Operation and Maintenance Management Method and System Based on Digital Twin

Abstract

The present invention relates to the technical field of operation and maintenance management, and particularly to an operation and maintenance management method based on digital twin, comprising the following steps: S1: By collecting the structural data and operation data of the actual operation and maintenance system, a corresponding digital twin model is constructed; S2: The operation data of the operation and maintenance system is collected in real time and transmitted to the cloud platform, and the real-time transmitted operation data is preprocessed in the cloud platform; S3: The preprocessed data is analyzed in real time through the digital twin model to detect the system operation status and abnormal conditions; S4: Based on historical data and real-time data, a fault prediction model is constructed using machine learning algorithms to predict possible future faults; S5: According to the fault prediction results and the system operation status, an optimized management plan is generated. Through precise data collection and real-time monitoring, potential fault hazards of equipment can be discovered in a timely manner, accidental shutdown of equipment is avoided, and production losses caused by faults are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of operation and maintenance management, and particularly to an operation and maintenance management method and system based on digital twin. Background Art

[0002] With the rapid development of digital and intelligent technologies, operation and maintenance management technologies are also constantly innovating. The operation and maintenance management method and system based on digital twin realize real-time monitoring, fault prediction and optimization management of the actual operation and maintenance system by constructing a virtual model, greatly improving the operation and maintenance efficiency and management level. However, the existing operation and maintenance management technologies still have some deficiencies in comprehensive monitoring, intelligent analysis and prediction, resulting in problems such as low operation and maintenance efficiency and slow response speed in actual applications.

[0003] After retrieval, an operation and maintenance management system with the publication number CN114637640B was disclosed, and the publication date was April 19, 2024. This patent provides an operation and maintenance management system, including a cloud platform, an operation and maintenance terminal and a management server. The management server has a monitoring module storage unit, a module acquisition unit, a monitoring module output unit, a device status judgment unit and a management side communication unit. This system realizes real-time monitoring and maintenance of cloud servers through a modular monitoring and warning mechanism. However, in this technical solution, the acquisition and operation of the monitoring module rely on the manual selection of operation and maintenance personnel, lacking automated and intelligent management means, which may lead to low operation and maintenance efficiency. In addition, the adaptability and compatibility of this system for multiple types of cloud servers need to be improved, and it is difficult to meet the complex and changeable operation and maintenance requirements.

[0004] After retrieval, an IT operation and maintenance management method and an IT operation and maintenance management device with the publication number CN118550573B were disclosed, and the publication date was October 29, 2024. This patent reflects the operation status and potential risks of the data center through data processing analysis and eigenvalue extraction, constructs a risk prediction model, and realizes risk warning and emergency response. However, in this technical solution, the data processing and analysis process is relatively complex, especially in the big data environment, there are problems with the real-time and accuracy of data processing. In addition, the risk prediction model of this system depends on historical data, and its ability to respond to newly emerging operation and maintenance problems and abnormal situations is limited, which may lead to insufficient flexibility and intelligence level of operation and maintenance management.

[0005] The above problems indicate that the existing operation and maintenance management technologies still have certain deficiencies in aspects such as automation, intelligence, real-time monitoring and risk prediction. Summary of the Invention

[0006] Based on the above objectives, the present invention provides an operation and maintenance management method based on digital twin, including the following steps:

[0007] S1: Build a corresponding digital twin model by collecting the structural data and operation data of the actual operation and maintenance system;

[0008] S2: Collect the operation data of the operation and maintenance system in real time and transmit it to the cloud platform, and preprocess the operation data transmitted in real time in the cloud platform;

[0009] S3: Perform real-time analysis on the preprocessed data through the digital twin model to detect the system operation status and abnormal conditions;

[0010] S4: Based on historical data and real-time data, use machine learning algorithms to build a fault prediction model to predict possible future faults. The input of the fault prediction model is historical data and real-time data, and the output is the fault prediction result;

[0011] S5: Generate an optimization management plan according to the fault prediction result and the system operation status, including system maintenance, resource scheduling and fault prevention measures.

[0012] Preferably, S1 specifically includes the following steps:

[0013] S1.1: Collect the structural data and operation data of the operation and maintenance system through sensors and data acquisition devices, including device status, environmental parameters and system performance indicators;

[0014] S1.2: Preprocess the collected data, including data cleaning, normalization and format conversion. Among them, data cleaning is used to remove invalid data and outliers in the data, normalization is used to normalize the data to the range of [0, 1], and format conversion is used to convert the preprocessed data into a unified format;

[0015] S1.3: Build a physical model according to the structural data of the operation and maintenance system, build a behavior model according to the operation data of the operation and maintenance system, and integrate the physical model and the behavior model to form a digital twin model.

[0016] Preferably, in S1.3, the physical model includes the geometric parameters, material properties and connection relationships of the device;

[0017] Use the following formula to describe the geometric parameters of the device:

[0018] ; where, is the geometric parameter of the th device, including the length, width, height, weight, material properties, connection relationships of the device;

[0019] The behavior model describes the operation status and performance indicators of the device. Use the following formula to describe the operation status of the device: ; where, is the The operating state of a device, and the specific parameters include the operating temperature, working current, working voltage, vibration frequency, and energy consumption of the device;

[0020] Integrate the physical model and the behavior model to form a digital twin model. It is expressed by the following formula: ; where is the digital twin model, composed of the physical model and the behavior model integrated together.

[0021] Preferably, S3 specifically includes the following steps:

[0022] S3.1: Extract key features from the preprocessed data, including device status, environmental parameters, and system performance indicators;

[0023] S3.2: Map the extracted features into the digital twin model;

[0024] S3.3: Based on the mapped data, perform status detection;

[0025] S3.4: Classify the detected abnormal situations.

[0026] Preferably, in S3.1, the feature extraction process is expressed by the following formula:

[0027] ; where is the th feature;

[0028] In S3.2, when mapping the extracted features into the digital twin model, the data mapping process is expressed by the following formula:

[0029] ; where is the th mapped data.

[0030] Preferably, in S3.3, based on the mapped data, the following formula is used for status detection:

[0031]

[0032] where the threshold is the preset demarcation value between the normal state and the abnormal state;

[0033] In S3.4,

[0034] When classifying the detected abnormal situations, the following formula is used for abnormal classification:

[0035] ;

[0036] Among them, threshold 1, threshold 2, threshold 3, threshold 4, etc. are preset abnormal type determination values.

[0037] Preferably, the specific steps of S4 include:

[0038] 4.1: Collect a large amount of preprocessed historical data and real-time data of the operation and maintenance system, including temperature, current, voltage, vibration frequency, and energy consumption;

[0039] 4.2: Construct a long short-term memory network based on the attention mechanism, including an input layer, an LSTM layer, an attention mechanism layer, and an output layer. Among them, the input layer: receives the preprocessed data, and the data format is [time step, number of features]. The time step represents the length of the time series, and the number of features represents the number of features in each time step;

[0040] The LSTM layer: contains multiple LSTM units, and each LSTM unit is composed of an input gate, an output gate, and a forget gate. The LSTM unit is used to capture the long-term dependencies in the time series data. The parameter settings of the LSTM layer are: input dimension: number of features;

[0041] Hidden layer dimension: 128;

[0042] Time step: 100;

[0043] The attention mechanism layer: introduces the attention mechanism to highlight the data of the key time steps in the time series through weighting;

[0044] The output layer: uses a fully connected layer to map the weighted hidden state to the fault prediction result. The activation function of the output layer is the sigmoid function, and the output range is [0, 1], representing the probability of the fault occurrence. The output data is the fault prediction result, including the fault type, fault location, and fault time;

[0045] 4.3: Establish a training set and a test set, and train and test the long short-term memory network based on the attention mechanism.

[0046] Preferably, in S4.3, the training set includes historical operation and maintenance data, and the data volume is , and the data format is , and each piece of data includes the historical operation parameters and fault labels of the device ;

[0047] The test set includes real-time operation and maintenance data, and the data volume is , and the data format is , and each piece of data includes the real-time operation parameters and fault labels of the device .

[0048] Preferably, in S5, the following steps are included:

[0049] S5.1: The cloud platform generates an optimization management plan based on the fault prediction result and the system operation status;

[0050] S5.2: The optimization management plan includes a system maintenance plan, a resource scheduling plan, and fault prevention measures;

[0051] S5.3: The specific system maintenance plan includes regularly checking and maintaining equipment, replacing aging or damaged components, and ensuring the normal operation of the equipment;

[0052] S5.4: The resource scheduling plan dynamically adjusts resource allocation according to the system operation status and prediction result to optimize system performance;

[0053] S5.5: The fault prevention measures include strengthening monitoring and alarms, promptly handling possible fault hazards, and avoiding the occurrence of faults.

[0054] Correspondingly, the embodiment of the present invention further provides an operation and maintenance management system based on digital twin for running the operation and maintenance management method based on digital twin described in the embodiment of the present invention, including a data acquisition module, a data preprocessing module, a digital twin module, a fault prediction module, an optimization management module, a cloud platform, and an operation and maintenance terminal. The data acquisition module, the data preprocessing module, the digital twin module, the fault prediction module, the optimization management module, and the operation and maintenance terminal are all electrically connected to the cloud platform;

[0055] The data acquisition module includes sensors and a data collector. The sensors are used to collect the structural data and operation data of the operation and maintenance system in real time, including equipment status, environmental parameters, and system performance indicators. The data collector is used to preliminarily process the data collected by the sensors and transmit the data to the cloud platform by wired or wireless means;

[0056] The cloud platform includes a data storage unit, a data processing unit, a communication unit, and a management unit. The data storage unit is used to store the data collected by the data acquisition module and the preprocessed data. The data processing unit is used to execute the functions of the data preprocessing module, the digital twin module, and the fault prediction module. The communication unit is used to transmit data with the data acquisition module and the operation and maintenance terminal. The management unit is used to manage each functional unit of the cloud platform;

[0057] The operation and maintenance terminal includes a display unit, an operation unit, and a communication unit. The display unit is used to display the working status of the operation and maintenance management system and the optimization management plan. The operation unit is used for operation and management by operation and maintenance personnel. The communication unit is used to transmit data with the cloud platform.

[0058] Advantages of the present invention:

[0059] 1. The operation and maintenance management method of digital twin can not only improve the predictability of equipment, but also achieve intelligent fault prevention and maintenance management. Through accurate data collection and real-time monitoring, potential fault hazards of equipment can be detected in a timely manner, avoiding unexpected equipment shutdowns and greatly reducing production losses caused by faults.

[0060] 2. Through accurate data collection, rigorous data preprocessing and multi-dimensional modeling, the efficient construction of the digital twin model is ensured, thus providing accurate and reliable information support for equipment operation and maintenance management, significantly improving the reliability of equipment, extending its service life, and reducing operation and maintenance costs.

[0061] 3. By integrating physical models and behavior models, the digital twin model can accurately simulate the physical structure and dynamic operating state of equipment. This simulation can not only help operation and maintenance personnel understand the working principle of equipment, but also predict the performance of equipment under different working conditions. For example, the physical model based on geometric parameters and material properties can help evaluate the durability of equipment in extreme environments such as high temperature and high pressure, while the behavior model can predict the temperature, vibration and other characteristics of equipment under specific operating conditions, providing real-time operation feedback.

[0062] 4. Through reasonable data extraction, mapping, detection and classification, a complete operation and maintenance management process is formed, which not only improves the accuracy and timeliness of equipment status monitoring, but also enhances the abnormal diagnosis and response capabilities, thus providing efficient and reliable support for equipment operation and maintenance.

[0063] 5. By establishing an accurate connection between the actual operating state of equipment and the virtual model, it provides strong support for subsequent status detection, fault diagnosis and predictive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings described below are only those of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0065] Figure 1 is the flowchart of the steps of the method of the present invention;

[0066] Figure 2 is the flowchart of the steps of S1 of the method of the present invention;

[0067] Figure 3 is the block diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it should be noted here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments. For some well-known technologies, those skilled in the art can also adopt other alternative methods for implementation; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0069] Please refer to Figures 1 - 3 , the embodiment of the present invention provides an operation and maintenance management method based on digital twin, including the following steps:

[0070] S1: By collecting the structure data and operation data of the actual operation and maintenance system, a corresponding digital twin model is constructed;

[0071] S2: The operation data of the operation and maintenance system is collected in real time and transmitted to the cloud platform, and the real-time transmitted operation data is preprocessed in the cloud platform;

[0072] S3: The preprocessed data is analyzed in real time through the digital twin model to detect the system operation status and abnormal conditions;

[0073] S4: Based on historical data and real-time data, a machine learning algorithm is used to construct a fault prediction model to predict possible future faults. The input of the fault prediction model is historical data and real-time data, and the output is the fault prediction result;

[0074] S5: According to the fault prediction result and the system operation status, an optimized management plan is generated, including system maintenance, resource scheduling, and fault prevention measures.

[0075] In a possible implementation manner, before starting operation and maintenance management, it is first necessary to construct a digital twin model by collecting the structure data and operation data of the operation and maintenance system. The structure data includes the geometric dimensions, material properties, composition structure, etc. of the equipment, while the operation data involves the operation status, environmental conditions, load conditions, etc. of the equipment. This digital twin model is the core in the subsequent operation and maintenance process and can provide accurate physical and behavioral data support for subsequent analysis and decision-making. Through the virtualization of the equipment, the digital twin model provides a basis for truly reflecting the equipment status for the entire operation and maintenance management.

[0076] Furthermore, during the normal operation of the equipment, it is crucial to collect operation data in real time. Through sensors and monitoring devices, the system can obtain real-time data. After these data are transmitted to the cloud platform, they undergo a preprocessing process, such as denoising, normalization, formatting, etc., to ensure the data quality for subsequent analysis. This step provides a high-quality and accurate data basis for subsequent real-time analysis and fault prediction.

[0077] Further, after the data preprocessing is completed, the digital twin model performs real-time analysis on the data, and then monitors the operating state of the system. In this process, the digital twin model provides an immediate feedback on the actual operating state of the device through virtual mapping, and can effectively detect abnormal situations occurring during operation. For example, when the temperature or vibration value of the device exceeds a predetermined threshold, the system can immediately identify potential failure risks. Through anomaly detection, early warnings can be issued in a timely manner to avoid the further development of equipment failures.

[0078] Further, based on historical data and the operation data collected in real time, a failure prediction model is constructed using machine learning algorithms (such as the LSTM network). The input of this model is historical data and real-time data. The model continuously optimizes the prediction accuracy through training on a large amount of historical data, so as to achieve the prediction of future failures. The failure prediction model can not only predict the time, location, and type of failures, but also identify potential risks in advance. The implementation of this step greatly improves the reliability of the equipment, avoids sudden failures, and reduces the downtime.

[0079] Further, based on the results of failure prediction and the real-time operating state of the equipment, the system generates an optimized management plan. These plans may include the regular maintenance plan of the equipment, the dynamic scheduling strategy of resources, and preventive measures for predicted failures, etc. Through these optimized management plans, it can help the operation and maintenance personnel prepare in advance and handle problems in a timely manner, thereby improving the operating efficiency of the equipment, extending its service life, and effectively reducing the maintenance cost.

[0080] In the embodiment of the present invention, S1 specifically includes the following steps:

[0081] S1.1: Collect the structural data and operation data of the operation and maintenance system through sensors and data acquisition devices, including equipment status, environmental parameters, and system performance indicators;

[0082] S1.2: Perform preprocessing on the collected data, including data cleaning, normalization, and format conversion. Among them, data cleaning is used to remove invalid data and outliers in the data, normalization is used to normalize the data to the range of [0, 1], and format conversion is used to convert the preprocessed data into a unified format;

[0083] S1.3: Construct a physical model according to the structural data of the operation and maintenance system, construct a behavior model according to the operation data of the operation and maintenance system, and integrate the physical model and the behavior model to form a digital twin model.

[0084] In a possible implementation, in step S1.1, the operation and maintenance system is comprehensively monitored through sensors and data acquisition devices to collect the structural data, environmental parameters, and system performance indicators of the device. The structural data includes the geometric characteristics, component parts, configuration, etc. of the device; the operation data involves the real-time state of the device, operation parameters (such as temperature, pressure, current, vibration, etc.), environmental factors (such as humidity, temperature, air pressure, etc.), and the performance indicators of the system (such as power, efficiency, etc.). These data are the basis for constructing the digital twin model to ensure that the obtained information truly reflects the working conditions and operating environment of the device.

[0085] Furthermore, in step S1.2, the data collected in S1.1 is cleaned, normalized, and format-converted. Data cleaning is to remove the invalid data (such as the error data generated by sensor failures) and outliers in the original data to ensure the accuracy and effectiveness of the data used for subsequent analysis. The normalization process maps the data to a standard range (usually [0, 1]) to facilitate the comparison between different features and model processing. Format conversion is to convert the cleaned and normalized data into a unified format to adapt to subsequent data analysis, modeling, and system integration. The preprocessed data has high quality and consistency and can provide reliable input for model construction.

[0086] Furthermore, in step S1.3, according to the preprocessed structural data and operation data, a physical model and a behavior model are respectively constructed. The physical model reflects the actual geometric structure, component attributes, and their mutual relationships of the device to ensure that the digital twin model can accurately reproduce the physical characteristics of the device. The behavior model constructs the dynamic behavior of the device by analyzing the operation data (such as real-time data like temperature, pressure, etc.) and shows the response characteristics of the device under different environmental conditions. By combining the physical model and the behavior model, a comprehensive digital twin model is formed, enabling the virtual model to simulate the behavior of the device in real time during dynamic operation and reflect its current state.

[0087] In the embodiment of the present invention, in S1.3, the physical model includes the geometric parameters, material properties, and connection relationships of the device;

[0088] The geometric parameters of the device are described using the following formula:

[0089] ; where is the geometric parameter of the th device, including the length, width, height, weight, material properties, and connection relationships of the device;

[0090] The behavior model describes the operating state and performance indicators of the device. The operating state of the device is described using the following formula: ; where is the The operating status of a device, and the specific parameters include the operating temperature, working current, working voltage, vibration frequency, and energy consumption of the device;

[0091] Integrate the physical model and the behavior model to form a digital twin model. It is represented by the following formula: ; where is the digital twin model, which is composed of the physical model and the behavior model integrated together.

[0092] In a possible implementation, the physical model includes the geometric parameters, material properties, and connection relationships of the device, and these elements together constitute the basic physical characteristics of the device. The geometric parameters of the device are described by the following formula: ; is the geometric parameter of the th device, including the length, width, height, weight, material properties, and connection relationships of the device. The geometric parameters provide the basic structural information of the device, the material properties reflect the physical properties of each component of the device (such as strength, temperature resistance, etc.), and the connection relationship describes how different components work together through physical contact or connection. These information help to construct a 3D model of the device and simulate the physical performance of the device under various environmental conditions.

[0093] The behavior model describes the performance of the device through its operating status, reflecting the dynamic performance of the device in actual operation. The specific parameters of the behavior model include the operating temperature, working current, working voltage, vibration frequency, and energy consumption of the device, etc. The operating status of the device is described by the following formula: ; where is the operating status of the th device, specifically including the real-time working conditions and performance data of the device. The behavior model reflects the operating characteristics of the device under different loads, environments, and working conditions through the real-time changes of these parameters, and helps to predict the possible faults or performance degradation trends of the device.

[0094] Combine the physical model and the behavior model to form a comprehensive digital twin model to characterize the real state of the device in actual operation: ; where is the digital twin model, which is composed of the physical model and the behavior model integrated together. Through this integration, the digital twin model can simultaneously display the physical structure and its dynamic operating status of the device, providing all-round support for further simulation, prediction, analysis, and decision-making.

[0095] In the embodiment of the present invention, S3 specifically includes the following steps:

[0096] S3.1: Extract key features from the preprocessed data, including device status, environmental parameters, and system performance metrics;

[0097] S3.2: Map the extracted features into the digital twin model;

[0098] S3.3: Perform status detection based on the mapped data;

[0099] S3.4: Classify the detected abnormal situations.

[0100] In a possible implementation, in S3.1, first, it is necessary to extract key features related to the device health and operating status from the original device data. These features include device status (such as temperature, pressure, rotational speed, etc.), environmental parameters (such as humidity, temperature, air quality, etc.), and system performance metrics (such as power consumption, vibration frequency, fault codes, etc.). The purpose of extracting these features is to find the core information from the massive data that can accurately reflect the device's operating health and potential risks. Usually, data processing and feature selection techniques are used, such as statistical analysis, Fourier transform, principal component analysis, etc. This step ensures the effectiveness of the data and the efficiency of processing, removing redundant or irrelevant noise information.

[0101] Furthermore, in S3.2, the key features extracted from the device operation data are mapped into the already constructed digital twin model. The digital twin model contains the physical model and behavior model of the device. By combining with the real-time data of the actual device, it can show the current operating status of the device. In this step, the extracted features are converted into the data format that the digital twin model can recognize to ensure the accuracy of the mapping. This mapping process is the key to realizing the interaction between the digital twin model and the real device, which can accurately reflect the actual situation of the device into the virtual model, thus providing a real and reliable data basis for subsequent analysis.

[0102] Furthermore, in S3.3, based on the data mapping in step S3.2, the system starts to detect and analyze the device status. This process uses the mapped data to monitor the device's health condition and detect whether there are abnormalities in real time. By comparing with the normal working range of the device, the system can find abnormal states during the device operation, such as too high temperature, excessive vibration, etc. The status detection uses algorithm models for data analysis, which may include machine learning models, statistical analysis methods, or rule-driven methods. The purpose of this step is to evaluate the device's operating status in real time and identify potential faults or performance problems in a timely manner.

[0103] Further, in S3.4, the system classifies the detected anomalies based on the results of the status detection. Anomalous situations may be caused by multiple factors, which may include equipment failures, environmental changes, external interferences, etc. Through classification, the system can further determine the severity of the anomaly, the possible causes, and the corresponding handling measures. For example, if it is detected that the equipment temperature is too high, the system can classify it as a possible cooling system failure or too high environmental temperature, and take corresponding operation and maintenance measures according to the classification results. This classification step is crucial for operation and maintenance decision-making, which helps operation and maintenance personnel quickly identify the cause of the anomaly and make targeted responses.

[0104] In the embodiment of the present invention, in S3.1, the feature extraction process is represented by the following formula:

[0105] ; where is the th feature;

[0106] In S3.2, the extracted features are mapped into the digital twin model, and the data mapping process is represented by the following formula:

[0107] ; where is the th mapped data.

[0108] In a possible implementation manner, in S3.1, key features are first extracted from the original device data. Through a series of mathematical and statistical methods, a set of feature sets representing the device status is generated. The feature set is represented by the formula: ; where is the th feature; these features include the temperature, pressure, vibration frequency, power consumption, etc. of the device. The purpose of feature extraction is to screen out the key information that can best represent the device health status from a large amount of original data. The selection of these features is based on previous experience, engineering knowledge, and data analysis, aiming to make the extracted features accurately reflect the device health status, performance changes, or failure risks.

[0109] Further, in S3.2, the features extracted in S3.1 are mapped into the digital twin model. The digital twin model is a mathematical model that virtually represents the device status and can simulate the behavior of the device under various operating conditions. The mapping process is represented by the formula: ; where is the th mapped data. The core of this step is to map the extracted features Convert it into a format that can be processed by the digital twin model and map it into the digital twin model through mathematical functions or algorithms. This process enables the real-time operation data of the device to be reflected in the virtual model, thereby simulating and analyzing the actual operation state of the device.

[0110] The feature extraction in S3.1 is closely connected to the data mapping process in S3.2. Each feature extracted through S3.1 will be mapped into the digital twin model as input to form mapping data . This connection relationship ensures the correspondence between the actual device operation state and the digital twin model, enabling the digital twin model to reflect the current health state, performance changes, and potential failure risks of the device. The accuracy of feature and data mapping directly affects the effectiveness of subsequent state detection and fault diagnosis.

[0111] Through mathematical methods and algorithms, the feature is converted into mapping data . This process can be regarded as a conversion from high-dimensional device state features to low-dimensional model representations. The feature extraction process obtains simplified high-dimensional features by analyzing a large amount of raw data, while the data mapping process converts these features into parameters or state variables of the digital twin model, thereby achieving the unity of virtual and real device states.

[0112] In the embodiment of the present invention, in S3.3, based on the mapped data, the following formula is used for state detection:

[0113] ;

[0114] where the threshold is the demarcation value preset for normal and abnormal states;

[0115] In S3.4,

[0116] classify the detected abnormal situations, and use the following formula for abnormal classification:

[0117] ;

[0118] where the thresholds 1, 2, 3, 4, etc. are the preset determination values for abnormal types.

[0119] In a possible implementation manner, in S3.3, by using the mapped data , the operation and maintenance system will determine whether the current state of the device is "normal" or "abnormal". The specific operation is to calculate the sum of the mapping data:

[0120] ;

[0121] In this step, the operation and maintenance system sums up all the mapped data. If the summation result is lower than the preset threshold, the device status is determined to be normal; if the summation result exceeds the threshold, it is determined to be abnormal. The threshold is set based on the device's historical operation data or expert experience, aiming to determine under what conditions the device exhibits an abnormal state. This step is crucial for the entire operation and maintenance management process because it can quickly identify the health status of the device and detect potential faults in a timely manner.

[0122] Furthermore, once the device status is determined to be abnormal, the next task is to classify the abnormality in detail. Step S3.4 uses a specific formula to classify the types of abnormalities:

[0123]

[0124] This process classifies the abnormalities based on specific characteristic values and their relationships with the preset thresholds. Different types of abnormalities may indicate different fault modes or device states. For example, changes in certain characteristic values may indicate too high temperature, while others may represent excessive device vibration. In this way, the operation and maintenance personnel can take different countermeasures for different types of abnormalities.

[0125] The status detection in S3.3 determines whether the device status is normal based on the summation result of the device mapped data, while the abnormality classification in S3.4 further refines the types of abnormalities. The connection between the two is that S3.3 first provides a rough determination (normal or abnormal) of the device status, and S3.4 further refines and analyzes the specific types of abnormalities after detecting an abnormality, thus providing more accurate guidance for subsequent treatment measures. The classification of abnormal states can help the operation and maintenance personnel accurately identify the problems and thus adopt targeted treatment solutions.

[0126] The overall status (normal or abnormal) of the device is obtained through S3.3. If the device is determined to be abnormal, S3.4 will enter a deeper level of analysis, that is, classify the abnormality by judging the characteristic values. The key to this process is the setting of the threshold. The reasonable selection of the threshold can effectively distinguish different types of abnormalities and reduce the possibility of misjudgment or missed judgment.

[0127] In the embodiment of the present invention, the specific steps of S4 include:

[0128] 4.1: Collect a large amount of preprocessed historical data and real-time data of the operation and maintenance system, including temperature, current, voltage, vibration frequency, and energy consumption;

[0129] 4.2: Construct a long short-term memory network based on the attention mechanism, including an input layer, an LSTM layer, an attention mechanism layer, and an output layer. Among them, the input layer: receives the preprocessed data, and the data format is [time step, number of features]. The time step represents the length of the time series, and the number of features represents the number of features at each time step;

[0130] The LSTM layer: contains multiple LSTM cells, and each LSTM cell consists of an input gate, an output gate, and a forget gate. The LSTM cell is used to capture the long-term dependencies in the time series data. The parameter settings of the LSTM layer are as follows: input dimension: number of features;

[0131] Hidden layer dimension: 128;

[0132] Time step: 100;

[0133] The attention mechanism layer: introduces the attention mechanism to highlight the data of the key time steps in the time series by weighting;

[0134] The output layer: uses a fully connected layer to map the weighted hidden state to the fault prediction result. The activation function of the output layer is the sigmoid function, and the output range is [0, 1], representing the probability of the occurrence of a fault. The output data is the fault prediction result, including the fault type, fault location, and fault time;

[0135] 4.3: Establish a training set and a test set to train and test the long short-term memory network based on the attention mechanism.

[0136] In a possible implementation manner, step 4.1 mainly collects historical data and real-time data in the operation and maintenance system. These data include the temperature, current, voltage, vibration frequency, and energy consumption of the device, etc. The goal of data collection is to obtain the operating state of the device at different time points, and provide input for fault prediction through these data. All the collected data needs to be preprocessed, and the processed data format is a time series, where the time step represents the length of the time series, and the number of features represents the number of features at each time step. By collecting and preprocessing the data, it is ensured that the data input into the model is of high quality and suitable for time series prediction.

[0137] The data collection and preprocessing in step 4.1 provide the necessary input for the LSTM network in 4.2. The preprocessed data can be input into the LSTM model in the form of a time series, enabling the model to effectively learn the operating state and potential fault information of the device at different times. The quality of the data directly affects the training effect of the LSTM network. Therefore, high-quality data collection and processing are the prerequisites for the model to successfully predict faults.

[0138] In step 4.2, a long short-term memory network (LSTM) based on the attention mechanism was constructed. The core of the LSTM network is to capture long-term dependencies in time series data. Especially for the operating state of the device and the changing trend of fault data, the LSTM network can effectively predict future fault trends. The LSTM layer processes the input data and memory state through its input gate, output gate, and forget gate. The LSTM settings here include the input dimension (number of features), hidden layer dimension (128), and time step (100). These parameters help the LSTM better capture the potential time patterns during the device operation.

[0139] By introducing the attention mechanism, the model can automatically learn and highlight those time steps that have the most influence on fault prediction, thus avoiding information loss or redundancy when dealing with long time series. The role of the attention mechanism is to weight different time steps of the input sequence, adjust the weights according to their contributions to the prediction result, and help the model pay more attention to the data at those critical moments.

[0140] After being processed by the LSTM and the attention mechanism, the output layer uses a fully connected layer to map the weighted hidden state to the fault prediction result. The activation function of the output layer is the sigmoid function, and its output range is [0, 1], representing the probability of a fault occurring. The output result includes specific information such as fault type, fault location, and fault time.

[0141] In 4.2, the LSTM layer and the attention mechanism layer work together. The LSTM layer is mainly responsible for capturing the time dependencies in the data, while the attention mechanism can assign higher weights to important time steps according to the contribution degree of historical data. This collaborative effect enhances the model's understanding ability of time series and can effectively address the challenges of long time series in device fault prediction.

[0142] In step 4.3, the collected historical data needs to be divided into a training set and a test set. The training set is used to train the LSTM model, and the test set is used to verify the prediction ability of the model. Through the training process, the LSTM model can adjust its internal parameters to better adapt to the operating modes and fault modes of different devices. During the training process, the network optimizes its prediction accuracy by minimizing the loss function. Usually, the cross-entropy loss function is used to evaluate the accuracy of fault prediction.

[0143] In 4.3, the establishment of the training set and the test set enables the LSTM model to conduct targeted learning and evaluation. The training set helps the model capture the operating mode of the device, while the test set examines the actual prediction ability of the model. Through continuous training and adjustment, the LSTM model can improve the accuracy of its fault prediction, thus providing more reliable prediction results for the operation and maintenance personnel.

[0144] In an embodiment of the present invention, in S4.3, the training set includes historical operation and maintenance data, and the data volume is , and the data format is , and each piece of data includes the historical operation parameters and fault labels of the device ;

[0145] The test set includes real-time operation and maintenance data, and the data volume is , and the data format is: , and each piece of data includes the real-time operation parameters and fault labels of the device .

[0146] In a possible implementation manner, the training set contains historical operation and maintenance data, and its data volume is , and the data format is , and each piece of data includes the historical operation parameters and fault labels of the device ; among them, the historical operation parameters of the device may include multiple features such as temperature, vibration, current, power consumption, etc., and these features reflect the state of the device at different time points. The fault label is marked according to the actual fault situation of the device, and is usually a binary classification value (for example, 0 means no fault, 1 means a fault occurs) or a multi-classification value (indicating different types of faults). The historical data in the training set provides sufficient samples for the LSTM model, helping the model learn different feature patterns of the device in normal operation and fault states, so as to effectively predict future device faults.

[0147] The test set includes real-time operation and maintenance data, and the data volume is , and the data format is: , and each piece of data includes the real-time operation parameters and fault labels of the device . These data are collected under the current operating state of the device, and are usually the data provided by the device real-time monitoring system. The role of the test set is to verify the performance of the trained LSTM model in actual applications, especially its prediction ability on unknown data (real-time data). Since real-time data usually fluctuates, the model needs to have strong generalization ability to make accurate fault predictions even in unfamiliar device states.

[0148] The training process is carried out by inputting the data in the training set Learn by minimizing a loss function (such as the cross-entropy loss function) to adjust the parameters (such as weights and biases) of the LSTM network, so that the model can fit historical operation and maintenance data and learn the complex relationship between normal and faulty states. After training, use the test set to verify the model and evaluate its prediction effect on real-time data. By comparing the prediction results with the actual fault labels The difference between them can be used to evaluate the performance of the model and further optimize it accordingly.

[0149] In the embodiment of the present invention, in S5, the following steps are included:

[0150] S5.1: The cloud platform generates an optimization management plan according to the fault prediction result and the system operation state;

[0151] S5.2: The optimization management plan includes a system maintenance plan, a resource scheduling plan, and fault prevention measures;

[0152] S5.3: The specific system maintenance plan includes regularly checking and maintaining equipment, replacing aging or damaged components to ensure the normal operation of the equipment;

[0153] S5.4: The resource scheduling plan dynamically adjusts resource allocation according to the system operation state and prediction results to optimize the system performance;

[0154] S5.5: The fault prevention measures include strengthening monitoring and alarms, promptly handling possible fault hazards, and avoiding the occurrence of faults.

[0155] In a possible implementation manner, in S5.1, the cloud platform analyzes by receiving the results from the fault prediction model and the current system operation state of the device, combines real-time data and historical data, and generates a globally optimized management plan. This plan aims to prevent and minimize device faults and covers subsequent maintenance, resource scheduling, and fault prevention, etc. The role of the cloud platform is to centrally process data from multiple devices and systems and analyze the optimal management plan through its powerful computing ability.

[0156] In S5.2, the generated optimization management plan includes three main parts: a system maintenance plan, a resource scheduling plan, and fault prevention measures. Each part is formulated based on the fault prediction result and the operation state of the device to ensure efficient and timely intervention and adjustment during the operation of the system.

[0157] The system maintenance plan is in S5.3, which specifically includes regularly inspecting and maintaining equipment, replacing aging or damaged components to ensure the continuous operation of the equipment. Through real-time monitoring of the equipment status and analysis of historical fault data, the maintenance plan can conduct early inspections and replacements of the key components of the equipment, reducing the risk of equipment aging or failure. The formulation of the maintenance plan can not only extend the service life of the equipment but also reduce the impact of sudden failures on production.

[0158] In S5.4, the resource scheduling scheme is dynamically adjusted according to the operating status and prediction results of the system. The cloud platform can monitor the load and resource usage of the equipment in real time and dynamically adjust the allocation of human resources, material resources, equipment, and funds according to changes in the system operating conditions (such as equipment load, usage frequency, fault prediction, etc.). By optimizing resource allocation, the overall performance of the system can be effectively improved, ensuring that key equipment can operate smoothly under high load and avoiding equipment overload or downtime caused by insufficient resources.

[0159] In S5.5, the fault prevention measures are to avoid the occurrence of faults and enhance the early warning ability of the system. The cloud platform, based on the real-time monitored operation data and fault prediction results, timely strengthens the monitoring and alarm of the equipment to ensure that the operation and maintenance personnel can react in a timely manner at the initial stage when there are potential fault hazards in the equipment, preventing the further deterioration of the faults. For example, when signs such as abnormal temperature or current fluctuations occur in the equipment, the system will automatically issue a warning and recommend that the operation and maintenance personnel conduct further inspections or adjustments.

[0160] Correspondingly, the embodiment of the present invention further provides an operation and maintenance management system based on digital twins for running the operation and maintenance management method based on digital twins in the embodiment of the present invention, including a data acquisition module, a data preprocessing module, a digital twin module, a fault prediction module, an optimization management module, a cloud platform, and an operation and maintenance terminal. The data acquisition module, data preprocessing module, digital twin module, fault prediction module, optimization management module, and operation and maintenance terminal are all electrically connected to the cloud platform;

[0161] The data acquisition module includes sensors and data collectors. The sensors are used to collect the structural data and operation data of the operation and maintenance system in real time, including equipment status, environmental parameters, and system performance indicators; the data collectors are used to perform preliminary processing on the data collected by the sensors and transmit the data to the cloud platform by wired or wireless means;

[0162] The cloud platform includes a data storage unit, a data processing unit, a communication unit, and a management unit. The data storage unit is used to store the data collected by the data acquisition module and the preprocessed data. The data processing unit is used to execute the functions of the data preprocessing module, digital twin module, and fault prediction module. The communication unit is used to transmit data with the data acquisition module and the operation and maintenance terminal. The management unit is used to manage each functional unit of the cloud platform;

[0163] The operation and maintenance terminal described above includes a display unit, an operation unit, and a communication unit. The display unit is used to display the working status of the operation and maintenance management system and the optimization management plan. The operation unit is used for operation and management by operation and maintenance personnel. The communication unit is used for data transmission with the cloud platform.

[0164] The present invention covers any alternatives, modifications, equivalent methods, and solutions made within the spirit and scope of the present invention. To enable the public to have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention. However, those skilled in the art can fully understand the present invention without the description of these details. In addition, well-known methods, processes, procedures, components, and circuits are not described in detail to avoid unnecessary confusion to the essence of the present invention.

[0165] The above description is only a preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A digital-twin-based operation and maintenance management method, characterized in that It includes the following steps: S1: Build a corresponding digital twin model by collecting the structural data and operation data of the actual operation and maintenance system; Among them, the digital twin model includes a physical model and a behavior model: The physical model includes the length, width, height, weight, material properties, and connection relationships of the equipment; The behavior model describes the operating state and performance indicators of the equipment, including the operating temperature, working current, working voltage, vibration frequency, and energy consumption of the equipment; Integrate the physical model and the behavior model to form a digital twin model; S2: Collect the operation data of the operation and maintenance system in real time and transmit it to the cloud platform, and preprocess the real-time transmitted operation data in the cloud platform; S3: Perform real-time analysis on the preprocessed data through the digital twin model to detect the system operating state and abnormal conditions; The specific steps in S3 include the following: S3.1: Extract key features from the preprocessed data, including equipment status, environmental parameters, and system performance indicators; S3.2: Map the extracted features into the digital twin model; S3.3: Based on the mapped data, perform status detection; S3.4: Classify the detected abnormal conditions; In S3.1, the feature extraction process is represented by the following formula: F = {f1, f2,..., f k}; where fi i is the i-th feature; In S3.2, when mapping the extracted features into the digital twin model, the data mapping process is represented by the following formula: D = {d1, d2,..., d l}; where di i is the i-th mapped data; In S3.3, based on the mapped data, the following formula is used for status detection: Among them, the threshold is the demarcation value preset for the normal state and the abnormal state; In S3.4, When classifying the detected abnormal conditions, the following formula is used for abnormal classification: Among them, threshold 1, threshold 2, threshold 3, and threshold 4 are the preset determination values for abnormal types; S4: Based on historical data and real-time data, build a fault prediction model using machine learning algorithms to predict possible future faults. The input of the fault prediction model is historical data and real-time data, and the output is the fault prediction result; S5: Generate an optimized management plan according to the fault prediction result and the system operating state, including system maintenance, resource scheduling, and fault prevention measures.

2. The operation and maintenance management method based on digital twin according to claim 1, wherein, The specific steps of S4 include: 4.1: Collect a large amount of preprocessed historical data and real-time data of the operation and maintenance system, including temperature, current, voltage, vibration frequency, and energy consumption; 4.2: Build a long short-term memory network based on the attention mechanism, including an input layer, an LSTM layer, an attention mechanism layer, and an output layer. Among them, the input layer: receives the preprocessed data, and the data format is [time step, number of features]. The time step represents the length of the time series, and the number of features represents the number of features in each time step; The LSTM layer: contains multiple LSTM units, and each LSTM unit consists of an input gate, an output gate, and a forget gate. The LSTM unit is used to capture the long-term dependencies in the time series data. The parameter settings of the LSTM layer are: input dimension: number of features; Hidden layer dimension: 128; Time step: 100; The attention mechanism layer: introduces the attention mechanism to highlight the data of key time steps in the time series through weighting; Output layer: Use a fully connected layer to map the weighted hidden state to the fault prediction result. The activation function of the output layer is the sigmoid function, and the output range is [0, 1], representing the probability of a fault occurring. The output data is the fault prediction result, including the fault type, fault location, and fault time. 4.3: Establish a training set and a test set, and train and test the long short-term memory network based on the attention mechanism.

3. The operation and maintenance management method based on digital twin according to claim 2, wherein, In S4.3, the training set includes historical operation and maintenance data with a data volume of N, and the data format is X = {x1, x2,..., x N}, and each data x i includes the historical operation parameters of the device and the fault label y i ; The test set includes real-time operation and maintenance data, with the data volume being M and the data format being X' = {x'1, x'2, …, x' m}, and each data x' i includes the real-time operation parameters of the device and the fault label y' i .

4. The operation and maintenance management method based on digital twin according to claim 1, wherein In S5, it includes the following steps: S5.1: The cloud platform generates an optimized management plan based on the fault prediction result and the system operation status. S5.2: The optimized management plan includes a system maintenance plan, a resource scheduling plan, and fault prevention measures. S5.3: The specific system maintenance plan includes regularly inspecting and maintaining equipment, replacing aging or damaged components to ensure the normal operation of the equipment. S5.4: The resource scheduling plan dynamically adjusts resource allocation according to the system operation status and prediction result to optimize system performance. S5.5: The fault prevention measures include strengthening monitoring and alerts, and promptly handling potential fault hazards to avoid the occurrence of faults.

5. A digital-twin-based operation and maintenance management system for running the digital-twin-based operation and maintenance management method according to any one of claims 2-4, characterized in that, It includes a data acquisition module, a data preprocessing module, a digital twin module, a fault prediction module, an optimization management module, a cloud platform, and an operation and maintenance terminal. The data acquisition module, data preprocessing module, digital twin module, fault prediction module, optimization management module, and operation and maintenance terminal are all electrically connected to the cloud platform. The data acquisition module includes sensors and a data collector. The sensors are used to collect the structural data and operation data of the operation and maintenance system in real time, including equipment status, environmental parameters, and system performance indicators. The data collector is used to perform preliminary processing on the data collected by the sensors and transmit it to the cloud platform by wired or wireless means. The cloud platform includes a data storage unit, a data processing unit, a communication unit, and a management unit. The data storage unit is used to store the data collected by the data acquisition module and the preprocessed data. The data processing unit is used to execute the functions of the data preprocessing module, digital twin module, and fault prediction module. The communication unit is used to transmit data with the data acquisition module and the operation and maintenance terminal. The management unit is used to manage each functional unit of the cloud platform. The operation and maintenance terminal includes a display unit, an operation unit, and a communication unit. The display unit is used to display the working status of the operation and maintenance management system and the optimized management plan. The operation unit is used for operation and management by operation and maintenance personnel. The communication unit is used to transmit data with the cloud platform.

Images