Transient optimization method and system for variable load turbine airfoil configuration

Abstract

The application discloses a variable load turbine air film layout transient optimization method and system, the method comprising: constructing a transient temperature database; extracting the low-dimensional potential vector of the transient temperature in the transient temperature database, and constructing a prediction network to establish the mapping of the air film hole layout geometric parameters and the variable load working condition parameters to the low-dimensional potential vector; obtaining a joint prediction model, and extracting a turbine blade temperature response characteristic index based on the joint prediction model; constructing a multi-objective optimization model based on the turbine blade temperature response characteristic index, and obtaining a Pareto front solution set through optimization of the multi-objective optimization model; calculating the transient temperature rise local gradient of each scheme in the Pareto front solution set, and determining the optimal air film cooling layout scheme based on the transient temperature rise local gradient. The scheme of the application realizes air film layout optimization considering excellent cooling performance and high structural reliability, and significantly improves the design efficiency and service life of the turbine blade under variable load working conditions.

Description

Technical Field

[0001] This application generally relates to the fields of gas turbine blade cooling and artificial intelligence technology. More specifically, this application relates to a transient optimization method and system for variable load turbine film cooling layout. Background Technology

[0002] With a high proportion of new energy sources connected to the grid, gas turbines frequently operate under varying load conditions to undertake peak shaving and frequency regulation tasks. Load changes cause complex and unsteady evolution of the parameters of the mainstream gas and coolant inside the turbine, leading to temperature fluctuations on the blade surface and becoming a major source of increased thermal stress on the blades. Currently, film cooling technology is widely used in engineering to protect the blades, but the layout of the film cooling holes directly affects the mixing of coolant and mainstream gas and the transient temperature rise characteristics. Therefore, under frequent varying load conditions, focusing only on steady-state cooling performance is insufficient to fully guarantee the service safety of the blades.

[0003] Currently, film cooling optimization technologies are mainly divided into the following two categories: Firstly, there is optimization based on numerical simulation: this mainly involves constructing a fluid-thermal coupled numerical calculation model for optimization (such as parameterized optimization of the entire blade distribution). Although this method can obtain temperature information across the entire field, the computational cost is extremely high, which cannot meet the needs of numerous iterative calls in transient optimization under varying loads.

[0004] Secondly, optimization based on AI feature dimensionality reduction and rapid prediction: In order to reduce computational costs, AI dimensionality reduction prediction methods based on POD mode decomposition have been developed in recent years. However, existing methods are mostly limited to local prediction of cooling efficiency of single-row, flat-plate film cooling holes, and have not been extended to the real blade level, and cannot reflect the cold air interference and coupling effect between multiple rows of holes in the entire blade.

[0005] In summary, the existing technology has the following obvious shortcomings: First, there is a lack of efficient methods for predicting transient temperatures at the blade level: existing technologies struggle to capture the complex flow-thermal coupling characteristics of the entire blade while reducing computational dimensionality, and cannot balance computational cost and accuracy in predicting transient temperature fields.

[0006] Second, there is a lack of consideration for transient thermal safety under varying loads: existing optimization objectives are generally limited to steady-state conditions and fail to effectively take into account the transient thermal safety requirements during load changes. Air film layouts obtained solely through steady-state optimization are highly susceptible to safety hazards such as excessively high localized transient temperature rises during dynamic load changes.

[0007] In view of this, there is an urgent need to provide a transient optimization scheme for the variable load turbine film cooling layout to solve the above problems, achieve efficient reconstruction of the transient temperature field of the entire blade, and effectively reduce the risk of local transient temperature rise during the variable load process while ensuring steady-state cooling performance. Summary of the Invention

[0008] In order to at least solve one or more of the technical problems mentioned above, this application proposes a transient optimization scheme for variable load turbine air film layout in several aspects.

[0009] In a first aspect, this application provides a transient optimization method for variable-load turbine film cooling layout, comprising: parameterizing the film cooling layout of turbine blades, sampling and generating multiple layout geometric schemes, and extracting transient temperature data of each layout geometric scheme under variable-load conditions based on transient fluid-thermal coupling simulation to construct a transient temperature database; using an autoencoder model to extract low-dimensional latent vectors of transient temperatures from the transient temperature database, and constructing a prediction network to establish a mapping between the geometric parameters of the film cooling layout and the parameters of the variable-load conditions to the low-dimensional latent vectors; jointly adjusting the prediction network and the decoder of the autoencoder model to obtain a joint prediction model, and extracting turbine blade temperature response characteristic indicators accordingly, the turbine blade temperature response characteristic indicators including average temperature rise rate, temperature rise rate uniformity, average rated temperature, and rated temperature uniformity; constructing a multi-objective optimization model based on the turbine blade temperature response characteristic indicators, and obtaining a Pareto front solution set through optimization using the multi-objective optimization model; calculating the transient local gradient of temperature rise for each scheme in the Pareto front solution set, and determining the optimal film cooling layout scheme based on the transient local gradient of temperature rise.

[0010] In a second aspect, this application provides a transient optimization system for variable-load turbine film gas layout. The system employs the transient optimization method for variable-load turbine film gas layout as described in any embodiment of the first aspect. The system includes: a database construction module for parameterizing the turbine blade film gas layout, sampling and generating multiple layout geometry schemes, and extracting transient temperature data of each layout geometry scheme under variable-load conditions based on transient flow-heat coupling simulation to construct a transient temperature database; and a feature dimensionality reduction and mapping module for extracting low-dimensional latent vectors of transient temperatures from the transient temperature database using an autoencoder model, and constructing a prediction network to establish the film gas pore layout geometry parameters and variable-load condition parameters to the low-dimensional latent vectors. The system includes: a vector mapping module; a model joint and index extraction module, used to jointly adjust the prediction network and the decoder of the autoencoder model to obtain a joint prediction model, and extract turbine blade temperature response characteristic indices, including average temperature rise rate, temperature rise rate uniformity, average rated temperature, and rated temperature uniformity; a multi-objective optimization module, used to construct a multi-objective optimization model based on the turbine blade temperature response characteristic indices, and to obtain the Pareto front solution set through optimization using the multi-objective optimization model; and an optimal solution determination module, used to calculate the transient local temperature rise gradient of each solution in the Pareto front solution set, and to determine the optimal film cooling layout scheme based on the transient local temperature rise gradient.

[0011] Through the transient optimization scheme for variable-load turbine film cooling layout provided above, this embodiment of the application introduces an autoencoder and a prediction network to construct a joint prediction model, achieving efficient dimensionality reduction and rapid prediction of complex transient temperature fields, greatly reducing the computational cost in the optimization process. Simultaneously, multi-objective optimization is performed by comprehensively considering steady-state temperature and transient temperature rise indices. After obtaining the Pareto front, rigorous screening is further conducted using the local gradient of transient temperature rise, effectively eliminating schemes with the risk of localized thermal stress concentration. This not only significantly improves the design efficiency of the film cooling layout but also ensures that the final scheme balances excellent steady-state cooling performance and transient variable-load performance, significantly improving the structural safety and service life of turbine blades under complex operating conditions.

[0012] Furthermore, in some embodiments, the local morphological parameters of the air film pores are fixed during the parameterization process, and adjustments are made only to the spatial layout position and spacing, effectively reducing the dimensionality of the design space. Simultaneously, an innovative B-spline function is used to establish a mathematical relationship between radial spacing and circumferential position, requiring only a few control variables to generate a smooth, continuous, and highly flexible non-uniform air film pore layout. This dimensionality-reducing and highly flexible parameterization strategy not only effectively avoids unreasonable designs such as pore overlap but also significantly improves the efficiency and convergence of subsequent sampling and optimization.

[0013] Furthermore, in some embodiments, minimizing the average temperature rise rate and the average rated temperature is used as a dual optimization objective. This ensures that the designed film cooling layout provides excellent steady-state cooling while effectively suppressing drastic temperature fluctuations during transient load changes. Simultaneously, the uniformity of temperature and temperature rise rate is used as a strict constraint, effectively preventing localized overheating or excessive temperature differences on the blade surface. This optimization mechanism comprehensively considers both the thermal protection efficiency of the turbine blade and the consistency of the temperature field distribution, fundamentally reducing the threat posed by localized thermal stress concentration to the blade's structural safety and service life.

[0014] Furthermore, in some embodiments, the risk of local thermal stress concentration during variable load processes is precisely quantified by calculating the transient temperature rise local gradient at two-dimensional coordinates on the turbine blade surface. After obtaining the Pareto front, a rigorous secondary screening of the solution set is performed using the maximum permissible local gradient as a threshold, effectively eliminating potential solutions that are highly likely to cause thermal fatigue and structural cracks, thus significantly improving the operational safety of the blade. Simultaneously, it allows for flexible selection of the final solution set that emphasizes steady-state performance or transient performance, based on actual engineering preferences, giving the design process extremely high flexibility and engineering practical value. Attached Figure Description

[0015] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of this application are illustrated by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts, wherein: Figure 1 An exemplary flowchart of a transient optimization method for a variable load turbine film structure according to an embodiment of this application is shown; Figure 2 An exemplary flowchart illustrating the parameterization of the turbine blade film gas layout according to an embodiment of this application is shown; Figure 3(a) shows a schematic diagram of the circumferential distribution of the air film pores in an embodiment of this application; Figure 3(b) shows a diagram illustrating the radial spacing of the air film perforation rows according to an embodiment of this application; Figure 3(c) shows a schematic diagram of the B-spline control curve of an embodiment of this application; Figure 4(a) shows a schematic diagram of the key temperature parameter curves changing over time under variable load conditions according to an embodiment of this application; Figure 4(b) shows a schematic diagram of the key flow parameters changing over time under variable load conditions according to an embodiment of this application; Figure 5(a) shows a fluid domain model diagram including the main flow and cold flow channels according to an embodiment of this application; Figure 5(b) shows a solid domain model diagram including internal heat conduction and film outflow in an embodiment of this application; Figure 6 This paper illustrates a flowchart of CAE-based dimensionality reduction of transient temperature data features of turbine blades according to an embodiment of this application. Figure 7 This paper presents a comparison chart of the actual temperature field, CAE reconstructed temperature field, and reconstruction relative error distribution of the test set samples in an embodiment of this application. Figure 8 This paper presents a flowchart of the end-to-end fast prediction process for the transient temperature response characteristics of turbine blades based on the CAE-DNN joint model according to an embodiment of this application. Figure 9(a) shows a comparison of the actual temperature field, the temperature field predicted by the CAE-DNN model, and the distribution of the relative prediction error of a test set sample in an embodiment of this application. Figure 9(b) shows a comparison of the true temperature field, the CAE-DNN model predicted temperature field, and the prediction relative error distribution of another test set sample in this application embodiment; Figure 10 The diagram illustrates the population distribution and Pareto front evolution of the multi-objective evolutionary algorithm according to an embodiment of this application under different iteration generations. Figure 11An exemplary flowchart illustrating an embodiment of this application shows how to determine the optimal film cooling layout scheme based on the local gradient of transient temperature rise. Figure 12(a) shows a comparison of the average blade temperature over time in an embodiment of this application; Figure 12(b) shows a comparison of the steady-state circumferential temperature distribution curve and surface temperature cloud map of the embodiment of this application under rated conditions; Figure 12(c) shows a comparison of the transient temperature rise rate circumferential distribution curve and the surface temperature rise rate cloud map of an embodiment of this application; Figure 13 An exemplary structural block diagram of a transient optimization system for a variable load turbine air film layout according to an embodiment of this application is shown. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be understood that the terms "comprising" and "including" as used in the specification and claims of this application indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0018] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this specification and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this specification and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0019] Figure 1 An exemplary flowchart of a transient optimization method 100 for a variable load turbine air film layout according to an embodiment of this application is shown.

[0020] like Figure 1 As shown, in step S110, the turbine blade air film layout is parameterized, multiple layout geometry schemes are sampled and generated, and transient temperature data of each layout geometry scheme under variable load conditions are extracted based on transient flow-thermal coupling simulation to construct a transient temperature database.

[0021] For details on the parameterization process of the turbine blade air film layout in the embodiments of this application, please refer to [link / reference]. Figure 2 .

[0022] Figure 2 An exemplary flowchart illustrating the parameterization of the turbine blade air film layout according to an embodiment of this application is shown.

[0023] like Figure 2 As shown, in step S210, the aperture shape, aperture diameter, compound angle, and tilt angle of the air film perforation array remain unchanged; only its spatial layout position and spacing are adjusted. In step S220, the circumferential distribution position of each air film perforation array is determined by setting the initial circumferential position variable of the first air film perforation array and sequentially defining the circumferential spacing variable between subsequent adjacent air film perforation arrays. In step S230, the radial spacing of the air film perforation array is set as a B-spline function with respect to the circumferential distribution position. In step S240, by determining the control variable for the radial spacing, the determined circumferential distribution position of the air film perforation array is substituted into the B-spline function to obtain the radial spacing corresponding to each air film perforation array.

[0024] In the embodiments of this application, during the execution of step S210, as shown in FIG3(a), the turbine blade to be optimized has two cooling chambers, front and rear, and multiple air film holes are distributed on the blade surface. Since the heat load on the leading edge and pressure side of the blade is higher and the cooling holes are more densely distributed, this application prioritizes the layout optimization of at least one air film hole (e.g., the R1-R8 holes shown in the figure) in this area.

[0025] In the embodiments of this application, during step S210, the circumferential position of the leading edge of the blade is defined as 0, and the circumferential position of the trailing edge is defined as 1. The circumferential position variable of the first exhaust film hole R1 is defined as x1, and the circumferential spacing variable between the i-th exhaust film hole and the (i-1)-th exhaust film hole is sequentially defined as ∆x. i As shown in Figure 3(b), the circumferential distribution position of each hole row can be uniquely determined by the initial position of the first row and the relative spacing variables of the subsequent hole rows.

[0026] In the embodiments of this application, the expression for the radial spacing of the air film perforation array is: ,in, The radial spacing of the air film pores. This refers to the circumferential distribution of the air film pores. Let i be the control variable for the i-th radial spacing. This represents the total number of control variables for radial spacing. Let be the i-th k-th order B-spline basis function.

[0027] In the embodiments of this application, in order to control the size of the variables and ensure the smoothness of the curve, the coordinate variables of three control points (P1, P2, P3) of the B-spline curve are selected as control variables. As shown in Figure 3(c), the control points P1, P2, and P3 are distributed around the circumferential positions of approximately 0.33, 0.50, and 0.77, respectively, and the radial spacing varies between -0.10 and 0.05.

[0028] In the embodiments of this application, the radial spacing parameterization method can also be replaced by a parameterization method based on freeform surface deformation technology or non-uniform rational B-spline curves.

[0029] In the embodiments of this application, during the setting of variable load conditions, firstly, a combination of operating parameters that varies with time is set. Then, during the transient fluid-thermal coupling simulation, each operating parameter in the combination of operating parameters is made to increase linearly from the initial low load state to the rated load state over time.

[0030] In the embodiments of this application, the combination of operating parameters includes turbine main flow rate, main flow rate inlet temperature, front chamber cold air flow rate, rear chamber cold air flow rate, front chamber cold air inlet temperature, and rear chamber cold air inlet temperature.

[0031] As shown in Figures 4(a) and 4(b), in the embodiments of this application, the total transient simulation duration is set to ∆t = 4s. During the process of time t changing from 0s to 4s: the turbine main flow... The flow rate increased linearly from 1.0525 kg / s to 1.0567 kg / s; the mainstream inlet temperature... The cold air flow rate in the front cavity increased linearly from 2004K to 2012K. The flow rate of the cold air in the rear cavity increased linearly from 0.04851 kg / s to 0.04871 kg / s. The flow rate increased linearly from 0.03088 kg / s to 0.03100 kg / s; the inlet temperature of the front chamber cold air increased linearly. Rear cavity cold air inlet temperature Both increased linearly from 879K to 883K.

[0032] In the embodiments of this application, as shown in Figures 5(a) and 5(b), a number of geometric parameter schemes are generated in the design space using the Latin hypercube sampling method. For each geometric scheme, a three-dimensional model of the turbine blade, including the fluid domain and the solid domain, is established and meshed. Based on the transient fluid-thermal coupling numerical simulation method, the set variable load boundary conditions are applied for calculation. The transient temperature values ​​of the blade surface at each time step during the variable load process are extracted for each scheme, thereby constructing a transient temperature database of the turbine blade under variable load conditions.

[0033] After completing step S110, in step S120, an autoencoder model is used to extract the low-dimensional latent vector of transient temperature from the transient temperature database, and a prediction network is constructed to establish the mapping between the geometric parameters of the film vent layout and the parameters of the variable load condition to the low-dimensional latent vector.

[0034] In the embodiments of this application, during the extraction of low-dimensional latent vectors of transient temperatures from a transient temperature database using an autoencoder model, firstly, the transient temperature data at each time step is standardized into a two-dimensional matrix (e.g., with a spatial resolution of 128×128). Secondly, the two-dimensional matrix is ​​input into the autoencoder model for training. By minimizing the reconstruction error, the autoencoder model learns the core spatial distribution features of the transient temperature data. Thirdly, after the autoencoder model is trained, the encoder of the autoencoder model is used to compress the transient temperature data to obtain low-dimensional latent vectors that retain the core spatial distribution features.

[0035] In the embodiments of this application, the autoencoder model employs a convolutional autoencoder (CAE).

[0036] like Figure 6 As shown, the two-dimensional matrix is ​​input into the designed convolutional autoencoder (CAE) network for training. By minimizing the reconstruction error, the network is forced to learn the intrinsic distribution law of the temperature field. After training, the transient temperature data is compressed into a low-dimensional latent vector C (e.g., compressed to 16-64 dimensions, preferably 32 dimensions) using the encoder, preserving the core spatial distribution characteristics of the temperature field.

[0037] like Figure 7 As shown, the average relative error of the CAE model reconstruction is within 1%, which can accurately preserve the core spatial distribution characteristics of the turbine blade temperature field.

[0038] In other embodiments of this application, the convolutional autoencoder (CAE) may be replaced by a variational autoencoder or a principal component analysis network.

[0039] In other embodiments of this application, the geometric parameters of the air film aperture layout are the aforementioned geometric parameter vector, which includes the initial circumferential position variable, the circumferential spacing variable between adjacent rows, and the radial spacing control variable.

[0040] In other embodiments of this application, the variable load operating parameters are the aforementioned combination of operating parameters including turbine main flow rate, main flow rate inlet temperature, front chamber cold air flow rate, rear chamber cold air flow rate, front chamber cold air inlet temperature, and rear chamber cold air inlet temperature.

[0041] In the embodiments of this application, during the process of constructing a prediction network to establish the mapping between the geometric parameters of the film orifice layout and the parameters of the variable load condition to low-dimensional latent vectors, firstly, the geometric parameters of the film orifice layout and the parameters of the variable load condition are concatenated and used as the input to the prediction network. Secondly, the low-dimensional latent vectors extracted by the autoencoder model are used as the output of the prediction network. Thirdly, a fully connected prediction network is constructed, and the prediction network is trained based on the input and output of the prediction network, thereby establishing the mapping relationship between the geometric parameters of the film orifice layout and the parameters of the variable load condition to the low-dimensional latent vectors.

[0042] In the embodiments of this application, firstly, the geometric parameters of the 11 air-supported film vent layout are concatenated with the variable load parameters at four specific times, serving as the input to the prediction network. Secondly, the low-dimensional latent vector C extracted by the CAE model is used as the output. Thirdly, a fully connected deep neural network (DNN) is constructed and trained to establish a nonlinear mapping relationship between the input parameters and low-dimensional features.

[0043] In other embodiments of this application, the prediction network may also employ Gaussian process regression, support vector regression, or gradient boosting tree.

[0044] After completing step S120, in step S130, the prediction network and the decoder of the autoencoder model are jointly adjusted to obtain a joint prediction model, and the turbine blade temperature response characteristic index is extracted accordingly.

[0045] In the embodiments of this application, during step S130, firstly, the pre-trained prediction network and the decoder of the autoencoder model are concatenated to form an end-to-end prediction network. Secondly, using the geometric parameters of the film cooling orifice layout and the parameters of the variable load condition as inputs, and using the error between the reconstructed physical temperature field and the actual temperature field as the overall loss function, the end-to-end prediction network is jointly adjusted to obtain a joint prediction model. Thirdly, the joint prediction model is used to predict the transient temperature field under any given film cooling orifice layout, and the turbine blade temperature response characteristic index is obtained through post-processing calculation.

[0046] In the embodiments of this application, such as Figure 8 As shown, during step S130, firstly, the pre-trained prediction network (DNN) is concatenated with the decoder part of the CAE to form a complete end-to-end prediction network. Secondly, using the geometric parameters of the film vent layout and the variable load condition parameters as inputs, the error between the finally reconstructed physical temperature field and the actual temperature field is directly used as the overall loss function to jointly fine-tune the parameters of the entire network. This joint fine-tuning makes the latent vectors predicted by the prediction network more consistent with the reconstruction requirements of the decoder.

[0047] As shown in Figures 9(a) and 9(b), the model after fine-tuning has a determination coefficient higher than 0.98 on the test set, and the average relative error is within 1%. Furthermore, this joint model is used to quickly predict the transient temperature field under any given film gas configuration, and the temperature response characteristics of the turbine blade are obtained through post-processing calculations.

[0048] In the embodiments of this application, the temperature response characteristics of the turbine blade include the average temperature rise rate, the uniformity of the temperature rise rate, the average temperature under rated conditions, and the uniformity of the temperature under rated conditions.

[0049] After completing step S130, in step S140, a multi-objective optimization model is constructed based on the temperature response characteristic index of the turbine blade, and the Pareto front solution set is obtained by optimization through the multi-objective optimization model.

[0050] In the embodiments of this application, during the construction of a multi-objective optimization model based on the temperature response characteristics of turbine blades, firstly, the geometric parameters of the film cooling vent layout are used as optimization variables. Secondly, minimizing the average temperature rise rate and the average rated temperature are used as the optimization objectives of the multi-objective optimization model. Thirdly, constraints for the multi-objective optimization model are constructed based on the uniformity of the temperature rise rate and the uniformity of the rated temperature.

[0051] In the embodiments of this application, the objective function of the multi-objective optimization model is: ,in, The objective function of the multi-objective optimization model is... This represents the average rate of temperature rise. The average temperature is the temperature under rated conditions, and N is the number of temperature data points. The temperature of the j-th temperature data point under rated termination conditions. The temperature is the j-th temperature data point in the initial running state.

[0052] In the embodiments of this application, the constraints of the multi-objective optimization model are as follows: ,in, The constraints for the multi-objective optimization model are as follows: To ensure uniformity of temperature rise rate, For rated temperature uniformity, Let j be the temperature rise rate at the j-th temperature data point. A threshold is set for the uniformity of the temperature rise rate. The threshold value is set for the temperature uniformity under rated conditions.

[0053] In the embodiments of this application, the threshold values ​​for temperature rise rate uniformity and rated state temperature uniformity are set according to actual needs and historical experience, and this application does not impose any restrictions on them.

[0054] In the embodiments of this application, a non-dominated sorting genetic algorithm (NSGA-II) with an elitist strategy is used to iteratively optimize the model, obtaining a Pareto optimal front solution that meets the conditions through selection, crossover, and mutation operations. This Pareto optimal front solution is a set of candidate air-film layout schemes that meet the performance requirements and achieve the best balance between steady-state cooling and transient protection.

[0055] In other embodiments of this application, the non-dominated sorting genetic algorithm can also be replaced by a multi-objective particle swarm optimization algorithm or a decomposition-based multi-objective evolutionary algorithm.

[0056] like Figure 10 As shown, the Pareto front clearly reveals the competitive trade-off between steady-state rated temperature and transient temperature rise rate.

[0057] After completing step S140, in step S150, the transient local temperature rise gradient of each scheme in the Pareto front solution set is calculated, and the optimal air film cooling layout scheme is determined based on the transient local temperature rise gradient.

[0058] In the embodiments of this application, the transient temperature rise local gradient is used to quantify the spatial drastic change in the temperature rise rate at various locations on the blade surface. The expression for the transient temperature rise local gradient is: ,in, For the transient temperature rise local gradient, Let j be the temperature rise rate at the j-th temperature data point. , The temperature of the j-th temperature data point under rated termination conditions. Let (x, y) be the temperature of the j-th temperature data point in the initial operating state, and (x, y) be the two-dimensional coordinates of the turbine blade surface.

[0059] In the embodiments of this application, the specific process of determining the optimal film cooling layout scheme based on the local gradient of transient temperature rise can be found in [reference needed]. Figure 11 .

[0060] Figure 11 An exemplary flowchart illustrating an embodiment of this application is shown, which determines the optimal air film cooling layout scheme based on the local gradient of transient temperature rise.

[0061] like Figure 11As shown, in step S1110, it is determined whether each solution in the Pareto front solution set satisfies a preset condition at any time during the transient process and at any position on the turbine blade surface. In response to the existence of a solution in the Pareto front solution set that satisfies the preset condition at a corresponding time during the transient process or at a corresponding position on the turbine blade surface, the corresponding solution is discarded in step S1120. In response to the existence of a solution in the Pareto front solution set that does not satisfy the preset condition at any time during the transient process or at any position on the turbine blade surface, the corresponding solution is stored in the remaining solution set in step S1130.

[0062] Next, in step S1140, it is determined whether the preset engineering preference prioritizes steady-state performance or transient performance. If the preset engineering preference prioritizes steady-state performance, in step S1150, the solution with the smallest average rated state temperature is selected from the remaining solution set as the optimal film cooling layout. If the preset engineering preference prioritizes transient performance, in step S1160, the solution with the smallest average temperature rise rate is selected from the remaining solution set as the optimal film cooling layout.

[0063] In the embodiments of this application, the preset condition is: the transient temperature rise local gradient is greater than the maximum allowable local gradient threshold.

[0064] In the embodiments of this application, the maximum permissible local gradient threshold is set according to actual needs and historical experience, and this application does not impose any restrictions on it.

[0065] In an embodiment of this application, in response to a solution in the Pareto front solution set triggering the preset local danger condition (i.e., a temperature rise abruptly exceeding a threshold) in a microscopic local region, in step S320, it is determined that the solution has an extremely high risk of local thermal stress concentration and blade cracking, and the solution is forcibly removed.

[0066] In an embodiment of this application, in response to a solution in the Pareto front solution set not triggering the preset local danger condition (i.e., the temperature rise gradient is always gentle and controllable) throughout the entire time period and the entire spatial domain, in step S330, the solution is determined to be a safe and usable alternative solution and stored in the remaining solution set.

[0067] In the embodiments of this application, given that each alternative solution in the remaining solution set has simultaneously met the global performance requirements and local security constraints, the pre-set engineering preferences will be further obtained and judged to determine the final output solution.

[0068] In the embodiments of this application, in response to the preset engineering preference that focuses on steady-state performance (i.e., for units that operate at full load for a long time, the ultimate basic cooling effect is pursued), in step S350, the solution with the smallest average rated state temperature is selected from the remaining solution set and used as the globally optimal air film cooling layout scheme.

[0069] In the embodiments of this application, in response to the preset engineering preference that focuses on transient performance (i.e., for units that need to frequently participate in grid peak shaving, the pursuit of the highest fatigue life during load change), in step S360, the solution with the smallest average temperature rise rate is selected from the remaining solution set and used as the globally optimal air film cooling layout scheme.

[0070] As shown in Figures 12(a)-12(c), the globally optimal solution determined after secondary screening reduces the average rated temperature of the blades by more than 10K, while the average transient temperature rise rate decreases by more than 0.25K / s. The distribution curves show a significant decrease in temperature level in the densely packed film cooling area, and the temperature rise rate at each circumferential position exhibits a relatively uniform decrease, achieving a dual improvement in steady-state cooling capacity and resistance to transient thermal shock.

[0071] In summary, through the transient optimization scheme for variable-load turbine film cooling layout provided above, this application embodiment achieves efficient dimensionality reduction and rapid prediction of complex transient temperature fields by introducing an autoencoder and a prediction network to construct a joint prediction model, greatly reducing the computational cost in the optimization process. Simultaneously, multi-objective optimization is performed by comprehensively considering steady-state temperature and transient temperature rise indices. After obtaining the Pareto front, rigorous screening is further conducted using the local gradient of transient temperature rise, effectively eliminating schemes with the risk of local thermal stress concentration. This not only significantly improves the design efficiency of the film cooling layout but also ensures that the final scheme balances excellent steady-state cooling performance and transient variable-load performance, significantly improving the structural safety and service life of turbine blades under complex operating conditions.

[0072] Furthermore, in some embodiments, the local morphological parameters of the air film pores are fixed during the parameterization process, and adjustments are made only to the spatial layout position and spacing, effectively reducing the dimensionality of the design space. Simultaneously, an innovative B-spline function is used to establish a mathematical relationship between radial spacing and circumferential position, requiring only a few control variables to generate a smooth, continuous, and highly flexible non-uniform air film pore layout. This dimensionality-reducing and highly flexible parameterization strategy not only effectively avoids unreasonable designs such as pore overlap but also significantly improves the efficiency and convergence of subsequent sampling and optimization.

[0073] Furthermore, in some embodiments, minimizing the average temperature rise rate and the average rated temperature is used as a dual optimization objective. This ensures that the designed film cooling layout provides excellent steady-state cooling while effectively suppressing drastic temperature fluctuations during transient load changes. Simultaneously, the uniformity of temperature and temperature rise rate is used as a strict constraint, effectively preventing localized overheating or excessive temperature differences on the blade surface. This optimization mechanism comprehensively considers both the thermal protection efficiency of the turbine blade and the consistency of the temperature field distribution, fundamentally reducing the threat posed by localized thermal stress concentration to the blade's structural safety and service life.

[0074] Furthermore, in some embodiments, the risk of local thermal stress concentration during variable load processes is precisely quantified by calculating the transient temperature rise local gradient at two-dimensional coordinates on the turbine blade surface. After obtaining the Pareto front, a rigorous secondary screening of the solution set is performed using the maximum permissible local gradient as a threshold, effectively eliminating potential solutions that are highly likely to cause thermal fatigue and structural cracks, thus significantly improving the operational safety of the blade. Simultaneously, it allows for flexible selection of the final solution set that emphasizes steady-state performance or transient performance, based on actual engineering preferences, giving the design process extremely high flexibility and engineering practical value.

[0075] This application also provides a transient optimization system for variable load turbine air film layout. It can use the aforementioned transient optimization method 100 for variable load turbine air film layout to perform transient optimization of variable load turbine air film layout, or other methods can be used for transient optimization of variable load turbine air film layout. This application does not limit this.

[0076] Figure 13 An exemplary structural block diagram of a transient optimization system for a variable load turbine air film layout according to an embodiment of this application is shown.

[0077] like Figure 13 As shown, the system 1300 includes a database construction module 1310, a feature dimensionality reduction and mapping module 1320, a model joint and index extraction module 1330, a multi-objective optimization module 1340, and an optimal solution determination module 1350. In the embodiments of this application, the database construction module 1310, the feature dimensionality reduction and mapping module 1320, the model joint and index extraction module 1330, the multi-objective optimization module 1340, and the optimal solution determination module 1350 can be separate units or can be set in the same controller; this application does not impose any restrictions here.

[0078] Specifically, the database construction module 1310 is used to parameterize the air film layout of turbine blades, sample and generate multiple layout geometry schemes, and extract transient temperature data of each layout geometry scheme under variable load conditions based on transient flow-heat coupling simulation to construct a transient temperature database.

[0079] Specifically, the feature dimensionality reduction and mapping module 1320 is used to extract the low-dimensional latent vector of transient temperature in the transient temperature database using an autoencoder model, and to construct a prediction network to establish the mapping of the geometric parameters of the air film vent layout and the parameters of the variable load condition to the low-dimensional latent vector.

[0080] Specifically, the model joint and index extraction module 1330 is used to jointly adjust the prediction network and the decoder of the autoencoder model to obtain a joint prediction model, and extract the temperature response characteristic index of the turbine blade accordingly. The temperature response characteristic index of the turbine blade includes the average temperature rise rate, the uniformity of the temperature rise rate, the average temperature under rated conditions, and the uniformity of the temperature under rated conditions.

[0081] Specifically, the multi-objective optimization module 1340 is used to construct a multi-objective optimization model based on the temperature response characteristic index of the turbine blade, and to obtain the Pareto front solution set through optimization using the multi-objective optimization model.

[0082] Specifically, the optimal solution determination module 1350 is used to calculate the transient local temperature rise gradient of each solution in the Pareto front solution set, and determine the optimal air film cooling layout solution based on the transient local temperature rise gradient.

[0083] When system 1300 performs transient optimization of the variable load turbine air-film layout using the aforementioned transient optimization method 100, the aforementioned steps S110 are executed through the database construction module 1310, S120 through the feature dimensionality reduction and mapping module 1320, S130 through the model joint and index extraction module 1330, S140 through the multi-objective optimization module 1340, and S150 through the optimal solution determination module 1350. The specific execution process can be found above and will not be repeated here.

[0084] While numerous embodiments of this application have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of this application. It should be understood that various alternatives to the embodiments of this application described herein may be employed in the practice of this application. The appended claims are intended to define the scope of protection of this application and therefore cover equivalents or alternatives within the scope of these claims.

Claims

1. A transient optimization method for variable load turbine film gas layout, characterized in that, include: The air film layout of turbine blades is parameterized, multiple layout geometry schemes are sampled and generated, and transient temperature data of each layout geometry scheme under variable load conditions are extracted based on transient fluid-thermal coupling simulation to construct a transient temperature database. An autoencoder model is used to extract the low-dimensional latent vector of transient temperature from the transient temperature database, and a prediction network is constructed to establish the mapping between the geometric parameters of the film vent layout and the parameters of the variable load condition to the low-dimensional latent vector. The prediction network and the decoder of the autoencoder model are jointly adjusted to obtain a joint prediction model, and the turbine blade temperature response characteristic index is extracted accordingly. The turbine blade temperature response characteristic index includes the average temperature rise rate, the uniformity of the temperature rise rate, the average temperature under rated conditions, and the uniformity of the temperature under rated conditions. A multi-objective optimization model is constructed based on the temperature response characteristics of the turbine blade, and the Pareto front solution set is obtained by optimization through the multi-objective optimization model. Calculate the transient local temperature rise gradient of each scheme in the Pareto front solution set, and determine the optimal film cooling layout scheme based on the transient local temperature rise gradient.

2. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, The following steps are performed during the parameterization of the turbine blade film gas configuration: Keep the pore shape, pore diameter, compound angle and tilt angle of the air film pore row unchanged, and only adjust its spatial layout and spacing; By setting the initial circumferential position variable of the first exhaust film hole and defining the circumferential spacing variable between subsequent adjacent air film hole rows in sequence, the circumferential distribution position of each air film hole row is determined. The radial spacing of the air film pore array is set as a B-spline function with respect to the circumferential distribution position, where the expression for the radial spacing of the air film pore array is: ,in, The radial spacing of the air film pores. This refers to the circumferential distribution of the air film pores. Let i be the control variable for the i-th radial spacing. This represents the total number of control variables for radial spacing. Let i be the k-th order B-spline basis function; By determining the control variable for the radial spacing, the circumferential distribution position of the determined air film pore rows is substituted into the B-spline function to obtain the radial spacing corresponding to each air film pore row.

3. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, The variable load condition is set through the following steps: Set a combination of operating parameters that vary over time, including turbine main flow rate, main flow inlet temperature, front chamber cold air flow rate, rear chamber cold air flow rate, front chamber cold air inlet temperature, and rear chamber cold air inlet temperature. In the transient flow-thermal coupling simulation process, each operating parameter in the operating parameter combination is made to increase linearly from the initial low load state to the rated load state over time.

4. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, In the process of extracting the low-dimensional latent vector of transient temperature from the transient temperature database using an autoencoder model, the following steps are performed: The transient temperature data at each time step are standardized into a two-dimensional matrix; The two-dimensional matrix is ​​input into the autoencoder model for training. By minimizing the reconstruction error, the autoencoder model learns the core spatial distribution features of the transient temperature data. After the autoencoder model is trained, the transient temperature data is compressed using the encoder of the autoencoder model to obtain a low-dimensional latent vector that retains the core spatial distribution features.

5. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, In the process of constructing the prediction network to establish the mapping between the geometric parameters of the film vent layout and the parameters under varying load conditions to the low-dimensional latent vector, the following steps are performed: The geometric parameters of the air film pore layout are concatenated with the parameters of the variable load condition and used as the input to the prediction network. The low-dimensional latent vector extracted by the autoencoder model is used as the output of the prediction network; A fully connected prediction network is constructed, and the prediction network is trained based on the input and output of the prediction network to establish the mapping relationship between the geometric parameters of the air film vent layout and the parameters of the variable load condition to the low-dimensional latent vector.

6. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, In the process of jointly adjusting the prediction network and the decoder of the autoencoder model to obtain a joint prediction model, and extracting the temperature response characteristic index of the turbine blade accordingly, the following steps are performed: The pre-trained prediction network is concatenated with the decoder of the autoencoder model to form an end-to-end prediction network; Using the geometric parameters of the air film pore layout and the parameters of the variable load condition as inputs, and the error between the reconstructed physical temperature field and the real temperature field as the overall loss function, the end-to-end prediction network is jointly adjusted to obtain the joint prediction model. The joint prediction model is used to predict the transient temperature field under any given air film layout, and the temperature response characteristics of the turbine blade are obtained through post-processing calculation.

7. The transient optimization method for variable load turbine film gas layout according to claim 1 or 6, characterized in that, In the process of constructing a multi-objective optimization model based on the temperature response characteristics of the turbine blade, the following steps are performed: The geometric parameters of the air film pore layout are used as optimization variables; The optimization objectives of the multi-objective optimization model are minimizing the average rate of temperature rise and the average rated state temperature, wherein the objective function of the multi-objective optimization model is: ,in, The objective function of the multi-objective optimization model is... This represents the average rate of temperature rise. The average temperature is the temperature under rated conditions, and N is the number of temperature data points. The temperature of the j-th temperature data point under rated termination conditions. The temperature of the j-th temperature data point in the initial running state; The constraints of the multi-objective optimization model are constructed based on the uniformity of temperature rise rate and the uniformity of rated state temperature. The constraints of the multi-objective optimization model are as follows: ,in, The constraints for the multi-objective optimization model are as follows: To ensure uniformity of temperature rise rate, For rated temperature uniformity, Let j be the temperature rise rate at the j-th temperature data point. A threshold is set for the uniformity of the temperature rise rate. The threshold value is set for the temperature uniformity under rated conditions.

8. The transient optimization method for variable load turbine film gas layout according to claim 1, characterized in that, The expression for the transient temperature rise local gradient is: ,in, For the transient temperature rise local gradient, Let j be the temperature rise rate at the j-th temperature data point. , The temperature of the j-th temperature data point under rated termination conditions. Let (x, y) be the temperature of the j-th temperature data point in the initial operating state, and (x, y) be the two-dimensional coordinates of the turbine blade surface.

9. The transient optimization method for variable load turbine film gas layout according to claim 1 or 8, characterized in that, In determining the optimal film cooling layout based on the local gradient of transient temperature rise, the following steps are performed: Determine whether each solution in the Pareto front solution set satisfies a preset condition at any moment in the transient process and at any position on the turbine blade surface. The preset condition is: the local gradient of the transient temperature rise is greater than the maximum allowable local gradient threshold. In response to the existence of Pareto front solution set, if the corresponding solution satisfies the preset condition at the corresponding time of the transient process or at the corresponding position on the turbine blade surface, the corresponding solution is discarded. In response to the fact that the corresponding solution in the Pareto front solution set does not meet the preset condition at any moment of the transient process and at any position on the turbine blade surface, the corresponding solution is stored in the remaining solution set. Determine whether the preset engineering preference emphasizes steady-state performance or transient performance; In response to the preset engineering preference that emphasizes steady-state performance, the solution with the smallest average rated state temperature is selected from the remaining solution set as the optimal air film cooling layout scheme. In response to a preset engineering preference that emphasizes transient performance, the solution with the smallest average temperature rise rate is selected from the remaining solution set as the optimal film cooling layout scheme.

10. A transient optimization system for variable load turbine film gas layout, characterized in that, The transient optimization of the variable-load turbine air-film layout is performed using the transient optimization method for the variable-load turbine air-film layout as described in any one of claims 1-9, wherein the system comprises: The database construction module is used to parameterize the air film layout of turbine blades, sample and generate multiple layout geometry schemes, and extract transient temperature data of each layout geometry scheme under variable load conditions based on transient flow-heat coupling simulation to build a transient temperature database. The feature dimensionality reduction and mapping module is used to extract the low-dimensional latent vector of transient temperature from the transient temperature database using an autoencoder model, and to construct a prediction network to establish the mapping between the geometric parameters of the film vent layout and the parameters of the variable load condition to the low-dimensional latent vector. The model joint and index extraction module is used to jointly adjust the prediction network and the decoder of the autoencoder model to obtain a joint prediction model, and extract the temperature response characteristic index of the turbine blade accordingly. The temperature response characteristic index of the turbine blade includes the average temperature rise rate, the uniformity of the temperature rise rate, the average temperature under rated conditions, and the uniformity of the temperature under rated conditions. The multi-objective optimization module is used to construct a multi-objective optimization model based on the temperature response characteristic index of the turbine blade, and to obtain the Pareto front solution set by optimization through the multi-objective optimization model. The optimal solution determination module is used to calculate the transient local temperature rise gradient of each solution in the Pareto front solution set, and determine the optimal air film cooling layout solution based on the transient local temperature rise gradient.

Images