High-speed shaft brake state detection method, device, equipment and storage medium

Abstract

The application discloses a high-speed shaft brake state detection method, device and equipment and a storage medium, belongs to the technical field of machine learning, and the application obtains thermal imaging data of a contact part of a high-speed shaft brake and a brake disc collected by a thermal imaging sensor and a vibration data sequence collected by a vibration sensor, extracts features from the thermal imaging data, obtains thermal imaging spatial features, extracts features from the vibration data sequence, obtains vibration time features, fuses the thermal imaging spatial features and the vibration time features, obtains multi-modal fusion features, inputs the multi-modal fusion features into a high-speed shaft brake state recognition model, obtains a state detection result output by the high-speed shaft brake state recognition model, and solves the technical problem that the reliability of high-speed shaft brake fault diagnosis by using temperature signals collected by a single sensor is relatively low.

Description

Technical Field

[0001] This application relates to the field of machine learning technology, and in particular to a method, apparatus, device, and storage medium for detecting the state of a high-speed shaft brake. Background Technology

[0002] If foreign objects get stuck in the gap between the brake disc and the friction pads of the high-speed shaft brake during unit operation, these faults often manifest as abnormal vibration data or abnormal brake surface temperature.

[0003] In related technologies, common fault detection methods usually use temperature sensors to detect whether the temperature of the brake pads exceeds the temperature threshold to determine whether the high-speed brake is operating abnormally. However, the reliability of high-speed axle brake fault diagnosis based on temperature signals collected by a single sensor is low, and it may even lead to incorrect judgment results under some complex working conditions. Summary of the Invention

[0004] The main objective of this application is to provide a method, device, and storage medium for detecting the condition of a high-speed shaft brake, aiming to solve the problem of low reliability in current high-speed shaft brake fault diagnosis using temperature signals collected by a single sensor.

[0005] To achieve the above objectives, in a first aspect, this application provides a method for detecting the state of a high-speed shaft brake, the method comprising:

[0006] Acquire thermal imaging data of the contact area between the high-speed axle brake and the brake disc collected by a thermal imaging sensor and vibration data sequence collected by a vibration sensor; the vibration signal data sequence includes multiple vibration data points ordered along time.

[0007] Feature extraction is performed on thermal imaging data to obtain spatial features of thermal imaging, and feature extraction is performed on vibration data sequences to obtain temporal features of vibration.

[0008] By fusing spatial features from thermal imaging and temporal features from vibration, multimodal fusion features are obtained.

[0009] The multimodal fusion features are input into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model.

[0010] Optionally, feature extraction is performed on the thermal imaging data to obtain spatial features of the thermal imaging, including:

[0011] Feature extraction of thermal imaging data is performed based on spatial attention mechanism to obtain spatial features of thermal imaging.

[0012] Optionally, feature extraction is performed on the thermal imaging data based on a spatial attention mechanism to obtain spatial features of the thermal imaging, including:

[0013] Feature extraction is performed on thermal imaging data to obtain a feature map matrix;

[0014] Determine the spatial attention vector for each feature in the feature map matrix to obtain the spatial attention matrix;

[0015] Based on the spatial attention matrix and the feature map matrix, a spatial feature enhancement map is obtained;

[0016] Feature extraction is performed on the spatial feature enhancement map to obtain the spatial features of thermal imaging.

[0017] Optionally, feature extraction is performed on the vibration data sequence to obtain vibration time features, including:

[0018] The vibration time features are obtained by extracting features from vibration data sequences based on the time attention mechanism.

[0019] Optionally, feature extraction is performed on the vibration data sequence based on a time attention mechanism to obtain vibration time features, including:

[0020] Spectral attention processing is performed on the vibration data sequence to obtain the spectral attention weights of each spectral component in the vibration data sequence;

[0021] Based on the spectral attention weights of each spectral component, the vibration data sequence is filtered to obtain vibration time characteristics.

[0022] Optionally, thermal imaging spatial features and vibration temporal features are fused to obtain multimodal fusion features, including:

[0023] Determine the first weight of the spatial features of thermal imaging and the second weight of the temporal features of vibration;

[0024] Based on the first and second weights, the spatial features of thermal imaging and the temporal features of vibration are weighted and fused to obtain multimodal fusion features.

[0025] Optionally, before inputting the multimodal fusion features into the high-speed axle brake state recognition model and obtaining the state detection results output by the high-speed axle brake state recognition model, the method further includes:

[0026] A training sample set is obtained, which includes multimodal fusion samples labeled with fault information. The multimodal fusion samples are obtained by fusing thermal imaging spatial feature samples and vibration temporal feature samples.

[0027] A support vector machine classifier is trained using the training sample set to obtain a high-speed axle brake state recognition model.

[0028] Secondly, in order to achieve the above objectives, this application further provides a high-speed shaft brake condition detection device, the high-speed shaft brake condition detection device comprising:

[0029] The acquisition module is used to acquire thermal imaging data of the contact area between the high-speed shaft brake and the brake disc collected by the thermal imaging sensor and vibration data sequence collected by the vibration sensor; the vibration signal data sequence includes multiple vibration data ordered along time.

[0030] The feature extraction module is used to extract features from thermal imaging data to obtain spatial features of thermal imaging, and to extract features from vibration data sequences to obtain temporal features of vibration.

[0031] The fusion module is used to fuse spatial features of thermal imaging and temporal vibration features to obtain multimodal fusion features;

[0032] The detection module is used to input multimodal fusion features into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model.

[0033] Thirdly, in order to achieve the above objectives, this application further provides a high-speed shaft brake state detection device, comprising: a processor, a memory, and a high-speed shaft brake state detection program stored in the memory, wherein the high-speed shaft brake state detection program is executed by the processor to implement the steps of the above-described high-speed shaft brake state detection method.

[0034] Fourthly, in order to achieve the above objectives, this application further provides a computer-readable storage medium storing a high-speed shaft brake state detection program, which, when executed by a processor, implements the above-described high-speed shaft brake state detection method.

[0035] It is easy to see that this application uses a pre-trained high-speed axle brake state recognition model to detect the working state of the high-speed axle brake. By extracting features from the acquired thermal imaging data and vibration data sequences, the extracted thermal imaging spatial features and vibration temporal features are fused, and the resulting multimodal fusion features are input into the high-speed axle brake state recognition model for working state detection. This achieves comprehensive detection of the temperature and vibration states of the high-speed axle brake during operation, solves the problem of inaccurate detection of the working state of the high-speed axle brake under some complex conditions, and improves the working reliability, safety and operating efficiency of the high-speed axle brake. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the high-speed shaft brake condition detection device of this application;

[0037] Figure 2 This is a flowchart illustrating the first embodiment of the high-speed shaft brake state detection method of this application;

[0038] Figure 3 This is a detailed flowchart of step S200 in the first embodiment of the high-speed shaft brake state detection method of this application;

[0039] Figure 4 This is a detailed flowchart of step S210 in the first embodiment of the high-speed shaft brake state detection method of this application;

[0040] Figure 5 This is a detailed flowchart of step S220 in the first embodiment of the high-speed shaft brake state detection method of this application;

[0041] Figure 6 This is a detailed flowchart of step S300 in the first embodiment of the high-speed shaft brake state detection method of this application;

[0042] Figure 7 This is a flowchart illustrating the second embodiment of the high-speed shaft brake state detection method of this application.

[0043] 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

[0044] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0045] If foreign objects get stuck in the gap between the brake disc and the friction pads of the high-speed shaft brake during unit operation, these faults often manifest as abnormal vibration data or abnormal brake surface temperature.

[0046] Currently, common fault detection methods typically use temperature thresholds as a judgment marker. Temperature sensors are used to detect whether the temperature of the brake pads exceeds the temperature threshold, thereby determining whether the high-speed brake is operating abnormally. However, the reliability of high-speed axle brake fault diagnosis based on temperature signals collected by a single sensor is low, and it may even lead to incorrect judgment results under certain complex operating conditions.

[0047] The following embodiments of this application will describe the high-speed shaft brake state detection method, apparatus, device, and storage medium used in the technical implementation of this application:

[0048] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a high-speed shaft brake status detection device in the hardware operating environment involved in the embodiments of this application.

[0049] like Figure 1As shown, the high-speed shaft brake status detection device may include: a processor 1001, such as a CPU, a user interface 1003, a memory 1005, and a communication bus 1002. The communication bus 1002 is used to establish communication between these components. The user interface 1003 may include a voice pickup module, such as a microphone array; optionally, the user interface 1003 may also be a display screen or an input unit such as a keyboard. The memory 1005 may be a high-speed RAM or a stable, non-volatile memory, such as a disk storage device. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0050] It is understood that the high-speed axle brake condition detection device may also include a network interface 1004, which may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). Optionally, the high-speed axle brake condition detection device may also include RF (Radio Frequency) circuitry, sensors, audio circuitry, a Wi-Fi module, etc.

[0051] Those skilled in the art will understand that Figure 1 The structure of the high-speed shaft brake condition detection device shown does not constitute a limitation on the high-speed shaft brake condition detection device. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0052] Based on, but not limited to, the hardware structure of the high-speed shaft brake condition detection device described above, this application provides a first embodiment of a high-speed shaft brake condition detection method. (Refer to...) Figure 2 , Figure 2 A flowchart illustrating the first embodiment of the high-speed shaft brake state detection method of this application is shown.

[0053] It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0054] In this embodiment, the high-speed shaft brake state detection method includes:

[0055] Step S100: Obtain thermal imaging data of the contact area between the high-speed shaft brake and the brake disc collected by the thermal imaging sensor and vibration data sequence collected by the vibration sensor.

[0056] The vibration signal data sequence includes multiple vibration data points ordered along time.

[0057] In this embodiment, fault detection is mainly performed on the contact area between the high-speed brake and the brake disc. For example, if the clearance of the high-speed axle brake is inconsistent during installation, it will cause inconsistent wear thickness of the friction pads and brake disc in various contact directions, resulting in a significant abnormal surface temperature at the contact area between the brake and the brake disc. Furthermore, if a foreign object enters and becomes stuck in the gap between the brake and the brake disc, it will cause abnormal changes in the vibration data during operation. Therefore, this embodiment targets the contact area between the high-speed brake and the brake disc by acquiring the temperature distribution at this location during brake operation using a thermal imaging sensor and the vibration data at this location using a vibration sensor. The temperature distribution is represented by thermal imaging data, and the vibration data is represented by a vibration data sequence.

[0058] Step S200: Extract features from thermal imaging data to obtain thermal imaging spatial features, and extract features from vibration data sequence to obtain vibration temporal features.

[0059] In this embodiment, thermal imaging spatial features are a series of characteristics extracted from thermal imaging data that describe the spatial distribution of surface temperature of an object. These features provide information about temperature changes and distribution in different regions of the thermal imaging image. Vibration temporal features are a series of features extracted from the time domain of the vibration signal to describe the changes of the vibration signal over time. These features provide information about the overall dynamic behavior of the vibration signal, which helps in analyzing the system's operating status and detecting faults or anomalies.

[0060] Specifically, different feature extraction methods are used for the acquired thermal imaging data and vibration time features. For example, for thermal imaging data, edge features with drastic temperature changes can be extracted, which helps to capture changes in temperature distribution. For vibration data sequences, nonlinear transients can be extracted to analyze the energy changes of nonlinear events during vibration.

[0061] Understandably, the goal of feature extraction from thermal imaging and vibration data sequences is to reduce data complexity, retain key information, and provide useful input for subsequent model training or analysis.

[0062] Step S300: Fuse thermal imaging spatial features and vibration temporal features to obtain multimodal fusion features.

[0063] In this step, the multimodal fusion feature is a comprehensive feature obtained by fusing thermal imaging spatial features and vibration temporal features. By fusing thermal imaging spatial features and vibration temporal features, it can automatically adapt to different data distributions and give full play to the advantages of the two modes, thereby improving the performance and anti-interference ability of fault detection.

[0064] Step S400: Input the multimodal fusion features into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model.

[0065] In this embodiment, the high-speed axle brake state recognition model is a classifier based on machine learning or deep learning, designed to automatically analyze multimodal fusion features to identify the operating state of the high-speed axle brake.

[0066] Specifically, the multimodal fused features are input into the high-speed axle brake state recognition model. This model utilizes its pre-learned patterns and feature associations to classify the current multimodal fused features. This model can be trained using machine learning classification algorithms such as Support Vector Machine (SVM) or Random Forest. The output of the high-speed axle brake state recognition model typically includes different categories or labels representing the different states the high-speed axle brake may be in, such as normal and abnormal states.

[0067] Understandably, the application of a high-speed shaft brake condition recognition model can improve fault detection capabilities, help identify potential faults or anomalies early, thereby optimizing equipment maintenance and operation plans, reducing potential fault risks, and improving system reliability and stability.

[0068] It is easy to see that this embodiment uses a pre-trained high-speed axle brake state recognition model to detect the working state of the high-speed axle brake. By extracting features from the acquired thermal imaging data and vibration data sequences, the extracted thermal imaging spatial features and vibration temporal features are fused, and the resulting multimodal fusion features are input into the high-speed axle brake state recognition model for working state detection. This achieves comprehensive detection of the temperature and vibration states of the high-speed axle brake during operation, solving the problem of inaccurate detection of the working state of the high-speed axle brake under some complex conditions, and improving the working reliability, safety, and operating efficiency of the high-speed axle brake.

[0069] Furthermore, refer to Figure 3 In one specific implementation, step S200 includes:

[0070] Step S210: Extract features from thermal imaging data based on spatial attention mechanism to obtain thermal imaging spatial features.

[0071] Step S220: Based on the time attention mechanism, feature extraction is performed on the vibration data sequence to obtain vibration time features.

[0072] In this embodiment, spatial attention and temporal attention mechanisms are two variations of attention mechanisms, used to process thermal imaging data and vibration data sequences, respectively. They can dynamically adjust the weights at different locations or time points, enabling the model to focus on different parts of the input.

[0073] Specifically, this embodiment extracts the spatial features of thermal imaging data through a spatial attention mechanism and extracts the temporal features of vibration through a temporal attention mechanism.

[0074] Understandably, spatial attention and temporal attention mechanisms can be used to focus more effectively on specific spatial locations in thermal imaging data and key time points in vibration data sequences, thereby improving the effectiveness of feature extraction.

[0075] Specifically, refer to Figure 4 In one specific implementation, step S210 includes:

[0076] Step S211: Extract features from the thermal imaging data to obtain a feature map matrix.

[0077] Step S212: Determine the spatial attention vector for each feature in the feature map matrix to obtain the spatial attention matrix.

[0078] Step S213: Obtain the spatial feature enhancement map based on the spatial attention matrix and the feature map matrix.

[0079] Step S214: Extract features from the spatial feature enhancement map to obtain thermal imaging spatial features.

[0080] In this embodiment, thermal imaging data is input into a feature extraction network for feature extraction, thereby obtaining a feature map output by the feature extraction network. This feature map can be represented by a matrix, i.e., a feature map matrix. Furthermore, the feature extraction network is connected to a spatial attention module, which determines the spatial attention vector for each feature in the feature map matrix, thus obtaining a spatial attention matrix. The spatial attention vector reflects the importance of each feature in the feature map, and can enhance key feature information in the feature map matrix.

[0081] Understandably, this spatial attention module can employ a global attention mechanism, that is, determine the spatial attention vector of a feature based on the similarity between each feature (e.g., corresponding to a 2×2 image region) and all other features. Alternatively, the spatial attention module can also employ a local attention mechanism, thereby determining the spatial attention vector of each feature (e.g., corresponding to a 2×2 image region) based on the similarity between it and its surrounding neighboring features.

[0082] After obtaining the spatial attention matrix, each element in the feature map matrix is ​​multiplied by the element in the spatial attention matrix at the corresponding position, thereby enhancing each feature in the feature map matrix and obtaining a spatial feature enhancement map.

[0083] The enhanced spatial feature map is then fed into a feature extraction convolutional neural network for convolution, pooling, and other feature extraction processes to obtain the spatial features of the thermal imaging. Understandably, during the feature extraction process of thermal imaging images, the spatial attention matrix can enhance the local features of the image, allowing the feature extraction convolutional neural network to focus more on important regions. This helps improve the recognition of details of temperature distribution and local structural features in thermal imaging data.

[0084] In a specific example, a spatial feature enhancement map is used as input to a feature extraction convolutional neural network (CNN). Convolutional layers then extract features from the enhanced map through convolution operations. The convolution operation slides the kernel across the enhanced map, calculating local features at different locations in the image. After the convolutional layers, activation functions such as ReLU can be added to introduce non-linearity, allowing the CNN to better learn and represent complex features. Pooling layers then reduce the dimensionality of the feature map, decreasing computational burden while preserving key information. By stacking convolutional layers, activation functions, and pooling layers multiple times, a deep network structure is gradually built. Finally, a fully connected layer is added on top of the feature extraction CNN to map the features extracted by the convolutional layers to the output layer, which outputs the final thermal imaging spatial features.

[0085] Understandably, since the importance of features varies in thermal imaging data of high-speed axle brakes, important salient features are often only manifested in local areas. In other words, some local areas contain more information. By using a spatial attention mechanism to calculate local areas in thermal imaging data, the convolutional neural network can pay more attention to the more important local features, thereby better realizing the extraction of spatial features in thermal imaging and improving the accuracy of fault detection by the high-speed axle brake condition recognition model.

[0086] Furthermore, refer to Figure 5 In one specific implementation, step S220 includes:

[0087] Step S221: Perform spectral attention processing on the vibration data sequence to obtain the spectral attention weights of each spectral component in the vibration data sequence.

[0088] Step S222: Based on the spectral attention weights of each spectral component, the vibration data sequence is filtered to obtain vibration time characteristics.

[0089] In this embodiment, the spectral attention processing mechanism determines the spectral attention weight of spectral components by focusing on the importance of different frequency components in the vibration data sequence. Furthermore, it automatically adjusts the filter parameters based on the spectral attention weights and the spectral characteristics of the vibration data sequence to achieve adaptive filtering of the time series data, thereby preserving more important spectral features to obtain the vibration temporal characteristics.

[0090] It is worth mentioning that the vibration time series can be input into a long short-term memory neural network to predict the vibration time series for the next 10 seconds.

[0091] Specifically, Long Short-Term Memory (LSTM) neural networks are deep learning models suitable for processing sequence data, particularly adept at capturing long-term dependencies within sequences. In one example, multiple vibration time series are used as input to an LSM network. The LSM network gradually learns the dynamic features of the sequences, captures long-term dependencies, and predicts future vibration time series based on these relationships.

[0092] Furthermore, refer to Figure 6 In a specific real-time mode, step S300 includes:

[0093] Step S310: Determine the first weight of the thermal imaging spatial features and the second weight of the vibration time features.

[0094] Step S320: Based on the first weight and the second weight, the thermal imaging spatial features and vibration temporal features are weighted and fused to obtain multimodal fusion features.

[0095] In this embodiment, the first weight refers to the weight assigned to the thermal imaging spatial features, representing the relative importance of the thermal imaging spatial features in multimodal fusion. This weight determines the proportion of thermal imaging spatial features in the final multimodal features. The second weight refers to the weight assigned to the vibration temporal features, representing the relative importance of the vibration temporal features in multimodal fusion. This weight determines the proportion of vibration temporal features in the final multimodal features.

[0096] In this embodiment, the weights are typically calculated automatically using an attention mechanism. Specifically, when the high-speed shaft brake is operating normally, the spatial features of thermal imaging may be more discriminative and should be given a higher weight. However, when the equipment malfunctions, the temporal features of vibration may be more discriminative and should be given a higher weight.

[0097] In this embodiment, thermal imaging spatial features and vibration time features are input into the attention layer for automatic calculation, and the first weight and the second weight are learned.

[0098] In a specific example, the first and second weights within the attention layer are initialized to a set of random initial values, and the spatial features of thermal imaging and the temporal features of vibration are passed to the attention layer. The attention layer then weights the input data according to the current attention weights. Subsequently, a loss function is calculated based on the difference between the output data of the attention layer and the real data. The attention weights are updated based on the gradient of the loss function, adjusting the first and second weights of the attention layer to minimize the value of the loss function, thereby outputting the most suitable first and second weights.

[0099] In another specific example, thermal imaging spatial features and vibration temporal features are used as inputs, and the thermal imaging spatial features and vibration temporal features are weighted and fused using Formula 3 to obtain multimodal fusion features.

[0100] Formula 3: Where w1 + w2 = 1;

[0101] Where x is the feature vector of the attention mechanism of multi-state fusion, w1 and w2 are the first weight and the second weight assigned to the thermal imaging spatial feature and the vibration temporal feature, respectively, k is the thermal imaging spatial feature vector of the k-th layer, and t is the vibration temporal feature at time t.

[0102] During feature fusion, the vibration time features output by the Long Short-Term Memory Neural Network (LSTM) can be fused with the thermal imaging spatial features output by the feature extraction convolutional neural network (CNN). Specifically, a multimodal fusion connection operation can be used to merge the feature vectors of both into a more comprehensive representation.

[0103] Furthermore, before fusion, the vibration temporal features and thermal imaging spatial features need to be preprocessed to ensure they have the same time step or sampling rate. Preprocessing can be achieved through interpolation sampling.

[0104] Understandably, by introducing an adaptive attention mechanism, the system can automatically adjust the weights of thermal imaging spatial features and vibration temporal features in multimodal fusion based on the operating state of the high-speed axle brake. When the high-speed axle brake is operating normally, the attention mechanism tends to assign higher weights to thermal imaging spatial features because, under normal conditions, thermal imaging can more effectively distinguish the operating state. However, when a fault occurs, the attention automatically adjusts to emphasize vibration temporal features, as the vibration signal is more discriminative in this situation. This adaptive weight learning scheme helps improve the system's sensitivity to changes in the state of the high-speed axle brake, enhancing the accurate monitoring and diagnosis capabilities of the equipment's health status.

[0105] Furthermore, refer to Figure 7 , Figure 7A flowchart illustrating the second embodiment of the high-speed shaft brake state detection method of this application is shown.

[0106] In this embodiment, before step S400, the following steps are also included:

[0107] Step S500: Obtain a training sample set, which includes multimodal fusion samples labeled with fault information; the multimodal fusion samples are obtained by fusing thermal imaging spatial feature samples and vibration time feature samples.

[0108] Step S600: Train a support vector machine classifier using the training sample set to obtain a high-speed axle brake state recognition model.

[0109] In this embodiment, the training sample set consists of multimodal fusion samples, which are obtained by fusing thermal imaging spatial feature samples and vibration temporal feature samples. The thermal imaging spatial feature samples consist of thermal imaging spatial features under different states, and the vibration temporal feature samples consist of vibration temporal features under different states.

[0110] Specifically, thermal imaging datasets under different operating conditions are input into the thermal imaging feature extraction model. After preprocessing, spatial attention calculation, and convolutional neural network processing, the corresponding thermal imaging spatial feature sets are obtained. Similarly, vibration data sequence sets under different operating conditions are input into the vibration feature extraction model. After training the model using a long short-term memory neural network, the corresponding vibration temporal feature sets are obtained.

[0111] Next, the obtained thermal imaging spatial feature set and vibration temporal feature set are automatically learned through an attention mechanism to obtain a multimodal fusion sample set. Understandably, the multimodal fusion sample set contains multimodal fusion features from different operating states. Subsequently, the multimodal fusion features in the multimodal fusion sample set are labeled with corresponding tags, which contain corresponding fault information, specifically fault states and non-fault states.

[0112] Subsequently, the multimodal fusion samples are used as the training set and fed into a support vector machine for training, resulting in a high-speed axle brake state recognition model. Understandably, the output of the high-speed axle brake state recognition model is influenced by the labels. That is, if the fault information in the labels includes both fault and non-fault states, the high-speed axle brake state recognition model performs binary classification of the training samples, meaning the model can determine whether a newly input detection sample is faulty or not.

[0113] Understandably, the accuracy of support vector machine training results is affected by bandwidth parameters and penalty factors. This embodiment can optimize bandwidth parameters and penalty factors through cross-validation and network search, thereby improving the accuracy of support vector machine training results.

[0114] Based on the same inventive concept, this application also provides a high-speed shaft brake state detection device, which includes:

[0115] The acquisition module is used to acquire thermal imaging data of the contact area between the high-speed shaft brake and the brake disc collected by the thermal imaging sensor and vibration data sequence collected by the vibration sensor; the vibration signal data sequence includes multiple vibration data ordered along time.

[0116] The feature extraction module is used to extract features from thermal imaging data to obtain spatial features of thermal imaging, and to extract features from vibration data sequences to obtain temporal features of vibration.

[0117] The fusion module is used to fuse spatial features of thermal imaging and temporal vibration features to obtain multimodal fusion features;

[0118] The detection module is used to input multimodal fusion features into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model.

[0119] It should be noted that the various embodiments of the high-speed shaft brake state detection device in this embodiment, and the technical effects they achieve, can be referred to the various implementation methods of the high-speed shaft brake state detection method in the foregoing embodiments, and will not be repeated here.

[0120] Furthermore, this application also proposes a computer storage medium storing a high-speed shaft brake state detection program. When executed by a processor, the high-speed shaft brake state detection program implements the steps of the high-speed shaft brake state detection method described above. Therefore, it will not be repeated here. Additionally, the beneficial effects of using the same method will not be repeated here either. For technical details not disclosed in the computer-readable storage medium embodiments of this application, please refer to the description of the method embodiments of this application. As an example, program instructions can be deployed to execute on a single computing device, or on multiple computing devices located at one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0121] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0122] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0124] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for detecting the state of a high-speed shaft brake, characterized in that, The methods include: Acquire thermal imaging data of the contact area between the high-speed axle brake and the brake disc collected by a thermal imaging sensor and vibration data sequence collected by a vibration sensor; the vibration signal data sequence includes multiple vibration data points ordered along time. Feature extraction is performed on the thermal imaging data to obtain thermal imaging spatial features, and feature extraction is performed on the vibration data sequence to obtain vibration temporal features; By fusing the aforementioned thermal imaging spatial features and vibration temporal features, multimodal fusion features are obtained; The multimodal fusion features are input into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model. The step of extracting features from the thermal imaging data to obtain thermal imaging spatial features includes: Feature extraction is performed on the thermal imaging data to obtain a feature map matrix; Determine the spatial attention vector for each feature in the feature map matrix to obtain the spatial attention matrix; Based on the spatial attention matrix and the feature map matrix, a spatial feature enhancement map is obtained; Feature extraction is performed on the spatial feature enhancement map to obtain the thermal imaging spatial features.

2. The high-speed shaft brake state detection method according to claim 1, characterized in that, The step of extracting features from the vibration data sequence to obtain vibration time features includes: The vibration time features are obtained by extracting features from the vibration data sequence based on the time attention mechanism.

3. The high-speed shaft brake state detection method according to claim 2, characterized in that, The step of extracting features from the vibration data sequence based on a time attention mechanism to obtain the vibration time features includes: The vibration data sequence is subjected to spectral attention processing to obtain the spectral attention weight of each spectral component in the vibration data sequence. Based on the spectral attention weights of each spectral component, the vibration data sequence is filtered to obtain vibration time characteristics.

4. The high-speed shaft brake state detection method according to claim 1, characterized in that, The fusion of the thermal imaging spatial features and vibration temporal features to obtain multimodal fusion features includes: Determine the first weight of the spatial features of thermal imaging and the second weight of the temporal features of vibration; Based on the first weight and the second weight, the thermal imaging spatial features and vibration temporal features are weighted and fused to obtain the multimodal fusion features.

5. The method for detecting the state of a high-speed shaft brake according to claim 1, characterized in that, Before inputting the multimodal fusion features into the high-speed axle brake state recognition model and obtaining the state detection result output by the high-speed axle brake state recognition model, the method further includes: A training sample set is obtained, which includes multimodal fusion samples labeled with fault information; the multimodal fusion samples are obtained by fusing thermal imaging spatial feature samples and vibration temporal feature samples. A support vector machine classifier is trained using the training sample set to obtain the state recognition model of the high-speed axle brake.

6. A high-speed shaft brake condition detection device, characterized in that, The high-speed shaft brake status detection device includes: The acquisition module is used to acquire thermal imaging data of the contact area between the high-speed shaft brake and the brake disc collected by the thermal imaging sensor and vibration data sequence collected by the vibration sensor; the vibration signal data sequence includes multiple vibration data ordered along time. The feature extraction module is used to extract features from the thermal imaging data to obtain thermal imaging spatial features, and to extract features from the vibration data sequence to obtain vibration temporal features. The fusion module is used to fuse the thermal imaging spatial features and vibration temporal features to obtain multimodal fusion features; The detection module is used to input multimodal fusion features into the high-speed axle brake state recognition model to obtain the state detection results output by the high-speed axle brake state recognition model. The step of extracting features from the thermal imaging data to obtain thermal imaging spatial features includes: Feature extraction is performed on the thermal imaging data to obtain a feature map matrix; Determine the spatial attention vector for each feature in the feature map matrix to obtain the spatial attention matrix; Based on the spatial attention matrix and the feature map matrix, a spatial feature enhancement map is obtained; Feature extraction is performed on the spatial feature enhancement map to obtain the thermal imaging spatial features.

7. A high-speed shaft brake condition detection device, characterized in that, include: A processor, a memory, and a high-speed shaft brake state detection program stored in the memory, wherein the high-speed shaft brake state detection program is executed by the processor to implement the steps of the high-speed shaft brake state detection method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A high-speed shaft brake state detection program is stored on a computer-readable storage medium, which, when executed by a processor, implements the high-speed shaft brake state detection method as described in any one of claims 1 to 5.

Images