Intelligent fault delimitation detection method and system based on terminal computing power host management operation platform
By applying intelligent fault delimited detection method on the terminal computing power host management operation platform, combining data preprocessing, domain knowledge fusion and CNN model optimization, the empirical dependence and real-time problems of traditional fault diagnosis methods are solved, and efficient and accurate fault diagnosis and rapid response are achieved.
Patent Information
- Application Number
- CN202510280928.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-11
- Publication Date
- 2025-06-20
AI Technical Summary
Traditional fault diagnosis methods rely on manual experience and rules, and have problems such as strong experience dependence, difficulty in formulating rules, poor real-time and poor adaptability, making it difficult to effectively deal with faults in complex systems.
The intelligent fault delimited detection method based on the terminal computing power host management operation platform is adopted, and intelligent fault diagnosis is achieved through steps such as data acquisition and preprocessing, domain knowledge fusion, CNN model construction and optimization, model training and evaluation, alarm convergence and root cause positioning, etc., combined with convolutional neural network and domain knowledge.
It improves the accuracy and efficiency of fault diagnosis, shortens fault response time, improves the reliability and stability of the system, and reduces storage and transmission costs, meeting the needs of high concurrency.
Smart Images
Figure CN120179446A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and specifically to an intelligent fault location detection method and system based on a terminal computing power host management and operation platform. Background Art
[0002] With the rapid development of information technology, terminal computing power hosts have been widely used in various industries. However, with the increase in system complexity, fault diagnosis and location have become increasingly difficult. Traditional fault diagnosis methods mainly rely on manual experience and rules, and have the following problems: 1. Strong dependence on manual experience: Traditional methods highly rely on the experience of operators, lack automation and intelligence, resulting in low diagnosis efficiency and being easily affected by human factors. 2. Difficulty in rule formulation: With the expansion of system scale, fault modes are diverse and complex, making it difficult to formulate comprehensive and effective rules, and potential fault types are easily overlooked. 3. Poor real-time performance: Traditional methods lack effective real-time analysis capabilities when dealing with large amounts of data, resulting in long fault response times and affecting the stability and reliability of the system. 4. Poor adaptability: Facing the dynamic changes of the system and new types of faults, traditional methods lack self-learning and adaptation capabilities and are difficult to cope with constantly changing fault modes.
[0003] To solve the above problems, in recent years, fault diagnosis methods based on artificial intelligence have gradually received attention. Among them, the convolutional neural network (CNN) has been introduced into the field of fault diagnosis due to its excellent performance in image processing and pattern recognition. CNN can automatically extract features from raw data, reduce manual intervention, and improve the accuracy and efficiency of diagnosis. However, existing CNN-based fault diagnosis methods still have some deficiencies: 1. High model complexity: CNN models usually contain a large number of parameters, with large computational amounts, high deployment and operation costs, and are difficult to meet real-time requirements. 2. High data demand: The training of CNN models requires a large amount of labeled data, and in practical applications, it may be difficult and costly to obtain a large amount of labeled data. 3. Lack of domain knowledge integration: Existing methods mainly rely on data-driven, lack in-depth understanding of the working principles and fault modes of equipment, resulting in limited generalization ability and accuracy of the models. Summary of the Invention
[0004] The purpose of the present invention is to provide an intelligent fault location detection method and system based on a terminal computing power host management and operation platform to solve the problems raised in the above background art.
[0005] To achieve the above purpose, the present invention provides the following technical solution: An intelligent fault location detection method based on a terminal computing power host management and operation platform, the method comprising the following steps:
[0006] Data collection and preprocessing: Collect operation data from the terminal computing host, including sensor data and log files, and clean, denoise, and normalize the data to improve data quality;
[0007] Domain knowledge integration: Combine the working principle and failure modes of the device to construct feature engineering, extract valuable feature information for fault diagnosis, and embed domain knowledge into the subsequent design of the convolutional neural network model to enhance the model's diagnostic ability and accuracy;
[0008] CNN model construction and optimization: Design a convolutional neural network architecture suitable for fault diagnosis, including convolutional layers, pooling layers, and fully connected layers, and optimize the model performance by adjusting the network structure and hyperparameters;
[0009] Model training and evaluation: Use the preprocessed data to train the convolutional neural network model, evaluate the generalization ability and accuracy of the model using methods such as cross-validation, and improve its performance in practical applications by continuously adjusting the model parameters;
[0010] Alarm convergence and root cause localization: When a fault occurs, the model generates alarm information, and quickly locates the fault root cause by analyzing the alarm propagation path to shorten the fault response time and improve the reliability and stability of the system.
[0011] Preferably, in the CNN model construction and optimization step, the following optimization measures are also included:
[0012] The activation function uses the ReLU activation function to alleviate the gradient disappearance problem and accelerate the training process;
[0013] Use L2 regularization and Dropout techniques to prevent model overfitting and improve generalization ability;
[0014] Perform data augmentation by rotation, scaling, and translation to expand the training dataset and enhance the robustness of the model;
[0015] Adopt the Adam optimizer, combine momentum and adaptive learning rate to improve training efficiency and convergence speed.
[0016] Preferably, in the alarm convergence and root cause localization step, the following steps are also included:
[0017] Preprocess the generated alarm information to remove duplicate and invalid alarms;
[0018] Construct a fault tree or fault propagation graph based on the alarm propagation path and domain knowledge to assist in quickly locating the fault root cause;
[0019] Establish a fault prediction model through historical data and fault mode analysis to early warn of potential faults.
[0020] Preferably, it further includes performance optimization and deployment steps, specifically including:
[0021] Adopt structured pruning technology to compress the model scale and reduce the consumption of computing resources;
[0022] Optimize the computing efficiency of the convolutional layer to achieve acceleration of MNN operators;
[0023] Deploy the optimized model to the terminal device to ensure its efficient operation and meet the requirements of real-time fault diagnosis.
[0024] Preferably, the method realizes beneficial effects such as improved alarm speed, compressed model volume, reduced memory occupancy, improved fault response efficiency, and enhanced concurrent processing ability by optimizing the model structure and accelerating the calculation, providing a solid technical support for intelligent operation and maintenance and accurate fault location.
[0025] An intelligent fault boundary detection system based on a terminal computing power host management and operation platform is applied to an intelligent fault boundary detection method based on a terminal computing power host management and operation platform. The system includes:
[0026] A data acquisition module for collecting operation data from the terminal computing power host, including sensor data and log files;
[0027] A data preprocessing module for cleaning, denoising, and normalizing the collected data to improve data quality;
[0028] A domain knowledge fusion module that combines the working principle and fault mode of the device to construct a feature engineering, extracts valuable feature information for fault diagnosis, and embeds domain knowledge into the subsequent convolutional neural network model design;
[0029] A CNN model construction and optimization module that designs a convolutional neural network architecture suitable for fault diagnosis, including convolutional layers, pooling layers, and fully connected layers, and optimizes the model performance by adjusting the network structure and hyperparameters;
[0030] A model training and evaluation module that trains the convolutional neural network model using the preprocessed data and evaluates the generalization ability and accuracy of the model using the cross-validation method;
[0031] An alarm convergence and root cause location module that generates alarm information according to the output result of the CNN model when a fault occurs and quickly locates the root cause of the fault by analyzing the alarm propagation path;
[0032] A performance optimization and deployment module for compressing and optimizing the CNN model to meet the requirements of real-time fault diagnosis and deploying the optimized model to the terminal device.
[0033] Preferably, the CNN model construction and optimization module further includes:
[0034] Adopt the ReLU activation function to alleviate the gradient vanishing problem and accelerate the training process;
[0035] Use L2 regularization and Dropout techniques to prevent model overfitting and improve generalization ability;
[0036] Perform data augmentation by means of rotation, scaling, translation, etc. to expand the training dataset and enhance the robustness of the model;
[0037] Adopt the Adam optimizer, combine momentum and adaptive learning rate to improve training efficiency and convergence speed.
[0038] Preferably, the alarm convergence and root cause location module further includes:
[0039] Preprocess the generated alarm information to remove duplicate and invalid alarms;
[0040] Construct a fault tree or a fault propagation graph according to the alarm propagation path and domain knowledge to assist in quickly locating the root cause of the fault;
[0041] Establish a fault prediction model through historical data and fault mode analysis to early warn of potential faults.
[0042] Preferably, the performance optimization and deployment module further includes:
[0043] Adopt structured pruning technology to compress the model size and reduce the consumption of computing resources;
[0044] Optimize the computing efficiency of the convolutional layer to achieve MNN operator acceleration;
[0045] Deploy the optimized model to the terminal device to ensure that the model can run efficiently and meet the requirements of real-time fault diagnosis;
[0046] Monitor the system performance and dynamically adjust and optimize the model according to actual requirements.
[0047] Compared with the prior art, the beneficial effects of the present invention are:
[0048] The intelligent fault delimitation detection method and system based on the terminal computing power host management and operation platform proposed by the present invention realize a 24.6% increase in the alarm speed by optimizing the model structure and accelerating the calculation; adopt structured pruning technology, the model volume is compressed by 69.3%, reducing the storage and transmission costs; optimize the computing efficiency of the convolutional layer, the memory occupancy is reduced by 72%, improving the operation efficiency of the system; by quickly locating the root cause of the fault, the fault response time is shortened, and the reliability and stability of the system are improved; the optimized model can simultaneously process more fault diagnosis tasks to meet the requirements of high concurrency. Brief Description of the Drawings
[0049] Figure 1 This is the flowchart of the method of the present invention. Detailed Embodiments
[0050] In order to clearly and completely describe the objectives, technical solutions of the present invention, and make the advantages more clearly understood, the following further elaborates on the embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of them, and are only used to explain the embodiments of the present invention, not to limit the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.
[0051] Embodiment 1. Please refer to Figure 1 , the present invention provides a technical solution: an intelligent fault location detection method based on a terminal computing power host management and operation platform, the method comprising the following steps:
[0052] 1. Data collection and preprocessing:
[0053] Data collection: Collect operation data from the terminal computing power host, including sensor data and log files.
[0054] Data preprocessing: Clean, denoise, and normalize the collected data to improve data quality and lay a foundation for subsequent analysis.
[0055] 2. Domain knowledge integration:
[0056] Feature engineering: Combine the working principle and fault mode of the device to construct feature engineering and extract valuable feature information for fault diagnosis.
[0057] Knowledge embedding: Embed domain knowledge into the model design to improve the diagnostic ability and accuracy of the model.
[0058] 3. CNN model construction and optimization:
[0059] 1) Model design: We designed a CNN architecture suitable for fault diagnosis, mainly including the following layers:
[0060] A. Convolutional Layer: Used to extract local features of the input data.
[0061] B. Pooling Layer: Used to reduce the dimension of the data and reduce the amount of calculation.
[0062] C. Fully Connected Layer: Used to integrate the previous features and output the final diagnostic result.
[0063] After each convolutional layer, there is a pooling layer following closely to gradually reduce the spatial dimension of the data and extract higher-level features.
[0064] 2) Model Optimization: To improve the performance and efficiency of the model, we have taken the following optimization measures:
[0065] A. Activation Function: The ReLU (Rectified Linear Unit) activation function is adopted, which can effectively alleviate the problem of gradient vanishing and accelerate the training process.
[0066] B. Regularization Techniques: L2 regularization and Dropout techniques are used to prevent the model from overfitting and improve the generalization ability.
[0067] C. Data Augmentation: The training dataset is extended by means of rotation, scaling, translation, etc. to enhance the robustness of the model.
[0068] D. Optimization Algorithm: The Adam optimizer is adopted, which combines momentum and adaptive learning rate to improve the training efficiency and convergence speed.
[0069] 4. Model Training and Evaluation:
[0070] Training: The CNN model is trained using the preprocessed data, and methods such as cross-validation are adopted to evaluate the generalization ability and accuracy of the model.
[0071] Evaluation: By continuously adjusting the model parameters, the performance of the model in practical applications is improved.
[0072] 5. Alarm Convergence and Root Cause Location:
[0073] Alarm Generation: When a fault occurs, the model generates alarm information.
[0074] Root Cause Analysis: By analyzing the alarm propagation path, the root cause of the fault is quickly located, the fault response time is shortened, and the reliability and stability of the system are improved.
[0075] 6. Performance Optimization and Deployment:
[0076] Model Compression: Structured pruning techniques are adopted to compress the model scale and reduce the consumption of computing resources.
[0077] Acceleration Implementation: Optimize the computing efficiency of the convolutional layer to achieve acceleration of the MNN operator.
[0078] Deployment: The optimized model is deployed to the terminal device to ensure its efficient operation and meet the requirements of real-time fault diagnosis.
[0079] Example 2. Based on Example 1, an intelligent fault location and detection system based on the terminal computing power host management and operation platform is proposed, which is applied to an intelligent fault location and detection method based on the terminal computing power host management and operation platform. The system includes:
[0080] A data acquisition module for collecting operation data from the terminal computing power host, including sensor data and log files;
[0081] A data preprocessing module for cleaning, denoising, and normalizing the collected data to improve data quality;
[0082] A domain knowledge fusion module that combines the working principle and fault mode of the device to construct a feature engineering, extracts valuable feature information for fault diagnosis, and embeds domain knowledge into the subsequent design of the convolutional neural network model;
[0083] A CNN model construction and optimization module that designs a convolutional neural network architecture suitable for fault diagnosis, including convolutional layers, pooling layers, and fully connected layers, and optimizes the model performance by adjusting the network structure and hyperparameters; It also includes:
[0084] Adopt the ReLU activation function to alleviate the gradient vanishing problem and accelerate the training process;
[0085] Use L2 regularization and Dropout techniques to prevent model overfitting and improve generalization ability;
[0086] Perform data augmentation through rotation, scaling, translation, etc. to expand the training dataset and enhance the robustness of the model;
[0087] Adopt the Adam optimizer, combine momentum and adaptive learning rate to improve training efficiency and convergence speed.
[0088] A model training and evaluation module that uses the preprocessed data to train the convolutional neural network model and adopts the cross-validation method to evaluate the generalization ability and accuracy of the model;
[0089] An alarm convergence and root cause location module that generates alarm information according to the output result of the CNN model when a fault occurs, and quickly locates the root cause of the fault by analyzing the alarm propagation path; It also includes:
[0090] Preprocess the generated alarm information to remove duplicate and invalid alarms;
[0091] Construct a fault tree or fault propagation graph according to the alarm propagation path and domain knowledge to assist in quickly locating the root cause of the fault;
[0092] Establish a fault prediction model through historical data and fault mode analysis to early warn of potential faults.
[0093] A performance optimization and deployment module, which is used to compress and optimize the CNN model to meet the requirements of real-time fault diagnosis and deploy the optimized model to the terminal device; it also includes:
[0094] Adopt structured pruning technology to compress the model scale and reduce the consumption of computing resources;
[0095] Optimize the computing efficiency of the convolutional layer and achieve MNN operator acceleration;
[0096] Deploy the optimized model to the terminal device to ensure that the model can run efficiently and meet the requirements of real-time fault diagnosis;
[0097] Monitor the system performance and dynamically adjust and optimize the model according to the actual requirements.
[0098] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. An intelligent fault demarcation detection method based on a terminal computing host management and operation platform, characterized in that: The method comprises the following steps: Data collection and preprocessing: Collecting operation data from the terminal computing host, including sensor data and log files, and cleaning, denoising and normalizing the data to improve data quality; Domain knowledge fusion: Combine the working principle and failure mode of the equipment to build feature engineering, extract feature information that is valuable for fault diagnosis, and embed domain knowledge into the subsequent convolutional neural network model design to improve the model's diagnostic ability and accuracy; CNN model construction and optimization: Design a convolutional neural network architecture suitable for fault diagnosis, including convolutional layers, pooling layers, and fully connected layers, and optimize model performance by adjusting network structure and hyperparameters; Model training and evaluation: Use the preprocessed data to train the convolutional neural network model, use cross-validation and other methods to evaluate the generalization ability and accuracy of the model, and continuously adjust the model parameters to improve its performance in practical applications; Alarm convergence and root cause location: When a fault occurs, the model generates alarm information and quickly locates the root cause of the fault by analyzing the alarm propagation path, thereby shortening the fault response time and improving the reliability and stability of the system.
2. According to claim 1, an intelligent fault demarcation detection method based on a terminal computing power host management and operation platform is characterized in that: The CNN model construction and optimization steps also include the following optimization measures: The activation function uses the ReLU activation function to alleviate the gradient vanishing problem and speed up the training process; Use L2 regularization and Dropout technology to prevent model overfitting and improve generalization ability; Perform data augmentation through rotation, scaling, and translation to expand the training data set and enhance the robustness of the model; The Adam optimizer is used, combined with momentum and adaptive learning rate, to improve training efficiency and convergence speed.
3. According to claim 1, an intelligent fault demarcation detection method based on a terminal computing power host management and operation platform is characterized in that: The alarm convergence and root cause location steps also include the following steps: Pre-process the generated alarm information to remove duplicate and invalid alarms; Based on the alarm propagation path and domain knowledge, a fault tree or fault propagation diagram is constructed to assist in quickly locating the root cause of the fault; Through historical data and failure mode analysis, a failure prediction model is established to provide early warning of potential failures.
4. According to claim 1, an intelligent fault demarcation detection method based on a terminal computing power host management and operation platform is characterized in that: It also includes performance optimization and deployment steps, including: Use structured pruning technology to compress the model size and reduce computing resource consumption; Optimize the computational efficiency of the convolutional layer and accelerate the MNN operator; Deploy the optimized model to the terminal device to ensure that it can run efficiently and meet the needs of real-time fault diagnosis.
5. According to claim 1, an intelligent fault demarcation detection method based on a terminal computing power host management and operation platform is characterized in that: The method optimizes the model structure and accelerates the calculation to achieve the beneficial effects of increasing alarm speed, compressing model volume, reducing memory usage, improving fault response efficiency and enhancing concurrent processing capabilities, providing solid technical support for intelligent operation and maintenance and accurate fault location.
6. An intelligent fault demarcation detection system based on a terminal computing power host management and operation platform, applied to an intelligent fault demarcation detection method based on a terminal computing power host management and operation platform as described in any one of claims 1 to 5 above, characterized in that: The system comprises: Data collection module, used to collect operation data from the terminal computing host, including sensor data and log files; The data preprocessing module cleans, denoises and normalizes the collected data to improve data quality; The domain knowledge fusion module combines the working principle and failure mode of the equipment to build feature engineering, extract feature information that is valuable for fault diagnosis, and embed the domain knowledge into the subsequent convolutional neural network model design; CNN model construction and optimization module, which designs a convolutional neural network architecture suitable for fault diagnosis, including convolutional layers, pooling layers, and fully connected layers, and optimizes model performance by adjusting network structure and hyperparameters; A model training and evaluation module, which uses the preprocessed data to train the convolutional neural network model and uses a cross-validation method to evaluate the generalization ability and accuracy of the model; The alarm convergence and root cause location module generates alarm information based on the output results of the CNN model when a fault occurs, and quickly locates the root cause of the fault by analyzing the alarm propagation path; The performance optimization and deployment module is used to compress and optimize the CNN model to meet the needs of real-time fault diagnosis and deploy the optimized model to the terminal device.
7. The intelligent fault demarcation detection system based on the terminal computing power host management and operation platform according to claim 6 is characterized in that: The CNN model building and optimization module also includes: Use ReLU activation function to alleviate the gradient vanishing problem and speed up the training process; Use L2 regularization and Dropout technology to prevent model overfitting and improve generalization ability; Perform data augmentation through rotation, scaling, translation, etc. to expand the training data set and enhance the robustness of the model; The Adam optimizer is used, combined with momentum and adaptive learning rate, to improve training efficiency and convergence speed.
8. The intelligent fault demarcation detection system based on the terminal computing power host management and operation platform according to claim 6 is characterized in that: The alarm convergence and root cause location module also includes: Pre-process the generated alarm information to remove duplicate and invalid alarms; Construct a fault tree or fault propagation diagram based on the alarm propagation path and domain knowledge to assist in quickly locating the root cause of the fault; Through historical data and failure mode analysis, a failure prediction model is established to provide early warning of potential failures.
9. The intelligent fault demarcation detection system based on the terminal computing power host management and operation platform according to claim 6 is characterized in that: The performance optimization and deployment module also includes: Use structured pruning technology to compress the model size and reduce computing resource consumption; Optimize the computational efficiency of the convolutional layer and accelerate the MNN operator; Deploy the optimized model to the terminal device to ensure that the model can run efficiently and meet the needs of real-time fault diagnosis; Monitor system performance and dynamically adjust and optimize the model based on actual needs.