Digital twin middle station and self-healing decision-making system for oil and gas equipment management

Abstract

The invention relates to the technical field of digital manufacturing, and discloses a digital twin middle station and self-healing decision system for oil and gas equipment management, which comprises a multi-source sensing terminal, a twin optimization module, a fault topology diagnosis module and an early warning execution terminal, a multi-modal synchronous fusion mechanism is constructed, and when the state of oil and gas equipment is monitored, a heterogeneous data alignment standard is formulated, and a dynamic frequency normalization rule is set for different equipment types, so that the definiteness of multi-source sensing data fusion is ensured; meanwhile, vibration, temperature, pressure and sound wave data are synchronously collected in real time, millisecond-level detection can be achieved, the data time difference problem can be eliminated, the accuracy of equipment state feature extraction is guaranteed, and the feature fusion error is further reduced; and generating a self-adaptive correction coefficient in real time by calculating the dynamic deviation of the physical model and the data model.

Description

Technical Field

[0001] The present invention relates to the field of digital manufacturing technology, and particularly to a digital twin middle platform and a self-healing decision-making system for oil and gas equipment management. Background Art

[0002] Oil and gas resources are extremely precious properties. It is crucial to develop and utilize oil and gas resources reasonably and safely. The safety and reliability of oil and gas equipment are directly related to the utilization of oil and gas resources and the personal safety of mining personnel. Once a failure of oil and gas equipment occurs, it will not only cause economic losses, but may also cause major casualties and environmental damage.

[0003] The main defects of existing power equipment operation status evaluation and prediction methods include: 1. Defects of traditional monitoring technologies: (1) Data fragmentation: The single-class sensor acquisition mode cannot synchronously fuse multi-dimensional parameters such as vibration, temperature, pressure, and sound waves, resulting in distorted extraction of equipment state characteristics.

[0004] (2) Response hysteresis: The artificial-dependent fault diagnosis process causes delays in decision execution and is difficult to meet the safety requirements of minute-level response.

[0005] (3) Model rigidity: The digital twin constructed by static physical equations cannot adapt to parameter drift caused by equipment aging and working condition fluctuations.

[0006] 2. Bottlenecks of intelligent diagnosis technologies: (1) Disconnected driving modes: The physical model and the data-driven model operate independently, and there is a dual failure risk under sensor anomalies and complex working conditions.

[0007] (2) Missing propagation paths: The fault propagation topological relationship between equipment components is ignored in the diagnosis process, and the system-level risk conduction effect cannot be quantified.

[0008] (3) Non-optimized decision-making: The maintenance strategy lacks a multi-objective collaborative optimization mechanism for resource paths, safety risks, and cost efficiency.

[0009] 3. Bottlenecks of industry core technologies: (1) Contradiction of asynchronous and distorted data: Due to the sampling time difference of multi-source sensing data exceeding 0.5 seconds, the dynamic feature fusion error rate is as high as more than 15%, exceeding the safety threshold of 5%; (2) Contradiction of dynamic mismatch of the twin body: The static model cannot track the real-time state changes of the equipment, and the parameter deviation exceeds 20% under complex working conditions of sand-containing fluids, leading to the risk of misjudgment; (3)Safety-efficiency imbalance contradiction: The manual decision-making mode is difficult to simultaneously meet the safety standards and the 90% cost control target. The actual passing rate of maintenance costs is less than 70%, and the gap in fault response time reaches 24 times.

[0010] Therefore, in view of the above problems, the digital twin middle platform and self-healing decision-making system for oil and gas equipment management provided by the present invention can integrate an innovative system of multi-modal sensing fusion, dual-track model collaboration, topology propagation analysis, and intelligent optimization decision-making, and construct an intelligent management system for oil and gas equipment with minute-level self-healing ability. Summary of the Invention

[0011] (I) Technical problems to be solved In view of the deficiencies of the prior art, the present invention provides a digital twin middle platform and self-healing decision-making system for oil and gas equipment management, which solves the problems raised in the above background technology.

[0012] (II) Technical solutions To achieve the above objectives, the present invention provides the following technical solutions: A digital twin middle platform and self-healing decision-making system for oil and gas equipment management, including: a multi-source perception terminal, a twin body optimization module, a fault topology diagnosis module, and a warning execution terminal; The multi-source perception terminal sets the equipment global monitoring area and formulates multi-modal data fusion standards, deploys vibration sensors, temperature sensors, pressure sensors, and acoustic wave sensors at multiple nodes in the monitoring area, performs real-time timestamp alignment and frequency normalization processing on multi-source heterogeneous sensing data, determines whether there are data asynchronous conflicts in the equipment operation state, and performs real-time data flow correction after discovering data fragmentation; The twin body optimization module collects real-time equipment physical parameters and real-time working condition data, generates a basic digital twin body according to the fluid dynamics equation and the material stress model, calculates the twin body state deviation value through the LSTM-Attention network, triggers the dual-track fusion algorithm in real time when the model parameter drift exceeds the threshold, and dynamically updates the twin body weight coefficient when the model is mismatched; The fault topology diagnosis module constructs a topological relationship diagram of equipment components, performs multi-level division on the fault propagation path, automatically detects the fault probability of component nodes, quantifies the fault correlation of pumps, valves, and pipelines in real time, and generates risk warning instructions in real time according to the fault conduction path eigenvalue; The warning execution terminal issues hierarchical warning instructions through an audible and visual alarm device when the system determines that the fault exceeds the limit, the model is mismatched, or the execution is abnormal, and synchronously drives the edge controller to intervene and feedback the manual intervention signal.

[0013] Preferably, it further includes: Edge-cloud collaborative architecture: Deploy fog computing nodes at the edge layer to preprocess sensor data in real time; Deploy a digital twin simulation platform in the cloud for large-scale model training and decision optimization; Realize data synchronization between the edge and the cloud through the OPC UA protocol.

[0014] Preferably, the system includes the following steps: S1. Collect real-time operation data of oil and gas equipment through a multi-source heterogeneous sensor network to generate a device multi-modal operation data set; the multi-modal operation data includes vibration waveform data, temperature distribution data, pressure pulsation data, and acoustic wave spectrum data; S2. Based on the device multi-modal operation data, construct a dynamic digital twin, and perform real-time correction of the digital twin parameters through a dual-rail fusion mechanism of physical model driving and data driving to generate an optimized digital twin model; S3. Use the optimized digital twin model to perform multi-scale diagnostic analysis of equipment faults, and combine the fault propagation tree model and the graph neural network topology relationship recognition to generate equipment fault feature vectors and fault location information; S4. Based on the fault feature vectors and fault location information, use the ant colony optimization algorithm to generate a self-healing decision-making plan, and the self-healing decision-making plan includes a maintenance resource scheduling path, a device parameter adaptive adjustment strategy, and a safety isolation instruction; S5. Execute the self-healing decision-making plan, drive the actuator through the edge control terminal to dynamically adjust the oil and gas equipment, and feedback the execution status to the digital twin in real time to complete the closed-loop self-healing control.

[0015] Preferably, the S1 includes: S11. Collect pipeline vibration waveform data through distributed fiber Bragg grating sensors, collect device temperature distribution data through infrared thermal imagers, collect pressure pulsation data through high-frequency pressure transmitters, and collect acoustic wave spectrum data through acoustic emission sensors; S12. Perform timestamp synchronization and sampling frequency normalization processing on the multi-modal operation data: When the sampling frequencies of the vibration waveform data and the acoustic wave spectrum data are inconsistent, use the cubic spline interpolation algorithm to adjust the sampling frequency of the acoustic wave spectrum data to be the same as that of the vibration waveform data; generate a time-aligned multi-modal operation data set: Where is vibration waveform data, is the infrared thermal imaging temperature distribution data, is the pressure pulsation data, is the acoustic wave spectrum data.

[0016] Preferably, S2 includes: S21. Construct a digital twin basic model based on physical mechanisms, including hydrodynamic equations, heat conduction equations, and material stress and deformation equations; S22. Input the multi-modal operation dataset into the LSTM-Attention neural network, and output the device state correction coefficient driven by data ; S23. Update the digital twin parameters through a dual-track fusion mechanism: Optimized parameter = Physical model parameter where is the device state correction coefficient driven by data; Generate an optimized digital twin model .

[0017] Preferably, S3 includes: S31. Analyze the device topology relationship based on the fault propagation tree model, and construct a fault propagation directed graph where the vertex represents the device component, and the edge represents the fault propagation path; S32. Input the multi-modal operation dataset and the fault propagation directed graph into the graph convolutional neural network, and output the fault probability matrix ; S33. When the fault probability of any component in exceeds the threshold , extract the vibration-temperature-pressure joint feature vector of this component .

[0018] Preferably, S4 includes: S41. Establish a self-healing decision optimization objective function: where is the repair time, is the resource cost, is the risk coefficient, is the weight coefficient; S42. Use the ant colony optimization algorithm to solve the objective function: Initialize the pheromone matrix of the repair path ; Dynamically adjust the path weight through an adaptive pheromone update strategy: Among them is the pheromone increment of the optimal path, is the pheromone concentration of the path volatilization coefficient, and t is the current iteration number; Output the self-healing decision scheme Path sequence, parameter adjustment value, isolation instruction .

[0019] Preferably, the dynamic digital twin adopts a four-layer architecture: Device layer: Integrate the sensor network and actuators; Transport layer: Realize data edge preprocessing and cloud transmission through 5G and industrial Internet of Things protocols; Model layer: Include physical model library, machine learning model library and real-time simulation engine; Application layer: Provide fault diagnosis, decision generation and visual interaction interface.

[0020] Preferably, the dual-track fusion mechanism in S23 further includes: S231. When the output deviation between the physical model and the data-driven model exceeds the fault tolerance threshold, start the Bayesian evidence fusion algorithm: Among them is the output of the physical model, is the output of the data-driven model, is the likelihood of the physical model; S232. Use as the weight to update the digital twin parameters.

[0021] Preferably, S1 further includes a special leakage detection module: S13. Perform wavelet packet decomposition on the acoustic wave spectrum data and extract the energy characteristics in the frequency band of 80 - 200 kHz ; S14. When exceeds the baseline value and the pressure pulsation data shows a negative pressure wave mutation, generate a pipeline leakage alarm .

[0022] (III) Beneficial effects Compared with the prior art, the present invention provides a digital twin middle platform and a self-healing decision system for oil and gas equipment management, having the following beneficial effects: 1. In the present invention, by constructing a multimodal synchronous fusion mechanism, when monitoring the state of oil and gas equipment, by formulating heterogeneous data alignment standards and setting dynamic frequency normalization rules for different equipment types, the clarity of multi-source sensing data fusion is ensured; at the same time, vibration, temperature, pressure and acoustic wave data are collected in real time and synchronized, and the data time difference problem can be detected and eliminated in milliseconds, ensuring the accuracy of equipment state feature extraction and further reducing the feature fusion error.

[0023] 2. In the present invention, by establishing a dual-track drive-topology diagnosis model, when analyzing equipment faults, by calculating the dynamic deviation between the physical model and the data model, an adaptive correction coefficient is generated in real time; when the sensor is abnormal or the working condition changes suddenly, the conduction path between components can be analyzed through the fault propagation tree, enabling the system to identify the fault correlation relationships ignored by traditional diagnosis, and when the fault propagation path is abnormal, the diagnosis direction can be corrected in real time through the topology network, ensuring the accuracy of fault location under complex working conditions.

[0024] 3. In the present invention, by designing a risk-balanced decision-making engine, when generating a self-healing plan, based on the real-time state of the equipment and the fault propagation topology, the maintenance efficiency, resource cost and safety risk are synchronized and coordinated through a multi-objective optimization algorithm; when the safety score of the maintenance path is lower than the industry standard, the plan can be optimized in real time through the dynamic decision weight adjustment mechanism, enabling the system to achieve the coordinated balance of safety and efficiency for different fault scenarios, reducing the risk of one-sidedness in manual decision-making, and further improving the comprehensive efficiency of equipment management. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a schematic diagram of the overall system architecture of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Please refer to Figure 1 , the digital twin middle platform and self-healing decision-making system for oil and gas equipment management, comprises: a multi-source sensing terminal, a twin body optimization module, a fault topology diagnosis module and an early warning execution terminal; The multi-source perception terminal sets the device's global monitoring area and formulates multi-modal data fusion standards. It deploys vibration sensors, temperature sensors, pressure sensors, and acoustic wave sensors at multiple nodes in the monitoring area, performs real-time timestamp alignment and frequency normalization on multi-source heterogeneous sensing data, determines whether data asynchronous conflicts occur in the device operation state, and performs real-time data stream correction after detecting data fragmentation; The twin optimization module generates a basic digital twin by collecting the device's physical parameters and real-time working condition data in real time, and calculates the twin state deviation value through the LSTM-Attention network. When the model parameter drift exceeds the threshold, it triggers the dual-track fusion algorithm in real time and dynamically updates the twin weight coefficient when the model is mismatched; The fault topology diagnosis module constructs a topology relationship diagram of device components, hierarchically divides the fault propagation path, automatically detects the fault probability of component nodes, quantifies the fault correlation of pumps, valves, and pipelines in real time, and generates risk warning instructions in real time according to the fault conduction path eigenvalue; The warning execution terminal issues graded warning instructions through the sound and light alarm device when the system determines that the fault exceeds the limit, the model is mismatched, or the execution is abnormal, and synchronously drives the edge controller to intervene and feedback the manual intervention signal.

[0028] It also includes: Edge-cloud collaborative architecture: Fog computing nodes are deployed at the edge layer to preprocess sensor data in real time; A digital twin simulation platform is deployed in the cloud for large-scale model training and decision optimization; Data synchronization between the edge and the cloud is achieved through the OPC UA protocol.

[0029] The system includes the following steps: S1. Collect the real-time operation data of oil and gas equipment through a multi-source heterogeneous sensor network to generate a device multi-modal operation data set; the multi-modal operation data includes vibration waveform data, temperature distribution data, pressure pulsation data, and acoustic wave spectrum data; S2. Based on the device multi-modal operation data, construct a dynamic digital twin, and perform real-time correction on the digital twin parameters through a dual-track fusion mechanism driven by physical models and data to generate an optimized digital twin model; S3. Use the optimized digital twin model to perform multi-scale diagnostic analysis of device faults, combine the fault propagation tree model and the graph neural network topology relationship recognition to generate device fault feature vectors and fault location information; S4. Based on the fault feature vectors and fault location information, use the ant colony optimization algorithm to generate a self-healing decision-making scheme, and the self-healing decision-making scheme includes the maintenance resource scheduling path, the device parameter adaptive adjustment strategy, and the safety isolation instruction; S5. Execute the self-healing decision-making plan, drive the actuator through the edge control terminal to dynamically adjust the oil and gas equipment, and feedback the execution status to the digital twin in real time to complete the closed-loop self-healing control.

[0030] S1 includes: S11. Collect pipeline vibration waveform data through distributed fiber optic Bragg grating sensors, collect equipment temperature distribution data through infrared thermal imagers, collect pressure pulsation data through high-frequency pressure transmitters, and collect acoustic wave spectrum data through acoustic emission sensors; S12. Perform timestamp synchronization and sampling frequency normalization processing on the multi-modal operation data: When the sampling frequencies of the vibration waveform data and the acoustic wave spectrum data are inconsistent, use the cubic spline interpolation algorithm to adjust the sampling frequency of the acoustic wave spectrum data to be the same as that of the vibration waveform data; generate a time-aligned multi-modal operation data set: where is vibration waveform data, is the infrared thermal imaging temperature distribution data, is the pressure pulsation data, is the acoustic wave spectrum data.

[0031] S2 includes: S21. Build a digital twin basic model based on physical mechanisms, including hydrodynamic equations, heat conduction equations, and material stress and deformation equations; S22. Input the multi-modal operation data set into the LSTM-Attention neural network and output the data-driven device status correction coefficient ; S23. Update the digital twin parameters through a dual-track fusion mechanism: Optimized parameter = physical model parameter where, is the data-driven device status correction coefficient; Generate an optimized digital twin model .

[0032] S3 includes: S31. Analyze the device topology relationship based on the fault propagation tree model and build a fault propagation directed graph where the vertex represents the device component, and the edge represents the fault propagation path; S32. Input the multi-modal operation data set With the fault propagation directed graph Input the graph convolutional neural network and output the fault probability matrix ; S33. When The fault probability of any component in exceeds the threshold , extract the vibration-temperature-pressure joint feature vector of this component , and generate fault location information .

[0033] S4 includes: S41. Establish the self-healing decision optimization objective function: Where is the maintenance time, is the resource cost, is the risk coefficient, is the weight coefficient; S42. Use the ant colony optimization algorithm to solve the objective function: Initialize the pheromone matrix of the maintenance path ; Dynamically adjust the path weight through the adaptive pheromone update strategy: Where is the pheromone increment of the optimal path, is the pheromone concentration of the path Volatility coefficient, t is the current iteration number; Output the self-healing decision plan Path sequence, parameter adjustment value, isolation instruction .

[0034] The dynamic digital twin adopts a four-layer architecture: Device layer: Integrate the sensor network and the actuator; Transport layer: Realize data edge preprocessing and cloud transmission through 5G and industrial Internet of Things protocols; Model layer: Include the physical model library, machine learning model library and real-time simulation engine; Application layer: Provide fault diagnosis, decision generation and visual interaction interface.

[0035] The dual-track fusion mechanism in S23 also includes: S231. When the output deviation between the physical model and the data-driven model exceeds the fault tolerance threshold, start the Bayesian evidence fusion algorithm: Where is the output of the physical model, For the output of the data-driven model, is the likelihood of the physical model; S232. Using as the weight to update the digital twin parameters.

[0036] S1 also includes a dedicated leakage detection module: S13. Perform wavelet packet decomposition on the acoustic wave spectrum data to extract the energy characteristics in the frequency band of 80 - 200 kHz ; S14. When exceeds the baseline value and the pressure pulsation data shows a negative pressure wave mutation, generate a pipeline leakage alarm .

[0037] Example 1: Early warning of compressor bearing failure in a land treatment plant: In the centrifugal compressor system of a certain onshore natural gas treatment plant, when the sulfur content in the inlet gas suddenly increases to 300 ppm, the system detects abnormal axial vibration through the three-axis vibration sensor on the compressor main shaft. The infrared thermal imager shows that the bearing housing temperature rises from 85 °C to 102 °C, and the high-frequency pressure transmitter records a pressure volatility of 0.8 MPa / s. The multi-source perception terminal uses the cubic spline interpolation algorithm to unify the frequencies of the acoustic wave data and the vibration data, eliminates the time difference of 0.48 seconds to generate a synchronous data set. The twin optimization module, based on the characteristics of the sulfur corrosion working condition, predicts the bearing stress value of 128 MPa through the physical model, while the LSTM-Attention network analyzes the vibration-temperature correlation characteristics and outputs δ = 0.19. After Bayesian fusion, the stress value is corrected to 152 MPa. The fault topology diagnosis module constructs a conduction chain of "bearing failure → rotor imbalance → shell rupture". The graph convolutional network outputs a bearing fault probability of 94% and a rotor imbalance probability of 81%, warning of the risk of shell bursting 14 minutes in advance. The self-healing decision engine generates a multi-objective optimization plan: the ant colony algorithm plans the shortest repair path of 283 meters, the command speed is reduced from 9800 rpm to 8500 rpm, and the emergency nitrogen purge is started synchronously. The edge controller completes the speed adjustment within 1.1 seconds through the Profinet protocol, and the audible and visual alarm starts the rotating red light alarm. After adjustment, the vibration value drops to 4.1 g, the bearing temperature stabilizes at 88 °C, and the pressure volatility returns to zero; compared with the 189-minute response time of the traditional SCADA system, this system completes fault suppression in 11 minutes, avoids a single accident loss of $1.2 million, and improves the equipment availability rate to 99.4%.

[0038] Example 2: Self-healing of the seal failure of the oil transfer pump on an offshore platform: In the high-pressure external transmission system of a deep-water oil and gas platform in the South China Sea, when sand-containing crude oil flows through the main transfer pump, the system captures abnormal vibrations of 7.5g in real time through a piezoelectric vibration sensor on the pump body bearing housing. At the same time, the distributed fiber optic temperature array detects that the temperature in the seal cavity soars from 82°C to 96°C, and the leakage characteristic energy value collected by the acoustic emission sensor in the 80 - 200 kHz frequency band reaches 18 dB. The multi-source perception terminal immediately starts millisecond-level timestamp alignment, unifies the sampling frequencies of vibration, temperature, and acoustic wave data to 12 kHz, and eliminates the 0.55-second transmission time difference to generate a synchronous data set. The dynamic twin construction module, based on the characteristics of the sand-containing working condition, analyzes the strong correlation characteristics between the vibration spectrum and the leakage acoustic wave through the LSTM-Attention network, outputs a correction coefficient of 0.22, and combines the Navier-Stokes equation to perform dual-track fusion correction on the thermal conductivity, increasing the base value of 48 W / m·K to 58.56 W / m·K, reducing the temperature prediction error from 21.3% to 4.2%. The fault topology diagnosis module constructs a three-level propagation path of "seal failure → bearing wear → coupling fracture". The graph convolutional neural network outputs a seal ring fault probability of 91% and a bearing damage probability of 79%. The system warns of the risk of coupling fracture 17 minutes in advance. The self-healing decision engine immediately generates an optimal solution: the ant colony algorithm plans a 324-meter repair path, commands the pump speed to drop from 3100 rpm to 2750 rpm, and closes the inlet and outlet valves according to the API 670 standard. The edge controller completes the speed reduction operation within 0.9 seconds, and the audible and visual alarm triggers a 110 dB strobe alarm. After adjustment, the vibration value drops to 3.8g, the temperature drops back to 79°C, and the leakage rate is controlled from 22 L / h to 4 L / h. Compared with the 210-minute response time of the traditional DCS system, this system completes closed-loop self-healing in 9 minutes, avoiding a loss of $830,000 for a single unplanned shutdown.

[0039] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the said element.

[0040] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Digital twin middle platform and self-healing decision-making system for oil and gas equipment management, characterized in that, Including: Multi-source perception terminal, twin optimization module, fault topology diagnosis module and warning execution terminal; The multi-source perception terminal sets the device's global monitoring area, formulates multi-modal data fusion standards, deploys vibration sensors, temperature sensors, pressure sensors and acoustic wave sensors at multiple nodes in the monitoring area, performs timestamp alignment and frequency normalization processing on multi-source heterogeneous sensing data in real time, judges whether there is data asynchronous conflict in the device operation state, and performs data flow correction in real time after detecting data fragmentation; The twin optimization module collects device physical parameters and real-time working condition data in real time, generates a basic digital twin according to the hydrodynamic equation and the material stress model, calculates the twin state deviation value through the LSTM-Attention network, triggers the dual-track fusion algorithm in real time when the model parameter drift exceeds the threshold, and dynamically updates the twin weight coefficient when the model is mismatched; The fault topology diagnosis module constructs a device component topology relationship diagram, divides the fault propagation path into multiple levels, automatically detects the fault probability of component nodes, quantifies the fault correlation degree of pumps, valves and pipelines in real time, and generates a risk warning instruction according to the fault conduction path eigenvalue in real time; The warning execution terminal issues a hierarchical warning instruction through the sound and light alarm device when the system determines that the fault exceeds the limit, the model is mismatched, or the execution is abnormal, and synchronously drives the edge controller to intervene and feedback the manual intervention signal.

2. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 1, characterized in that, Also including: Edge-cloud collaborative architecture: Deploy fog computing nodes at the edge layer to preprocess sensor data in real time; Deploy a digital twin simulation platform in the cloud for large-scale model training and decision optimization; Realize data synchronization between the edge and the cloud through the OPC UA protocol.

3. Digital twin middle platform and self-healing decision-making system for oil and gas equipment management, characterized in that, The system includes the following steps: S1. Collect the real-time operation data of the oil and gas equipment through the multi-source heterogeneous sensor network to generate a device multi-modal operation data set; the multi-modal operation data includes vibration waveform data, temperature distribution data, pressure pulsation data and acoustic wave spectrum data; S2. Based on the device multi-modal operation data, construct a dynamic digital twin, and perform real-time correction on the digital twin parameters through the physical model-driven and data-driven dual-track fusion mechanism to generate an optimized digital twin model; S3. Use the optimized digital twin model to perform multi-scale diagnostic analysis of device faults, combine the fault propagation tree model and the graph neural network topology relationship recognition to generate device fault feature vectors and fault location information; S4. Based on the fault feature vector and fault location information, use the ant colony optimization algorithm to generate a self-healing decision-making scheme, and the self-healing decision-making scheme includes the maintenance resource scheduling path, the device parameter adaptive adjustment strategy and the safety isolation instruction; S5. Execute the self-healing decision-making scheme, drive the actuator to perform dynamic adjustment on the oil and gas equipment through the edge control terminal, and feedback the execution state to the digital twin in real time to complete the closed-loop self-healing control.

4. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, characterized in that The S1 includes: S11. Collect pipeline vibration waveform data through distributed fiber optic Bragg grating sensors, collect equipment temperature distribution data through infrared thermal imagers, collect pressure pulsation data through high-frequency pressure transmitters, and collect acoustic wave spectrum data through acoustic emission sensors; S12. Perform timestamp synchronization and sampling frequency normalization processing on multi-modal operation data: When the sampling frequencies of the vibration waveform data and the acoustic wave spectrum data are inconsistent, use the cubic spline interpolation algorithm to adjust the sampling frequency of the acoustic wave spectrum data to be the same as that of the vibration waveform data; generate a time-aligned multi-modal operation data set: Among them, is vibration waveform data, is infrared thermal imaging temperature distribution data, is pressure pulsation data, is acoustic wave spectrum data.

5. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, characterized in that, The S2 includes: S21. Build a digital twin basic model based on physical mechanisms, including hydrodynamic equations, heat conduction equations, and material stress and deformation equations; S22. Input the multi-modal operation data set into the LSTM-Attention neural network to output a data-driven device status correction coefficient ; S23. Update the digital twin body parameters through a dual-track fusion mechanism: Optimized parameter = Physical model parameter Among them, is the data-driven device status correction coefficient; Generate an optimized digital twin model .

6. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, wherein The S3 includes: S31. Analyze the device topology relationship based on the fault propagation tree model and construct a fault propagation directed graph Among them, vertices represent device components, and edges represent fault propagation paths; S32. Input the multi-modal operation data set and the fault propagation digraph into the graph convolutional neural network, and output the fault probability matrix ; S33. When the failure probability of any component exceeds the threshold , extract the vibration-temperature-pressure joint feature vector of this component , and generate fault location information .

7. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, characterized in that, The S4 includes: S41. Establish a self-healing decision optimization objective function: wherein is the repair time, is the resource cost, is the risk coefficient, is the weight coefficient; S42. Use the ant colony optimization algorithm to solve the objective function: Initialize the pheromone matrix of the repair path ; Dynamically adjust the path weights through an adaptive pheromone update strategy: Among them is the pheromone increment of the optimal path, is the pheromone concentration of the path is the evaporation coefficient, and t is the current iteration number; Output self-healing decision-making plan Path sequence, parameter adjustment value, isolation instruction 。 8. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, characterized in that, The dynamic digital twin body adopts a four-layer architecture: Device layer: Integrate the sensor network and actuators; Transport layer: Realize data edge preprocessing and cloud transmission through 5G and industrial Internet of Things protocols; Model layer: Include a physical model library, a machine learning model library, and a real-time simulation engine; Application layer: Provide fault diagnosis, decision generation, and visual interaction interfaces.

9. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 5, characterized in that, The dual-track fusion mechanism in S23 also includes: S231. When the output deviation between the physical model and the data-driven model exceeds the fault tolerance threshold, start the Bayesian evidence fusion algorithm: wherein is the output of the physical model, is the output of the data-driven model, is the likelihood of the physical model; S232. Use as the weight to update the digital twin parameters.

10. The digital twin middle platform and self-healing decision-making system for oil and gas equipment management according to claim 3, wherein, The S1 also includes a dedicated leakage detection module: S13. Perform wavelet packet decomposition on the acoustic wave spectrum data and extract the energy characteristics in the frequency band of 80 - 200 kHz ; S14. When exceeds the baseline value and the pressure pulsation data shows a sudden change in negative pressure wave, a pipeline leakage alarm is generated .

Images