Case management system and method based on lightweight neural network
By adopting lightweight neural network technology in the chassis management system, intelligent management and analysis of chassis data is solved, and the problems of low processing efficiency and lack of self-learning ability in the existing technology are achieved, efficient and accurate chassis status monitoring and fault prediction are achieved, and stable operation and performance optimization of the computer are supported.
Patent Information
- Application Number
- CN202411894074.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2025-05-30
AI Technical Summary
When existing chassis management technologies process the growing chassis data and complex data structures, the processing efficiency cannot meet the requirements of real-time, and lack self-learning and adaptability, making it difficult to adapt to complex and changeable data environments, resulting in inaccurate classification results and affecting the accuracy of management decisions.
The chassis management system based on lightweight neural network is adopted to realize intelligent management and analysis of chassis sensor data and resource usage data through data preprocessing, model training and real-time data analysis. The system includes a data acquisition module, a data preprocessing module, a chassis analysis model, a visualization module and an early warning module. It classifies and analyzes real-time data through a lightweight neural network model, predicts faults and outputs early warnings.
It improves the efficiency and accuracy of chassis management, realizes real-time monitoring and fault prediction of the internal state of the chassis, supports stable operation and performance optimization of the computer, and reduces operational costs and maintenance difficulties.
Smart Images
Figure CN120066826A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of server chassis management, and particularly relates to a chassis management system and method based on a lightweight neural network. Background Art
[0002] In the context of the rapid development of modern information technology, the complexity and integration of computer hardware are constantly increasing. As an important part of computer hardware, a large number of sensors and resources are integrated inside the chassis for real-time monitoring and recording of the operating status of the computer. These sensor data and resource usage data are crucial for ensuring the stable operation of the computer, optimizing performance, and preventing potential failures.
[0003] The current mainstream chassis management technologies, although to a certain extent, achieve the monitoring and management of sensor data, resource usage data, etc. inside the chassis, still have some obvious disadvantages in practical applications: Traditional chassis management technologies often use fixed rules and algorithms for data processing. When facing the increasing chassis data and complex data structures, the processing efficiency often fails to meet the real-time requirements. This may cause administrators to be unable to obtain the real-time status of the chassis in a timely manner, thereby affecting the timely discovery and handling of potential problems. The classification of chassis data is often based on simple threshold judgments or pattern matching, making it difficult to cope with complex and changing data environments. This may lead to inaccurate classification results or even misjudgments, thereby affecting the decision-making accuracy of chassis management. The mainstream chassis management technologies usually lack the ability of self-learning and adaptation, and cannot be dynamically adjusted and optimized according to the actual operating status of the chassis. This makes the management strategy tend to be too rigid and difficult to adapt to the changing chassis environment and user needs. Traditional chassis management technologies require a large amount of computing resources and storage space to support data processing and analysis, which increases the operating cost and maintenance difficulty of the system. Especially in large data centers or cloud computing environments, this resource consumption problem is particularly prominent, and it lacks good integration and scalability, making it difficult to seamlessly dock and work collaboratively with other management systems or tools. This limits the expansion of chassis management functions and the expansion of the application scope.
[0004] Therefore, how to improve the traditional chassis management method to combine the actual needs of chassis management, achieve intelligent management and analysis of sensor data and resource usage data inside the chassis, improve the efficiency and accuracy of chassis management, and provide strong support for the stable operation and performance optimization of the computer is a technical problem that urgently needs to be solved at present. Summary of the Invention
[0005] The purpose of the present invention is to provide a chassis management system and method based on a lightweight neural network, which is used to improve the traditional chassis management method, combine the actual needs of chassis management, realize the intelligent management and analysis of sensor data and resource usage data in the chassis, improve the efficiency and accuracy of chassis management, and provide strong support for the stable operation and performance optimization of the computer.
[0006] To solve the above technical problems, the technical solutions adopted by the present invention are as follows:
[0007] A chassis management method based on a lightweight neural network includes the following steps:
[0008] S1: Obtain the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and perform data preprocessing on the collected data;
[0009] S2: Create and deploy a chassis analysis model, input the historical environmental parameter data and historical resource usage data after data preprocessing into the chassis analysis model, train the chassis analysis model, verify the trained chassis analysis model, and determine whether the verification result meets the preset model performance requirements. If so, execute step S3; if not, retrain the model until the model meets the preset performance requirements;
[0010] S3: Real-time collect the environmental parameter data and resource usage data inside the chassis through the data acquisition module, and input them into the trained chassis analysis model. Classify and analyze the real-time collected environmental parameter data and resource usage data through the trained chassis analysis model, and transmit the classification and analysis results to the visualization module;
[0011] S4: The chassis analysis model predicts the faults of the chassis based on the classification and analysis results of step S3, and outputs the fault prediction results;
[0012] S5: Transmit the fault prediction results to the visualization module and the warning module, display the fault prediction results through the visualization module, and give warnings for the predicted faults through the warning module;
[0013] S6: The chassis administrator timely obtains the classified and analyzed chassis environmental parameter data and resource usage data, and makes corresponding adjustments to the resource configuration and operating environment of the chassis based on the displayed classification and analysis results and the predicted faults warned by the warning module.
[0014] Preferably, the specific process of data preprocessing for the collected data in step S1 is as follows:
[0015] S11: Remove the noise data from the collected historical environmental parameter data and historical resource usage data, and process the outliers among them;
[0016] S12: Integrate the historical environmental parameter data and historical resource usage data from different sources after removing noise and outliers, and perform data transformation on the integrated data to convert all data into a unified data form;
[0017] S13: Perform dimensionality reduction on the data after data transformation, and convert the data after data transformation into data on a set of new coordinate axes where each dimension is pairwise independent through linear transformation.
[0018] Preferably, the formula for converting the data into a unified data form in step S12 is:
[0019] X ij * = (X ij - μ j ) / σ j ;
[0020] Among them, X ij * is the converted data, X ij is the original data before conversion, μ j is the mean of the jth feature, and σ j is the standard deviation of the jth feature.
[0021] Preferably, the specific process of performing dimensionality reduction on the data after data transformation in step S13 is as follows:
[0022] S131: Establish a data matrix of the data after data transformation, and calculate the covariance matrix of the data matrix. The data elements at each position in the covariance matrix are:
[0023]
[0024] Among them, C ij is the covariance of the random variables xi and xj. N is the number of data samples, n is the data dimension, μi is the mean of the data in the i dimension, and μj is the mean of the data in the j dimension;
[0025] S132: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors;
[0026] S133: Calculate the contribution rate of each component for each eigenvalue in step S132, and select the data of the eigenvalues whose contribution rate reaches a specified ratio.
[0027] Preferably, the chassis analysis model in step S2 is a lightweight neural network model. The lightweight neural network model includes a depthwise convolutional layer and a pointwise convolutional layer. The depthwise convolutional layer includes multiple convolutional kernels, and each convolutional kernel processes one channel of the feature data. The pointwise convolutional layer linearly combines the multiple feature maps output by the depthwise convolutional layer to generate a final output feature map.
[0028] Preferably, the specific process of training and validating the chassis analysis model is as follows:
[0029] S21: Extract part of the data from the historical environmental parameter data and historical resource usage data as pre-training data, input it into the chassis analysis model, and freeze all the weights of the model;
[0030] S22: Add a custom top classifier, optimizer, loss function, and evaluation metrics, compile the model, and then input the historical environmental parameter data and historical resource usage data into the chassis analysis model for training;
[0031] S23: Adjust the learning rate and batch size during the training of the chassis analysis model to optimize the model;
[0032] S24: Extract the validation set from the historical environmental parameter data and historical resource usage data, input it into the optimized chassis analysis model, and calculate metrics such as model accuracy and recall rate to evaluate the model performance.
[0033] In a second aspect, a chassis management system based on a lightweight neural network is provided, which is used to implement any one of the above-mentioned chassis management methods based on a lightweight neural network. It includes a data acquisition module, a data preprocessing module, a model creation module, a chassis analysis model, a visualization module, and an early warning module. The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the chassis analysis model, the model creation module is connected to the chassis analysis model, and the chassis analysis model is connected to the visualization module and the early warning module;
[0034] The data acquisition module is used to obtain the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and collect the environmental parameter data and resource usage data inside the chassis in real time;
[0035] The data preprocessing module is used to preprocess the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and the environmental parameter data and resource usage data collected inside the chassis in real time;
[0036] The model creation module is used to create and deploy the chassis analysis model;
[0037] The chassis analysis model is used to classify and analyze the environmental parameter data and resource usage data collected in real time, predict the faults of the chassis based on the classification and analysis results, output the fault prediction results, and transmit the classification and analysis results to the visualization module;
[0038] The visualization module is used to display the classification and analysis results and the fault prediction results;
[0039] The early warning module is used to give early warnings about the predicted faults.
[0040] The beneficial effects of the present invention include:
[0041] The chassis management system and method based on a lightweight neural network provided by the present invention obtain the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, input the preprocessed data into the chassis analysis model for training, collect the environmental parameter data and resource usage data inside the chassis in real time through the data acquisition module, classify and analyze the environmental parameter data and resource usage data collected in real time by the trained chassis analysis model, and predict the faults of the chassis, and output the fault prediction results to the visualization module; transmit the fault prediction results to the visualization module and the early warning module, and the chassis administrator makes corresponding adjustments to the resource configuration and operating environment of the chassis based on the predicted faults. The above process combines the actual requirements of chassis management, realizes the intelligent management and analysis of the sensor data and resource usage data inside the chassis, improves the efficiency and accuracy of chassis management, and provides strong support for the stable operation and performance optimization of the computer.
[0042] First, by removing the noise data in the collected historical environmental parameter data and historical resource usage data, processing the outliers therein, performing integration processing, and performing data conversion on the integrated processed data to convert all data into a unified data form, and performing dimensionality reduction processing on the data after data conversion, the data preprocessing process effectively avoids the influence of the noise data and abnormal data therein on the subsequent model training and prediction processes, and improves the performance of the chassis analysis model.
[0043] Second, by establishing a data matrix of the data after data conversion, calculating the covariance matrix of the data matrix, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors; calculating the contribution rate of each component for each eigenvalue, and selecting the data of the eigenvalues whose contribution rate reaches a specified ratio realizes the dimensionality reduction processing of the data, which is convenient for subsequent classification and analysis of the data by the chassis analysis model, and improves the accuracy of the classification and analysis results.
[0044] Finally, extract some data as pre-training data and input it into the chassis analysis model, and freeze all the weights of the model; add a custom top classifier, optimizer, loss function, and evaluation metrics, and compile the model. Then, train, optimize, and validate the compiled chassis analysis model, and calculate metrics such as model accuracy and recall to evaluate the model performance. This process ensures that the model meets the specified performance requirements. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] Figure 1 FIG. is a schematic flowchart of the chassis management method based on a lightweight neural network according to the present invention.
[0046] Figure 2 FIG. is a schematic flowchart of the dimensionality reduction process for the data after data conversion according to the present invention.
[0047] Figure 3 FIG. is a schematic architecture diagram of the chassis management system based on a lightweight neural network according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0048] The following further describes the present invention in detail with reference to the Figures 1 to 3 accompanying drawings:
[0049] Referring to the Figure 1 accompanying drawings, a chassis management method based on a lightweight neural network includes the following steps:
[0050] S1: Obtain historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and perform data preprocessing on the collected data;
[0051] S2: Create and deploy a chassis analysis model, input the historical environmental parameter data and historical resource usage data after data preprocessing into the chassis analysis model, train the chassis analysis model, and verify the trained chassis analysis model. Then, determine whether the verification result meets the preset model performance requirements. If so, execute step S3; if not, retrain the model until the model meets the preset performance requirements;
[0052] S3: Real-time collect the environmental parameter data and resource usage data inside the chassis through the data acquisition module, and input them into the trained chassis analysis model. Classify and analyze the real-time collected environmental parameter data and resource usage data through the trained chassis analysis model, and transmit the classification and analysis results to the visualization module;
[0053] S4: The chassis analysis model predicts the faults of the chassis based on the classification and analysis results of step S3, and outputs the fault prediction results;
[0054] S5: Transmit the fault prediction result to the visualization module and the early warning module, display the fault prediction result through the visualization module, and give an early warning of the predicted fault through the early warning module;
[0055] S6: The chassis administrator timely obtains the classified and analyzed chassis environment parameter data and resource usage data, and makes corresponding adjustments to the resource configuration and operating environment of the chassis based on the displayed classification and analysis results and the predicted faults warned by the early warning module.
[0056] In this embodiment, first, obtain the historical environment parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and perform data preprocessing to remove noise data and abnormal data, so as to avoid the influence of noise data and abnormal data on the subsequent model training process. Input the preprocessed data into the chassis analysis model for training. Real-time collect the environment parameter data and resource usage data inside the chassis through the data acquisition module, classify and analyze the real-time collected environment parameter data and resource usage data through the trained chassis analysis model, predict the faults of the chassis, and output the fault prediction result to the visualization module; transmit the fault prediction result to the visualization module and the early warning module, and the chassis administrator makes corresponding adjustments to the resource configuration and operating environment of the chassis based on the predicted faults. The above process combines the actual requirements of chassis management, realizes the intelligent management and analysis of the sensor data and resource usage data inside the chassis, improves the efficiency and accuracy of chassis management, and provides strong support for the stable operation and performance optimization of the computer.
[0057] Embodiment 2
[0058] Based on Embodiment 1, the specific process of data preprocessing for the collected data in step S1 is as follows:
[0059] S11: Remove the noise data from the collected historical environment parameter data and historical resource usage data, and process the outliers among them;
[0060] S12: Integrate the historical environment parameter data and historical resource usage data from different sources after removing noise and outliers, and perform data conversion on the integrated data to convert all data into a unified data form;
[0061] S13: Perform dimensionality reduction processing on the data after data conversion, and convert the data after data conversion into data on a set of new coordinate axes where each dimension is independent of each other through linear transformation.
[0062] The data preprocessing process, which removes the noise data from the collected historical environmental parameter data and historical resource usage data, processes the outliers therein, performs integration processing, and converts the integrated data, converts all data into a unified data form, and performs dimensionality reduction processing on the data after the data conversion, effectively avoids the influence of the noise data and abnormal data therein on the subsequent model training and prediction processes, and improves the performance of the chassis analysis model.
[0063] In this embodiment, the formula for converting the data into a unified data form in step S12 is:
[0064] X ij *=(X ij -μ j ) / σ j ;
[0065] Wherein, X ij * is the converted data, X ij is the original data before conversion, μ j is the mean of the jth feature, and σ j is the standard deviation of the jth feature.
[0066] In another implementation manner of this embodiment, as shown in Figure 2 the specific process of performing dimensionality reduction processing on the data after the data conversion in step S13 is as follows:
[0067] S131: Establish a data matrix of the data after the data conversion, and calculate the covariance matrix of the data matrix. The data elements at each position in the covariance matrix are:
[0068]
[0069] Wherein, C ij is the covariance of the random variables xi and xj, N is the number of data samples, n is the data dimension, μi is the mean of the data in the i dimension, and μj is the mean of the data in the j dimension;
[0070] S132: Perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors;
[0071] S133: Calculate the contribution rate of each component for each eigenvalue in step S132, and select the data of the eigenvalues whose contribution rate reaches a specified ratio.
[0072] By establishing the data matrix of the data after data conversion as described above, calculating the covariance matrix of the data matrix, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors; calculating the contribution rate of each component for each eigenvalue, and selecting the data of the eigenvalues whose contribution rate reaches a specified ratio, the process of dimensionality reduction of the data is realized, which facilitates subsequent classification and analysis of the data through the chassis analysis model, and improves the accuracy of the classification and analysis results.
[0073] Embodiment 3
[0074] Based on Embodiment 1 or Embodiment 2, the chassis analysis model in step S2 is a lightweight neural network model. The lightweight neural network model includes a depthwise convolutional layer and a pointwise convolutional layer. The depthwise convolutional layer includes multiple convolutional kernels, and each convolutional kernel processes one channel of the feature data. The pointwise convolutional layer linearly combines the multiple feature maps output by the depthwise convolutional layer to generate the final output feature map. The depthwise convolutional layer has three channels, and each channel convolutional layer has three independent convolutional kernels, and each convolutional kernel performs a convolutional operation on the input data of its corresponding channel. By processing the data of one channel with each convolutional kernel, the amount of calculation and the number of parameters are reduced. The pointwise convolutional layer has a 1x1 convolutional kernel, and performs weighted summation on the multiple feature maps in the channel dimension to obtain the final output feature map. The lightweight neural network model greatly reduces the amount of calculation and the number of parameters while maintaining the same output data size through the depthwise convolutional layer and the pointwise convolutional layer, making the chassis analysis model more efficient.
[0075] In this embodiment, the specific process of training and validating the chassis analysis model is as follows:
[0076] S21: Extract part of the data from the historical environmental parameter data and historical resource usage data as pre-training data, and input it into the chassis analysis model, and freeze all the weights of the model to keep them unchanged during the training process.
[0077] S22: Add a custom top classifier, optimizer, loss function and evaluation metrics and compile the model, and then input the historical environmental parameter data and historical resource usage data into the chassis analysis model for training, and the loss function is the cross-entropy loss function.
[0078] S23: Adjust the learning rate and batch size to optimize the model during the training process of the chassis analysis model. During the training process, it is necessary to monitor the changes of the loss function and accuracy to ensure that the model is continuously improved during the training process.
[0079] S24: Extract the validation set input from the historical environmental parameter data and historical resource usage data to optimize the chassis analysis model, and calculate indicators such as model accuracy and recall to evaluate the model performance. Save the trained model to a file for loading in subsequent use. Usually, save the weights and structure of the model for subsequent prediction and deployment.
[0080] By extracting the above partial data as pre-training data and inputting it into the chassis analysis model, and freezing all the weights of the model; adding a custom top classifier, optimizer, loss function, and evaluation metrics and compiling the model, training, optimizing, and validating the compiled chassis analysis model, and calculating indicators such as model accuracy and recall to evaluate the model performance, the process ensures that the model meets the specified performance requirements, enabling the model to perform subsequent data classification and analysis and accurate prediction of faults.
[0081] A chassis management system based on a lightweight neural network is used to implement any one of the chassis management methods based on a lightweight neural network, see Figure 3 As shown, it includes a data acquisition module, a data preprocessing module, a model creation module, a chassis analysis model, a visualization module, and an early warning module. The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the chassis analysis model, the model creation module is connected to the chassis analysis model, and the chassis analysis model is connected to the visualization module and the early warning module. The data acquisition module is used to obtain the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and to collect the environmental parameter data and resource usage data inside the chassis in real time. The data preprocessing module is used to preprocess the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis and the environmental parameter data and resource usage data collected inside the chassis in real time. The model creation module is used to create and deploy the chassis analysis model; the chassis analysis model is used to classify and analyze the environmental parameter data and resource usage data collected in real time, predict the faults of the chassis based on the classification and analysis results, output the fault prediction results, and transmit the classification and analysis results to the visualization module. The visualization module is used to display the classification and analysis results and the fault prediction results, and the early warning module is used to give early warnings for the predicted faults.
[0082] In summary, the chassis management system and method based on a lightweight neural network provided by the present invention obtain historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, input the preprocessed data into the chassis analysis model for training, collect the environmental parameter data and resource usage data inside the chassis in real time through the data acquisition module, classify and analyze the real-time collected environmental parameter data and resource usage data through the trained chassis analysis model, predict the faults of the chassis, and output the fault prediction results to the visualization module; transmit the fault prediction results to the visualization module and the warning module, and the chassis administrator makes corresponding adjustments to the resource configuration and operating environment of the chassis based on the predicted faults. The above process combines the actual requirements of chassis management, realizes the intelligent management and analysis of sensor data and resource usage data inside the chassis, improves the efficiency and accuracy of chassis management, and provides strong support for the stable operation and performance optimization of the computer.
[0083] The data preprocessing process, which removes the noise data from the collected historical environmental parameter data and historical resource usage data, processes the outliers therein, performs integration processing, and converts the integrated data, converts all data into a unified data form, and performs dimensionality reduction processing on the data after data conversion, effectively avoids the influence of the noise data and abnormal data therein on the subsequent model training and prediction processes, and improves the performance of the chassis analysis model. By establishing a data matrix of the data after data conversion, calculating the covariance matrix of the data matrix, performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors; calculating the contribution rate of each component for each eigenvalue, and selecting the data of the eigenvalues whose contribution rate reaches a specified ratio realizes the dimensionality reduction processing of the data, which is convenient for subsequent classification and analysis of the data through the chassis analysis model, and improves the accuracy of the classification and analysis results. By extracting part of the data as pre-training data and inputting it into the chassis analysis model, and freezing all the weights of the model; adding a custom top classifier, optimizer, loss function, and evaluation metrics, and compiling the model, training, optimizing, and validating the compiled chassis analysis model, and calculating metrics such as model accuracy and recall rate to evaluate the model performance ensures that the model meets the specified performance requirements.
Claims
1. A chassis management method based on lightweight neural network, characterized in that: The following steps are involved: S1: Acquire historical environmental parameter data and historical resource usage data collected by the data collection module inside the chassis, and perform data preprocessing on the collected data; S2: Create and deploy a chassis analysis model, input the historical environmental parameter data and historical resource usage data after data preprocessing into the chassis analysis model, train the chassis analysis model, and verify the trained chassis analysis model, and determine whether the verification result meets the preset model performance requirements. If so, execute step S3, if not, retrain the model until the model meets the preset performance requirements; S3: Collecting the environmental parameter data and resource usage data inside the chassis in real time through the data acquisition module, and inputting them into the trained chassis analysis model; classifying and analyzing the environmental parameter data and resource usage data collected in real time through the trained chassis analysis model, and transmitting the classification and analysis results to the visualization module; S4: the chassis analysis model predicts the failure of the chassis based on the classification and analysis results of step S3, and outputs the failure prediction result; S5: transmitting the fault prediction result to the visualization module and the early warning module, displaying the fault prediction result through the visualization module, and issuing an early warning for the predicted fault through the early warning module; S6: The chassis administrator promptly obtains the classified and analyzed chassis environmental parameter data and resource usage data, and makes corresponding adjustments to the chassis resource configuration and operating environment based on the displayed classification and analysis results and the predicted faults of the warning module.
2. The chassis management method based on lightweight neural network according to claim 1 is characterized in that: The specific process of preprocessing the collected data in step S1 is as follows: S11: removing noise data from the collected historical environmental parameter data and historical resource usage data, and processing abnormal values therein; S12: integrating the historical environmental parameter data and historical resource usage data from different sources after removing noise and outliers, and converting the integrated data to a unified data format; S13: Perform dimensionality reduction processing on the data after data conversion, and convert the data after data conversion into a set of data on a new coordinate axis with independent dimensions in pairs through linear transformation.
3. The chassis management method based on lightweight neural network according to claim 2 is characterized in that: The formula for converting data into a unified data form in step S12 is: X ij *=(X ij -m j ) / s j ; Among them, X ij * is the converted data, X ij is the original data before conversion, μ j is the mean of the jth feature, σ j is the standard deviation of the jth feature.
4. The chassis management method based on lightweight neural network according to claim 2 is characterized in that: The specific process of performing dimensionality reduction processing on the data after data conversion in step S13 is as follows: S131: Establish a data matrix of the data after data conversion, and calculate the covariance matrix of the data matrix. The data elements at each position in the covariance matrix are: Among them, C ij is the covariance of random variables xi and xj, N is the number of data samples, n is the data dimension, μi is the mean of the data in the i dimension, and μj is the mean of the data in the j dimension; S132: performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues and corresponding eigenvectors; S133: Calculate the contribution rate of each component for each eigenvalue in step S132, and select the data of the eigenvalues whose contribution rate reaches a specified ratio.
5. The chassis management method based on lightweight neural network according to claim 4 is characterized in that: The chassis analysis model in step S2 is a lightweight neural network model, which includes a channel-by-channel convolution layer and a point-by-point convolution layer. The channel-by-channel convolution layer includes multiple convolution kernels, each convolution kernel processes a channel of feature data, and the point-by-point convolution layer linearly combines multiple feature maps output by the channel-by-channel convolution layer to generate a final output feature map.
6. The chassis management method based on lightweight neural network according to claim 1 is characterized in that: The specific process of training and validating the chassis analysis model is as follows: S21: extracting some data from the historical environmental parameter data and the historical resource usage data as pre-training data, inputting the data into the chassis analysis model, and freezing all weights of the model; S22: adding a customized top classifier, optimizer, loss function and evaluation index and compiling the model, and then inputting historical environmental parameter data and historical resource usage data into the chassis analysis model for training; S23: During the training of the chassis analysis model, adjusting the learning rate and the batch size to optimize the model; S24: Extract validation sets from historical environmental parameter data and historical resource usage data to input into the optimized chassis analysis model, and calculate indicators such as model accuracy and recall rate to evaluate model performance.
7. A chassis management system based on a lightweight neural network, used to implement a chassis management method based on a lightweight neural network as described in any one of claims 1 to 6, characterized in that: It includes a data acquisition module, a data preprocessing module, a model creation module, a chassis analysis model, a visualization module and an early warning module, wherein the data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the chassis analysis model, the model creation module is connected to the chassis analysis model, and the chassis analysis model is connected to the visualization module and the early warning module; The data acquisition module is used to obtain the historical environmental parameter data and historical resource usage data collected by the data acquisition module inside the chassis, and to collect the environmental parameter data and resource usage data inside the chassis in real time; The data preprocessing module is used to preprocess the historical environmental parameter data and historical resource usage data collected by the data collection module inside the chassis and the environmental parameter data and resource usage data collected in real time inside the chassis; The model creation module is used to create and deploy a chassis analysis model; The chassis analysis model is used to classify and analyze the environmental parameter data and resource usage data collected in real time, and predict the failure of the chassis based on the classification and analysis results, output the failure prediction results, and transmit the classification and analysis results to the visualization module; The visualization module is used to display the classification and analysis results and the fault prediction results; The early warning module is used to issue an early warning for the predicted fault.