Gearbox output shaft bearing life prediction method, device and equipment

Abstract

The application discloses a gearbox output shaft bearing life prediction method, device and equipment, and relates to the technical field of power batteries. The method comprises the following steps: converting the segmented gearbox output shaft bearing vibration signal into a vibration image, and extracting a sample vibration image; determining a physical model life index based on a life prediction physical model, labeling a preset number of sample vibration images to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images; extracting global features and local features of the query set and extracting global features and local features of the support set; fusing the global features and local features of the query set and the global features and local features of the support set to obtain a fusion feature pair; and predicting the life of the gearbox output shaft bearing based on the fusion feature pair to obtain a life prediction result. The advantages of deep learning and physical models are combined to predict the life of the gearbox output shaft bearing, thereby improving the accuracy and reliability of the prediction.

Description

Technical Field

[0001] This application relates to the field of power battery technology, and in particular to a method, apparatus and equipment for predicting the life of gearbox output shaft bearings. Background Technology

[0002] The output shaft bearing of a transmission is one of the key components of an automotive drivetrain, and its performance directly affects the overall vehicle's operating efficiency and reliability. Early bearing failure often leads to serious mechanical damage and even safety accidents. Therefore, predicting and monitoring the lifespan of transmission output shaft bearings is of significant practical importance.

[0003] Currently, traditional empirical formula methods for predicting the lifespan of transmission output shaft bearings have limited accuracy and are difficult to adapt to complex and ever-changing actual working conditions. They rely excessively on historical data and are not accurate enough for predicting bearing lifespan under new engines or special working conditions. Traditional finite element analysis methods use engineering software for mechanical analysis to predict the stress and strain distribution of transmission output shaft bearings under specific working conditions. This includes model building, boundary setting, model debugging, and operational calculations, which requires a large amount of simulation resources, has a long cycle, and high computational costs, making it unsuitable for real-time monitoring and rapid prediction. Traditional machine learning methods are highly dependent on feature engineering, making it difficult to handle high-dimensional data and capture complex patterns in time series, and their accuracy needs to be improved. Traditional deep learning methods use deep learning models such as convolutional neural networks (CNN) and recurrent neural networks (RNN) to automatically extract complex features of vibration signals for lifespan prediction, but most of them rely on a large amount of labeled data for training, which is often difficult to obtain in practical applications.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, apparatus, and device for predicting the life of a gearbox output shaft bearing, aiming to solve the technical problem of low accuracy in predicting the life of gearbox output shaft bearings in the prior art.

[0006] To achieve the above objectives, this application provides a method for predicting the life of a gearbox output shaft bearing, the method comprising:

[0007] Vibration signals at different times during the entire life cycle of the gearbox output shaft bearing are acquired, the segmented vibration signals are converted into corresponding vibration images, and sample vibration images are extracted from the vibration images.

[0008] Based on the physical model for life prediction of the gearbox output shaft bearing, the physical model life index of the gearbox output shaft bearing is determined.

[0009] Based on the physical model lifetime index, a preset number of sample vibration images are labeled to obtain a query set consisting of labeled vibration images and a support set consisting of unlabeled vibration images.

[0010] Extract global and local features from labeled vibration images in the query set, and extract global and local features from unlabeled vibration images in the support set;

[0011] The global and local features of the query set and the global and local features of the support set are fused to obtain fused feature pairs;

[0012] Based on the fused feature pairs, the lifespan of the gearbox output shaft bearing is predicted, and the lifespan prediction results are obtained.

[0013] In one embodiment, before determining the physical model life index of the gearbox output shaft bearing based on the physical model for life prediction, the method further includes:

[0014] Obtain the performance parameters of the gearbox output shaft bearing. The performance parameters should include at least the rated dynamic load, actual dynamic load, life index, engine speed, reliability correction factor, and life correction factor.

[0015] Based on the first correspondence between performance parameters and physical model life indicators, a physical model for predicting the life of the gearbox output shaft bearing is constructed.

[0016] In one embodiment, the steps of extracting global and local features of labeled vibration images in the query set and extracting global and local features of unlabeled vibration images in the support set include:

[0017] The labeled vibration images in the query set and the unlabeled vibration images in the support set are input into the local feature extractor to obtain the local features of the query set and the local features of the support set.

[0018] The labeled vibration images in the query set and the unlabeled vibration images in the support set are input into the global feature extractor to obtain the global features of the query set and the global features of the support set.

[0019] In one embodiment, the local feature extractor includes a GAT layer, a feature transformation layer, a regularization layer, and a pooling layer connected in sequence. The step of inputting labeled vibration images from the query set and unlabeled vibration images from the support set into the local feature extractor to obtain local features of the query set and local features of the support set includes:

[0020] The labeled vibration images from the query set and the unlabeled vibration images from the support set are respectively used as the first input images and input into the GAT layer for feature extraction to obtain the corresponding initial local features;

[0021] The initial local features are input into the feature transformation layer for dimension transformation, and the transformed initial local features are input into the regularization layer and pooling layer for shape adjustment, thus obtaining the local features of the query set and the local features of the support set.

[0022] In one embodiment, the global feature extractor includes a filtering layer, a convolutional layer, a max pooling layer, a DenseNet layer, a feature transformation layer, a regularization layer, and an average pooling layer connected in sequence. The steps of inputting labeled vibration images from the query set and unlabeled vibration images from the support set into the global feature extractor to obtain the global features of the query set and the global features of the support set include:

[0023] The labeled vibration images from the query set and the unlabeled vibration images from the support set are respectively used as the second input images and input into the filtering layer for filtering to obtain the corresponding high-quality images.

[0024] High-quality images are input into convolutional and max-pooling layers for feature extraction and dimensionality reduction to obtain the corresponding initial global features.

[0025] The initial global features are passed to the feature transformation layer for dimensionality transformation through the DenseNet Layer. The transformed initial local features are then input into the regularization layer and the average pooling layer to obtain the global features of the query set and the global features of the support set.

[0026] In one embodiment, the step of fusing the global and local features of the query set and the global and local features of the support set to obtain fused feature pairs includes:

[0027] The global and local features of the query set are merged to obtain the fused features of the query set.

[0028] The global and local features of the support set are merged to obtain the fused features of the support set.

[0029] The fusion features of the query set and the fusion features of the support set are concatenated to obtain a fusion feature pair.

[0030] In one embodiment, the step of predicting the lifespan of the gearbox output shaft bearing based on fused feature pairs to obtain the lifespan prediction result includes:

[0031] The fused feature pairs are input to the prediction analysis unit to obtain the probability distribution of the lifetime prediction result. The prediction analysis unit includes a first convolutional layer, a first normalization layer, a first activation layer, a first pooling layer, a second convolutional layer, a second normalization layer, a second activation layer, a second pooling layer, a Dropout layer, and a fully connected layer connected in sequence.

[0032] The lifetime prediction results are determined based on the probability distribution of the lifetime prediction results.

[0033] In one embodiment, after the step of predicting the lifespan of the gearbox output shaft bearing based on fused feature pairs and obtaining the lifespan prediction result, the method further includes:

[0034] Maintenance recommendations are determined based on the life prediction results of the gearbox output shaft bearing.

[0035] Furthermore, to achieve the above objectives, this application also proposes a gearbox output shaft bearing life prediction device, which includes:

[0036] The feature extraction module is used to acquire vibration signals at different times during the entire life cycle of the gearbox output shaft bearing, convert the segmented vibration signals into corresponding vibration images, and extract sample vibration images from the vibration images.

[0037] The feature extraction module is also used to determine the physical model life index of the gearbox output shaft bearing based on the physical model of the gearbox output shaft bearing.

[0038] The feature extraction module is also used to label a preset number of sample vibration images based on the physical model lifetime index, so as to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images.

[0039] The feature extraction module is also used to extract global and local features of labeled vibration images in the query set, and to extract global and local features of unlabeled vibration images in the support set.

[0040] The feature fusion module is used to fuse the global and local features of the query set and the global and local features of the support set to obtain fused feature pairs.

[0041] The life prediction module is used to predict the life of the gearbox output shaft bearing based on fused feature pairs, and obtain the life prediction result.

[0042] In addition, to achieve the above objectives, this application also proposes a gearbox output shaft bearing life prediction device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the gearbox output shaft bearing life prediction method described above.

[0043] In addition, to achieve the above objectives, the present invention also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the gearbox output shaft bearing life prediction method described above.

[0044] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the gearbox output shaft bearing life prediction method described above.

[0045] This application provides a method for predicting the lifespan of a gearbox output shaft bearing. The method involves acquiring vibration signals at different times throughout the entire lifespan of the gearbox output shaft bearing, converting the segmented vibration signals into corresponding vibration images, and extracting sample vibration images from these images. Based on a physical model for predicting the lifespan of the gearbox output shaft bearing, a physical model life index is determined. Based on the physical model life index, a preset number of sample vibration images are labeled to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images. Global and local features of the labeled vibration images in the query set are extracted, as are global and local features of the unlabeled vibration images in the support set. The global and local features of the query set and the support set are fused to obtain a fused feature pair. Based on the fused feature pair, the lifespan of the gearbox output shaft bearing is predicted to obtain the lifespan prediction result. This application converts vibration signals into vibration images, then extracts a portion of these vibration images as sample data for preliminary life prediction. A small number of vibration images labeled with physical model life indicators are used as the query set, while a large number of unlabeled vibration images are used as the support set. This process fully extracts the local and global features of the samples. Then, based on the fused feature information, predictive analysis is performed to obtain the final prediction result. This approach combines the advantages of deep learning and physical models, improving the accuracy and reliability of the prediction. It also fully extracts and fuses feature information, improving the model's accuracy and generalization ability. Furthermore, by conducting preliminary predictions and constructing a support set, based on the idea of ​​few-shot learning, the reliance on a large amount of labeled data is reduced, further improving the model's generalization ability. This solves the technical problem of low accuracy in predicting the life of gearbox output shaft bearings. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart illustrating an embodiment of the gearbox output shaft bearing life prediction method of this application.

[0049] Figure 2 This is a schematic diagram of SDP image conversion for the gearbox output shaft bearing life prediction method provided in Embodiment 1 of this application.

[0050] Figure 3 A schematic diagram of the local feature extractor and global feature extractor structure of the gearbox output shaft bearing life prediction method provided in Embodiment 1 of this application;

[0051] Figure 4 A schematic diagram of the predictive analysis unit structure for the gearbox output shaft bearing life prediction method provided in Embodiment 1 of this application;

[0052] Figure 5 This is a flowchart illustrating Embodiment 2 of the method for predicting the life of the gearbox output shaft bearing in this application;

[0053] Figure 6 A simplified flowchart illustrating the gearbox output shaft bearing life prediction method provided in Embodiment 2 of this application;

[0054] Figure 7 This is a schematic diagram of the module structure of the gearbox output shaft bearing life prediction device according to an embodiment of this application;

[0055] Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the gearbox output shaft bearing life prediction method in this application embodiment.

[0056] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0057] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0058] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0059] The main solution of this application embodiment is as follows: Vibration signals at different times throughout the entire lifespan of the gearbox output shaft bearing are acquired; the segmented vibration signals are converted into corresponding vibration images; sample vibration images are extracted from the vibration images; based on the physical model for life prediction of the gearbox output shaft bearing, the physical model life index of the gearbox output shaft bearing is determined; based on the physical model life index, a preset number of sample vibration images are labeled to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images; global and local features of the labeled vibration images in the query set are extracted, and global and local features of the unlabeled vibration images in the support set are extracted; the global and local features of the query set and the support set are fused to obtain a fused feature pair; based on the fused feature pair, the lifespan of the gearbox output shaft bearing is predicted to obtain the lifespan prediction result.

[0060] This application provides a solution that converts vibration signals into vibration images, then extracts a portion of these vibration images as sample data for preliminary life prediction. A small number of vibration images labeled with physical model life indicators are used as the query set, while a large number of unlabeled vibration images are used as the support set. This process fully extracts local and global features from the samples, and then performs predictive analysis based on the fused feature information to obtain the final prediction result. This approach combines the advantages of deep learning and physical models, improving the accuracy and reliability of the prediction. It also fully extracts and fuses feature information, enhancing the model's accuracy and generalization ability. Furthermore, by conducting preliminary predictions and constructing a support set, based on the idea of ​​few-shot learning, it reduces the reliance on a large amount of labeled data, further improving the model's generalization ability and solving the technical problem of low accuracy in predicting the life of gearbox output shaft bearings.

[0061] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a gearbox output shaft bearing life prediction device. This embodiment does not specifically limit it in this regard. The following uses a gearbox output shaft bearing life prediction device as an example to describe this embodiment and the following embodiments.

[0062] This application provides a method for predicting the lifespan of a gearbox output shaft bearing, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the gearbox output shaft bearing life prediction method of this application.

[0063] In this embodiment, the method for predicting the life of the gearbox output shaft bearing includes steps S10 to S60:

[0064] Step S10: Obtain vibration signals at different times during the entire life cycle of the gearbox output shaft bearing, convert the segmented vibration signals into corresponding vibration images, and extract sample vibration images from the vibration images.

[0065] It should be noted that this embodiment collects vibration signals at different times throughout the entire life cycle of the gearbox output shaft bearing for predicting the bearing's lifespan. The vibration signal data can be angular acceleration data collected by sensors, or it can be flexibly adjusted according to actual conditions; no specific limitation is made therein.

[0066] Additionally, it should be noted that the collected vibration signals need to be segmented. The length of the segment can be set according to actual needs and is not specifically limited. In this embodiment, the collected vibration signals are segmented according to a length of 2560.

[0067] It is understandable that the segmented vibration signal can be converted into a vibration image using the SDP conversion method, or other suitable conversion methods can be used; no specific limitation is made in this regard.

[0068] It should be understood that data extraction from the vibration images is necessary to obtain sample data for lifespan prediction, i.e., sample vibration images. Random sampling can typically be used for extraction, but it can also be flexibly adjusted according to actual needs; there is no specific limitation on this method. The number of images extracted is usually μN, where N is the total number of vibration images and μ is the set extraction ratio. The specific value of the extraction ratio can be set according to the actual situation; there is no specific limitation on this method. This process extracts a certain number of vibration signals as sample vibration images.

[0069] In one feasible implementation, the SDP conversion method is used to convert the segmented vibration signal into a vibration image.

[0070] It should be noted that the vibration image obtained at this time is an SDP image.

[0071] It is understandable that, given two selected time intervals of l, the amplitudes of the angular acceleration of the gearbox output shaft bearing are respectively D. n D n+l By using SDP transformation, the amplitude D of the gearbox output shaft bearing angular acceleration at time n can be obtained. n Converting to polar coordinates, let's denote a point as P{r(n), δ(n), φ(n)}, where r(n) is the radius of point P in polar coordinates, i.e., the distance between point P and the pole O, called the polar radius of point P; δ(n) is the angle by which point P rotates counterclockwise along the polar axis OX; and φ(n) is the angle by which point P rotates clockwise along the polar axis OX, called the polar angle of point P. Figure 2 As shown.

[0072] It can be seen that the characteristic parameters of the polar radius and polar angle together determine the position of point P in polar coordinates, satisfying the calculation relationship:

[0073]

[0074] Among them, D min D represents the minimum amplitude of the gearbox output shaft bearing angular acceleration over a time series. max Let θ be the maximum amplitude of the angular acceleration of the gearbox output shaft bearing in the time series, l be the time interval, and κ be the angle magnification factor. It is usually necessary to reasonably select the initial value of the mirror symmetry plane rotation angle θ to avoid excessive image overlap or insufficient symmetry. The angle magnification factor κ should be smaller than the mirror symmetry plane rotation angle θ. The time series l should be within the range of 1 to 10; subtle differences between time series can be reflected by its different values. The parameter settings during SDP conversion of the vibration signal under different conditions have a certain impact on the properties of the converted image. Specific parameters can be set according to the actual situation to ensure the quality of the SDP image of the vibration signal. In this embodiment, the mirror symmetry plane rotation angle θ can be set to 60°, the angle magnification factor κ to 40°, and the time interval l to 5. Based on this, an SDP image (vibration image) of the vibration signal of the gearbox output shaft bearing is obtained. The changes in the frequency components of the vibration signal will be reflected in the SDP image, enabling the differentiation of different vibration signals.

[0075] The original vibration signal is converted into structured data (SDP image) by the signal processing unit, which not only preserves the key information of the signal, but also facilitates subsequent data analysis and processing.

[0076] Step S20: Based on the physical model for life prediction of the gearbox output shaft bearing, determine the physical model life index of the gearbox output shaft bearing.

[0077] It should be noted that the physical model for predicting the lifespan of the transmission output shaft bearing is the same physical model used to predict the lifespan of the transmission output shaft bearing. The physical model's lifespan index focuses on the performance parameters of the transmission output shaft bearing. The lifespan prediction physical model calculates and evaluates the lifespan of the transmission output shaft bearing. Generally, transmission output shaft bearings are typically rolling bearings.

[0078] In one feasible implementation, the step of constructing the physical model life index of the gearbox output shaft bearing may include: obtaining the performance parameters of the gearbox output shaft bearing; and constructing a life prediction physical model of the gearbox output shaft bearing based on a first correspondence between the performance parameters and the physical model life index.

[0079] It should be noted that the lifespan of a transmission output shaft bearing typically requires comprehensive consideration of its design, operation, and performance evaluation requirements, including various influencing factors such as workload, operating conditions, lubrication method, lubrication status, material properties, and geometry. This embodiment utilizes performance parameters to construct a lifespan prediction physical model. The selected performance parameters include at least rated dynamic load, actual dynamic load, lifespan index, engine speed, reliability correction factor, and lifespan correction factor. These parameters provide comprehensive input data for the lifespan prediction and performance evaluation of the transmission output shaft bearing.

[0080] It is understandable that the first correspondence between performance parameters and physical model lifetime indices, i.e., the calculation formula satisfied by the physical model lifetime indices, is as follows:

[0081]

[0082] In the formula, L2 is the physical model lifetime index, expressed in hours, and C d It is the rated dynamic load of the gearbox output shaft bearing, F equ ε is the actual dynamic load of the gearbox output shaft bearing, ε is the life index, n is the engine speed (in revolutions per minute), β1 is the reliability correction factor for the gearbox output shaft bearing, and β2 is the life correction factor for the gearbox output shaft bearing.

[0083] Actual dynamic load F of the gearbox output shaft bearing equ This refers to the load that a bearing bears under actual working conditions. It represents the stress condition of the gearbox output shaft bearing and affects the fatigue life of the gearbox output shaft bearing material. It can be expressed by the formula F. equ =X·F r +Y·F a Perform the calculation, where F r For radial load, F a X represents the axial load, X is the radial load factor, and Y is the axial load factor. The rated dynamic load C of the gearbox output shaft bearing. d The life index (ε) is a theoretical value determined by the bearing manufacturer based on the bearing type, structural dimensions, material properties, and manufacturing process. It represents the maximum load the bearing can withstand under specific conditions without permanent deformation and is provided in the bearing's product catalog or technical documentation. The life index ε is a constant related to the bearing material properties and is usually provided by the bearing manufacturer. The value of the life index may differ for different types of rolling bearings; for example, the life index for ball bearings is typically 3, while the life index for roller bearings is typically 10 / 3, depending on the type and size of the rollers. Engine speed n typically affects the operating condition of the gearbox output shaft bearing, influencing its average pressure and lubrication status.

[0084] The reliability correction factor β1 for the gearbox output shaft bearing is related to reliability and is usually provided by the bearing manufacturer. Specific values ​​can be found in Table 1 (this is just an example; other suitable values ​​can also be set). 90% reliability refers to the expected fatigue life reached or exceeded by 90% of a sufficiently large number of identical bearings operating under the same conditions, also known as the rated life L. 10 Correspondingly, 95% reliability refers to the expected fatigue life (also known as rated life L5) of a sufficiently large number of identical bearings operating under the same conditions, where 95% of the bearings do not experience fatigue failure. 10 As a measure of bearing life, the definition This can be used to determine the bearing size sufficient to avoid fatigue failure. For a given load, the bearing size calculated using high reliability (e.g., 99%) and a reliability correction factor may result in an oversized bearing. Higher reliability leads to larger calculated bearing sizes; therefore, the reliability and reliability correction factor need to be flexibly adjusted according to the actual application. The life correction factor β2 for the gearbox output shaft bearing is obtained from the relationship curve between the dynamic viscosity ratio of the lubricating oil, the lubricating oil contamination coefficient, and the fatigue load limit value of the gearbox output shaft bearing, and is provided by the bearing manufacturer.

[0085] Table 1

[0086] reliability(%) Rated life <![CDATA[Reliability correction factor β1]]> 90 <![CDATA[L 10 ]]> 1 95 <![CDATA[L5]]> 0.62 96 <![CDATA[L4]]> 0.53 97 <![CDATA[L3]]> 0.44 98 <![CDATA[L2]]> 0.33 99 <![CDATA[L1]]> 0.21

[0087] It should be understood that bearing life is actually affected by a combination of factors, such as bearing type, lubrication conditions, bearing installation accuracy, fatigue characteristics of bearing materials, crankshaft dynamics, firing order of multi-cylinder engines, bearing geometry, lubrication conditions, bearing clearance, material properties, and environmental factors. Therefore, in practical applications, when constructing a physical model for predicting the life of the gearbox output shaft bearing, factors such as bearing fatigue life, wear life, and lubrication life can also be considered. The calculation formulas used in the model can be flexibly adjusted according to the actual situation to ensure the reliability and performance optimization of the gearbox output shaft bearing design.

[0088] Step S30: Based on the physical model lifetime index, a preset number of sample vibration images are labeled to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images.

[0089] It should be noted that in this embodiment, the physical model lifetime index is used to annotate a small number of sample vibration images. Annotated vibration images are the sample vibration images that have been annotated, while unannotated vibration images are the sample vibration images that have not been annotated.

[0090] Additionally, it should be noted that the preset quantity is the number of labeled samples set, which is usually small and can be flexibly adjusted according to actual needs. There is no specific limit to this. The number of labeled vibration images is much smaller than the number of unlabeled vibration images.

[0091] It is understood that in this embodiment, a small number of labeled vibration images are used as the query set, and a large number of other unlabeled vibration images are used as the support set. In other words, the query set consists of labeled vibration images, and the support set consists of unlabeled vibration images.

[0092] Step S40: Extract the global and local features of the labeled vibration images in the query set, and extract the global and local features of the unlabeled vibration images in the support set.

[0093] It should be noted that this embodiment constructs a local feature extractor and a global feature extractor respectively. The local feature extractor is used to extract local features of the labeled vibration images in the query set and supports local features of the unlabeled vibration images in the set. The global feature extractor is used to extract global features of the labeled vibration images in the query set and supports global features of the unlabeled vibration images in the set.

[0094] In one feasible implementation, step S40 may include:

[0095] Step S401: Input the labeled vibration images in the query set and the unlabeled vibration images in the support set into the local feature extractor to obtain the local features of the query set and the local features of the support set.

[0096] refer to Figure 3 The local feature extractor comprises a GAT (Graph Attention Network) layer, a feature transformation layer, a regularization layer, and a pooling layer connected in sequence. In this embodiment, two consecutive GAT layers (GATBlock) are used.

[0097] In one feasible implementation, the step of inputting labeled vibration images from the query set and unlabeled vibration images from the support set into a local feature extractor to obtain local features of the query set and local features of the support set includes: inputting labeled vibration images from the query set and unlabeled vibration images from the support set into a GAT layer as first input images for feature extraction to obtain corresponding initial local features; inputting the initial local features into a feature transformation layer for dimension transformation, and inputting the transformed initial local features into a regularization layer and a pooling layer for shape adjustment to obtain local features of the query set and local features of the support set.

[0098] It should be noted that the first input image is the image that is input into the local feature extractor for feature extraction. In this embodiment, both the query set and the support set images need to be input into the local feature extractor for feature extraction. Therefore, the labeled vibration images in the query set and the unlabeled vibration images in the support set are respectively used as the first input images.

[0099] Understandably, the first input image is first fed into the GAT layer for feature extraction. During this process, each pixel of the first input image is treated as a node in the graph, and a corresponding adjacency matrix is ​​constructed. The adjacency matrix represents the connection relationships between nodes. Since the GAT layer is a GNN (Graph Neural Network) based on an attention mechanism, it can dynamically assign weights to the neighbors of each node to aggregate information and update the node's features. Thus, the initial local features of the first input image can be extracted through two GAT layers. The two GAT layers in this embodiment enhance the image's feature representation capability and can more comprehensively aggregate information. The first GAT layer extracts preliminary local features, while the second GAT module further extracts higher-level local features based on these features. Each GAT layer uses an attention mechanism to weighted aggregate the neighbors of a node. Successively set GAT layers can further weight and aggregate these aggregation results, thereby capturing more complex relationships between nodes. Next, the initial local features are input into a feature transformation layer for dimensionality transformation, typically by reducing the dimensionality of the initial local features to facilitate subsequent processing. The dimensionality reduction method can be chosen according to actual needs; for example, Flatten dimensionality reduction can be used to flatten multidimensional data into one-dimensional data. Then, the transformed initial local features are input into a regularization layer and a pooling layer. The regularization layer uses batch normalization to reduce overfitting, and the pooling layer uses average pooling, which can adjust the pooling window size according to the size of the input data to maintain consistency in feature dimensionality. Finally, the output of the pooling layer is the local feature of the first input image. In other words, by using labeled vibration images from the query set as the first input image, the local features of the labeled vibration images in the query set can be obtained; similarly, by using unlabeled vibration images from the support set as the first input image, the local features of the unlabeled vibration images in the support set can be obtained.

[0100] Step S402: Input the labeled vibration images in the query set and the unlabeled vibration images in the support set into the global feature extractor to obtain the global features of the query set and the global features of the support set.

[0101] refer to Figure 4The global feature extractor includes a filtering layer, a convolutional layer, a feature transformation layer, a regularization layer, two DenseNet layers, and two pooling layers. The first pooling layer is a max pooling layer, and the second pooling layer is an average pooling layer. The two DenseNet layers are set consecutively, that is, the filtering layer, convolutional layer, max pooling layer, DenseNetLayer layer, feature transformation layer, regularization layer, and average pooling layer are connected in sequence.

[0102] It should be noted that the second input image is the image that is input into the global feature extractor for feature extraction. In this embodiment, both the query set and the support set images need to be input into the global feature extractor for feature extraction. Therefore, the labeled vibration images in the query set and the unlabeled vibration images in the support set are used as the second input images, respectively.

[0103] Understandably, the second input image is first input to the filtering layer for filtering processing to remove noise and improve the overall image quality, thereby outputting a higher quality image. During this process, the filtering method can be selected according to actual needs. For example, this embodiment uses Gaussian filtering, and the output of the filtering layer is:

[0104]

[0105] In the formula, g represents the output of the Gaussian filter, and w i b represents the filter weights. i σ represents the filter bias, x represents the input data, i.e., the second input image, and σ represents the filter standard deviation.

[0106] It should be understood that inputting the high-quality image into the convolutional and max-pooling layers can effectively extract features from the high-quality image and reduce the dimensionality of the feature map, achieving feature extraction and dimensionality reduction of the high-quality image. At this point, the corresponding features, i.e., the initial global features, can be initially obtained. The convolutional layer uses a 1×1 convolutional kernel. Subsequently, the initial global features are input into the DenseNet layer. The DenseNet layer can utilize the densely connected convolutional network structure to achieve efficient feature reuse, reduce the gradient vanishing problem, and enable the network to learn features more deeply. At the same time, using consecutive DenseNet layers can continuously fuse features from different layers, which helps the model capture richer and more diverse feature representations, enhances the model's expressive power, and can effectively transfer and retain feature information. Thus, the initial global features are passed to the feature transformation layer, which performs dimensionality transformation on the initial local features, usually by reducing the dimensionality of the initial local features, to facilitate subsequent processing.

[0107] In few-shot learning, due to the limited amount of training data, the model is prone to overfitting. To improve this problem, a regularization layer is added to reduce overfitting. For example, during training, a portion of features are randomly discarded to increase the model's randomness and robustness, improve its generalization ability, and avoid overfitting. This process can be represented as follows:

[0108] y=X⊙D

[0109] In the formula, X represents the input feature matrix, ⊙ represents element-wise multiplication, D is the regularization matrix, and represents the Bernoulli distribution calculated based on the regularized probability distribution. An average pooling layer is also included, which reduces the computational cost of the model while preserving the mean information in the feature map. Therefore, after passing through the regularization layer and the average pooling layer, the final output of the global feature extractor is obtained, which is the local feature of the second input image. In other words, by using the labeled vibration images from the query set as the second input image, the global features of the labeled vibration images in the query set can be obtained; similarly, by using the unlabeled vibration images from the support set as the second input image, the global features of the unlabeled vibration images from the support set can be obtained.

[0110] By using a few-shot learning mechanism, the reliance on a large amount of labeled data can be reduced, enabling the model to learn effective feature representations under limited data conditions.

[0111] Step S50: The global and local features of the query set and the global and local features of the support set are fused to obtain fused feature pairs;

[0112] In one feasible implementation, step S50 may include steps S501 to S503:

[0113] Step S501: Merge the global features and local features of the query set to obtain the fused features of the query set;

[0114] It is understandable that fused features are features resulting from the fusion of global and local features. In this embodiment, the global and local features of the query set are merged to generate the fused features of the query set. The calculation formula is as follows:

[0115] f s (x)=Cat(f sp (x),f sa (x))

[0116] In the formula, f s (x) represents the fusion feature of the query set, f sa (x) represents the global feature of the query set, f sp (x) represents a local feature of the query set, and the Cat() operation merges the features in the last dimension.

[0117] Step S502: Merge the global features and local features of the support set to obtain the fused features of the support set;

[0118] It is understood that this embodiment merges the global and local features of the support set to generate the fused features of the support set, and the calculation formula is as follows:

[0119] f q (x)=Cat(f qp (x),f qa (x))

[0120] In the formula, f q (x) represents the fusion feature of the support set, f qa (x) represents the global feature of the support set, f qp (x) represents a local feature of the support set, and the Cat() operation merges the features in the last dimension.

[0121] Step S503: Concatenate the fusion features of the query set and the fusion features of the support set to obtain a fusion feature pair.

[0122] Understandably, the fused features of the query set and the fused features of the support set are concatenated into feature pairs, i.e., fused feature pairs, as shown below:

[0123] f con =C(f) s (x),f q (x))

[0124] In the formula, f con f represents the fused feature pair. s (x) represents the fusion feature of the query set, f q (x) represents the fusion feature of the support set.

[0125] Step S60: Based on the fused feature pairs, the lifespan of the gearbox output shaft bearing is predicted to obtain the lifespan prediction result.

[0126] It should be noted that the final lifetime prediction result is obtained by using fusion features. The lifetime prediction result is usually the predicted remaining lifetime, which can be expressed in numerical form, such as 75% or 80%.

[0127] In one feasible implementation, step S60 may include steps S601 to S602:

[0128] Step S601: Input the fused feature pairs into the prediction analysis unit to obtain the probability distribution of the lifetime prediction results;

[0129] It should be noted that the fused feature pairs are input into the predictive analysis unit, which then predicts the lifespan of the gearbox output shaft bearing to generate a probability distribution of its remaining lifespan. (Reference) Figure 3 The prediction and analysis unit includes a first convolutional layer, a first normalization layer, a first activation layer, a first pooling layer, a second convolutional layer, a second normalization layer, a second activation layer, a second pooling layer, a Dropout layer, and a fully connected layer, connected in sequence.

[0130] The first convolutional layer has a kernel size of 5×5 and a number of 64 kernels. It can extract important features from the fused feature pairs. The first normalization layer can normalize the features input to the first convolutional layer, reduce internal covariate bias, and accelerate the training process. The first activation layer uses the ReLU activation function and outputs the features normalized by the first normalization layer. The first pooling layer is a max pooling layer. After the max pooling layer, the second convolutional layer can further extract deeper features. Then, it passes through the second normalization layer and the second activation layer, followed by the second pooling layer (average pooling layer) to reduce the feature dimensionality. Finally, it passes through the Dropout layer and the fully connected layer for predictive analysis, providing the corresponding probability distribution.

[0131] In this process, cross-entropy is used as the loss function, which measures the difference between two probability distributions. The parameters are continuously adjusted using the loss function to ultimately obtain the trained predictive analysis unit, namely the life prediction model for the gearbox output shaft bearing. This allows us to obtain the probability distribution of the life prediction results.

[0132] By optimizing the loss function to adjust the model parameters, the accuracy of predictions and the convergence of the model can be further enhanced.

[0133] Step S602: Determine the lifetime prediction result based on the probability distribution of the lifetime prediction result.

[0134] Understandably, by finding the category with the highest probability based on the probability distribution of remaining lifespan, the final lifespan prediction result can be obtained.

[0135] This embodiment provides a method for predicting the lifespan of a gearbox output shaft bearing. The method involves acquiring vibration signals at different times throughout the entire lifespan of the gearbox output shaft bearing, converting the segmented vibration signals into corresponding vibration images, and extracting sample vibration images from these images. Based on a physical model for predicting the lifespan of the gearbox output shaft bearing, a physical model lifespan index is determined. Based on the physical model lifespan index, a preset number of sample vibration images are labeled to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images. Global and local features of the labeled vibration images in the query set are extracted, as are global and local features of the unlabeled vibration images in the support set. The global and local features of the query set and the support set are fused to obtain a fused feature pair. Based on the fused feature pair, the lifespan of the gearbox output shaft bearing is predicted to obtain the lifespan prediction result. Vibration signals are converted into vibration images. Then, a portion of these vibration images are extracted as sample data for preliminary lifetime prediction. A small number of vibration images labeled with physical model lifetime indices are used as the query set, while a large number of unlabeled vibration images are used as the support set. This process fully extracts the local and global features of the samples. Then, predictive analysis is performed based on the fused feature information to obtain the final prediction result. This approach combines the advantages of deep learning and physical models, improving the accuracy and reliability of predictions. It also fully extracts and fuses feature information, improving the model's accuracy and generalization ability. Furthermore, by conducting preliminary predictions and constructing the support set, based on the idea of ​​few-shot learning, the dependence on a large amount of labeled data is reduced, further enhancing the model's generalization ability.

[0136] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 5 Step S60 may be followed by step S70:

[0137] Step S70: Based on the life prediction results of the gearbox output shaft bearing, determine maintenance recommendations;

[0138] It should be noted that after obtaining the life prediction results of the gearbox output shaft bearing, relative maintenance suggestions are given based on the prediction results.

[0139] When the remaining bearing life is 75% or greater, maintenance recommendations typically include both preventative maintenance and performance monitoring. At this stage, the bearing still has a long lifespan, so the focus should be on preventative maintenance, such as regular lubrication, checking the bearing's cleanliness and lubrication condition to ensure there are no signs of wear or damage. Furthermore, performance monitoring of the bearing should be implemented or intensified, recording key indicators such as bearing temperature and vibration to detect potential problems early.

[0140] When the remaining bearing life is greater than or equal to 50% but less than 75%, maintenance recommendations typically include two aspects: enhanced performance monitoring and a bearing replacement plan. Enhanced performance monitoring can involve increasing monitoring frequency and using more advanced monitoring technologies, such as online vibration analysis, to more accurately predict bearing wear. For the bearing replacement plan, consideration can begin to be given to developing a replacement schedule, including determining the replacement time, budget, and required resources.

[0141] When the remaining service life is greater than or equal to 25% but less than 50%, maintenance recommendations typically include two aspects: periodic inspection and design improvement. Periodic inspection involves shortening the inspection cycle and checking the bearing's condition more frequently, including checking bearing wear, lubrication, and the operating environment. Design improvement involves considering design modifications to the bearing or its operating environment to extend its service life.

[0142] When the remaining lifespan is less than 25%, maintenance recommendations typically include two aspects: an emergency replacement plan and spare parts inventory. An emergency replacement plan can be developed to ensure timely replacement before the bearing reaches the end of its lifespan. Furthermore, ensuring sufficient spare parts inventory allows for rapid replacement in the event of bearing failure, minimizing downtime.

[0143] This embodiment provides a method for predicting the life of a gearbox output shaft bearing. The vibration signal is converted into a vibration image, and then a portion of the vibration images are extracted as sample data for preliminary life prediction. A small number of vibration images labeled with physical model life indicators are used as the query set, and a large number of unlabeled vibration images are used as the support set. The local and global features of the samples are fully extracted, and then predictive analysis is performed based on the fused feature information to obtain the final prediction result. Based on the result, relative maintenance suggestions are given to ensure the safe use of the gearbox output shaft bearing.

[0144] For example, to help understand the implementation process of the gearbox output shaft bearing life prediction method obtained by combining this embodiment with the above-described embodiment two, please refer to... Figure 6 , Figure 6 A simplified flowchart of a method for predicting the life of a gearbox output shaft bearing is provided, specifically:

[0145] Step 1: Data preparation, collecting vibration signals at different times during the entire life cycle of the gearbox output shaft bearing, and segmenting them;

[0146] Step 2: The signal processing unit converts the collected vibration signals into vibration images;

[0147] Step 3: The data extraction unit extracts samples from a large number of vibration images through random sampling;

[0148] Step 4: Input the extracted vibration images into the physical life analysis unit for preliminary life prediction.

[0149] Step 5: Using a small number of vibration images labeled with the physical model's life index as the query set and a large number of unlabeled vibration images as the support set, input the query set and support set into the fusion feature extraction unit to extract the local and global features of each dataset.

[0150] Step 6: Fuse the extracted global and local features of the query set and the global and local features of the support set to obtain fused feature pairs;

[0151] Step 7: Input the fused feature pairs into the predictive analysis unit, perform predictive analysis, give the probability distribution of lifetime prediction, and determine the final lifetime prediction result.

[0152] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the gearbox output shaft bearing life prediction method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0153] This application also provides a gearbox output shaft bearing life prediction device, please refer to... Figure 7 The gearbox output shaft bearing life prediction device includes:

[0154] The feature extraction module 10 is used to acquire vibration signals at different times during the entire life cycle of the gearbox output shaft bearing, convert the segmented vibration signals into corresponding vibration images, and extract sample vibration images from the vibration images.

[0155] The feature extraction module 10 is also used to determine the physical model life index of the gearbox output shaft bearing based on the physical model for life prediction of the gearbox output shaft bearing.

[0156] The feature extraction module 10 is also used to annotate a preset number of sample vibration images based on the physical model lifetime index, so as to obtain a query set composed of annotated vibration images and a support set composed of unannotated vibration images.

[0157] The feature extraction module 10 is also used to extract global and local features of the labeled vibration images in the query set, and to extract global and local features of the unlabeled vibration images in the support set.

[0158] The feature fusion module 20 is used to fuse the global and local features of the query set and the global and local features of the support set to obtain fused feature pairs.

[0159] The life prediction module 30 is used to predict the life of the gearbox output shaft bearing based on the fused feature pairs, and obtain the life prediction result.

[0160] In one feasible implementation, the feature extraction module 10 is also used to obtain the performance parameters of the gearbox output shaft bearing, the performance parameters including at least the rated dynamic load, actual dynamic load, life index, engine speed, reliability correction factor and life correction factor;

[0161] Based on the first correspondence between performance parameters and physical model life indicators, a physical model for predicting the life of the gearbox output shaft bearing is constructed.

[0162] In one feasible implementation, the feature extraction module 10 is further configured to input the labeled vibration images in the query set and the unlabeled vibration images in the support set into the local feature extractor respectively to obtain the local features of the query set and the local features of the support set.

[0163] The labeled vibration images in the query set and the unlabeled vibration images in the support set are input into the global feature extractor to obtain the global features of the query set and the global features of the support set.

[0164] In one feasible implementation, the local feature extractor includes a GAT layer, a feature transformation layer, a regularization layer and a pooling layer connected in sequence. The feature extraction module 10 is also used to input the labeled vibration images in the query set and the unlabeled vibration images in the support set as the first input images into the GAT layer for feature extraction to obtain the corresponding initial local features.

[0165] The initial local features are input into the feature transformation layer for dimension transformation, and the transformed initial local features are input into the regularization layer and pooling layer for shape adjustment, thus obtaining the local features of the query set and the local features of the support set.

[0166] In one feasible implementation, the global feature extractor includes a filter layer, a convolutional layer, a max pooling layer, a DenseNet Layer, a feature transformation layer, a regularization layer, and an average pooling layer connected in sequence. The feature extraction module 10 is also used to input the labeled vibration images in the query set and the unlabeled vibration images in the support set as second input images into the filter layer for filtering processing to obtain the corresponding high-quality images.

[0167] High-quality images are input into convolutional and max-pooling layers for feature extraction and dimensionality reduction to obtain the corresponding initial global features.

[0168] The initial global features are passed to the feature transformation layer for dimensionality transformation through the DenseNet Layer. The transformed initial local features are then input into the regularization layer and the average pooling layer to obtain the global features of the query set and the global features of the support set.

[0169] In one feasible implementation, the feature fusion module 20 is further used to merge the global features and local features of the query set to obtain the fused features of the query set.

[0170] The global and local features of the support set are merged to obtain the fused features of the support set.

[0171] The fusion features of the query set and the fusion features of the support set are concatenated to obtain a fusion feature pair.

[0172] In one feasible implementation, the lifetime prediction module 30 is further used to input the fused feature pairs into the prediction analysis unit to obtain the probability distribution of the lifetime prediction result. The prediction analysis unit includes a first convolutional layer, a first normalization layer, a first activation layer, a first pooling layer, a second convolutional layer, a second normalization layer, a second activation layer, a second pooling layer, a Dropout layer, and a fully connected layer connected in sequence.

[0173] The lifetime prediction results are determined based on the probability distribution of the lifetime prediction results.

[0174] In one feasible implementation, the life prediction module 30 is also used to determine maintenance recommendations based on the life prediction results of the gearbox output shaft bearing.

[0175] The gearbox output shaft bearing life prediction device provided in this application employs the gearbox output shaft bearing life prediction method described in the above embodiments, which can solve the technical problem of low accuracy in predicting the life of gearbox output shaft bearings. Compared with the prior art, the beneficial effects of the gearbox output shaft bearing life prediction device provided in this application are the same as those of the gearbox output shaft bearing life prediction method provided in the above embodiments, and other technical features in the gearbox output shaft bearing life prediction device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0176] This application provides a gearbox output shaft bearing life prediction device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the gearbox output shaft bearing life prediction method in the above embodiment 1.

[0177] The following is for reference. Figure 8The diagram illustrates a structural schematic suitable for implementing the gearbox output shaft bearing life prediction device according to embodiments of this application. The gearbox output shaft bearing life prediction device in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The gearbox output shaft bearing life prediction device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0178] like Figure 8 As shown, the gearbox output shaft bearing life prediction device may include a processing unit 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the gearbox output shaft bearing life prediction device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the gearbox output shaft bearing life prediction device to communicate wirelessly or wiredly with other devices to exchange data. Although a gearbox output shaft bearing life prediction device with various systems is shown in the figure, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0179] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0180] The gearbox output shaft bearing life prediction device provided in this application, employing the gearbox output shaft bearing life prediction method described in the above embodiments, can solve the technical problem of low accuracy in gearbox output shaft bearing life prediction. Compared with the prior art, the beneficial effects of the gearbox output shaft bearing life prediction device provided in this application are the same as those of the gearbox output shaft bearing life prediction method provided in the above embodiments, and other technical features in this gearbox output shaft bearing life prediction device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0181] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0182] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0183] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the gearbox output shaft bearing life prediction method in the above embodiments.

[0184] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0185] The aforementioned computer-readable storage medium may be included in the gearbox output shaft bearing life prediction device; or it may exist independently and not be assembled into the gearbox output shaft bearing life prediction device.

[0186] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the gearbox output shaft bearing life prediction device, the gearbox output shaft bearing life prediction device performs the following actions: acquires vibration signals at different times throughout the entire life cycle of the gearbox output shaft bearing, converts the segmented vibration signals into corresponding vibration images, and extracts sample vibration images from the vibration images; determines the physical model life index of the gearbox output shaft bearing based on the physical model life index; annotates a preset number of sample vibration images based on the physical model life index to obtain a query set composed of annotated vibration images and a support set composed of unannotated vibration images; extracts global and local features from the annotated vibration images in the query set and extracts global and local features from the unannotated vibration images in the support set; fuses the global and local features of the query set and the support set to obtain a fused feature pair; and predicts the life of the gearbox output shaft bearing based on the fused feature pair to obtain a life prediction result.

[0187] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0188] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0189] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0190] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described gearbox output shaft bearing life prediction method, thereby solving the technical problem of low accuracy in gearbox output shaft bearing life prediction. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the gearbox output shaft bearing life prediction method provided in the above embodiments, and will not be repeated here.

[0191] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the gearbox output shaft bearing life prediction method described above.

[0192] The computer program product provided in this application can solve the technical problem of low accuracy in predicting the life of gearbox output shaft bearings. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the gearbox output shaft bearing life prediction method provided in the above embodiments, and will not be repeated here.

[0193] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for predicting the life of a gearbox output shaft bearing, characterized in that, The method includes: Vibration signals at different times during the entire life cycle of the gearbox output shaft bearing are acquired, the segmented vibration signals are converted into corresponding vibration images, and sample vibration images are extracted from the vibration images. Based on the physical model for life prediction of the gearbox output shaft bearing, the physical model life index of the gearbox output shaft bearing is determined. Based on the physical model lifetime index, a preset number of sample vibration images are labeled to obtain a query set composed of labeled vibration images and a support set composed of unlabeled vibration images. Extract the global and local features of the labeled vibration images in the query set, and extract the global and local features of the unlabeled vibration images in the support set; The global and local features of the query set and the global and local features of the support set are fused to obtain fused feature pairs; Based on the fused feature pairs, the lifespan of the gearbox output shaft bearing is predicted, and the lifespan prediction result is obtained.

2. The method as described in claim 1, characterized in that, Before the step of determining the physical model life index of the gearbox output shaft bearing based on the life prediction physical model of the gearbox output shaft bearing, the method further includes: Obtain the performance parameters of the gearbox output shaft bearing, including at least the rated dynamic load, actual dynamic load, life index, engine speed, reliability correction factor, and life correction factor; Based on the first correspondence between performance parameters and physical model life indicators, a physical model for predicting the life of the gearbox output shaft bearing is constructed.

3. The method as described in claim 1, characterized in that, The steps of extracting global and local features of the labeled vibration images in the query set, and extracting global and local features of the unlabeled vibration images in the support set, include: The labeled vibration images in the query set and the unlabeled vibration images in the support set are respectively input into the local feature extractor to obtain the local features of the query set and the local features of the support set. The labeled vibration images in the query set and the unlabeled vibration images in the support set are input into the global feature extractor to obtain the global features of the query set and the global features of the support set.

4. The method as described in claim 3, characterized in that, The local feature extractor includes a GAT layer, a feature transformation layer, a regularization layer, and a pooling layer connected in sequence. The step of inputting the labeled vibration images in the query set and the unlabeled vibration images in the support set into the local feature extractor to obtain the local features of the query set and the local features of the support set includes: The labeled vibration images in the query set and the unlabeled vibration images in the support set are respectively used as the first input images and input into the GAT layer for feature extraction to obtain the corresponding initial local features; The initial local features are input into the feature transformation layer for dimension transformation, and the transformed initial local features are input into the regularization layer and the pooling layer for shape adjustment to obtain the local features of the query set and the local features of the support set.

5. The method as described in claim 3, characterized in that, The global feature extractor includes a filtering layer, a convolutional layer, a max pooling layer, a DenseNet layer, a feature transformation layer, a regularization layer, and an average pooling layer connected in sequence. The step of inputting the labeled vibration images in the query set and the unlabeled vibration images in the support set into the global feature extractor to obtain the global features of the query set and the global features of the support set includes: The labeled vibration images in the query set and the unlabeled vibration images in the support set are respectively input into the filtering layer as second input images for filtering processing to obtain the corresponding high-quality images; The high-quality image is input into the convolutional layer and the max pooling layer for feature extraction and dimensionality reduction to obtain the corresponding initial global features; The initial global features are passed to the feature transformation layer for dimensionality transformation through the DenseNet Layer, and the transformed initial local features are input into the regularization layer and the average pooling layer to obtain the global features of the query set and the global features of the support set.

6. The method as described in claim 1, characterized in that, The step of fusing the global and local features of the query set and the global and local features of the support set to obtain fused feature pairs includes: The global and local features of the query set are merged to obtain the fused features of the query set; The global and local features of the support set are merged to obtain the fused features of the support set; The fusion features of the query set and the fusion features of the support set are concatenated to obtain the fusion feature pair.

7. The method as described in claim 1, characterized in that, The step of predicting the lifespan of the gearbox output shaft bearing based on the fused feature pairs to obtain the lifespan prediction result includes: The fused features are input into the prediction analysis unit to obtain the probability distribution of the lifetime prediction result. The prediction analysis unit includes a first convolutional layer, a first normalization layer, a first activation layer, a first pooling layer, a second convolutional layer, a second normalization layer, a second activation layer, a second pooling layer, a Dropout layer, and a fully connected layer connected in sequence. The lifetime prediction result is determined based on the probability distribution of the lifetime prediction result.

8. The method according to any one of claims 1 to 7, characterized in that, The step of predicting the lifespan of the gearbox output shaft bearing based on the fused feature pair and obtaining the lifespan prediction result further includes: Based on the life prediction results of the gearbox output shaft bearing, maintenance recommendations are determined.

9. A device for predicting the life of a gearbox output shaft bearing, characterized in that, The device includes: The feature extraction module is used to acquire vibration signals at different times during the entire life cycle of the gearbox output shaft bearing, convert the segmented vibration signals into corresponding vibration images, and extract sample vibration images from the vibration images. The feature extraction module is also used to determine the physical model life index of the gearbox output shaft bearing based on the physical model for life prediction of the gearbox output shaft bearing. The feature extraction module is also used to annotate a preset number of sample vibration images based on the physical model lifetime index, so as to obtain a query set composed of annotated vibration images and a support set composed of unannotated vibration images. The feature extraction module is also used to extract global and local features of the labeled vibration images in the query set, and to extract global and local features of the unlabeled vibration images in the support set. The feature fusion module is used to fuse the global and local features of the query set and the global and local features of the support set to obtain fused feature pairs. The life prediction module is used to predict the life of the gearbox output shaft bearing based on the fused feature pair, and obtain the life prediction result.

10. A device for predicting the life of a gearbox output shaft bearing, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the gearbox output shaft bearing life prediction method as described in any one of claims 1 to 8.

Images