A method, system, and medium for gas turbine rotor blade temperature field prediction

Abstract

The application relates to a gas turbine rotor blade temperature field prediction method, system and medium, wherein the method comprises the following steps: constructing a high-fidelity finite element discrete model of a gas turbine rotor blade, obtaining a snapshot matrix of a blade temperature field through high-fidelity finite element thermodynamic simulation calculation based on the high-fidelity finite element discrete model, establishing a reduced-order basis according to a reduced-order method, constructing a temperature field based on the reduced-order basis and a constitutive equation, constructing an initial temperature field prediction model based on the temperature field according to a preset depth operator network framework, iteratively training a neural network in the initial temperature field prediction model through a training set, and obtaining a target temperature field prediction model, and predicting a temperature field of a current gas turbine blade based on the target temperature field prediction model according to a boundary condition of the current gas turbine rotor blade to obtain a real-time prediction result of the blade temperature field, thereby solving the problem of low temperature field prediction efficiency in a large number of working conditions in the related art.

Description

Technical Field

[0001] This application relates to the technical field of runoff data reconstruction, and in particular to a method, system and medium for predicting the temperature field of gas turbine rotor blades. Background Technology

[0002] Gas turbines are characterized by their compact structure, high output power, fast start-up speed, and low carbon emissions. They are widely used in fields such as power, shipbuilding, and aerospace. Gas turbine rotors need to operate for a long time in a high-speed, high-temperature, and high-pressure environment, which requires stringent design and manufacturing processes. They reflect a country's comprehensive industrial strength.

[0003] Currently, the operating environment of rotor blades in mainstream in-service gas turbines exceeds 1200℃. Therefore, the blade surface cannot be directly exposed to the impact of high-temperature combustion gases. This necessitates the design of highly complex internal cooling gas channels, maintaining the entire system in a delicate gas-solid coupling equilibrium. Precise control of the cooling gas flow and the intake velocity of the high-temperature combustion gases becomes crucial for the real-time operation control of gas turbine rotor blades. However, existing mature finite element method calculations cannot quickly calculate physical models with complex structures and boundary conditions, making it difficult to predict the temperature field under numerous operating conditions.

[0004] Currently, for related technologies, gas turbines suffer from low temperature field prediction efficiency due to the long calculation time of numerical models for rotor blades and the large number of operating conditions. No effective solution has yet been proposed. Summary of the Invention

[0005] This application provides a method, system, and medium for predicting the temperature field of gas turbine rotor blades, which at least solves the problem of low temperature field prediction efficiency in related technologies under a large number of operating conditions, and no effective solution has yet been proposed.

[0006] In a first aspect, embodiments of this application provide a method for predicting the temperature field of a gas turbine rotor blade, the method comprising:

[0007] A high-fidelity finite element discrete model of a gas turbine rotor blade is constructed. Based on the high-fidelity finite element discrete model, a snapshot matrix of the blade temperature field is obtained through high-fidelity finite element thermodynamic simulation.

[0008] Based on the snapshot matrix, a reduced-order basis is established according to the order reduction method. Based on the reduced-order basis and the constitutive equation, a temperature field is constructed. Based on the temperature field, an initial temperature field prediction model is constructed according to a preset deep operator network framework.

[0009] The neural network in the initial temperature field prediction model is iteratively trained using the training set until the preset convergence condition is met, thereby obtaining the target temperature field prediction model.

[0010] Based on the current boundary conditions of the gas turbine rotor blades, and using the target temperature field prediction model, the temperature field of the current gas turbine blades is predicted, and the real-time prediction results of the blade temperature field are obtained.

[0011] In one embodiment, the construction of a high-fidelity finite element discrete model of a gas turbine rotor blade includes:

[0012] A high-fidelity finite element discrete model of a gas turbine rotor blade was constructed using the 3D modeling software UG and Matlab, and a txt file was obtained by analyzing the discrete nodes.

[0013] The high-fidelity finite element discrete model is a discretized modeling model. The quality of the finite element mesh is maintained above a preset value. The discretized modeling adopts a two-dimensional hexahedral structure, and each mesh contains 15-25 nodes.

[0014] In one embodiment, obtaining the snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation based on the high-fidelity finite element discrete model includes:

[0015] Based on the high-fidelity finite element discrete model, the temperature field distribution data of the blade under different operating conditions is obtained through high-fidelity finite element thermodynamic simulation calculation.

[0016] The temperature field distribution data are aggregated into a matrix to generate a snapshot matrix of the blade temperature field, wherein the number of the snapshot matrices exceeds a preset number.

[0017] In one embodiment, before iteratively training the neural network in the initial temperature field prediction model using a training set, the method further includes:

[0018] Based on the temperature field prediction control equation, the training data of the initial temperature field prediction model is obtained.

[0019] In one embodiment, obtaining training data for the initial temperature field prediction model based on the predictive control equation of the temperature field includes:

[0020] Based on the snapshot matrix of the temperature field, the temperature field distribution data of the blade under different operating conditions is obtained. The temperature field distribution data includes the time-temperature variation curve, the time-pressure variation curve, and the time-speed variation curve.

[0021] Training data for the initial temperature field prediction model is obtained from the temperature field distribution data through the predictive control equation. The training data includes a training set and a validation set, with the training set accounting for 3 / 5 of the training data and the validation set accounting for 2 / 5 of the training data.

[0022] In one embodiment, the step of iteratively training the neural network in the initial temperature field prediction model using a training set until a preset convergence condition is met to obtain the target temperature field prediction model includes:

[0023] The input and output data are normalized, and the training set data is transformed using the map minmax function;

[0024] Define the structure of a neural network, which includes the number of nodes in the input layer, hidden layer, and output layer;

[0025] Set the parameters of the neural network, which include the transfer function, training function, and learning function;

[0026] The neural network in the initial temperature field prediction model is iteratively trained using the training set until the preset convergence condition is met.

[0027] After training is completed, the initial temperature field prediction model is validated using a validation set to check its performance on unseen data. If the relative error of the validation result is within a preset range, the model is considered effective. If the relative error of the validation result is outside the preset range, the algorithm needs to be adjusted and training continued until the convergence condition is met to obtain the target temperature field prediction model.

[0028] In one embodiment, the number of input layer nodes is 2-5, the number of hidden layer nodes is 3-6, and the number of output layer nodes is 2-5.

[0029] The transfer function from the input layer to the hidden layer is logsig, the transfer function from the hidden layer to the output layer is tansig, the training function is trainlm, and the learning function is learngdm;

[0030] The iterative training includes 400-600 iterations and a training error target of 1×10⁻⁶. -5 The learning rate is 0.01-0.03, and the minimum number of failed confirmations is 30-60.

[0031] In one embodiment, the step of predicting the temperature field of the current gas turbine blade based on the target temperature field prediction model, according to the current boundary conditions of the gas turbine rotor blade, and obtaining the real-time prediction result of the temperature field, includes:

[0032] The boundary conditions of the gas turbine rotor blades are collected from field sensors. These boundary conditions include the main gas inlet and outlet pressures, temperatures, and rotor speed. The boundary conditions are used as input data for the target temperature field prediction model.

[0033] The boundary conditions are preprocessed, including data normalization.

[0034] The preprocessed boundary conditions are input into the target temperature field prediction model, which outputs the real-time prediction results of the temperature field. The real-time prediction results of the temperature field characterize the temperature distribution at different locations of the blade under the current operating conditions.

[0035] Secondly, embodiments of this application provide a gas turbine rotor blade temperature field prediction system, the system comprising a snapshot matrix module, an initial temperature field prediction model construction module, a target temperature field prediction model construction module, and a real-time prediction result module, wherein:

[0036] The snapshot matrix module is used to obtain a snapshot matrix of the blade temperature field based on a pre-defined high-fidelity finite element discrete model of the gas turbine rotor blade through high-fidelity finite element thermodynamic simulation calculations. The snapshot matrix is ​​used to construct the temperature field. The high-fidelity finite element discrete model is a discretized model, and the quality of the finite element mesh is maintained above a preset value. The discretized model adopts a three-dimensional hexahedral structure, and the preset value is 0.2-0.4. Each mesh contains 15-25 nodes. The snapshot matrix reflects the time-temperature, time-speed, and time-pressure curves at different positions of the blade.

[0037] The module for constructing the initial temperature field prediction model is used to establish a reduced-order basis based on the snapshot matrix according to the order reduction method, construct a temperature field based on the reduced-order basis and the constitutive equation, and construct an initial temperature field prediction model based on the temperature field according to the preset deep operator network framework. The constitutive equation is used to describe the physical properties and behavior of the material.

[0038] The module for constructing the target temperature field prediction model is used to iteratively train the neural network in the initial temperature field prediction model using a training set until a preset convergence condition is met, thereby obtaining the target temperature field prediction model.

[0039] According to the predictive control equation of the temperature field, the training data for the initial temperature field prediction model is obtained, including: obtaining temperature field distribution data of the blade under different operating conditions based on the snapshot matrix of the temperature field, wherein the temperature field distribution data includes time-temperature variation curves, time-pressure variation curves, and time-speed variation curves; and obtaining training data for the initial temperature field prediction model from the temperature field distribution data through the predictive control equation; the number of snapshot matrices exceeds a preset number, which is 500-1500.

[0040] The real-time prediction result module is used to predict the temperature field of the current gas turbine rotor blades based on the target temperature field prediction model, according to the current boundary conditions of the gas turbine rotor blades, wherein the boundary conditions include the main gas inlet and outlet pressures, temperatures, and rotor speed, and to obtain the real-time prediction result of the blade temperature field.

[0041] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a gas turbine rotor blade temperature field prediction method as described in the first aspect above.

[0042] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for predicting the temperature field of a gas turbine rotor blade as described in the first aspect above.

[0043] The gas turbine rotor blade temperature field prediction method, system, and medium provided in this application embodiment have at least the following technical effects.

[0044] A high-fidelity finite element discrete model of a gas turbine rotor blade is constructed. Based on this model, a snapshot matrix of the blade temperature field is obtained through high-fidelity finite element thermodynamic simulation. Using this snapshot matrix, a reduced-order basis is established using a reduction-order method. Based on this reduced-order basis and constitutive equations, a temperature field is constructed. Based on this temperature field, an initial temperature field prediction model is built using a pre-defined deep operator network framework. The neural network in the initial temperature field prediction model is iteratively trained using a training set until a pre-defined convergence condition is met, yielding a target temperature field prediction model. Based on the current boundary conditions of the gas turbine rotor blade and the target temperature field prediction model, the temperature field of the current gas turbine blade is predicted, obtaining real-time prediction results. This achieves real-time prediction of the blade temperature field, ensuring improved gas turbine control efficiency and unit operation safety, and enhancing the prediction efficiency and accuracy of the blade temperature field. It also addresses the problem of low temperature field prediction efficiency in related technologies under numerous operating conditions.

[0045] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0047] Figure 1This is a flowchart of a method for predicting the temperature field of a gas turbine rotor blade according to an embodiment of this application;

[0048] Figure 2 This is a flowchart illustrating step S101 according to an exemplary embodiment;

[0049] Figure 3 This is a flowchart illustrating step S103 according to an exemplary embodiment;

[0050] Figure 4 This is a schematic diagram illustrating a neural network architecture according to an exemplary embodiment;

[0051] Figure 5 This is a schematic diagram of the temperature field error curve of a verification dataset according to an exemplary embodiment;

[0052] Figure 6 A block diagram of a system for predicting the temperature field of a gas turbine rotor blade is shown according to an exemplary embodiment;

[0053] Figure 7 A block diagram of an electronic device illustrated according to an exemplary embodiment. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0055] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0056] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0057] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0058] High-fidelity numerical models, such as the finite element method and the finite volume method, characterize physical laws based on physical governing equations and through a combination of numerical techniques. They provide relatively accurate representations of the thermodynamic response and evolution of high-temperature components. However, the computational cost of high-fidelity numerical models often depends on the type of problem being analyzed, the discrete degrees of freedom, constraints, and boundary conditions. These numerical models are typically inefficient when simulating the thermodynamic coupling behavior of complex high-temperature components. Therefore, using traditional high-fidelity numerical methods for thermodynamic state prediction is not an effective technique, as its computational complexity cannot meet the requirements for online monitoring and fault diagnosis of high-temperature components during operation.

[0059] Currently, an effective technique to address the inefficiency of numerical computation is model reduction. Generally speaking, model reduction is a relative term; a reduced-order model (ROM) is a lower-order approximation model built upon a high-fidelity full-order model (FOM). Unlike the black-box model of data-driven modeling, reduced-order models, like full-order models, are based on physical governing equations. They capture the dominant patterns and physical essence of physical states through physical feature extraction, thereby constructing a lower-order model to approximate the mechanical behavior of the original full-order model.

[0060] Therefore, currently, for related technologies, gas turbines suffer from low temperature field prediction efficiency due to the long calculation time of rotor blade numerical models and the problem of low efficiency under many operating conditions, and no effective solution has yet been proposed.

[0061] In a first aspect, embodiments of this application provide a method for predicting the temperature field of gas turbine rotor blades. Figure 1 This is a flowchart for predicting the temperature field of a gas turbine rotor blade, such as... Figure 1 As shown, a method for predicting the temperature field of a gas turbine rotor blade includes:

[0062] Step S101: Based on the pre-built high-fidelity finite element discrete model of the gas turbine rotor blade, obtain the snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation calculation.

[0063] Step S102: Based on the snapshot matrix, establish a reduced-order basis according to the order reduction method. Based on the reduced-order basis and constitutive equation, construct the temperature field. Based on the temperature field, construct the initial temperature field prediction model according to the preset deep operator network framework.

[0064] Step S103: Iteratively train the neural network in the initial temperature field prediction model using the training set until the preset convergence condition is met, and obtain the target temperature field prediction model.

[0065] Step S104: Based on the current boundary conditions of the gas turbine rotor blades and the target temperature field prediction model, predict the temperature field of the current gas turbine blades and obtain the real-time prediction results of the blade temperature field.

[0066] In summary, the proposed method provides a method for predicting the temperature field of a gas turbine rotor blade. It involves constructing a high-fidelity finite element discrete model of the gas turbine rotor blade, and obtaining a snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation based on this model. Based on the snapshot matrix, a reduced-order basis is established using a reduction-order method. A temperature field is constructed based on the reduced-order basis and constitutive equations. An initial temperature field prediction model is then constructed based on this temperature field and a pre-defined deep operator network framework. The neural network in the initial temperature field prediction model is iteratively trained using a training set until a pre-defined convergence condition is met, resulting in a target temperature field prediction model. Based on the current boundary conditions of the gas turbine rotor blade and the target temperature field prediction model, the temperature field of the current gas turbine blade is predicted, obtaining real-time prediction results. This method achieves real-time prediction of the blade temperature field, ensuring improved gas turbine control efficiency and unit operation safety, and enhancing the prediction efficiency and accuracy of the blade temperature field. It also addresses the problem of low temperature field prediction efficiency in related technologies under numerous operating conditions.

[0067] In one embodiment, before obtaining the snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation based on a pre-defined high-fidelity finite element discrete model of the gas turbine rotor blade, the method further includes:

[0068] A high-fidelity finite element discrete model of a gas turbine rotor blade was constructed using the 3D modeling software UG and Matlab, and a txt file was obtained by analyzing the discrete nodes.

[0069] The high-fidelity finite element discrete model is a discretized modeling method. The quality of the finite element mesh is maintained above the preset value. The discretized modeling adopts a two-dimensional hexahedral structure, and each mesh contains 15-25 nodes.

[0070] Optionally, a high-fidelity finite element discretization model is used, which ensures that the quality of the finite element mesh is maintained above a preset value, such as 0.2-0.4; in this embodiment, 0.3 is used. The discretized model adopts a two-dimensional hexahedral structure, meaning each mesh contains 15-25 nodes; in this embodiment, 20 nodes are used. A high-precision model facilitates subsequent accurate calculations and the convergence of mechanical curves. Specifically, discretization modeling is a method of transforming a continuous physical system into a set of discrete elements to facilitate numerical simulation and analysis. In finite element analysis, discretization means dividing the geometry of the blade into multiple small elements (i.e., finite elements), each represented by a set of discrete nodes. To ensure the accuracy and reliability of the calculation results, the quality of the finite element mesh needs to be maintained above a certain level. The mesh quality factors (such as mesh distortion, element shape, etc.) need to meet certain standards. In this case, a two-dimensional hexahedral structure is chosen, with each mesh containing approximately 20 nodes. This structure is chosen to ensure high accuracy of the model in subsequent calculations and to facilitate the convergence of mechanical curves.

[0071] Constructing a high-fidelity finite element discrete model is to ensure that the temperature field of the gas turbine rotor blades under different operating conditions can be accurately simulated, thereby providing reliable data support for subsequent temperature field prediction models and improving the accuracy of blade temperature field prediction.

[0072] Figure 2 This is a flowchart illustrating step S101 according to an exemplary embodiment, as follows: Figure 2 As shown, step S101 involves obtaining a snapshot matrix of the blade temperature field through high-fidelity finite element discrete model of the gas turbine rotor blades using high-fidelity finite element thermodynamic simulation. This specifically includes the following steps:

[0073] Step S1011: Based on the high-fidelity finite element discrete model, obtain the temperature field distribution data of the blade under different operating conditions through high-fidelity finite element thermodynamic simulation calculation.

[0074] Step S1012: Summarize the temperature field distribution data into a matrix to generate a snapshot matrix of the blade temperature field, wherein the number of snapshot matrices exceeds the preset number.

[0075] Optionally, a pre-constructed high-fidelity finite element discrete model is used for thermodynamic simulation calculations. The temperature distribution of the blades under various operating conditions is obtained. This process generates a large amount of temperature field data, i.e., snapshots. Snapshots include time-temperature, time-rotation speed, and time-pressure variation curves at different locations on the blades. After completing the thermodynamic simulation calculations, all obtained temperature field distribution data are summarized into a matrix, which is the snapshot matrix. The number of snapshot matrices exceeds a preset number, which can be 500-1500. In this embodiment, 1000 snapshots are used, which helps to more accurately reconstruct the temperature field when establishing a reduced-order basis, reducing the difference between the results and the finite element calculations. The snapshot matrix is ​​an important data source for subsequent temperature field reconstruction and prediction model training.

[0076] In one embodiment, step S102 involves establishing a reduced-order basis based on the snapshot matrix using a reduction-order method, constructing a temperature field based on the reduced-order basis and constitutive equations, and building an initial temperature field prediction model based on the temperature field using a preset deep operator network framework. Specifically, this includes:

[0077] The main features are extracted from the snapshot matrix obtained from the high-fidelity finite element discrete model, and a reduced basis is established using the Orthogonal Eigenfactor Decomposition (POD) method. POD is a data-driven reduction technique that reduces computational complexity by mapping high-dimensional data to a low-dimensional subspace. The core idea of ​​this method is to extract a set of orthogonal basis vectors from the snapshot matrix. These basis vectors can represent the main features of the original data with fewer dimensions. The basis vectors extracted by the POD reduction method are called the reduced basis, and these basis vectors can capture the main trends of temperature field changes. By using the reduced basis, high-dimensional temperature field data can be compressed into a lower-dimensional space, thereby simplifying subsequent computational tasks and improving efficiency. After obtaining the reduced basis, the temperature field can be quickly reconstructed by combining it with the constitutive equations of the blade material (equations describing the physical properties and behavior of the material), as it only requires a small number of reduced basis vectors to approximate the entire temperature field. The temperature field reconstructed based on the reduced basis is used as input to a neural network, and a pre-defined deep operator network framework is used to construct the temperature field of the blade during service online. The deep operator network framework refers to the structure and training method of a neural network, which includes parameters such as the number of layers, the number of nodes in each layer, the activation function, and the loss function.

[0078] Step S102 reduces the dimensionality of the temperature field data through a reduction-order method, thereby significantly reducing the computational resource requirements and improving computational efficiency. The main features extracted by the POD reduction method can better reflect the essential changing trends of the temperature field, making the model built based on these features have stronger generalization ability and better able to handle temperature field predictions under different operating conditions. By combining the reduced-order basis and constitutive equations to construct a temperature field model, prediction speed can be improved while ensuring model accuracy, enabling the model to provide prediction results more quickly in practical applications. This greatly improves computational efficiency and prediction speed while maintaining model accuracy.

[0079] In one embodiment, before iteratively training the neural network in the initial temperature field prediction model using a training set, the method further includes: obtaining training data for the initial temperature field prediction model based on the temperature field prediction control equation. Specifically, this includes:

[0080] Based on the snapshot matrix of the temperature field, the temperature field distribution data of the blade under different operating conditions is obtained. The temperature field distribution data includes the time-temperature variation curve, the time-pressure variation curve, and the time-speed variation curve.

[0081] Training data for the initial temperature field prediction model is obtained from the temperature field distribution data through the predictive control equation. The training data includes a training set and a validation set, with the training set accounting for 3 / 5 of the training data and the validation set accounting for 2 / 5 of the training data.

[0082] Optionally, temperature field snapshot matrices of the blade under different operating conditions are obtained through thermodynamic simulation calculations using a high-fidelity finite element discrete model. These snapshot matrices contain the temperature distribution of the blade under different operating conditions, including curves showing the changes in time versus temperature, time versus pressure, and time versus rotational speed. Temperature field distribution data of the blade under different operating conditions is extracted from these snapshot matrices. Based on the extracted temperature field distribution data, training data required for the initial temperature field prediction model is generated using predictive control equations. Predictive control equations are a method of transforming actual physical phenomena into a data form that can be learned by the model. The generated training data is divided into a training set and a validation set in a 3:2 ratio. The training set is used to train the model, and the validation set is used to verify the model's generalization ability. For example, if there are 5 temperature curves, 3 temperature curves are used as the training set, and the remaining 2 curves are used as the validation set, with one curve being an interpolation curve and the other an extrapolation curve in the validation dataset.

[0083] By extracting temperature field distribution data from the snapshot matrix and transforming it into the format required for model training, the data generated by finite element simulation calculations is utilized to the maximum extent, ensuring the richness and representativeness of the training data. Training data is generated from the temperature field distribution data using predictive control equations, enabling the model to learn the patterns of temperature field changes with time, pressure, rotational speed, and other factors, thereby improving prediction accuracy.

[0084] Figure 3 This is a flowchart illustrating step S103 according to an exemplary embodiment, as follows: Figure 3 As shown, step S103 involves iteratively training the neural network in the initial temperature field prediction model using the training set until a preset convergence condition is met, thereby obtaining the target temperature field prediction model. Specifically, this includes the following steps:

[0085] Step S1031: Normalize the input and output data by transforming the training set data using the map minmax function.

[0086] Step S1032: Define the structure of the neural network. The structure of the neural network includes the number of nodes in the input layer, hidden layer, and output layer.

[0087] Step S1033: Set the parameters of the neural network. The parameters of the neural network include the transfer function, training function, and learning function.

[0088] Step S1034: Iteratively train the neural network in the initial temperature field prediction model using the training set until the preset convergence condition is met.

[0089] Step S1035: After training is completed, the initial temperature field prediction model is validated using a validation set to check its performance on unseen data. If the relative error of the validation result is within a preset range, the model is considered effective. If the relative error of the validation result is outside the preset range, the algorithm needs to be adjusted and training continued until the convergence condition is met to obtain the target temperature field prediction model.

[0090] Alternatively, the training set can be transformed using the map minmax function, scaling the input and output data to a fixed range. Normalization can prevent large values ​​in the data from adversely affecting neural network training, and it can also accelerate convergence. Figure 4 This is a schematic diagram illustrating a neural network architecture according to an exemplary embodiment, such as... Figure 4As shown, the structure of the neural network is defined, including the number of nodes in the input layer, hidden layer, and output layer. The input layer has 2-5 nodes (3 in this example), the hidden layer has 3-6 nodes (4 in this example), and the output layer has 2-5 nodes (3 in this example). The parameters of the neural network are set, including the transfer function, training function, and learning function. The transfer function from the input layer to the hidden layer is logsig, which introduces non-linearity to the model, enabling it to learn more complex patterns. The transfer function from the hidden layer to the output layer is tansig. The training function is trainlm, suitable for small to medium-sized datasets, and converges to the optimal solution relatively quickly. The learning function is learngdm, which helps overcome local minima and improves training stability. Iterative training refers to the process of gradually reducing the prediction error by repeatedly adjusting the weights. Each iteration updates the weights based on the gradient of the loss function until the preset convergence condition is met. Iterative training includes 400-600 iterations (500 in this example), with a training error target of 1×10⁻⁶. -5 The learning rate is 0.01-0.03, and this embodiment uses 0.01. The minimum number of failed confirmations is 30-60, and this embodiment uses 50.

[0091] Figure 5 This is a schematic diagram illustrating the temperature field error curve of a verification dataset according to an exemplary embodiment, such as... Figure 5 As shown, after training, the initial temperature field prediction model is validated using a validation set to check its performance on unseen data. If the relative error of the validation results is within a preset range, the model is considered effective; if the relative error is outside the preset range, the algorithm needs to be adjusted, and training continues until the convergence condition is met to obtain the target temperature field prediction model. It is worth noting that the in-sample training dataset and the out-of-sample validation training dataset are compared with temperature fields and field sensors built using mature industry software such as ANSYS and ABAQUS based on finite element methods. If the relative error of the validation results is within the allowable range, the method is effective; otherwise, the algorithm needs to be readjusted. Figure 5 During the training, when the model was trained under interpolation conditions, the relative error of the samples was within 4%. However, as the training time progressed and extrapolation conditions were used for training, the relative error of the samples increased to approximately 10%. Overall, the error level of the trained model remained within an acceptable range, demonstrating good adaptability and robustness.

[0092] Step S103 utilizes the advantage of being grid-independent, similar to classic gridless methods. Combining this with intrinsic orthogonal decomposition (IOD) constructs a database to efficiently predict and reconstruct the temperature field of gas turbine rotor blades. The constructed neural network framework is non-invasive, thus the neural grid exhibits strong robustness and adaptability. Furthermore, the research object can be transformed into other objects with complex geometries and boundary conditions, such as gas turbine rotors, outer cylinders, and inner cylinders. Subsequently, a thermo-coupled neural network prediction model can be developed based on this architecture. Through reasonable data normalization, network structure design, parameter settings, and training processes, the model can more accurately learn the changing patterns of the temperature field, thereby improving prediction accuracy. By using training and validation sets, the model not only performs well on training data but also maintains stable prediction performance on unseen data, enhancing its robustness.

[0093] In one embodiment, step S104 involves predicting the temperature field of the current gas turbine blades based on the target temperature field prediction model, according to the current boundary conditions of the gas turbine rotor blades, and obtaining the real-time prediction result of the blade temperature field. Specifically, this includes:

[0094] The boundary conditions of the gas turbine rotor blades are collected from field sensors. These boundary conditions include the main gas inlet and outlet pressures, temperatures, and rotor speeds. The boundary conditions are then used as input data for the target temperature field prediction model.

[0095] The boundary conditions are preprocessed, including data normalization.

[0096] The preprocessed boundary conditions are input into the target temperature field prediction model, which outputs the real-time prediction results of the temperature field. The real-time prediction results of the temperature field characterize the temperature distribution at different locations on the blade under the current operating conditions.

[0097] Optionally, real-time operating data of the gas turbine rotor blades is collected from field sensors. This data includes, but is not limited to, the pressure and temperature at the main gas inlet and outlet, and the rotor speed. This data serves as input to the temperature field prediction model, reflecting the operating status under current conditions. The collected boundary condition data undergoes preprocessing, primarily data normalization. Normalization ensures that the format of the input data is consistent with the data used during model training, avoiding inaccurate predictions due to differences in data magnitude. Normalization can use the same normalization method as the training data, such as the map minmax function. The normalized boundary condition data is then input into the target temperature field prediction model. The model processes the input data based on patterns learned during training and outputs the blade temperature field distribution under the current operating conditions.

[0098] By inputting real-time data collected by field sensors into the model, the temperature field distribution of the blades under the current operating conditions can be obtained. The model is constructed using techniques such as order reduction and neural networks, giving it strong robustness and adaptability, enabling it to make good predictions even when operating conditions change.

[0099] The principle of this application is as follows: First, the time-temperature, time-speed, and time-pressure change curves at different positions of the blade during the start-up process calculated by the finite element model are processed into training samples for a neural network. The number of neural network nodes and training parameters are set, and a neural network prediction model is established. By dividing the training set and validation set of the training samples, the neural network is trained and then validated with the validation set to obtain a high-precision prediction model of the time-temperature change curves at different positions of the rotor blade during the gas turbine start-up process. This solves the problem that the long calculation time of the blade numerical model makes it impossible to predict a large number of operating conditions. This provides a basis for real-time monitoring of the blade operating temperature field on site and provides a guarantee for improving the gas turbine control efficiency and unit operation safety.

[0100] In summary, the proposed method provides a method for predicting the temperature field of a gas turbine rotor blade. It involves constructing a high-fidelity finite element discrete model of the gas turbine rotor blade, and obtaining a snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation based on this model. Based on the snapshot matrix, a reduced-order basis is established using a reduction-order method. A temperature field is constructed based on the reduced-order basis and constitutive equations. An initial temperature field prediction model is then constructed based on this temperature field and a pre-defined deep operator network framework. The neural network in the initial temperature field prediction model is iteratively trained using a training set until a pre-defined convergence condition is met, resulting in a target temperature field prediction model. Based on the current boundary conditions of the gas turbine rotor blade and the target temperature field prediction model, the temperature field of the current gas turbine blade is predicted, obtaining real-time prediction results. This method achieves real-time prediction of the blade temperature field, ensuring improved gas turbine control efficiency and unit operation safety, and enhancing the prediction efficiency and accuracy of the blade temperature field. It also addresses the problem of low temperature field prediction efficiency in related technologies under numerous operating conditions.

[0101] Secondly, embodiments of this application provide a system for predicting the temperature field of gas turbine rotor blades. Figure 6 This is a block diagram illustrating a system for predicting the temperature field of a gas turbine rotor blade according to an exemplary embodiment. Figure 6 As shown, the system includes a snapshot matrix module 610, an initial temperature field prediction model construction module 620, a target temperature field prediction model construction module 630, and a real-time prediction result module 640, wherein:

[0102] The snapshot matrix module 610 is used to obtain the snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation calculation based on a pre-defined high-fidelity finite element discrete model of the gas turbine rotor blade.

[0103] The module 620 for constructing an initial temperature field prediction model is used to establish a reduced basis based on the snapshot matrix and the order reduction method, construct a temperature field based on the reduced basis and the constitutive equation, and construct an initial temperature field prediction model based on the temperature field and a preset deep operator network framework.

[0104] The target temperature field prediction model module 630 is used to iteratively train the neural network in the initial temperature field prediction model through the training set until the preset convergence condition is met, and obtain the target temperature field prediction model.

[0105] The real-time prediction result module 640 is used to predict the temperature field of the current gas turbine blade based on the boundary conditions of the current gas turbine rotor blade and the target temperature field prediction model, and obtain the real-time prediction result of the blade temperature field.

[0106] In summary, the gas turbine rotor blade temperature field prediction system provided in this application addresses the problem of low temperature field prediction efficiency under numerous operating conditions in related technologies through a snapshot matrix module 610, an initial temperature field prediction model construction module 620, a target temperature field prediction model construction module 630, and a real-time prediction result module 640. Specifically, a high-fidelity finite element discrete model of the gas turbine rotor blade is constructed. Based on this model, a snapshot matrix of the blade temperature field is obtained through high-fidelity finite element thermodynamic simulation. Based on the snapshot matrix, a reduced-order basis is established using a reduction-order method. Based on the reduced-order basis and constitutive equations, a temperature field is constructed. Based on this temperature field, an initial temperature field prediction model is constructed according to a preset deep operator network framework. The neural network in the initial temperature field prediction model is iteratively trained using a training set until a preset convergence condition is met, obtaining the target temperature field prediction model. Based on the current boundary conditions of the gas turbine rotor blade and the target temperature field prediction model, the temperature field of the current gas turbine blade is predicted, obtaining the real-time prediction result of the blade temperature field. This technology enables real-time prediction of blade temperature fields, ensuring improved gas turbine control efficiency and unit operation safety, and enhancing the prediction efficiency and accuracy of blade temperature fields. It also addresses the problem of low temperature field prediction efficiency in related technologies across a wide range of operating conditions.

[0107] It should be noted that the gas turbine rotor blade temperature field prediction system provided in this embodiment is used to implement the above-described embodiments, and details already described will not be repeated. As used above, terms such as "module," "unit," and "subunit" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the above embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0108] Thirdly, embodiments of this application provide an electronic device, Figure 7 This is a block diagram illustrating an electronic device according to an exemplary embodiment. (e.g.) Figure 7 As shown, the electronic device may include a processor 71 and a memory 72 storing computer program instructions.

[0109] Specifically, the processor 71 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0110] The memory 72 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 72 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 72 may include removable or non-removable (or fixed) media. Where appropriate, the memory 72 may be internal or external to a data processing device. In a particular embodiment, the memory 72 is non-volatile memory. In a particular embodiment, the memory 72 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0111] The memory 72 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 71.

[0112] The processor 71 reads and executes computer program instructions stored in the memory 72 to implement any of the gas turbine rotor blade temperature field prediction methods in the above embodiments.

[0113] In one embodiment, a device for predicting the temperature field of a gas turbine rotor blade may further include a communication interface 73 and a bus 70. Wherein, as Figure 7 As shown, the processor 71, memory 72, and communication interface 73 are connected through bus 70 and complete communication with each other.

[0114] The communication interface 73 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of this application. The communication port 73 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0115] Bus 70 includes hardware, software, or both, that couples together components of a device for predicting the temperature field of a gas turbine rotor blade. Bus 70 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 70 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 70 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0116] Fourthly, embodiments of this application provide a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements a method for predicting the temperature field of a gas turbine rotor blade provided in the first aspect.

[0117] The readable storage medium may be more specifically adopted, including but not limited to: portable disk, hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0118] In a possible implementation, the present invention can also be implemented as a program product comprising program code, which, when the program product is run on a terminal device, causes the terminal device to perform the steps of implementing the gas turbine rotor blade temperature field prediction method provided in the first aspect.

[0119] The program code for executing the present invention can be written in any combination of one or more programming languages. The program code can be executed entirely on the user device, partially on the user device, as a standalone software package, partially on the user device and partially on a remote device, or entirely on a remote device.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0121] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method of gas turbine rotor blade temperature field prediction, characterized by, The method includes: Based on a pre-defined high-fidelity finite element discrete model of a gas turbine rotor blade, a snapshot matrix of the blade temperature field is obtained through high-fidelity finite element thermodynamic simulation. This snapshot matrix is ​​used to construct the temperature field. The high-fidelity finite element discrete model is a discretized model, with the quality of the finite element mesh maintained above a preset value. The discretized model uses a three-dimensional hexahedral structure, and the preset value is 0.2-0.

4. Each mesh contains 15-25 nodes. The snapshot matrix reflects the time-temperature, time-speed, and time-pressure curves at different locations on the blade. Based on the snapshot matrix, a reduced-order basis is established according to the order reduction method. Based on the reduced-order basis and the constitutive equation, a temperature field is constructed. Based on the temperature field, an initial temperature field prediction model is constructed according to a preset deep operator network framework. The constitutive equation is used to describe the physical properties and behavior of the material. According to the predictive control equation of the temperature field, the training data of the initial temperature field prediction model is obtained, including: obtaining temperature field distribution data of the blade under different operating conditions based on the snapshot matrix of the temperature field, wherein the temperature field distribution data includes time-temperature variation curves, time-pressure variation curves, and time-speed variation curves; obtaining training data for the initial temperature field prediction model from the temperature field distribution data through the predictive control equation; the number of snapshot matrices exceeds a preset number, which is 500-1500; iteratively training the neural network in the initial temperature field prediction model using the training set until a preset convergence condition is met, thereby obtaining the target temperature field prediction model; Based on the current boundary conditions of the gas turbine rotor blades, including the main gas inlet and outlet pressures, temperatures, and rotor speed, the temperature field of the current gas turbine blades is predicted based on the target temperature field prediction model, and the real-time prediction results of the blade temperature field are obtained.

2. The method according to claim 1, characterized in that, The step of iteratively training the neural network in the initial temperature field prediction model using a training set until a preset convergence condition is met to obtain the target temperature field prediction model includes: The input and output data are normalized, and the training set data is transformed using the map minmax function; Define the structure of a neural network, which includes the number of nodes in the input layer, hidden layer, and output layer; Set the parameters of the neural network, which include the transfer function, training function, and learning function; The neural network in the initial temperature field prediction model is iteratively trained using the training set until the preset convergence condition is met. After training is completed, the initial temperature field prediction model is validated using a validation set to check its performance on unseen data. If the relative error of the validation result is within a preset range, the model is considered effective. If the relative error of the validation result is outside the preset range, the algorithm needs to be adjusted and training continued until the convergence condition is met to obtain the target temperature field prediction model.

3. The method according to claim 2, characterized in that, The number of nodes in the input layer is 2-5, the number of nodes in the hidden layer is 3-6, and the number of nodes in the output layer is 2-5. The transfer function from the input layer to the hidden layer is logsig, the transfer function from the hidden layer to the output layer is tansig, the training function is trainlm, and the learning function is learngdm; The iterative training includes 400-600 times of iteration numbers, 1x10 -5 -3 of learning rate, and 30-60 times of minimum confirmation failure numbers.

4. The method according to claim 1, characterized in that, The snapshot matrix of the blade temperature field, obtained through high-fidelity finite element thermodynamic simulation based on the high-fidelity finite element discrete model, includes: Based on the high-fidelity finite element discrete model, the temperature field distribution data of the blade under different operating conditions is obtained through high-fidelity finite element thermodynamic simulation calculation. The temperature field distribution data are aggregated into a matrix to generate a snapshot matrix of the blade temperature field, wherein the number of the snapshot matrices exceeds a preset number.

5. The method according to claim 1, characterized in that, Before obtaining the snapshot matrix of the blade temperature field through high-fidelity finite element thermodynamic simulation based on a pre-defined high-fidelity finite element discrete model of the gas turbine rotor blades, the method further includes: A high-fidelity finite element discrete model of a gas turbine rotor blade was constructed using the 3D modeling software UG and Matlab, and a txt file was obtained by analyzing the discrete nodes. The high-fidelity finite element discrete model is a discretized modeling model. The quality of the finite element mesh is maintained above a preset value. The discretized modeling adopts a two-dimensional hexahedral structure, and each mesh contains 15-25 nodes.

6. The method according to claim 1, characterized in that, The step of predicting the temperature field of the current gas turbine blades based on the boundary conditions of the current gas turbine rotor blades and the target temperature field prediction model, and obtaining the real-time prediction results of the temperature field, includes: The boundary conditions of the gas turbine rotor blades are collected from field sensors. These boundary conditions include the main gas inlet and outlet pressures, temperatures, and rotor speed. The boundary conditions are used as input data for the target temperature field prediction model. The boundary conditions are preprocessed, including data normalization. The preprocessed boundary conditions are input into the target temperature field prediction model, which outputs the real-time prediction results of the temperature field. The real-time prediction results of the temperature field characterize the temperature distribution at different locations of the blade under the current operating conditions.

7. A gas turbine rotor blade temperature field prediction system, characterized in that, The system includes a snapshot matrix module, an initial temperature field prediction model construction module, a target temperature field prediction model construction module, and a real-time prediction result module, wherein: The snapshot matrix module is used to obtain a snapshot matrix of the blade temperature field based on a pre-defined high-fidelity finite element discrete model of the gas turbine rotor blade through high-fidelity finite element thermodynamic simulation calculations. The snapshot matrix is ​​used to construct the temperature field. The high-fidelity finite element discrete model is a discretized model, and the quality of the finite element mesh is maintained above a preset value. The discretized model adopts a three-dimensional hexahedral structure, and the preset value is 0.2-0.

4. Each mesh contains 15-25 nodes. The snapshot matrix reflects the time-temperature, time-speed, and time-pressure curves at different positions of the blade. The module for constructing the initial temperature field prediction model is used to establish a reduced-order basis based on the snapshot matrix according to the order reduction method, construct a temperature field based on the reduced-order basis and the constitutive equation, and construct an initial temperature field prediction model based on the temperature field according to the preset deep operator network framework. The constitutive equation is used to describe the physical properties and behavior of the material. The module for constructing the target temperature field prediction model is used to iteratively train the neural network in the initial temperature field prediction model using a training set until a preset convergence condition is met, thereby obtaining the target temperature field prediction model. According to the predictive control equation of the temperature field, the training data for the initial temperature field prediction model is obtained, including: obtaining temperature field distribution data of the blade under different operating conditions based on the snapshot matrix of the temperature field, wherein the temperature field distribution data includes time-temperature variation curves, time-pressure variation curves, and time-speed variation curves; and obtaining training data for the initial temperature field prediction model from the temperature field distribution data through the predictive control equation; the number of snapshot matrices exceeds a preset number, which is 500-1500. The real-time prediction result module is used to predict the temperature field of the current gas turbine blade based on the target temperature field prediction model, according to the current boundary conditions of the gas turbine rotor blade, wherein the boundary conditions include the main gas inlet and outlet pressure, temperature and rotor speed, and to obtain the real-time prediction result of the blade temperature field.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a method for predicting the temperature field of a gas turbine rotor blade as described in any one of claims 1 to 6.

Images