A method, system, terminal and medium for diagnosing a press machine bearing fault

Abstract

The application belongs to the technical field of bearing fault diagnosis, and specifically discloses a press machine bearing fault diagnosis method, system, terminal and medium, which comprises the following steps: collecting multi-source signal data of a press machine bearing during operation; carrying out denoising and standardization processing on the signals, and extracting time domain and frequency domain features; fusing different sensor features to construct multi-dimensional feature representation; inputting the features into a deep learning model combining a convolutional neural network and a long short-term memory network for training and reasoning, and introducing a multi-core maximum mean difference strategy to realize feature distribution alignment; and based on a multi-label classification mode, carrying out fault recognition on real-time signals, and triggering an alarm when a prediction probability meets a preset condition. The application realizes multi-source data fusion modeling and accurate recognition of compound faults, has strong cross-condition adaptability and real-time diagnosis capability, and is suitable for intelligent operation and maintenance scenarios of press machine equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of bearing fault diagnosis, and particularly relates to a press bearing fault diagnosis method, system, terminal and medium. BACKGROUND

[0002] The press bearing on the high-speed stamping production line is a vital moving part in the press and other mechanical equipment, and its performance directly determines the stability and operating efficiency of the equipment. With the long-time and high-load operation of the press, the bearing may be damaged to different degrees, such as fatigue wear, cracks, surface spalling and other faults, under the continuous action of wear and stress from the working environment. If these faults are not detected in time and repaired, the bearing may fail, which may further cause serious faults of the equipment and even affect the normal operation of the entire production line. The bearing faults not only increase the maintenance cost, but also prolong the downtime, causing a significant decline in production efficiency, and even endanger the safety of personnel and equipment. Therefore, how to accurately and real-timely diagnose the bearing faults during the operation of the press is the key to ensuring the efficient, stable and safe operation of production.

[0003] The traditional bearing fault diagnosis method usually relies on the collection and analysis of vibration signals, and common methods include feature extraction and analysis based on time domain or frequency domain. These methods extract features in the signal, such as vibration amplitude, frequency components, periodic changes, etc., to judge the health status of the bearing. However, time domain and frequency domain feature analysis methods have their inherent limitations, mainly in that they cannot effectively capture the complex nonlinear and time-varying characteristics of bearing faults. Especially when the bearing fault has nonlinear and non-stationary characteristics, these traditional methods often fail to comprehensively and accurately reflect the actual fault condition of the bearing. In addition, the traditional methods are highly sensitive to noise, especially in complex working conditions and high-noise environments, which can easily lead to a decrease in diagnostic accuracy, thereby affecting the timeliness and reliability of fault detection. SUMMARY

[0004] The application provides a press bearing fault diagnosis method, system, terminal and medium to solve the problem that the traditional method of the prior art often fails to comprehensively and accurately reflect the actual fault condition of the bearing when the bearing fault has nonlinear and non-stationary characteristics.

[0005] The technical scheme adopted by the application is as follows:

[0006] In a first aspect, the application provides a press bearing fault diagnosis method, which comprises the following steps:

[0007] Step S1, collecting multi-source signal data of the press bearing during operation, the multi-source signal data including vibration signals, temperature signals and pressure signals;

[0008] Step S2, pre-processing the multi-source signals, the pre-processing including denoising and standardization of the signals;

[0009] to obtain a pre-processed time series signal set X = [x1, x2, …, xn], where n is the number of data points, xi is the pre-processed multi-source signal data at the i-th time point; n i

[0010] extracting time domain and frequency domain features from the pre-processed signals;

[0011] Step S3, fusing the features extracted from different signal sources to construct a multi-dimensional feature representation for representing the bearing state;

[0012] Step S4, inputting the multi-dimensional features into a deep learning model combining convolutional neural network and long short-term memory network for training, and introducing a multi-kernel maximum mean difference strategy into the deep learning model to align the subspace distribution of the input features under different working conditions;

[0013] the convolutional neural network is used to extract local spatial information of the input features;

[0014] the long short-term memory network is used to model the dependency relationship of the features over time;

[0015] the extracted features are sent to the long short-term memory network composed of multiple memory units for time series modeling through a fully connected layer, and the Softmax function is used in the output layer to convert the classification result into the prediction probability of each type of fault:

[0016]

[0017] where zk is the score of class k, K is the number of classes, is the prediction probability of class k,

[0018] where w i is the weighting coefficient, F i is the time-frequency feature of the i-th sensor;

[0019] Step S5, based on the trained deep learning model, real-time collected signals are subjected to fault recognition, and an alarm prompt is issued when the recognition result meets the preset condition.

[0020] Further, in step S2, the time domain feature extraction includes mean, standard deviation, skewness, kurtosis features, and the time domain feature is defined as:

[0021]

[0022] ​​​

[0023]

[0024]

[0025] wherein, is the statistical mean of the signal segment, is the statistical standard deviation of the signal segment, is the skewness of the signal X, is the kurtosis of the signal X;

[0026] The signal is analyzed in the frequency domain using a fast Fourier transform to extract the spectral mean, spectral standard deviation, spectral energy, and peak frequency characteristics.

[0027] Further, in step S3, a domain adaptation strategy based on multi-kernel maximum mean discrepancy is used, and the domain adaptation strategy is constructed by combining multiple different types and scales of kernel functions to construct a high-dimensional subspace mapping.

[0028] Further, the kernel function u is a linear combination of positive definite functions , which is represented by the weighted sum of multiple kernel functions:

[0029]

[0030] wherein, is the i-th positive definite kernel function, is a non-negative weighted coefficient of the positive definite kernel function .

[0031] Further, in step S5, the fault recognition adopts a multi-label classification strategy, combines a Binary Relevance strategy and a label correlation weight modeling mechanism, and incorporates the co-occurrence mode between various fault labels into the joint modeling process. A sliding time window mechanism is introduced in the model deployment stage to dynamically evaluate the continuously collected data. When the prediction probability of any category accumulates more than the set threshold within the window, or the concurrent probability of multiple categories simultaneously meets the warning condition, the system generates a composite fault alarm signal and performs priority classification prompt.

[0032] In a second aspect, the application provides a press machine bearing fault diagnosis system, which comprises:

[0033] A data acquisition component is configured to acquire multi-source signal data of the press machine bearing during operation, wherein the multi-source signal data comprises vibration signals, temperature signals, and pressure signals.

[0034] A preprocessing component is configured to perform denoising and standardization processing on the multi-source signals, generate a preprocessed time series signal set, and extract time domain and frequency domain features from the signal set.

[0035] a feature fusion component configured to weight and fuse time-frequency features from different sensors to construct a multi-dimensional feature representation for representing bearing conditions;

[0036] a diagnostic model component configured to input the multi-dimensional features into a deep learning model combining a convolutional neural network and a long short-term memory network for training and inference, the deep learning model introducing a multi-kernel maximum mean difference strategy to adapt to feature migration under different operating conditions;

[0037] a fault identification and alarm component configured to, after the diagnostic model completes classification and judgment on the real-time collected signals, issue an alarm prompt signal when the predicted probability of any fault type meets a preset condition, to prompt maintenance personnel to take maintenance measures.

[0038] In a third aspect, the present application provides a terminal, comprising:

[0039] a memory configured to store a diagnosis program for press machine bearing faults;

[0040] a processor configured to execute the diagnosis system for press machine bearing faults to implement the steps of the diagnosis method for press machine bearing faults as described in the first aspect.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium, the storage medium storing computer instructions, when the computer reads the computer instructions in the storage medium, the computer executes the diagnosis method for press machine bearing faults as described in the first aspect.

[0042] The above technical solutions can be seen that the advantages of the present application are:

[0043] (1) The deep fusion of multi-modal data is realized, and the accuracy and robustness of fault diagnosis are improved. By introducing multi-modal data fusion technology, the data from multiple heterogeneous sources such as vibration sensors and temperature sensors are effectively integrated, avoiding information loss and diagnosis deviation that may be caused by a single data source. In the fusion process, a unified data expression framework and a deep feature mapping mechanism are constructed to enhance the correlation and complementarity between different modal data, thereby realizing more comprehensive and accurate fault state recognition and significantly improving the adaptability and generalization ability of the diagnostic model under complex working conditions.

[0044] (2) The present application is based on a deep learning framework, fuses time-domain and frequency-domain feature information, and uses structures such as convolutional neural networks and long short-term memory networks to extract key features of bearing fault signals. It can effectively identify multiple typical fault types including inner ring fault, outer ring fault, rolling element fault and their combination faults, and the diagnostic model has good generalization ability and adapts to the variability of different loads, speeds and working conditions.

[0045] (3), the online monitoring and fault prediction mechanism is introduced, the sliding window processing and dynamic evaluation method are combined, the health state change of the bearing can be continuously perceived, the potential degradation trend can be identified in advance, and intelligent early warning and maintenance decision support of the stamping equipment are realized. The method effectively reduces the unplanned downtime caused by sudden failure, prolongs the service life of the equipment, significantly reduces the maintenance cost and improves the overall operation efficiency and safety of the production line. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the present application, the drawings needed to be used in the description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0047] Figure 1 Flowchart of the fault diagnosis method for press machine bearing;

[0048] Figure 2 Composite fault diagnosis framework diagram of the fault diagnosis method for press machine bearing. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0050] Please refer to Figure 1 and Figure 2 , the present application provides a fault diagnosis method for press machine bearing, comprising the following steps:

[0051] Step S1, collecting multi-source signal data of press machine bearing in the running process, the multi-source signal data including vibration signal, temperature signal and pressure signal;

[0052] Step S2, pre-processing the multi-source signal, the pre-processing including signal denoising and standardization operation;

[0053] obtaining the pre-processed time series signal set X=[x1,x2,…,x n ], wherein n is the number of data points, x i is the multi-source signal data of the i-th time point after pre-processing;

[0054] extracting time domain and frequency domain features from the pre-processed signal;

[0055] Step S3, the features extracted from different signal sources are fused to construct a multi-dimensional feature representation for representing the bearing state;

[0056] Step S4, the multi-dimensional features are input into a deep learning model combining convolutional neural network and long short-term memory network for training, and the deep learning model introduces a multi-kernel maximum mean difference strategy to align the subspace distribution of the input features under different working conditions;

[0057] The convolutional neural network is used to extract the local spatial information of the input features;

[0058] The long short-term memory network is used to model the dependence relationship of the features over time;

[0059] The extracted features are sent to the long short-term memory network composed of multiple memory units for time series modeling through the fully connected layer, and the Softmax function is used in the output layer to convert the classification result into the prediction probability of each type of fault:

[0060]

[0061] wherein zk is the score of class k, K is the number of classes, is the prediction probability of class k, ;

[0062] wherein w i is the weighting coefficient, F i is the time-frequency feature of the i-th sensor.

[0063] Step S5, based on the trained deep learning model, real-time collected signals are subjected to fault recognition, and an alarm prompt is issued when the recognition result meets the preset condition.

[0064] In some embodiments, in step S2, the time domain feature extraction includes mean, standard deviation, skewness, kurtosis features, and the time domain feature is defined as:

[0065]

[0066]

[0067]

[0068]

[0069] wherein is the statistical mean of the signal segment, is the statistical standard deviation of the signal segment, is the skewness of the signal X, is the kurtosis of the signal X;

[0070] The signals are analyzed in the frequency domain using fast Fourier transform, and the spectral mean, spectral standard deviation, spectral energy, and peak frequency characteristics are extracted.

[0071] In some embodiments, in step S3, the signals collected by different sensors are , where m represents the number of sensors, and the time-frequency characteristics extracted from each signal are F1, F2, …, Fm. m The time-frequency characteristics of these signals are combined into a multi-dimensional feature vector through a weighted fusion method:

[0072] .

[0073] In some preferred embodiments, the deep learning model used in step S4 includes a combination of convolutional neural networks (CNN) and long short-term memory networks (LSTM) to achieve joint modeling of multi-dimensional features. Specifically:

[0074] The convolutional neural network is used to extract the local spatial structure information of the input features. The network structure includes three convolutional layers connected in series, and the convolution kernel size of each convolutional layer is 3x3, which is used to extract local textures and change patterns at different scales. Each convolutional layer is connected with a rectified linear unit (ReLU) activation function to enhance the non-linear expression ability, and the feature map is compressed and down-sampled through the maximum pooling operation, thereby reducing the computational complexity and improving the hierarchical expression ability of the features.

[0075] The feature map output by the convolutional layer is dimensionally transformed and structurally reconstructed through a fully connected layer to adapt to the input requirements of the subsequent sequence modeling network.

[0076] The long short-term memory network (LSTM) receives the feature vector sequence output by the fully connected layer and models the time series to capture the dynamic association and evolution law of multi-dimensional features between different time steps, thereby enhancing the model's ability to identify the trend of bearing operating state changes.

[0077] To improve the generalization performance of the model under complex working conditions such as varying loads and varying speeds, the deep learning model further introduces a multi-kernel maximum mean difference (MK-MMD) strategy to perform subspace mapping and distribution alignment on the input features between the training domain and the target application domain. The MK-MMD mechanism realizes the minimization of cross-domain feature differences by constructing a linear combination of multiple positive definite kernel functions, thereby alleviating the problem of decreased diagnostic accuracy caused by working condition migration.

[0078] To adapt to the feature distribution shift caused by different working conditions (such as load and speed), a multi-kernel maximum mean discrepancy (MK-MMD) strategy is introduced to construct subspace mapping by combining multiple kernel functions, thereby enhancing the model's adaptability to inconsistent features between domains.

[0079] The kernel function u is a linear combination of positive definite functions , which is represented by the weighted sum of multiple kernel functions:

[0080]

[0081] wherein, is the ith positive definite kernel function, is the non-negative weighted coefficient of the positive definite kernel function .

[0082] In the output layer of the model structure, the Softmax function is used to convert the neural network output vector into the predicted probability value of each fault category, ensuring that the classification result has good interpretability and discriminability. Specifically, each node of the model output corresponds to a fault category, and the output value reflects the probability of occurrence of this category, which can be used to determine whether the sample has one or more fault features.

[0083] Further, in the last output layer of the model, the output of the model is converted into multi-class probability by the Softmax function:

[0084]

[0085] wherein, z k is the score of class k, K is the number of classes, is the predicted probability of class k.

[0086] In some further embodiments of the present application, the fault recognition process in step S5 adopts a multi-label classification strategy to solve the identification problem of multiple types of faults (such as inner ring fault, outer ring fault, and rolling element fault) that may exist simultaneously in the actual operation of the press machine bearing. This strategy can support parallel judgment of multiple potential fault modes, and improve the recognition ability of the system for complex fault states.

[0087] Specifically, this embodiment represents the fault state of each sample as a label vector, where each dimension corresponds to a predefined fault type. The system uses the Binary Relevance (BR) strategy to decompose the multi-label classification task into several independent binary classification sub-tasks, and trains a corresponding sub-classifier for each fault type.

[0088] The output layer structure is constructed as follows:

[0089]

[0090] wherein C is the number of fault categories, and the model can simultaneously identify multiple fault labels.

[0091] During the training process, a cross-entropy loss function is used as the objective function, the optimization method is an Adam optimizer, a validation set is set to control overfitting and optimize parameters, and the training goal is to minimize:

[0092]

[0093] wherein p i is the predicted probability of category i.

[0094] On this basis, to improve the model's expression ability for the dependency between labels, a label correlation modeling mechanism is introduced, parameters such as label co-occurrence probability and mutual information weight are introduced to capture the coupling features between multiple fault types, thereby improving the accuracy of composite fault recognition.

[0095] To enhance the stability and robustness of the diagnostic system in engineering applications, a sliding time window mechanism is introduced in the model deployment stage. The specific method is: the real-time collected sensor data is continuously fragmented according to a fixed window length (such as 5 seconds, 1024 frames, etc.), and the prediction results in each window are dynamically accumulated and evaluated. The system sets the fault alarm triggering rules as follows:

[0096] When the cumulative value of the prediction probability of a certain fault category in the sliding time window continuously exceeds the set threshold (such as 0.85), it is considered that the fault of this category has high confidence, and the system will determine that the fault exists and issue an alarm;

[0097] Or, in the same time window, if the prediction probabilities of multiple fault categories simultaneously exceed the corresponding thresholds, the system will determine that it is a composite fault state, and according to the prediction probability and weight calculation results of each category, the fault is prioritized and prompted to assist the maintenance personnel to arrange the repair sequence according to the severity.

[0098] In some embodiments, the present application provides a system for press machine bearing fault diagnosis, which is suitable for real-time monitoring and intelligent diagnosis of bearing operating state in high-speed stamping production line. The system comprises:

[0099] a data acquisition component for synchronously acquiring multi-source signal data related to the bearing during the operation of the press machine, the multi-source signals including but not limited to vibration signals, temperature signals and pressure signals; wherein the vibration signals are used to reflect the mechanical dynamic response characteristics, the temperature and pressure signals are used to provide auxiliary information of the operating environment and the load state, and the three types of signals are synchronously acquired by different types of industrial sensors at a preset sampling frequency;

[0100] A preprocessing component is configured to perform standardization processing on the collected multi-source signals, and the preprocessing process includes noise suppression, outlier rejection, and amplitude normalization operations. After preprocessing, the system divides each type of signal into equal-length time windows to construct a unified time series signal set. The component further extracts typical statistical features and spectral features from each signal segment, including mean, standard deviation, skewness, kurtosis, spectral energy, and peak frequency, to characterize the state changes of the bearing at different time periods.

[0101] A feature fusion component is configured to perform weighted fusion of features from multiple sensors to form a multi-dimensional feature vector representing the current bearing state. The weighting coefficients can be set based on sensor reliability, historical performance, or training phase learning results to enhance the discrimination ability and robustness of the fused features.

[0102] A diagnostic model component is configured to perform classification learning and fault identification on the fused feature vector. The model uses a composite structure combining convolutional neural networks (CNN) and long short-term memory networks (LSTM). CNN is used to extract local spatial features of the input features, and LSTM is used to model the dependence relationship of the features over time. To adapt to the distribution shift caused by multi-working-condition changes, the model further introduces a multi-kernel maximum mean difference (MK-MMD) strategy to align the feature distributions of the source domain and the target domain, thereby improving the generalization ability under unseen working conditions.

[0103] A fault identification and alarm component is configured to deploy the trained deep learning model to a field diagnosis platform, receive real-time collected signals, and perform fault identification operations. The component supports a multi-label classification strategy, with an independent output node for each fault mode. When the prediction probability of any fault label continuously exceeds a set threshold (e.g., 0.85) within a sliding time window, the system will automatically issue an alarm signal to the operator through a graphical interface or an industrial communication protocol, assisting them in making predictions and handling faults.

[0104] In some embodiments, the present application provides a terminal, comprising:

[0105] A memory is configured to store a diagnosis program for press machine bearing faults;

[0106] A processor is configured to perform the steps of the diagnosis method for press machine bearing faults when executing the diagnosis system for press machine bearing faults.

[0107] In some embodiments, the present application provides a computer-readable storage medium, which stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes the diagnosis method for press machine bearing faults.

[0108] It can be understood that the systems, apparatuses, modules or units illustrated by the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, and the computer can be specifically a personal computer, a laptop computer, a personal digital assistant, a tablet computer, a wearable device, or a combination of any of these devices.

[0109] In a typical configuration, a computer includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0110] The memory can include non-persistent memory in computer readable media, random access memory (RAM), and / or non-volatile memory such as read-only memory (ROM) or flash memory. The memory is an example of computer readable media.

[0111] Computer readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology for storing information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, disk storage, quantum memory, graphene-based storage medium or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition in this paper, computer readable media does not include transitory computer readable media, such as modulated data signals and carriers.

[0112] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation portal for user to choose authorization or refusal.

[0113] It is also to be noted that the terms "comprising", "including", and "having" or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, includes, or has a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a", "includes... a", "has... a", or the like, does not, without more constraints, exclude the existence of additional identical elements.

[0114] The above description of certain embodiments of the present specification has been presented for the purposes of illustration and description. Other embodiments are within the scope and spirit of the present specification. In some cases, the actions or steps recited in the claims can be performed in a different order and still accomplish the desired results. Additionally, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous.

[0115] The terminology used in this description of one or more embodiments of the specification is for the purpose of describing particular embodiments only and is not intended to be limiting of one or more embodiments of the specification. As used in this description of one or more embodiments of the specification and the accompanying claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term "and / or," as used herein, refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0116] It will be understood that, although the terms first, second, third, etc. can be used herein to describe various information, these terms are used only to distinguish one piece of information from another, but do not imply a relationship between the information. For example, without departing from the scope of one or more embodiments of the specification, the first information can be named the second information, and similarly, the second information can be named the first information. Depending on the context, the word "if' as used herein can be interpreted as meaning "when" or "in response to determining."

[0117] The foregoing description of preferred embodiments of one or more aspects of the present specification has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit one or more aspects of the present specification to the precise form disclosed. Modifications, equivalent substitutions, improvements, and other implementations are intended to be included within the scope of one or more aspects of the present specification.

Claims

1. A method of diagnosing a press machine bearing failure, characterized by, The method comprises the following steps: Step S1, collecting multi-source signal data of the press bearing during operation, the multi-source signal data comprising vibration signals, temperature signals and pressure signals; Step S2, pre-processing the multi-source signals, the pre-processing comprising denoising and standardization of the signals; The preprocessed time series signal set X=[x1,x2,…,x] is obtained. n ], where n is the number of data points, x i This refers to the preprocessed multi-source signal data at the i-th time point; extracting time domain and frequency domain features from the pre-processed signals; Step S3, fusing the features extracted from different signal sources to construct a multi-dimensional feature representation representing the bearing state; Step S4, inputting the multi-dimensional features into a deep learning model combining a convolutional neural network and a long short-term memory network, and introducing a multi-kernel maximum mean difference strategy into the deep learning model to align the subspace distribution of the input features under different working conditions; the convolutional neural network is used to extract local spatial information of the input features; the long short-term memory network is used to model the time-varying dependence of the features; the extracted features are sent to the long short-term memory network composed of multiple memory units for time series modeling through a fully connected layer, and the Softmax function is used in the output layer to convert the classification result into the prediction probability of each type of fault: wherein zk is the score of class k, K is the number of classes, is the predicted probability of class k, ; where w i is a weighting coefficient, F i is the time-frequency feature of the i-th sensor, and m represents the number of sensors. Step S5, based on the trained deep learning model, real-time collected signals are subjected to fault identification, and an alarm prompt is issued when the identification result meets the preset condition; the multi-label classification strategy is adopted for fault identification, the Binary Relevance strategy and the label correlation weight modeling mechanism are combined to include the co-occurrence mode between the labels of each type of fault into the joint modeling process, and the sliding time window mechanism is introduced in the model deployment stage to dynamically evaluate the continuously collected data, when the prediction probability of any category accumulates more than the set threshold within the window, or the concurrent probability of multiple categories meets the warning condition at the same time, the system generates a composite fault alarm signal and gives priority classification prompt.

2. The method of diagnosing a press bearing fault according to claim 1, characterized in that, In step S2, the time domain feature extraction includes mean, standard deviation, skewness and kurtosis features, and the time domain feature is defined as: wherein, is the statistical mean of the signal segment, is the statistical standard deviation of the signal segment, is the skewness of the signal X, is the kurtosis of the signal X; Fast Fourier Transform is used to analyze the frequency domain of the signal, and the spectral mean, spectral standard deviation, spectral energy and peak frequency features are extracted.

3. The method of diagnosing a press bearing fault according to claim 1, wherein, In step S3, a domain adaptation strategy based on multi-kernel maximum mean difference is used, and the domain adaptation strategy is constructed by combining multiple kernel functions of different types and scales to construct a high-dimensional subspace mapping.

4. The method of diagnosing a press bearing fault according to claim 3, characterized in that, The kernel function u is a positive definite function is a linear combination of kernel functions, represented by a weighted sum of multiple kernel functions: wherein, is the ith positive definite kernel function, is a positive definite kernel function is a non-negative weighting coefficient.

5. A terminal, characterized by comprising: It comprises: a memory for storing a diagnosis program of press bearing faults; a processor for executing the diagnosis system of the press bearing faults to implement the steps of the diagnosis method of the press bearing faults according to claim 1.

6. A computer-readable storage medium, characterized in that, The storage medium stores computer instructions, and when the computer reads the computer instructions in the storage medium, the computer executes the diagnosis method of the press bearing faults according to claim 1.

Images