fourier transformed using the Hamming window to obtain the spectrum analysis of the digital signal.

[0082] Step 4: Status indicator identification. For the characteristic frequencies of bearing faults, these features are extracted and enhanced from the time domain signal analysis. C functions are generated using MATLAB coder for frequency domain feature extraction of a given FFT spectrum, as well as predefined significance values ​​for peak extraction and C functions for the selected machine learning algorithm. To gain a deeper understanding of the system fault, a set of sliding bearing acoustic emission features is extracted from each data window, as shown in Table 1. Among them, 6 features are extracted from the frequency domain, including root mean square, mean, shape factor, kurtosis, skewness, impulse factor, crest factor, margin factor, variance, and median measurement. 10 features are extracted from the time domain signal analysis, including P1 to P6. The fault domain contains time domain and frequency domain features. After extracting all features, they are sorted by importance using a feature selection method based on variance analysis. P1 is the most important feature, followed by variance σ2 and RMS. In order to capture transient changes in the signal and enhance the main features, a three-layer neural network autoencoder consisting of an input layer, a hidden layer, and an output layer is introduced to enhance and reconstruct the data features.

[0083] Table 1. Acoustic emission feature set of sliding bearings:

[0084]

[0085] Step 5: Model training Based on the feature importance evaluation results, we selected a cross-validation trade-off value k=6 to verify the accuracy of each classification algorithm. By comparing the recognition accuracy of different machine learning techniques, see Table 2, we found that most of the misclassifications occurred during bearing failure. Further analysis showed that the refined decision tree model performed best when the mean, skewness, and median features were removed. Therefore, we decided to use the refined decision tree algorithm to train the model on the dataset. During the training process, we used pruning techniques, feature selection to prevent overfitting, and threshold optimization to optimize the performance of the decision tree model. Finally, the performance of the model was evaluated through the confusion matrix to fully understand its classification effect on different categories.

[0086] Table 2: Comparison of recognition accuracy of various machine learning technologies:

[0087] Serial number Classification Algorithms Accuracy 1 Refined decision tree 97.2％ 2 Bag-Tree Ensemble Classifier 96.6％ 3 Quadratic SVM 96.0％ 4 Naive Bayes 93.5％ 5 LR 92.6％

[0088] Step 6: Transfer learning. First, after obtaining the source domain model for sliding bearing fault diagnosis, fine-tune different types of sliding bearings to meet the specific fault diagnosis needs of various types of sliding bearings. Next, evaluate the performance of the fine-tuned model on the target domain dataset, and make adjustments and improvements based on the fault diagnosis results. By iteratively optimizing the model and migration strategy, the accuracy and robustness of sliding bearing fault diagnosis can be further improved.

[0089] After obtaining the final fault diagnosis model, all the collected feature values ​​are input into the model to test its performance, see Figure 6 The test results are as follows Figure 2 As shown in the figure, the known label represents the determined fault type, and the predicted label represents the fault type predicted by the model. The test samples include the following categories: Fault 1 is healthy state, Fault 2 is bearing wear, Fault 3 is bearing crack, Fault 4 is poor lubrication, and Fault 5 is the presence of impurities and particles. The test results show that the accuracy of the model remains basically consistent from the development stage to the implementation stage. The reduction of feature dimension improves the classification speed without affecting the accuracy of the model; the application of empirical cumulative distribution function enhances the stability and accuracy of the model. Specifically, the prediction accuracy of Fault 1 is 99.6%, Fault 2 is 98.9%, Fault 3 is 97.2%, Fault 4 is 93.9%, and Fault 5 is 96.4%. This shows that the sliding bearing fault diagnosis method based on online transfer learning can extract relevant features from the original data even with the least amount of data, and a wider range of sliding bearing faults can be diagnosed through a single type of sliding bearing data, ensuring the stability and accuracy of the diagnosis results. The method of the present invention can realize fault diagnosis of more types of sliding bearings based on fault source data of a single type of sliding bearing, and provide efficient monitoring and diagnosis services for the health status of related equipment.

[0090] Specific embodiments can be combined Figure 3 Description, such as Figure 3 The complete workflow of the implementation phase is shown. First, the acoustic signal of the sliding bearing is obtained by the acoustic emission sensor. The signal is transmitted to the data acquisition card through the preamplifier. 2048 samples are collected per second, and each sample is 16-bit pulse code modulation. After preprocessing, the fault features are extracted. The recording mode is selected to collect and store these feature data. The length of each record is one minute, and each record corresponds to a fault state. Therefore, more than three fault types can be added, such as combined faults that may occur in the normal life cycle of the bearing. After that, the collected features are trained on the fault diagnosis model using machine learning technology. The selected model is a refined decision tree model, which has the advantages of fast running speed and small memory usage.

[0091] Based on the above method, the present invention also discloses a system for constructing a sliding bearing fault diagnosis model, see Figure 7 ,include:

[0092] Data acquisition module: used to obtain the sound signal of the bearing;

[0093] Preprocessing module: used to preprocess the sound signal of the bearing to obtain a training data set;

[0094] Feature extraction module: used to obtain bearing fault feature data from the training data set;

[0095] Model generation module: used to use online transfer learning method to perform model training on training data set and bearing fault feature data to obtain sliding bearing fault diagnosis model.

[0096] The various modules of the system of the present invention cooperate with each other to realize non-contact fault diagnosis of sliding bearings with high diagnostic accuracy.

[0097] Embodiment 4:

[0098] This embodiment discloses a system for constructing a sliding bearing fault diagnosis model. In order to improve the fault diagnosis accuracy and robustness of different types of sliding bearings, the present invention fine-tunes the diagnosis models of different sliding bearings, evaluates and optimizes the model performance on the target domain data set, and ultimately achieves efficient fault diagnosis.

[0099] Figure 4 The schematic diagram of the structure of a sliding bearing fault diagnosis device based on online transfer learning is shown, which includes the following components:

[0100] 1. Sensor node: used to monitor the sound signals generated by the bearing in real time. The sensor needs to have high sensitivity and wide-band response characteristics, and can accurately capture weak sound signals. You can choose to use an acoustic emission sensor close to one side of the sliding bearing or an intelligent sound collection system based on a MEMS microphone.

[0101] 2. Data acquisition module: responsible for receiving the sound signals collected by the sensor nodes and converting them into digital signals for processing. This module requires high-speed and high-precision analog-to-digital conversion capabilities to ensure accurate acquisition and conversion of sound signals. You can choose a PCI-2 acoustic emission data acquisition card or other equipment that can collect and record bearing sound signals.

[0102] 3. Data processing module: Receives digital signals from the data acquisition module, performs filtering, noise reduction, analysis and feature extraction, and extracts characteristic parameters related to sliding bearing faults. This module needs to include a dedicated signal processing algorithm for bearing fault diagnosis to effectively extract fault features and reduce noise interference.

[0103] 4. Fault detection model: used to determine whether the bearing has abnormalities or potential faults. The model is trained and optimized based on the characteristic parameters of the acoustic emission signal and the fault labels in the training set. After training, it can perform fault diagnosis based on the characteristic parameters of the sound signal. The model needs to be fully trained and verified to ensure high-accuracy identification and judgment of different types of bearing faults.

[0104] 5. Test and display module: Receives and displays the diagnostic results output by the fault detection model and provides them to users or other devices. This module also supports data logging, alarm, and remote communication functions to help users monitor the bearing status in real time and take corresponding maintenance measures. Although the fault detection model can achieve the same function on the cloud server, deploying it on the edge device can respond to faults more quickly and reduce communication delays and dependencies.

[0105] An electronic device comprises: a processor; a memory for storing computer program instructions; and steps for implementing a method for constructing a sliding bearing fault diagnosis model when executing the computer program.

[0106] A storage medium stores computer program instructions. When the computer program instructions are loaded and run by a processor, the processor executes a method for constructing a sliding bearing fault diagnosis model.

[0107] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0109] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0110] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0111] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for constructing a sliding bearing fault diagnosis model, characterized in that: The following steps are involved: Get the sound signal of the bearing; Preprocess the sound signal of the bearing to obtain a training data set; The feature selection method is used to obtain bearing fault feature data from the training data set; The online transfer learning method is used to train the model on the training data set and bearing fault feature data to obtain the sliding bearing fault diagnosis model.

2. The method for constructing a sliding bearing fault diagnosis model according to claim 1, characterized in that: The sound signals of the bearing include: sound signals of healthy bearings, sound signals of bearing wear, sound signals of bearing cracks, sound signals of poor lubrication, and sound signals of the presence of impurities and particles.

3. The method for constructing a sliding bearing fault diagnosis model according to claim 1, characterized in that: The steps of preprocessing the sound signal of the bearing to obtain the training data set are specifically as follows: Normalizing the sound signal of the bearing to obtain a normalized sound signal; The environmental noise and mechanical interference in the normalized sound signal are removed to obtain the bearing fault sound signal; The bearing fault sound signal is converted from the time domain to the frequency domain through fast Fourier transform to obtain the training data set.

4. The method for constructing a sliding bearing fault diagnosis model according to claim 3, characterized in that: The step of normalizing the sound signal of the bearing to obtain a normalized sound signal is specifically as follows: Get the minimum and maximum values ​​in the sound signal of the bearing; Normalization is achieved by subtracting the minimum value from each signal value in the bearing sound signal and dividing the resultant value by the difference between the maximum value and the minimum value to obtain a normalized sound signal.

5. The method for constructing a sliding bearing fault diagnosis model according to claim 1, characterized in that: The steps of obtaining bearing fault feature data from the training data set are specifically as follows: Extract fault-related features from the time domain signals of the training data set and enhance the fault-related features; The enhanced fault-related features are sorted using a variance analysis-based method to obtain bearing fault feature data.

6. The method for constructing a sliding bearing fault diagnosis model according to claim 5, characterized in that: The steps of extracting fault-related features from the time domain signal of the training data set and enhancing the fault-related features are specifically as follows: Generate C functions using MATLAB Coder and use the C functions to extract fault-related features from the time domain signals of the training dataset; A three-layer neural network autoencoder consisting of an input layer, a hidden layer and an output layer is introduced to enhance and reconstruct fault-related features.

7. A sliding bearing fault diagnosis method, characterized in that: The following steps are involved: Get the sound signal of the bearing; The sound signal of the bearing is input into the sliding bearing fault diagnosis model described in any one of claims 1 to 6 to perform sliding bearing fault diagnosis.

8. A system for constructing a sliding bearing fault diagnosis model for implementing the method described in any one of claims 1 to 6, characterized in that: include: Data acquisition module: used to obtain the sound signal of the bearing; Preprocessing module: used to preprocess the sound signal of the bearing to obtain a training data set; Feature extraction module: used to obtain bearing fault feature data from the training data set; Model generation module: used to use online transfer learning method to perform model training on training data set and bearing fault feature data to obtain sliding bearing fault diagnosis model.

9. An electronic device, comprising: Processor; memory, electronic device used to store computer program instructions; characterized in that it is used to implement the steps of the method for constructing a sliding bearing fault diagnosis model as described in any one of claims 1-6 when executing the computer program.

10. A storage medium storing computer program instructions, characterized in that: When the computer program instructions are loaded and executed by the processor, the processor executes the method for constructing a sliding bearing fault diagnosis model according to any one of claims 1 to 6.

Images