Remote fault diagnosis method and system based on artificial intelligence and edge calculation

Abstract

The invention discloses a remote fault diagnosis method and system based on artificial intelligence and edge computing, and the method comprises the steps: data collection and processing: edge equipment collects the operation state data of the equipment in real time through a built-in sensor, and transmits the data to a monitoring island; real-time monitoring and fault diagnosis: the monitoring island receives real-time data of the edge equipment, monitors key equipment parameters through an adaptive threshold mechanism, triggers early warning once the data exceeds a preset threshold, and transmits abnormal data to the diagnosis island; algorithm optimization and deployment: adopting an incremental learning mechanism to realize gradual optimization of the model, and adjusting a fault mode according to feedback of the diagnosis island and expert experience; and equipment and data management: realizing storage, safety management and cross-site sharing of fault data through an equipment island and a data island part. The method is suitable for real-time monitoring and fault diagnosis of the multi-station auxiliary machines of the thermal power plant, and aims to improve the operation reliability of equipment, reduce the maintenance cost and realize intelligent management.

Description

Technical Field

[0001] The present invention relates to a remote fault diagnosis method and system, and in particular to a remote fault diagnosis method and system based on artificial intelligence and edge computing. Background Art

[0002] With the continuous development of modern power systems, equipment management and fault diagnosis of thermal power plants are facing increasingly severe challenges. Traditional equipment maintenance methods mainly rely on planned maintenance and manual inspections. Although this model played an important role in the early days, it gradually exposed its limitations in the face of the complex equipment structure and efficient operation requirements of modern large-scale power plants. Planned maintenance is prone to insufficient or excessive maintenance due to its fixed cycle, which may not only ignore potential faults, but also increase unnecessary operation and maintenance costs. In addition, manual inspections rely on the experience and judgment of operation and maintenance personnel, and are often difficult to detect in time when facing early hidden dangers of equipment, resulting in an increased risk of sudden failures.

[0003] Existing intelligent fault diagnosis technologies have alleviated the shortcomings of traditional models to a certain extent. By introducing big data analysis, machine learning, and expert systems, these diagnostic systems can analyze the operating status of equipment, predict potential problems, and provide data support for equipment optimization. However, the centralized data processing model still has the problem of insufficient response speed. When equipment from multiple plants and stations is connected to the central system for monitoring at the same time, the transmission and processing of large amounts of data will cause significant delays, limiting the system's ability to respond quickly to faults. In addition, the diversity and nonlinear characteristics of complex equipment failures often require targeted optimization and adjustment, and static algorithms are difficult to adapt to these changes, resulting in limited accuracy and flexibility in diagnosis. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a remote fault diagnosis method and system based on artificial intelligence and edge computing, which can improve the operating reliability of equipment, reduce maintenance costs and realize intelligent management.

[0005] Technical solution: The present invention comprises the following steps:

[0006] Data collection and processing: Edge devices collect equipment operating status data in real time through built-in sensors and transmit the data to the monitoring island;

[0007] Real-time monitoring and fault diagnosis: The monitoring island receives real-time data from edge devices and monitors key device parameters through an adaptive threshold mechanism. Once the data exceeds the preset threshold, an early warning is triggered and the abnormal data is transmitted to the diagnosis island.

[0008] Algorithm optimization and deployment: Adopt incremental learning mechanism to realize gradual optimization of the model and adjust the failure mode according to the feedback from the diagnosis island and expert experience;

[0009] Equipment and data management: The equipment island and data island departments are used to realize the storage, security management and cross-site sharing of fault data.

[0010] The edge device uses wavelet denoising and low-pass filtering to pre-process the collected data.

[0011] The edge device is connected to the monitoring system of the plant through an industrial communication protocol.

[0012] The diagnostic island performs in-depth analysis of data from multiple devices, integrating AI models to identify complex failure modes and predict trends.

[0013] After the diagnosis is completed, the diagnosis island transmits the diagnosis result to the algorithm island.

[0014] The equipment island records and manages the operating data, fault records and maintenance history of each plant and station equipment, and generates equipment health files and health indexes.

[0015] The data island provides system-wide data storage and governance.

[0016] The edge equipment, monitoring island and diagnosis island are deployed on the plant side.

[0017] The algorithm island, device island and data island are deployed on the regional side.

[0018] A remote fault diagnosis system based on artificial intelligence and edge computing, comprising edge devices, monitoring islands, diagnostic islands, algorithm islands, equipment islands and data islands, wherein the edge devices, monitoring islands and diagnostic islands are deployed on the plant side, and the algorithm islands, equipment islands and data islands are deployed on the regional side;

[0019] Edge devices are installed on key equipment to collect equipment operating status data in real time and transmit the pre-processed data to the monitoring island;

[0020] The monitoring island monitors the data of edge devices in real time according to the adaptive threshold mechanism. When it detects that the device status parameters exceed the threshold, it triggers a warning signal and transmits the abnormal data to the diagnosis island.

[0021] The diagnosis island uses deep learning models and spectrum analysis technology to identify complex fault patterns of abnormal data, predict the degradation trend of equipment, and feed back the diagnosis results to the algorithm island;

[0022] The algorithm island optimizes the diagnostic algorithm based on the incremental learning mechanism and expert feedback, and encrypts and transmits the optimized model to the diagnostic island and edge devices;

[0023] The equipment island generates an equipment health index and provides maintenance recommendations for each plant based on historical data analysis trends;

[0024] Data Island provides system-wide data storage and security management to ensure data integrity and security.

[0025] Beneficial effects: The present invention constructs a complete fault monitoring and diagnosis system through the collaborative work of edge devices, monitoring islands, diagnostic islands, algorithm islands, equipment islands and data islands, which can realize real-time data collection, intelligent diagnosis and continuous algorithm optimization; it is suitable for real-time monitoring and fault diagnosis of key auxiliary equipment at multiple sites in thermal power plants, aiming to improve the operating reliability of equipment, reduce maintenance costs and realize intelligent management. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a functional module flow chart of the remote fault diagnosis system of the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described below in conjunction with the accompanying drawings.

[0028] Example 1

[0029] The remote fault diagnosis method based on artificial intelligence and edge computing of this embodiment includes the following steps:

[0030] S1. Data collection and processing: The edge device of the present invention is installed on key equipment in thermal power plants, including typical auxiliary equipment such as boiler feed water pumps and condenser circulation pumps. The edge device collects equipment operating status data in real time through built-in sensors, including temperature, vibration, and pressure, etc. The data can be exchanged with the rotating machinery diagnostic monitoring system (TDM) and the plant-level monitoring information system (SIS) through industrial protocols such as Modbus and OPC. For example, when monitoring the feed water pump, the edge device collects bearing vibration and motor current to grasp the equipment operation status in real time.

[0031] The edge device communicates with the rotating machinery diagnostic monitoring management system (TDM) and the plant-level monitoring information system (SIS) in the power plant to obtain relevant data of the power plant equipment and perform data processing. The specific process is as follows:

[0032] Edge device data collection: Edge devices communicate with TDM and SIS systems through industrial communication protocols (Modbus, OPC or fieldbus, etc.). Vibration data, equipment status data and management data of rotating machinery are collected from the TDM system, including but not limited to bearing vibration signals, gear status and equipment operation time records, etc. These data can be used to identify possible mechanical problems of rotating machinery. Real-time operation data of each device, such as temperature, pressure, current and voltage parameters, are collected from the SIS system to fully understand the equipment operation status.

[0033] The edge device is equipped with a built-in data acquisition module that can directly receive data from the TDM and SIS systems. The signal is converted into a standardized digital signal through the signal processing unit in the edge computing device for subsequent processing and analysis.

[0034] Data preprocessing, in this example, the data preprocessing methods used include data denoising, data standardization and anomaly detection. In this example, a method combining wavelet denoising and low-pass filtering is used to denoise the original data. Specifically, the signal is decomposed in multiple layers through wavelet denoising to separate the noise part from the useful signal. For example, in the denoising process of vibration data, wavelet decomposition can be used to extract low-frequency signals and remove high-frequency noise generated during equipment operation; low-pass filtering suppresses noise components whose frequencies exceed the set threshold and retains effective low-frequency signals for further analysis of the equipment status.

[0035] In order to eliminate the dimensional differences between different types of data, the denoised data is standardized so that different data can maintain consistency in subsequent processing and analysis. Specifically, the edge device converts each data into the same order of magnitude or unified dimension to make the data comparable. For example, for temperature data, it is uniformly converted to degrees Celsius; for vibration data, it is converted to vibration speed in millimeters per second. In addition, data standardization also includes adjusting the values ​​of different parameters through normalization or standard deviation scaling methods so that they are between 0 and 1 or have a standard normal distribution with a mean of 0 and a standard deviation of 1. For example: when the rotation speed, vibration acceleration and temperature are collected, these data are standardized to be unified to the same dimension and range, so that the subsequent fault diagnosis model can process these data more effectively, thereby obtaining more reliable analysis results.

[0036] The sliding mean method and the 3σ principle are used to detect anomalies in the preprocessed data. Specifically, the sliding mean method detects the deviation of the data by sliding the short-term mean of the equipment parameters. For example, by calculating the sliding mean of the current signal, if the current fluctuation exceeds the set threshold, it can be determined that there may be an electrical fault. The 3σ principle is used to determine whether there are abnormal points. Specifically, if the data deviates from its historical mean by more than three standard deviations, the data point is considered to have potential anomalies. For example, in vibration data, if the vibration amplitude of certain data points is detected to significantly exceed the normal range, these data points will be marked as abnormal.

[0037] Through the above method, edge devices can efficiently collect, preprocess and detect various data on the operating status of equipment, reduce unnecessary data transmission pressure, improve the overall real-time and intelligence of the system, and provide strong data support for the in-depth diagnosis and optimization of multi-plant equipment.

[0038] S2. Real-time monitoring and fault diagnosis: Real-time monitoring and fault diagnosis focus on the collaborative work of the monitoring island and the diagnostic island, aiming to achieve real-time status monitoring, early warning and in-depth diagnostic analysis of key equipment. The monitoring island is deployed on the plant side, communicates with edge devices, obtains real-time data of the power plant to achieve real-time monitoring of power plant equipment. The diagnostic island is also deployed on the plant side, communicates with edge devices, and combines equipment fault diagnosis algorithms to perform more complex fault mode analysis and fault trend prediction on the equipment.

[0039] The monitoring island monitors the key equipment parameters of each plant in real time through an adaptive threshold mechanism. The monitoring island receives real-time data collected by edge devices, including parameters such as temperature, vibration, and pressure, and sets specific threshold ranges for different devices. For example, the normal operating temperature and pressure data of the condenser circulation pump will be used as standard thresholds. Once the detection data exceeds the preset value, the monitoring island will automatically issue an early warning. The monitoring island judges the operating status of the pump by the set temperature threshold (such as 40°C) and pressure threshold (such as 2MPa). If the temperature or pressure value deviates from the set range, the monitoring island will trigger a warning signal and transmit the abnormal data to the diagnostic island.

[0040] The monitoring island dynamically adjusts the thresholds of various parameters based on historical data. For example, the pressure range of the condenser circulation pump will automatically expand or shrink according to seasonal load fluctuations to ensure the accuracy of monitoring.

[0041] The diagnostic island is mainly responsible for aggregating and deeply analyzing the data of multiple devices to identify complex failure modes. The diagnostic island integrates artificial intelligence models to identify failure modes and trend predictions for abnormal data transmitted by edge devices and the monitoring island. The diagnostic island uses deep learning models to analyze the real-time data of the equipment and extract characteristic patterns to identify early fault signals. For example: in the diagnosis of boiler feed pumps, the diagnostic island receives abnormal vibration data from the monitoring island and uses spectrum analysis combined with convolutional neural network models to identify possible wear or failure risks of bearings, thereby predicting failure trends.

[0042] The diagnostic island predicts the deterioration trend of the equipment by retrospectively analyzing the historical operation data of the equipment. For example, if the low-frequency component of the spectrum of the vibration signal gradually increases, the system will mark it as a potential bearing wear risk and recommend maintenance personnel to inspect the equipment.

[0043] After completing the in-depth analysis, the diagnosis island transmits the diagnosis results to the algorithm island on the regional side. The algorithm island updates the algorithm based on the diagnosis results and expert experience, and completes the incremental training of the diagnosis model. Finally, the algorithm island sends the updated algorithm to the diagnosis island for subsequent fault detection. Taking the abnormal vibration of the boiler feed pump as an example, the fault analysis results of the diagnosis island show that the bearings show signs of wear. The algorithm island fine-tunes the model weights based on the fault characteristics and expert feedback to ensure that the model can accurately identify similar faults in subsequent detections.

[0044] S3. Algorithm optimization and deployment: The algorithm island on the regional side is responsible for the management and update optimization of the fault diagnosis algorithm. The algorithm island compares the performance of different model versions and selects the model with higher diagnosis accuracy as the final version.

[0045] The algorithm island receives the fault diagnosis results and related data transmitted by the diagnosis island, and optimizes and upgrades the diagnosis algorithm by combining the data records and expert experience of the edge devices and the monitoring island. The main task of the algorithm island is to learn and adapt to the latest changes in the fault mode to ensure the high accuracy and real-time performance of the entire system. Specifically, the algorithm island uses an incremental learning mechanism to achieve gradual optimization of the model, ensuring that the model can accurately identify different types of faults under changing working conditions.

[0046] The incremental learning mechanism is used to make small-scale updates based on the existing model, enabling the model to identify new fault modes or changes in operating conditions. The algorithm island collects equipment operation data from different plants and feedback from the diagnostic island, and adjusts the model parameters according to the real-time operating conditions. For example, when the vibration characteristics of the boiler feed pump are monitored to change, the algorithm island adjusts the model's neural network weights and bias values ​​through incremental learning to adapt to the new characteristics.

[0047] The algorithm island integrates an expert system, allowing domain experts to provide suggestions for improving fault modes and diagnostic algorithms based on their experience, especially in rare or complex fault scenarios. For example, for atypical frequency components appearing in the vibration spectrum, experts can mark them as specific fault types and introduce the mark into the model update process through the expert module of the algorithm island to ensure the robustness and accuracy of the model.

[0048] After collecting the feedback data from the diagnosis island and the annotation information from experts, the algorithm island uses an integrated deep learning framework (such as TensorFlow, PyTorch, etc.) to train the model and iteratively optimize the algorithm using large data sets. Model training uses a combination of batch training and stochastic gradient descent to improve training efficiency and prevent overfitting. The algorithm island also introduces transfer learning technology to migrate the fault characteristics of similar devices to the model of the target device to improve the diagnostic accuracy of new devices or new faults.

[0049] During the model training process, Algorithm Island uses cross-validation to divide the data set into multiple subsets and gradually verify the performance of the model on different data sets. This method ensures that the new algorithm not only performs well on the training data, but also guarantees high accuracy in the actual application data.

[0050] To further ensure the reliability of the diagnosis results, Algorithm Island introduces a manual verification process on the trained model. Experts review and verify the model's diagnosis results, especially those for rare faults and key equipment. By comparing historical cases with actual equipment conditions, experts confirm the accuracy of the model's diagnosis and ensure the effectiveness of the updated algorithm in key application scenarios.

[0051] During the model optimization process, Algorithm Island monitors multiple performance indicators such as diagnostic accuracy, recall rate, F1 score, etc. If an indicator does not meet the preset standard, the system will trigger additional training and adjustment. Through these performance monitoring, Algorithm Island ensures the high reliability and stability of the new model in fault detection.

[0052] When the optimized model passes verification and expert review, the algorithm island enters the algorithm deployment phase. The algorithm island uses an encrypted data transmission protocol to securely transmit the new model to the plant-side diagnostic island. The deployment process involves version control and encryption mechanisms to ensure the security and consistency of algorithm updates.

[0053] Algorithm Island establishes independent version control for each updated model to record the updated content and performance improvement effects. After each algorithm version is deployed, the previous version will be retained as a backup so that it can be quickly rolled back when an anomaly is detected. For example, when a plant station experiences an increase in false positives after a new model is deployed, the system can immediately roll back to the old version to ensure operational stability.

[0054] After the new model is deployed, the algorithm island monitors the performance of the new algorithm in actual working conditions through a real-time feedback mechanism and receives further feedback from the diagnostic island. The feedback data includes fault recognition rate, false alarm rate, and maintenance reports from operators of each plant. The algorithm island incorporates this feedback information into the closed-loop system of model optimization, continuously adjusting and optimizing the model to adapt it to the environment and operating conditions of different plants.

[0055] The Diagnosis Island summarizes the detection performance of each device under the new algorithm and sends it back to the Algorithm Island. For situations where the diagnosis effect is poor (such as working conditions with a high false alarm rate), the Algorithm Island further adjusts the sensitivity of the model or modifies the diagnosis logic to more accurately match the device status.

[0056] The algorithm island uses the feedback results of the diagnosis island and the maintenance information of the plant operators as the basis for model optimization. Through the closed-loop optimization mechanism, the algorithm island continuously updates the model to improve the long-term reliability and diagnostic accuracy of the overall system. For example, if the edge device frequently issues false warnings for the fan bearing, the algorithm island will automatically correct the threshold of the vibration data and readjust the classification boundary of the model to reduce false alarms.

[0057] The algorithm island adopts a continuous learning mechanism to enable the diagnostic model to adapt to different operating conditions. Specifically, the algorithm island will cluster the fault data over the past period of time, identify new fault modes or trends, and introduce them into the model training, so that the model can recognize the changing operating environment and fault characteristics. For example, in the continuous monitoring of the condenser circulating pump, if the vibration pattern changes significantly, the algorithm island will automatically identify this change and mark it as a new fault mode. By adding this fault mode to the model update, the system can identify the normal and abnormal states of the equipment under different loads, thereby achieving accurate monitoring.

[0058] The algorithm island regularly detects the environmental parameters (such as temperature, humidity, etc.) in which the plant equipment operates, and adjusts the input weights of the model according to these conditions. For example, a higher weight is given to temperature data in a high temperature environment, thereby maintaining high sensitivity in failures caused by thermal stress.

[0059] S4. Equipment and data management: The equipment island on the regional side is responsible for the management of various equipment in the power plant, recording and managing the operating data, fault records and maintenance history information of each plant and station equipment. Through the accumulation of a large amount of data, a comprehensive equipment operation file is established to provide support for trend analysis and fault prediction. Establish equipment health files to achieve equipment health analysis and fault prediction, and provide corresponding maintenance suggestions. Establish a unique file for each power plant equipment in the equipment island to record the historical operating parameters, number of faults and maintenance records of the equipment. For example, the file of the boiler feed pump will include historical vibration spectra, temperature change curves and related causes of failures. The equipment island generates an equipment health index based on these data and provides an operating life prediction for each equipment.

[0060] The equipment island will regularly analyze the equipment data of each plant and station to identify the changing trend of the equipment status. For example, by analyzing the historical pressure and vibration data of the condenser circulation pump, the equipment island identifies the increasing trend of abnormal pressure fluctuations and predicts possible sealing problems, thus notifying maintenance personnel to conduct inspections.

[0061] The Equipment Island will provide maintenance recommendations to plant operators based on the historical failure modes and health index of each type of equipment. For example, when the equipment health index drops below a preset threshold, the Equipment Island will automatically generate a maintenance plan and recommend that operators check and replace wearing parts during the next downtime.

[0062] The data island on the regional side is responsible for the data storage, governance and secure transmission across sites of the entire system. The data island ensures the data integrity and privacy of the system through encrypted storage and permission control, and supports data sharing between different plants and sites. As a data center, the data island coordinates the data flow of the entire system, making data management and analysis more centralized and efficient.

[0063] The data island can encrypt all transmitted and stored data, and ensure that only authorized personnel can access sensitive data through permission control. For example, equipment operation data and fault records are stored in the data island using encryption algorithms and have hierarchical access rights. Ordinary operators can only view data at their own site, while senior administrators can view data across sites.

[0064] Data Island supports data sharing among multiple plants to promote the improvement of diagnostic algorithms and the collaborative management of equipment status. For example, if a new type of failure occurs in the condenser circulation pump of a plant, Data Island will record and share the failure mode with other plants to help them identify similar problems in advance and perform preventive maintenance.

[0065] The equipment island and the data island work together to form a closed-loop optimization mechanism through in-depth analysis of equipment historical data and sharing of fault information across sites. After the fault is handled, the equipment island archives the results, and the data island is responsible for summarizing the feedback results from each plant and station and optimizing the overall fault management strategy of the system.

[0066] The equipment island feeds back the fault maintenance records and the changing trends of the equipment health index to the data island, which then aggregates and analyzes the fault information of different plants and stations. For example, after detecting similar vibration anomalies on boiler feed pumps at multiple plants and stations, the data island classifies such failure modes as common problems and lists them as key concerns, reminding all plants and stations to focus on checking this component during regular maintenance.

[0067] Example 2

[0068] The remote fault diagnosis system based on artificial intelligence and edge computing in this embodiment is specifically used for real-time monitoring, fault diagnosis and intelligent management of key auxiliary equipment at multiple sites in thermal power plants. The system includes six modules: edge devices, monitoring islands, diagnostic islands, algorithm islands, equipment islands and data islands, which respectively realize functions such as data acquisition, real-time monitoring, intelligent diagnosis, algorithm optimization, equipment management and data governance. Among them, edge devices, monitoring islands and diagnostic islands are deployed on the plant side, and algorithm islands, equipment islands and data islands are deployed on the regional side. Edge devices are installed on key equipment such as boiler feed pumps and condenser circulating pumps to collect temperature, vibration, pressure and other data in real time, and use wavelet denoising and low-pass filtering technology to pre-process the data and transmit it to the plant monitoring island. The monitoring island uses an adaptive threshold mechanism to dynamically adjust the monitoring standard. When the equipment status exceeds the threshold, it triggers an early warning and transmits abnormal data to the diagnostic island. The diagnostic island integrates convolutional neural networks and spectrum analysis technology to perform in-depth fault analysis on abnormal data, identify complex fault modes and predict equipment degradation trends.

[0069] On the regional side, the algorithm island optimizes the diagnostic algorithm through an incremental learning mechanism based on the diagnostic island feedback and expert experience to adapt to changes in different working conditions and ensure the high accuracy and robustness of the diagnostic model. The optimized algorithm is transmitted to the diagnostic island and edge devices of the plant station through encryption to support real-time diagnosis and fault prediction of the equipment. The equipment island is responsible for recording the health records and operating data of each plant and station equipment, generating equipment health indexes and providing trend analysis and maintenance recommendations. The data island ensures the security and integrity of data through encrypted storage and permission control, and supports cross-site data sharing and management, so that each plant station can share failure modes and maintenance experience to achieve closed-loop optimization of equipment management. Figure 1 As shown, the functional description of each functional module is as follows:

[0070] Edge devices: installed on key equipment in thermal power plants, such as boiler feed pumps and condenser circulation pumps, to collect real-time equipment status data such as temperature, vibration, and pressure. Edge devices use a variety of denoising and standardization processing technologies to ensure data accuracy and consistency, and connect to the plant monitoring system (such as TDM and SIS) through industrial communication protocols to achieve efficient data transmission and preprocessing.

[0071] Monitoring Island: The monitoring island is deployed on the plant side and communicates with edge devices in real time to monitor the parameters of key equipment. The monitoring island uses an adaptive threshold mechanism to dynamically adjust the monitoring standards, make real-time judgments on the temperature and pressure changes of equipment such as condenser circulation pumps, and issue a warning signal when the detection data exceeds the threshold.

[0072] Diagnostic Island: The diagnostic island is mainly responsible for in-depth analysis and fault diagnosis of equipment. Combined with AI models, the diagnostic island performs complex fault pattern recognition and trend prediction on data from the monitoring island. For example, the diagnostic island uses spectrum analysis and convolutional neural networks to identify potential wear or failure risks of equipment, thereby providing early warning and diagnosis.

[0073] Algorithm Island: The algorithm island is deployed on the regional side, receives the diagnostic results fed back by the diagnostic island, and continuously optimizes the diagnostic algorithm based on the latest fault characteristics and expert experience. The algorithm island uses incremental learning and transfer learning methods to ensure that the algorithm can adapt to changes in equipment conditions, and redeploys the updated algorithm to the diagnostic island on the plant side through encrypted transmission.

[0074] Equipment Island: The equipment island records and manages the equipment data of each plant and station to form equipment health files and operation history databases. Based on historical data and health indexes, the equipment island analyzes equipment trends and makes maintenance recommendations. For example, after detecting an abnormal pressure trend of a boiler feed pump, the equipment island automatically generates a maintenance reminder.

[0075] Data Island: Data Island is responsible for data governance of the entire system, ensuring data security and privacy, and supporting cross-site data sharing. Through encrypted storage and permission management, Data Island can share fault information and maintenance records across plants and sites, promoting collaborative management and optimization of equipment at each plant and site.

Claims

1. A remote fault diagnosis method based on artificial intelligence and edge computing, characterized in that: The following steps are involved: Data collection and processing: Edge devices collect equipment operating status data in real time through built-in sensors and transmit the data to the monitoring island; Real-time monitoring and fault diagnosis: The monitoring island receives real-time data from edge devices and monitors key device parameters through an adaptive threshold mechanism. Once the data exceeds the preset threshold, an early warning is triggered and the abnormal data is transmitted to the diagnosis island. Algorithm optimization and deployment: Adopt incremental learning mechanism to realize gradual optimization of the model and adjust the failure mode according to the feedback of diagnosis island and expert experience; Equipment and data management: The equipment island and data island departments are used to realize the storage, security management and cross-site sharing of fault data.

2. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 1, characterized in that: The edge device uses wavelet denoising and low-pass filtering to pre-process the collected data.

3. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 2, characterized in that: The edge device is connected to the monitoring system of the plant through an industrial communication protocol.

4. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 1, characterized in that: The diagnostic island performs in-depth analysis of data from multiple devices, integrating AI models to identify complex failure modes and predict trends.

5. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 4, characterized in that: After the diagnosis is completed, the diagnosis island transmits the diagnosis result to the algorithm island.

6. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 5, characterized in that: The equipment island records and manages the operating data, fault records and maintenance history of each plant and station equipment, and generates equipment health files and health indexes.

7. A remote fault diagnosis method based on artificial intelligence and edge computing according to claim 6, characterized in that: The data island provides system-wide data storage and governance.

8. The remote fault diagnosis method based on artificial intelligence and edge computing according to claim 1 is characterized in that: The edge equipment, monitoring island and diagnosis island are deployed on the plant side.

9. The remote fault diagnosis method based on artificial intelligence and edge computing according to claim 7 is characterized in that: The algorithm island, device island and data island are deployed on the regional side.

10. A remote fault diagnosis system based on artificial intelligence and edge computing, characterized in that: It includes edge devices, monitoring island, diagnosis island, algorithm island, equipment island and data island. Among them, edge devices, monitoring island and diagnosis island are deployed on the plant side, and algorithm island, equipment island and data island are deployed on the regional side. Edge devices are installed on key equipment to collect equipment operating status data in real time and transmit the pre-processed data to the monitoring island; The monitoring island monitors the data of edge devices in real time according to the adaptive threshold mechanism. When it detects that the device status parameters exceed the threshold, it triggers a warning signal and transmits the abnormal data to the diagnosis island. The diagnosis island uses deep learning models and spectrum analysis technology to identify complex fault patterns of abnormal data, predict the degradation trend of equipment, and feed back the diagnosis results to the algorithm island; The algorithm island optimizes the diagnostic algorithm based on the incremental learning mechanism and expert feedback, and encrypts and transmits the optimized model to the diagnostic island and edge devices; The equipment island generates an equipment health index and provides maintenance recommendations for each plant based on historical data analysis trends; Data Island provides system-wide data storage and security management.

Images