Gas turbine housing full life cycle performance prediction system based on digital twinning

Abstract

The invention relates to the technical field of gas turbines, in particular to a gas turbine housing full life cycle performance prediction system based on digital twinning. According to the technical scheme, the gas turbine housing full life cycle performance prediction system based on digital twinning comprises a digital twinning real-time monitoring module, a performance prediction analysis module, a fault early warning diagnosis module, a maintenance decision support module, an acoustic performance optimization module and a temperature control optimization module. The optical fiber grating sensor array, the acoustic fingerprint identification module and the vibration energy collector are adopted to form a multi-source data acquisition network, high-precision synchronous acquisition of three parameters of temperature, sound wave frequency and vibration energy is realized, data integrity is ensured, timeliness and accuracy of fault early warning are remarkably improved, the risk of non-planned shutdown is reduced, and the fault early warning efficiency is improved. And meanwhile, the decision-making efficiency of operators and the reliability of the system are improved through an immersive interactive interface.

Description

Technical Field

[0001] This invention relates to the field of gas turbine technology, and in particular to a gas turbine casing full life cycle performance prediction system based on digital twins. Background Technology

[0002] The gas turbine casing full life cycle performance prediction system is an intelligent platform based on digital twin and multidisciplinary simulation technology. By integrating design parameters, material data, operating conditions and environmental information, it constructs a virtual model of the casing, thereby realizing the simulation and prediction of the performance evolution of the casing from design, manufacturing, operation to decommissioning.

[0003] Existing gas turbine casing full life cycle performance prediction systems generally suffer from core defects such as weak real-time monitoring capabilities and lagging data analysis. They often only achieve single-point acquisition of basic parameters and lack a synchronous fusion mechanism for multi-source heterogeneous data. As a result, the system cannot build a complete physical state profile. For example, the acquisition frequency and timestamp of key parameters such as temperature, vibration, and acoustics are difficult to align precisely. There are significant delays in the data preprocessing stage, and insufficient edge computing capabilities result in anomaly detection response times exceeding minutes, which seriously restricts the timeliness of fault warnings. In addition, the visualization methods of traditional systems are limited to two-dimensional charts or simple three-dimensional models, lacking immersive interactive functions. Operators cannot intuitively perceive subtle changes in the equipment's operating status, leading to a delay in the discovery of potential problems. Maintenance decisions rely on experience judgment rather than data-driven decisions, ultimately increasing the risk of unplanned downtime and raising maintenance costs.

[0004] To address the aforementioned shortcomings, this solution employs a multi-source data acquisition network comprised of a fiber optic grating sensor array, an acoustic fingerprint recognition module, and a vibration energy harvester. This network enables high-precision synchronous acquisition of three parameters: temperature, sound frequency, and vibration energy. Timestamp alignment accuracy reaches the microsecond level, and the sampling frequency is uniformly set at 1kHz to ensure data integrity. The edge computing preprocessing unit integrates an FPGA accelerator card and an adaptive filter, enabling real-time data cleaning, compression, and preliminary analysis through hardware parallel computing. A lossless compression algorithm reduces the transmission bandwidth to 30% of the original data. Outlier removal and standardization are completed at the edge, shortening the response time to the millisecond level. The 3D visualization unit, based on a VR immersive headset and holographic projection device, uses voxel rendering technology to generate heat maps and airflow distribution maps, supporting multi-angle rotation and local magnification analysis. The temperature field is dynamically rendered with a red-blue gradient, and the sound wave propagation path is intuitively displayed in a ripple pattern. Vibration anomalies are located through a combination of flashing icons and physical vibration prompts from haptic feedback gloves, significantly improving the timeliness and accuracy of fault warnings and reducing the risk of unplanned downtime. Simultaneously, the immersive interactive interface enhances operator decision-making efficiency and system reliability. Summary of the Invention

[0005] To overcome the core shortcomings of existing gas turbine casing life-cycle performance prediction systems, such as weak real-time monitoring capabilities and lagging data analysis, these systems often only achieve single-point acquisition of basic parameters and lack a synchronous fusion mechanism for multi-source heterogeneous data. This results in the system being unable to construct a complete physical state profile. For example, the acquisition frequency and timestamps of key parameters such as temperature, vibration, and acoustics are difficult to align precisely, there are significant delays in the data preprocessing stage, and insufficient edge computing capabilities lead to anomaly detection response times exceeding minutes, severely restricting the timeliness of fault warnings. In addition, the visualization methods of traditional systems are limited to two-dimensional charts or simple three-dimensional models, lacking immersive interactive functions. Operators cannot intuitively perceive subtle changes in the equipment's operating status, leading to a delay in the discovery of potential problems. Maintenance decisions rely on experience judgment rather than data-driven approaches, ultimately resulting in increased risks of unplanned downtime and escalating maintenance costs.

[0006] The technical solution of this invention is: a gas turbine casing full life cycle performance prediction system based on digital twins, comprising the following modules: Digital twin real-time monitoring module: used for real-time state perception of physical systems and multi-source data fusion analysis through virtual mapping; Performance Prediction and Analysis Module: Used for full lifecycle performance prediction and optimization of flow field, thermal field, and acoustic field based on digital twin model; Fault early warning and diagnosis module: used to extract fault features, perform intelligent diagnosis, and execute multi-level early warnings based on AI algorithms; Maintenance Decision Support Module: Used for intelligent optimization of maintenance strategies and full lifecycle cost analysis based on digital twins; Acoustic performance optimization module: used for collaborative design of active noise source control and material optimization to address acoustic defects; Temperature control optimization module: used for high-precision temperature field control and dynamic management of thermal stress risk.

[0007] As a preferred option, the digital twin real-time monitoring module includes: A11: Multi-source data acquisition unit, including fiber optic grating sensor array, acoustic fingerprint recognition module and vibration energy harvester, used to integrate high-precision sensor array for real-time acquisition of multiple parameters; A12: Edge computing preprocessing unit, including FPGA acceleration card, adaptive filter and data compression module, used to realize real-time data cleaning, compression and preliminary edge analysis through FPGA acceleration and adaptive filtering; A13: 3D visualization unit, including VR immersive helmet, haptic feedback gloves and holographic projection device, used for 3D dynamic visualization of system operation status and abnormal haptic feedback using VR / holographic technology.

[0008] Preferably, the digital twin real-time monitoring module includes the following steps when it is in operation: S11: Deploy fiber Bragg grating sensor arrays, acoustic fingerprint recognition modules, and vibration energy harvesters at key locations on the gas turbine casing. By synchronously collecting three parameters—temperature, sound frequency, and vibration energy—the system ensures precise alignment of timestamps from multiple data sources and a uniform sampling frequency of 1kHz. S12: The FPGA acceleration card is used to perform real-time parallel processing of the raw data. An adaptive filter is used to eliminate 50Hz power frequency interference. The data compression module uses the lossless LZ4 algorithm to compress the transmission bandwidth to 30% of the original data. Preliminary outlier removal and data standardization are completed at the edge. S13: The processed data is mapped to a digital twin virtual model and visualized in three dimensions using a VR immersive headset. The temperature field is rendered with a red-blue gradient, the sound wave propagation path is displayed with dynamic ripples, and abnormal vibration points are marked with flashing icons. S14: The haptic feedback glove uses a piezoelectric sensor array to provide physical tactile feedback for abnormal points. It will detect when the temperature exceeds 55℃ or the vibration amplitude exceeds 0.5mm / s. 2 At that time, the corresponding area of ​​the glove generates a vibration feedback of 0.3 N·m, assisting the operator in quickly locating the fault area; S15: The holographic projection device generates heat maps and airflow distribution maps based on voxel rendering technology. It displays the temperature gradient and airflow velocity distribution of a 0.1m×0.1m grid inside the enclosure in real time through 4K resolution holographic projection, and supports multi-angle rotation viewing and local magnification analysis.

[0009] Preferably, the performance prediction and analysis module includes: A21: Flow field simulation unit, including CFD accelerated calculation unit, turbulence model optimizer and boundary layer monitoring probe, used to accurately predict airflow distribution and separation phenomena through CFD accelerated calculation and turbulence model optimization; A22: Thermal management unit, including infrared thermal imager array, phase change material temperature controller and thermal stress monitoring plate, used to combine infrared thermal imaging and phase change material for dynamic monitoring of temperature field and early warning of thermal stress risk; A23: Acoustic optimization unit, including active noise cancellation speaker array, acoustic metamaterial panel and sound field reconstruction algorithm module, used for noise source localization and broadband sound absorption optimization using active noise cancellation and metamaterial technology.

[0010] Preferably, the performance prediction and analysis module includes the following steps when it is working: S21: High-precision simulation of the internal flow field of the gas turbine casing is performed using a CFD acceleration computing unit. A three-dimensional turbulence model is established based on the Navier-Stokes equations. The k-ε parameters are adjusted through a turbulence model optimizer, and the flow field is updated in real time at 0.1 seconds / step under a GPU parallel computing architecture. S22: The infrared thermal imager array scans the surface of the housing at a frame rate of 120Hz. Combined with the phase change material temperature controller, the temperature field is dynamically adjusted. When the local temperature exceeds 45℃, the phase change material is automatically triggered to absorb heat and maintain the temperature stable within ±2℃. S23: The acoustic optimization unit locates noise sources through an active noise-canceling speaker array, uses beamforming algorithms to pinpoint noise source locations within the 20-20kHz frequency band with an error of no more than 5°, and attenuates noise by 15dB in specific frequency bands through an acoustic metamaterial panel; S24: The boundary layer monitoring probe captures real-time changes in the airflow boundary layer, measures the boundary layer thickness using a wall shear stress sensor, and optimizes ventilation system design parameters using a flow field reconstruction algorithm module. S25: The digital twin thermal model establishes the heat conduction equation based on finite element analysis, combines a PID optimization controller to predict the temperature field in real time, trains an LSTM model with historical data to predict the temperature field distribution in the next 10 minutes with an error controlled within ±1℃, and automatically adjusts the fan speed to maintain the optimal thermal environment.

[0011] Preferably, the fault early warning and diagnosis module includes: A31: Feature extraction unit, including a wavelet transform processor, a time-frequency analysis module, and a feature fusion chip, used for enhanced fault feature extraction and multi-source feature fusion through wavelet transform and time-frequency analysis; A32: Diagnostic reasoning unit, including an expert knowledge base, a deep learning engine, and a Bayesian network module, used to combine the expert knowledge base and the deep learning engine for dynamic assessment and accurate identification of fault probabilities; A33: Early warning execution unit, including intelligent air valve controller, audible and visual alarm matrix and emergency communication module, is used for fault emergency response and dual-link communication guarantee through intelligent air valve control and audible and visual alarm.

[0012] Preferably, the fault early warning and diagnosis module includes the following steps when it is working: S31: The wavelet transform processor performs multi-scale decomposition on sensor data, extracts fault features through the db4 wavelet basis, and identifies abnormal frequency components in non-stationary signals by combining the time-frequency analysis module; S32: The feature fusion chip uses a weighted fusion algorithm to intelligently weight the three parameters of temperature, vibration, and acoustics. The weighting coefficients are determined through optimization using a genetic algorithm. S33: The diagnostic reasoning unit combines an expert knowledge base with a deep learning engine. It uses a convolutional neural network to classify fused features, and a Bayesian network module dynamically evaluates the probability of failure. When the probability exceeds 80%, an early warning is triggered. S34: The intelligent damper controller in the early warning execution unit automatically switches to the backup channel when the main channel airflow is detected to be below 65000 m³ / h. 3 At a rate of / h, the electric air valve completes the switching within 0.5 seconds; S35: The sound and light alarm matrix activates a three-level alarm system. The first-level yellow warning is indicated by flashing LED lights, the second-level orange warning is indicated by a buzzer, and the third-level red warning is indicated by sending 5G+satellite dual-link alarm information to the operation and maintenance center through the emergency communication module, with a response time of no more than 2 seconds.

[0013] As a preferred option, the maintenance decision support module includes: A41: Status assessment unit, including a health assessment model, a remaining life predictor, and a maintenance cost analyzer, is used to quantitatively assess the system status by integrating multi-parameter health scores and remaining life predictions; A42: Decision optimization unit, including a genetic algorithm optimizer, a multi-objective optimization module, and an expert decision support system, used to automatically generate maintenance plans and balance cost and reliability through genetic algorithms and multi-objective optimization; A43: Knowledge Management Unit, including a fault case database, a maintenance knowledge graph, and a training simulation system, is used to build the fault case database and knowledge graph for intelligent recommendation of maintenance solutions and virtual training.

[0014] As a preferred option, the acoustic performance optimization module includes: A51: Noise source identification unit, including a microphone array, acoustic imager, and spectrum analysis module, used to generate noise distribution heatmaps and perform spectrum analysis through the microphone array and acoustic imaging; A52: Active control unit, including adaptive filter, active silencer and acoustic feedback controller, used for dynamic suppression of low-frequency noise using adaptive filtering and active silencing technology; A53: Material optimization unit, including acoustic metamaterials, damping coatings and sound insulation structures, used for customized sound absorption bands and vibration suppression through acoustic metamaterials and damping coatings.

[0015] As a preferred option, the temperature control optimization module includes: A61: Temperature monitoring unit, including an infrared thermal imager, a thermocouple array, and a temperature gradient sensor, used to integrate infrared thermal imaging and thermocouple array for non-contact global temperature measurement and gradient capture; A62: Control and execution unit, including intelligent air valve, variable frequency fan and cooling water circulation system, used for precise flow regulation and dynamic thermal management through intelligent air valve and variable frequency fan; A63: Model prediction unit, including digital twin thermal model, PID optimization controller and thermal stress analysis module, used for real-time prediction and adaptive control of temperature field based on digital twin thermal model and PID optimization.

[0016] The beneficial effects of this invention are: 1. Existing gas turbine casing lifecycle performance prediction systems generally suffer from core defects such as weak real-time monitoring capabilities and lagging data analysis. They often only achieve single-point acquisition of basic parameters and lack a mechanism for synchronous fusion of multi-source heterogeneous data. This results in the system's inability to construct a complete physical state profile. For example, the acquisition frequency and timestamps of key parameters such as temperature, vibration, and acoustics are difficult to align precisely. There are significant delays in the data preprocessing stage, and insufficient edge computing capabilities lead to anomaly detection response times exceeding minutes, severely restricting the timeliness of fault warnings. In addition, traditional systems' visualization methods are limited to two-dimensional charts or simple three-dimensional models, lacking immersive interactive functions. Operators cannot intuitively perceive subtle changes in equipment operating status, leading to delayed discovery of potential problems. Maintenance decisions rely on experience judgment rather than data-driven approaches, ultimately increasing the risk of unplanned downtime and raising maintenance costs. This solution uses a fiber optic grating sensor array, an acoustic fingerprint recognition module, and a vibration energy harvester to form a multi-source data acquisition network to achieve temperature, vibration, and acoustic data fusion. High-precision synchronous acquisition of three parameters: sound wave frequency, vibration energy, and time stamp alignment accuracy down to the microsecond level. The sampling frequency is uniformly set to 1kHz to ensure data integrity. The edge computing preprocessing unit integrates an FPGA acceleration card and an adaptive filter. Through hardware parallel computing, it realizes real-time data cleaning, compression, and preliminary analysis. The lossless compression algorithm compresses the transmission bandwidth to 30% of the original data. Outlier removal and standardization are completed at the edge, and the response time is shortened to the millisecond level. The three-dimensional visualization unit is based on a VR immersive helmet and a holographic projection device. It uses voxel rendering technology to generate heat maps and airflow distribution maps, supports multi-angle rotation viewing and local magnification analysis. The temperature field is dynamically rendered with red-blue gradient colors, and the sound wave propagation path is displayed intuitively in the form of ripples. Vibration anomalies are located through flashing icons and physical vibration prompts from tactile feedback gloves, which significantly improves the timeliness and accuracy of fault warnings, reduces the risk of unplanned downtime, and improves the decision-making efficiency of operators and the reliability of the system through an immersive interactive interface. 2. Existing gas turbine casing lifecycle performance prediction systems often only achieve short-term predictions of single parameters, lacking the ability to predict multi-physics fields such as flow, heat, and sound. They fail to reflect the dynamic evolution of equipment operating conditions, and the prediction models are mostly based on empirical formulas or simplified physical models, neglecting long-term influencing factors such as material aging and environmental changes. This leads to significant deviations between prediction results and actual operating conditions. Furthermore, traditional systems lack self-optimization mechanisms, and the prediction models cannot be dynamically updated based on real-time operating data, causing prediction accuracy to gradually decline over time. This prevents the system from identifying potential performance degradation trends in advance, making it difficult to formulate scientific maintenance plans, ultimately resulting in shortened equipment lifespan, increased maintenance costs, and even safety accidents. This solution constructs a multi-physics field coupled prediction system for flow, heat, and sound fields based on a digital twin model, employing CFD acceleration. The computational unit and turbulence model optimizer achieve high-precision flow field simulation. Combined with an infrared thermal imager array and a phase change material temperature controller, it enables dynamic monitoring of the temperature field and early warning of thermal stress risks. The acoustic optimization unit uses an active noise-canceling speaker array and acoustic metamaterial panels to locate noise sources and optimize broadband sound absorption. The boundary layer monitoring probe captures changes in the airflow boundary layer in real time and optimizes ventilation system design parameters in combination with the flow field reconstruction algorithm module. The digital twin thermal model establishes the heat conduction equation based on finite element analysis and uses a PID optimization controller to achieve real-time temperature field prediction. The LSTM model is trained using historical data to predict future temperature field distribution with an error controlled within ±1℃. This represents a leap from single-parameter prediction to multi-physics coupled prediction. The self-optimization mechanism of the digital twin model enables dynamic updating of the prediction model, significantly improving the accuracy and reliability of the full life cycle performance prediction. Attached Figure Description

[0017] Figure 1 The diagram shown is a schematic of the framework of a gas turbine casing full life cycle performance prediction system based on digital twins according to the present invention. Figure 2 The diagram shown is a schematic of the workflow of the real-time monitoring module of the digital twin-based gas turbine casing full life cycle performance prediction system of the present invention. Figure 3 The diagram shows the workflow of the performance prediction and analysis module of the gas turbine casing full life cycle performance prediction system based on digital twin of the present invention. Figure 4 The diagram shows the workflow of the fault early warning and diagnosis module of the gas turbine casing full life cycle performance prediction system based on digital twins according to the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Please see Figure 1-4This invention provides an embodiment: a gas turbine casing full life cycle performance prediction system based on digital twins, comprising the following modules: Digital twin real-time monitoring module: used for real-time state perception of physical systems and multi-source data fusion analysis through virtual mapping; Performance Prediction and Analysis Module: Used for full lifecycle performance prediction and optimization of flow field, thermal field, and acoustic field based on digital twin model; Fault early warning and diagnosis module: used to extract fault features, perform intelligent diagnosis, and execute multi-level early warnings based on AI algorithms; Maintenance Decision Support Module: Used for intelligent optimization of maintenance strategies and full lifecycle cost analysis based on digital twins; Acoustic performance optimization module: used for collaborative design of active noise source control and material optimization to address acoustic defects; Temperature control optimization module: used for high-precision temperature field control and dynamic management of thermal stress risk.

[0020] As a preferred option, the digital twin real-time monitoring module includes: A11: Multi-source data acquisition unit, including fiber optic grating sensor array, acoustic fingerprint recognition module and vibration energy harvester, used to integrate high-precision sensor array for real-time acquisition of multiple parameters; A12: Edge computing preprocessing unit, including FPGA acceleration card, adaptive filter and data compression module, used to realize real-time data cleaning, compression and preliminary edge analysis through FPGA acceleration and adaptive filtering; A13: 3D visualization unit, including VR immersive helmet, haptic feedback gloves and holographic projection device, used for 3D dynamic visualization of system operation status and abnormal haptic feedback using VR / holographic technology.

[0021] Preferably, the digital twin real-time monitoring module includes the following steps when it is in operation: S11: Deploy fiber Bragg grating sensor arrays, acoustic fingerprint recognition modules, and vibration energy harvesters at key locations on the gas turbine casing. By synchronously collecting three parameters—temperature, sound frequency, and vibration energy—the system ensures precise alignment of timestamps from multiple data sources and a uniform sampling frequency of 1kHz. S12: The FPGA acceleration card is used to perform real-time parallel processing of the raw data. An adaptive filter is used to eliminate 50Hz power frequency interference. The data compression module uses the lossless LZ4 algorithm to compress the transmission bandwidth to 30% of the original data. Preliminary outlier removal and data standardization are completed at the edge. S13: The processed data is mapped to a digital twin virtual model and visualized in three dimensions using a VR immersive headset. The temperature field is rendered with a red-blue gradient, the sound wave propagation path is displayed with dynamic ripples, and abnormal vibration points are marked with flashing icons. S14: The haptic feedback glove uses a piezoelectric sensor array to provide physical tactile feedback for abnormal points. It will detect when the temperature exceeds 55℃ or the vibration amplitude exceeds 0.5mm / s. 2 At that time, the corresponding area of ​​the glove generates a vibration feedback of 0.3 N·m, assisting the operator in quickly locating the fault area; S15: The holographic projection device generates heat maps and airflow distribution maps based on voxel rendering technology. It displays the temperature gradient and airflow velocity distribution of a 0.1m×0.1m grid inside the enclosure in real time through 4K resolution holographic projection, and supports multi-angle rotation viewing and local magnification analysis.

[0022] Preferably, the performance prediction and analysis module includes: A21: Flow field simulation unit, including CFD accelerated calculation unit, turbulence model optimizer and boundary layer monitoring probe, used to accurately predict airflow distribution and separation phenomena through CFD accelerated calculation and turbulence model optimization; A22: Thermal management unit, including infrared thermal imager array, phase change material temperature controller and thermal stress monitoring plate, used to combine infrared thermal imaging and phase change material for dynamic monitoring of temperature field and early warning of thermal stress risk; A23: Acoustic optimization unit, including active noise cancellation speaker array, acoustic metamaterial panel and sound field reconstruction algorithm module, used for noise source localization and broadband sound absorption optimization using active noise cancellation and metamaterial technology.

[0023] Preferably, the performance prediction and analysis module includes the following steps when it is working: S21: High-precision simulation of the internal flow field of the gas turbine casing is performed using a CFD acceleration computing unit. A three-dimensional turbulence model is established based on the Navier-Stokes equations. The k-ε parameters are adjusted through a turbulence model optimizer, and the flow field is updated in real time at 0.1 seconds / step under a GPU parallel computing architecture. S22: The infrared thermal imager array scans the surface of the housing at a frame rate of 120Hz. Combined with the phase change material temperature controller, the temperature field is dynamically adjusted. When the local temperature exceeds 45℃, the phase change material is automatically triggered to absorb heat and maintain the temperature stable within ±2℃. S23: The acoustic optimization unit locates noise sources through an active noise-canceling speaker array, uses beamforming algorithms to pinpoint noise source locations within the 20-20kHz frequency band with an error of no more than 5°, and attenuates noise by 15dB in specific frequency bands through an acoustic metamaterial panel; S24: The boundary layer monitoring probe captures real-time changes in the airflow boundary layer, measures the boundary layer thickness using a wall shear stress sensor, and optimizes ventilation system design parameters using a flow field reconstruction algorithm module. S25: The digital twin thermal model establishes the heat conduction equation based on finite element analysis, combines a PID optimization controller to predict the temperature field in real time, trains an LSTM model with historical data to predict the temperature field distribution in the next 10 minutes with an error controlled within ±1℃, and automatically adjusts the fan speed to maintain the optimal thermal environment.

[0024] Preferably, the fault early warning and diagnosis module includes: A31: Feature extraction unit, including a wavelet transform processor, a time-frequency analysis module, and a feature fusion chip, used for enhanced fault feature extraction and multi-source feature fusion through wavelet transform and time-frequency analysis; A32: Diagnostic reasoning unit, including an expert knowledge base, a deep learning engine, and a Bayesian network module, used to combine the expert knowledge base and the deep learning engine for dynamic assessment and accurate identification of fault probabilities; A33: Early warning execution unit, including intelligent air valve controller, audible and visual alarm matrix and emergency communication module, is used for fault emergency response and dual-link communication guarantee through intelligent air valve control and audible and visual alarm.

[0025] Preferably, the fault early warning and diagnosis module includes the following steps when it is working: S31: The wavelet transform processor performs multi-scale decomposition on sensor data, extracts fault features through the db4 wavelet basis, and identifies abnormal frequency components in non-stationary signals by combining the time-frequency analysis module; S32: The feature fusion chip uses a weighted fusion algorithm to intelligently weight the three parameters of temperature, vibration, and acoustics. The weighting coefficients are determined through optimization using a genetic algorithm. S33: The diagnostic reasoning unit combines an expert knowledge base with a deep learning engine. It uses a convolutional neural network to classify fused features, and a Bayesian network module dynamically evaluates the probability of failure. When the probability exceeds 80%, an early warning is triggered. S34: The intelligent damper controller in the early warning execution unit automatically switches to the backup channel when the main channel airflow is detected to be below 65000 m³ / h. 3 At a rate of / h, the electric air valve completes the switching within 0.5 seconds; S35: The sound and light alarm matrix activates a three-level alarm system. The first-level yellow warning is indicated by flashing LED lights, the second-level orange warning is indicated by a buzzer, and the third-level red warning is indicated by sending 5G+satellite dual-link alarm information to the operation and maintenance center through the emergency communication module, with a response time of no more than 2 seconds.

[0026] As a preferred option, the maintenance decision support module includes: A41: Status assessment unit, including a health assessment model, a remaining life predictor, and a maintenance cost analyzer, is used to quantitatively assess the system status by integrating multi-parameter health scores and remaining life predictions; A42: Decision optimization unit, including a genetic algorithm optimizer, a multi-objective optimization module, and an expert decision support system, used to automatically generate maintenance plans and balance cost and reliability through genetic algorithms and multi-objective optimization; A43: Knowledge Management Unit, including a fault case database, a maintenance knowledge graph, and a training simulation system, is used to build the fault case database and knowledge graph for intelligent recommendation of maintenance solutions and virtual training.

[0027] As a preferred option, the acoustic performance optimization module includes: A51: Noise source identification unit, including a microphone array, acoustic imager, and spectrum analysis module, used to generate noise distribution heatmaps and perform spectrum analysis through the microphone array and acoustic imaging; A52: Active control unit, including adaptive filter, active silencer and acoustic feedback controller, used for dynamic suppression of low-frequency noise using adaptive filtering and active silencing technology; A53: Material optimization unit, including acoustic metamaterials, damping coatings and sound insulation structures, used for customized sound absorption bands and vibration suppression through acoustic metamaterials and damping coatings.

[0028] As a preferred option, the temperature control optimization module includes: A61: Temperature monitoring unit, including an infrared thermal imager, a thermocouple array, and a temperature gradient sensor, used to integrate infrared thermal imaging and thermocouple array for non-contact global temperature measurement and gradient capture; A62: Control and execution unit, including intelligent air valve, variable frequency fan and cooling water circulation system, used for precise flow regulation and dynamic thermal management through intelligent air valve and variable frequency fan; A63: Model prediction unit, including digital twin thermal model, PID optimization controller and thermal stress analysis module, used for real-time prediction and adaptive control of temperature field based on digital twin thermal model and PID optimization.

[0029] Example 1 Background: The gas turbine enclosure system of a certain power plant has long suffered from problems such as lagging monitoring, untimely fault warnings, and high maintenance costs. The traditional system can only collect basic parameters at a single point and lacks the ability to fuse multi-source data, resulting in a fault detection cycle of several hours, more than 15 unplanned shutdowns per year, and losses of more than 2 million yuan per shutdown. To solve the above problems, a gas turbine enclosure full life cycle performance prediction system based on digital twins was introduced. Through virtual mapping, the system realizes real-time state perception of the physical system and multi-source data fusion analysis, constructs a multi-physics field coupled prediction model of flow field, thermal field and acoustic field, realizes intelligent extraction of fault features and multi-level early warning execution, and finally forms an intelligent optimization system for maintenance strategies and a full life cycle cost analysis system.

[0030] Implementation steps: S41: Fiber optic grating sensor arrays, acoustic fingerprint recognition modules, and vibration energy harvesters are installed at key locations on the gas turbine casing to achieve synchronous acquisition of three parameters: temperature, sound frequency, and vibration energy. The sampling frequency is uniformly 1kHz, and the timestamp alignment accuracy reaches the microsecond level. S42: Integrates an FPGA acceleration card and adaptive filter, enabling real-time data cleaning, compression, and preliminary analysis through hardware parallel computing. The lossless compression algorithm reduces the transmission bandwidth to 30% of the original data, and outlier removal and standardization are completed at the edge, reducing the response time to milliseconds. S43: The processed data is mapped onto a 3D virtual model, and a VR immersive headset is used to achieve red-blue gradient rendering of the temperature field, dynamic ripple display of the sound wave propagation path, and flashing icon marking of vibration anomalies. Combined with haptic feedback gloves, physical tactile cues are provided. S44: Flow field simulation is performed using a CFD-accelerated computing unit. A turbulence model optimizer adjusts the k-ε parameters to achieve real-time flow field updates at 0.1 seconds per step. An infrared thermal imager array scans the surface of the casing at a 120Hz frame rate, and a phase change material temperature controller maintains the temperature stable within ±2℃. The acoustic optimization unit uses beamforming algorithms to ensure noise source localization error does not exceed 5°, and acoustic metamaterial panels achieve 15dB noise attenuation. S45: The wavelet transform processor performs multi-scale decomposition of sensor data, and the feature fusion chip uses a genetic algorithm to optimize weight coefficients to achieve intelligent weighting of three-parameter features; the diagnostic reasoning unit performs fault classification through a convolutional neural network, and a Bayesian network dynamically evaluates the fault probability. When the probability exceeds 80%, a level three alarm is triggered. The intelligent air valve controller completes the backup channel switching within 0.5 seconds, and the emergency communication module sends alarm information to the operation and maintenance center through a 5G + satellite dual link. S46: The health assessment model integrates multi-parameter scoring to achieve quantitative assessment of system status; the genetic algorithm optimizer automatically generates maintenance plans; and the knowledge management unit uses a fault case database and maintenance knowledge graph to achieve intelligent recommendation of maintenance solutions and virtual training, ultimately forming a cost-reliability balanced maintenance strategy.

[0031] Data comparison table:

[0032] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A gas turbine casing life-cycle performance prediction system based on digital twins; characterized in that: It consists of the following modules: Digital twin real-time monitoring module: used for real-time state perception of physical systems and multi-source data fusion analysis through virtual mapping; Performance Prediction and Analysis Module: Used for full lifecycle performance prediction and optimization of flow field, thermal field, and acoustic field based on digital twin model; Fault early warning and diagnosis module: used to extract fault features, perform intelligent diagnosis, and execute multi-level early warnings based on AI algorithms; Maintenance Decision Support Module: Used for intelligent optimization of maintenance strategies and full lifecycle cost analysis based on digital twins; Acoustic performance optimization module: used for collaborative design of active noise source control and material optimization to address acoustic defects; Temperature control optimization module: used for high-precision temperature field control and dynamic management of thermal stress risk.

2. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 1, characterized in that: The digital twin real-time monitoring module includes: A11: Multi-source data acquisition unit, including fiber optic grating sensor array, acoustic fingerprint recognition module and vibration energy harvester, used to integrate high-precision sensor array for real-time acquisition of multiple parameters; A12: Edge computing preprocessing unit, including FPGA acceleration card, adaptive filter and data compression module, used to realize real-time data cleaning, compression and preliminary edge analysis through FPGA acceleration and adaptive filtering; A13: 3D visualization unit, including VR immersive helmet, haptic feedback gloves and holographic projection device, used for 3D dynamic visualization of system operation status and abnormal haptic feedback using VR / holographic technology.

3. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 2, characterized in that: The real-time monitoring module of the digital twin operates by including the following steps: S11: Deploy fiber Bragg grating sensor arrays, acoustic fingerprint recognition modules, and vibration energy harvesters at key locations on the gas turbine casing. By synchronously collecting three parameters—temperature, sound frequency, and vibration energy—the system ensures precise alignment of timestamps from multiple data sources and a uniform sampling frequency of 1kHz. S12: The FPGA acceleration card is used to perform real-time parallel processing of the raw data. An adaptive filter is used to eliminate 50Hz power frequency interference. The data compression module uses the lossless LZ4 algorithm to compress the transmission bandwidth to 30% of the original data. Preliminary outlier removal and data standardization are completed at the edge. S13: The processed data is mapped to a digital twin virtual model and visualized in three dimensions using a VR immersive headset. The temperature field is rendered with a red-blue gradient, the sound wave propagation path is displayed with dynamic ripples, and abnormal vibration points are marked with flashing icons. S14: The haptic feedback glove uses a piezoelectric sensor array to provide physical tactile feedback for abnormal points. It will detect when the temperature exceeds 55℃ or the vibration amplitude exceeds 0.5mm / s. 2 At that time, the corresponding area of ​​the glove generates a vibration feedback of 0.3 N·m, assisting the operator in quickly locating the fault area; S15: The holographic projection device generates heat maps and airflow distribution maps based on voxel rendering technology. It displays the temperature gradient and airflow velocity distribution of a 0.1m×0.1m grid inside the enclosure in real time through 4K resolution holographic projection, and supports multi-angle rotation viewing and local magnification analysis.

4. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 1, characterized in that: The performance predictive analytics module includes: A21: Flow field simulation unit, including CFD accelerated calculation unit, turbulence model optimizer and boundary layer monitoring probe, used to accurately predict airflow distribution and separation phenomena through CFD accelerated calculation and turbulence model optimization; A22: Thermal management unit, including infrared thermal imager array, phase change material temperature controller and thermal stress monitoring plate, used to combine infrared thermal imaging and phase change material for dynamic monitoring of temperature field and early warning of thermal stress risk; A23: Acoustic optimization unit, including active noise cancellation speaker array, acoustic metamaterial panel and sound field reconstruction algorithm module, used for noise source localization and broadband sound absorption optimization using active noise cancellation and metamaterial technology.

5. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 4, characterized in that: The performance predictive analytics module performs the following steps: S21: High-precision simulation of the internal flow field of the gas turbine casing is performed using a CFD acceleration computing unit. A three-dimensional turbulence model is established based on the Navier-Stokes equations. The k-ε parameters are adjusted through a turbulence model optimizer, and the flow field is updated in real time at 0.1 seconds / step under a GPU parallel computing architecture. S22: The infrared thermal imager array scans the surface of the housing at a frame rate of 120Hz. Combined with the phase change material temperature controller, the temperature field is dynamically adjusted. When the local temperature exceeds 45℃, the phase change material is automatically triggered to absorb heat and maintain the temperature stable within ±2℃. S23: The acoustic optimization unit locates noise sources through an active noise-canceling speaker array, uses beamforming algorithms to pinpoint noise source locations within the 20-20kHz frequency band with an error of no more than 5°, and attenuates noise by 15dB in specific frequency bands through an acoustic metamaterial panel; S24: The boundary layer monitoring probe captures real-time changes in the airflow boundary layer, measures the boundary layer thickness using a wall shear stress sensor, and optimizes ventilation system design parameters using a flow field reconstruction algorithm module. S25: The digital twin thermal model establishes the heat conduction equation based on finite element analysis, combines a PID optimization controller to predict the temperature field in real time, trains an LSTM model with historical data to predict the temperature field distribution in the next 10 minutes with an error controlled within ±1℃, and automatically adjusts the fan speed to maintain the optimal thermal environment.

6. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 1, characterized in that: The fault early warning and diagnosis module includes: A31: Feature extraction unit, including a wavelet transform processor, a time-frequency analysis module, and a feature fusion chip, used for enhanced fault feature extraction and multi-source feature fusion through wavelet transform and time-frequency analysis; A32: Diagnostic reasoning unit, including an expert knowledge base, a deep learning engine, and a Bayesian network module, used to combine the expert knowledge base and the deep learning engine for dynamic assessment and accurate identification of fault probabilities; A33: Early warning execution unit, including intelligent air valve controller, audible and visual alarm matrix and emergency communication module, is used for fault emergency response and dual-link communication guarantee through intelligent air valve control and audible and visual alarm.

7. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 6, characterized in that: The fault early warning and diagnosis module includes the following steps when it is working: S31: The wavelet transform processor performs multi-scale decomposition on sensor data, extracts fault features through the db4 wavelet basis, and identifies abnormal frequency components in non-stationary signals by combining the time-frequency analysis module; S32: The feature fusion chip uses a weighted fusion algorithm to intelligently weight the three parameters of temperature, vibration, and acoustics. The weighting coefficients are determined through optimization using a genetic algorithm. S33: The diagnostic reasoning unit combines an expert knowledge base with a deep learning engine. It uses a convolutional neural network to classify fused features, and a Bayesian network module dynamically evaluates the probability of failure. When the probability exceeds 80%, an early warning is triggered. S34: The intelligent damper controller in the early warning execution unit automatically switches to the backup channel when the main channel airflow is detected to be below 65000 m³ / h. 3 At a rate of / h, the electric air valve completes the switching within 0.5 seconds; S35: The sound and light alarm matrix activates a three-level alarm system. The first-level yellow warning is indicated by flashing LED lights, the second-level orange warning is indicated by a buzzer, and the third-level red warning is indicated by sending 5G+satellite dual-link alarm information to the operation and maintenance center through the emergency communication module, with a response time of no more than 2 seconds.

8. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 1, characterized in that: The maintenance decision support module includes: A41: Status assessment unit, including a health assessment model, a remaining life predictor, and a maintenance cost analyzer, is used to quantitatively assess the system status by integrating multi-parameter health scores and remaining life predictions; A42: Decision optimization unit, including a genetic algorithm optimizer, a multi-objective optimization module, and an expert decision support system, used to automatically generate maintenance plans and balance cost and reliability through genetic algorithms and multi-objective optimization; A43: Knowledge Management Unit, including a fault case database, a maintenance knowledge graph, and a training simulation system, is used to build the fault case database and knowledge graph for intelligent recommendation of maintenance solutions and virtual training.

9. The gas turbine casing full life cycle performance prediction system based on digital twin as described in claim 1, characterized in that: The acoustic performance optimization module includes: A51: Noise source identification unit, including a microphone array, acoustic imager, and spectrum analysis module, used to generate noise distribution heatmaps and perform spectrum analysis through the microphone array and acoustic imaging; A52: Active control unit, including adaptive filter, active silencer and acoustic feedback controller, used for dynamic suppression of low-frequency noise using adaptive filtering and active silencing technology; A53: Material optimization unit, including acoustic metamaterials, damping coatings and sound insulation structures, used for customized sound absorption bands and vibration suppression through acoustic metamaterials and damping coatings.

10. The gas turbine casing full life cycle performance prediction system based on digital twin according to claim 1, characterized in that: The temperature control optimization module includes: A61: Temperature monitoring unit, including an infrared thermal imager, a thermocouple array, and a temperature gradient sensor, used to integrate infrared thermal imaging and thermocouple array for non-contact global temperature measurement and gradient capture; A62: Control and execution unit, including intelligent air valve, variable frequency fan and cooling water circulation system, used for precise flow regulation and dynamic thermal management through intelligent air valve and variable frequency fan; A63: Model prediction unit, including digital twin thermal model, PID optimization controller and thermal stress analysis module, used for real-time prediction and adaptive control of temperature field based on digital twin thermal model and PID optimization.

Images