A method, apparatus, computer equipment, and medium for predicting the outlet temperature of an aero-engine combustor.

Abstract

This invention provides a method, apparatus, computer equipment, and medium for predicting the outlet temperature of an aero-engine combustor, relating to the field of performance evaluation and prediction technology. The method includes the following steps: collecting input parameters related to combustor operation and structure in batches to generate an optimal parameter set; constructing an operating condition vector based on combustor operating conditions; constructing a stable parameter set using the optimal parameter set and the operating condition vector; filtering all parameters in the stable parameter set and adjusting the weights corresponding to the parameters in the operating condition vector using an adaptive weight model; constructing a physical constraint coupling prediction model; training the physical constraint coupling prediction model based on physical constraints to generate a trained physical constraint coupling prediction model; and outputting the non-uniformity of the combustor outlet temperature distribution using this model. This solution can achieve high-precision prediction of the non-uniformity of the combustor outlet temperature distribution under different pressure, temperature, and equivalence ratio conditions.

Description

Technical Field

[0001] This invention relates to the field of performance evaluation and prediction technology, and in particular to a method, apparatus, computer equipment and medium for predicting the outlet temperature of an aero-engine combustion chamber. Background Technology

[0002] The temperature distribution at the combustor exit of an aero-engine has a critical impact on the heat load distribution of downstream turbine blades, component life, and overall engine performance. Therefore, indicators such as OTDF (Overhead Temperature Distribution Factor) and RTDF (Radial Temperature Distribution Factor) are commonly used to quantitatively evaluate the uniformity of the combustor exit temperature field. Lower temperature non-uniformity helps extend turbine life, improve overall engine efficiency, and reduce the risk of thermal stress.

[0003] Existing performance evaluation methods mainly rely on two approaches: one is to measure the outlet temperature field through bench tests to directly calculate indicators such as OTDF; the other is to use numerical simulation to predict the internal flow and combustion processes of the combustion chamber. Both methods have significant drawbacks: high testing costs and long cycles, large computational load in numerical simulations, and limited ability to extrapolate operating conditions.

[0004] To reduce testing and computational costs, existing technologies attempt to predict combustion chamber outlet temperature distribution by simplifying parameters and using empirical models. For example, performance evaluation models are established based on the functional relationship between the variance of the head gas coefficient and the non-uniformity of the outlet temperature. While these methods can reduce testing workload under certain conditions, their prediction accuracy and applicability are significantly limited due to considering only a single variable. In fact, the combustion chamber outlet temperature distribution depends not only on fuel supply non-uniformity but also on a combination of factors, including spray characteristics (droplet size, spray angle), flow organization parameters, cooling and dilution airflow distribution, inlet distortion, and wall cooling conditions. Existing single-parameter fitting methods cannot fully reflect these complex mechanisms. Furthermore, existing prediction models are mostly empirically fitted or driven by black-box data, lacking necessary physical constraints. For instance, models may exhibit unreasonable situations where the predicted OTDF increases as the non-uniformity of the head gas coefficient decreases, or the predicted results exceed the physically defined range. These problems weaken the reliability of the models in engineering design and operational condition extrapolation. In addition, most existing methods are designed for single-rig conditions and lack the ability to generalize across pressure, temperature, equivalence ratio and other operating conditions, making it difficult to meet the needs of combustion performance evaluation under complex operating conditions in the future.

[0005] Therefore, there is an urgent need to propose a prediction method that can integrate the influence of multiple parameters and has variable screening, adaptive weight adjustment mechanism and physical constraint embedding, so as to break through the limitations of traditional "empirical fitting", improve prediction accuracy and cross-condition generalization ability, thereby reducing the dependence on experiments and numerical simulations and better supporting combustion chamber design and verification. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a method for predicting the outlet temperature of an aero-engine combustor, to solve the technical problem that existing technologies are difficult to adapt to the combustion performance evaluation requirements under complex future operating conditions. The method includes:

[0007] Input parameters related to combustion chamber operation and structure are collected in batches to generate an optimal parameter set, combustion chamber operating conditions are obtained, and an operating condition vector is constructed based on the combustion chamber operating conditions. The operating condition vector is used to characterize the working state of the combustion chamber.

[0008] Using the preferred parameter set and the operating condition vector, a stable parameter set with a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions is constructed. All parameters in the stable parameter set are screened, and the weights corresponding to the parameters in the operating condition vector are adjusted by an adaptive weight model to generate a preprocessed sample set.

[0009] A physical constraint coupling prediction model is constructed, physical constraints are determined, and the physical constraint coupling prediction model is trained based on the physical constraints to generate a trained physical constraint coupling prediction model. The preprocessed sample set is input into the trained physical constraint coupling prediction model, and the non-uniformity of the combustion chamber outlet temperature distribution is output.

[0010] This invention also provides a device for predicting the outlet temperature of an aero-engine combustor, addressing the technical problem that existing technologies struggle to meet the demands of combustion performance evaluation under complex future operating conditions. The device includes:

[0011] The parameter acquisition module is used to collect input parameters related to combustion chamber operation and combustion chamber structure in batches, generate an optimal parameter set, obtain combustion chamber operating conditions, and construct an operating condition vector through the combustion chamber operating conditions, wherein the operating condition vector is used to characterize the working state of the combustion chamber;

[0012] The preprocessing and parameter filtering module is used to construct a stable parameter set that has a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions through the preferred parameter set and the operating condition vector, filter all parameters in the stable parameter set, and adjust the weights corresponding to the parameters in the operating condition vector through an adaptive weight model to generate a preprocessed sample set.

[0013] OTDF The prediction output module is used to construct a physical constraint coupled prediction model, determine physical constraints, train the physical constraint coupled prediction model based on the physical constraints, generate a trained physical constraint coupled prediction model, input the preprocessed sample set into the trained physical constraint coupled prediction model, and output the non-uniformity of the combustion chamber outlet temperature distribution.

[0014] This invention also provides 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 the above-mentioned method for predicting the outlet temperature of any aero-engine combustion chamber, thereby solving the technical problem in the prior art that it is difficult to adapt to the combustion performance evaluation requirements under complex future operating conditions.

[0015] This invention also provides a computer-readable storage medium storing a computer program that executes any of the above-described methods for predicting the outlet temperature of an aero-engine combustor, in order to solve the technical problem that the prior art is difficult to adapt to the needs of combustion performance evaluation under complex future operating conditions.

[0016] Compared with the prior art, the beneficial effects that at least one technical solution adopted in the embodiments of this specification can achieve include at least:

[0017] It can achieve high-precision prediction of the non-uniformity of combustion chamber outlet temperature distribution under different pressure, temperature, equivalence ratio and other conditions. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a method for predicting the outlet temperature of an aero-engine combustion chamber provided in an embodiment of the present invention;

[0020] Figure 2 This is a flowchart illustrating a method for predicting the outlet temperature of an aero-engine combustion chamber, as provided in an embodiment of the present invention.

[0021] Figure 3 This is a structural block diagram of a computer device provided in an embodiment of the present invention;

[0022] Figure 4 This is a structural block diagram of a device for predicting the outlet temperature of an aero-engine combustion chamber, provided in an embodiment of the present invention. Detailed Implementation

[0023] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0024] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] In this embodiment of the invention, a method for predicting the outlet temperature of an aero-engine combustion chamber is provided, such as... Figure 1 and Figure 2 As shown, the method includes:

[0026] Step S101: Collect input parameters related to combustion chamber operation and combustion chamber structure in batches, generate an optimal parameter set, obtain combustion chamber operating conditions, and construct an operating condition vector through the combustion chamber operating conditions, wherein the operating condition vector is used to characterize the working state of the combustion chamber;

[0027] Step S102: Construct a stable parameter set that has a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions through the preferred parameter set and the operating condition vector. Filter all parameters in the stable parameter set and adjust the weights corresponding to the parameters in the operating condition vector through an adaptive weight model to generate a preprocessed sample set.

[0028] Step S103: Construct a physical constraint coupling prediction model, determine the physical constraints, train the physical constraint coupling prediction model based on the physical constraints, generate a trained physical constraint coupling prediction model, input the preprocessed sample set into the trained physical constraint coupling prediction model, and output the non-uniformity of the combustion chamber outlet temperature distribution.

[0029] In specific implementation, the following steps are used to construct a stable parameter set that has a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions, using the preferred parameter set and the operating condition vector; to filter all parameters in the stable parameter set; and to adjust the weights corresponding to the parameters in the operating condition vector using an adaptive weight model to generate a preprocessed sample set:

[0030] All parameters in the preferred parameter set and the operating condition vector are standardized to generate a standardized parameter set. Based on the standardized parameter set, an extended feature set is constructed, wherein the extended feature set consists of the preferred parameter set, the operating condition vector, and interaction terms between each parameter in the preferred parameter set and each parameter in the operating condition vector. The extended feature set is then filtered using a stability selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set, wherein the variables in the stable parameter set have a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions. An adaptive weighting model is constructed, and the operating condition vector is input into the adaptive weighting model to output the weights corresponding to the parameters in the operating condition vector. The stable parameter set containing the weights is used as the preprocessed sample set.

[0031] In specific implementation, the extended feature set is screened using a stable selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set. The variables in this stable parameter set have a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions.

[0032] Based on the stable selection method, the selection frequency of each variable in the extended feature set is calculated, and variables with selection frequencies higher than a first preset threshold are selected to generate a first candidate set.

[0033] The variables in the extended feature set are grouped according to their physical correlation. Significant variable groups are identified using group sparse regression, and all variables within these significant variable groups are saved to a second candidate set. A surrogate model of the combustion chamber outlet temperature field is constructed. Using this surrogate model, the total sensitivity index of each variable in the extended feature set is calculated based on the sensitivity analysis method. Variables with a total sensitivity index higher than a second preset threshold are selected to generate a third candidate set. An initial variable set is generated from the first, second, and third candidate sets through intersection or weighted voting. Cross-condition consistency verification is performed on each variable in the initial variable set, and a stable parameter set is generated.

[0034] In practice, the adaptive weight model is constructed through the following steps:

[0035] A condition-dependent exponential weighting function is constructed, and an adaptive weighting model is built based on the exponential weighting function, wherein the exponential weighting function... ,in, For the working condition vector, i For the index of the input variable, The first in the working condition vector j Each working condition component j For the index of the operating condition components, Let the dimension be the working condition vector. The original weight of the i-th input variable in the stable parameter set under the operating condition vector r. Let be the bias coefficient of the i-th variable. Let be the sensitivity coefficient of the i-th input variable to the j-th working condition component; define a target loss function, and train the adaptive weight model using historical datasets through the target loss function, and determine the bias coefficients. and sensitivity coefficient The optimal value.

[0036] In practice, the physical constraint coupling prediction model is constructed through the following steps:

[0037] ,in, This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. For bias terms, i, j All are indices of input variables and 1≤i <j≤n , n The number of stable variables. Let be the value of the i-th input variable in the stable variable set. Let i be the second-order interaction feature formed by the i-th input variable and the j-th input variable. r Working condition vector ， For working condition vector r The i Normalized weight coefficients of each input variable. For the first i The input variable and the first j Interaction weights of input variables.

[0038] In practice, the physical constraints are determined through the following steps:

[0039] Determine monotonicity constraints so that the prediction results satisfy... ,in, For partial differential equations, The variance of the head residual gas coefficient. To determine the non-uniformity of the combustion chamber outlet temperature distribution; to establish boundedness constraints so that the prediction results satisfy... Determine the energy / dilution conservation constraint to ensure the outlet temperature rise. In accordance with the heat of fuel reaction and the cooling / dilution heat capacity, wherein the outlet temperature rise is consistent with the heat of fuel reaction and the cooling / dilution heat capacity. These are theoretical parameters based on the heat of fuel reaction, the flow rate of cooling and dilution gases, and the specific heat generated.

[0040] In practice, the following steps are taken to construct a physical constraint coupling prediction model and then evaluate the physical constraint coupling prediction model:

[0041] Obtain historical test datasets containing multiple batches and various operating conditions; divide all operating conditions in the historical test datasets into... K A mutually exclusive set of operating conditions, through the aforementioned K Calculate the cross-condition generalization performance index for each mutually exclusive set of operating conditions. K Physical consistency verification is performed on mutually exclusive working condition groups to generate a first physical consistency verification result; a data subset containing at least two independent test batches is selected from the historical test dataset, and a cross-batch stability index is calculated using the data subset; physical consistency verification is then performed on the data subset to generate a second physical consistency verification result; the physical consistency error is calculated using the first physical consistency verification result and the second physical consistency verification result; the physical constraint coupling prediction model is evaluated based on the cross-working condition generalization performance index, the cross-batch stability index, and the physical consistency error.

[0042] This invention provides a combustion chamber outlet temperature prediction method that combines a driving mechanism, variable selection, and adaptive weighting, and incorporates physical constraints. This method can achieve high-precision prediction of the non-uniformity of combustion chamber outlet temperature distribution under different pressures, temperatures, and equivalence ratios. By introducing an adaptive weight selection mechanism and a physical consistency constraint model, this invention overcomes the limitations of traditional single empirical fitting and black-box data-driven methods. While maintaining physical rationality, it improves prediction accuracy and cross-condition applicability, and reduces reliance on large-scale bench tests and high-cost numerical simulations while meeting design and verification requirements.

[0043] like Figure 2 As shown, in one embodiment of the present invention, the outlet temperature prediction method generally includes:

[0044] Step 1: Data preparation and parameter acquisition.

[0045] First, parameters and operating conditions are collected and input, including residual gas coefficient variance, average droplet size, spray cone angle, flow organization parameters (such as main swirl number or equivalent mixing intensity coefficient), dilution air distribution coefficient (i.e., the proportion of dilution orifice flow rate to total intake air), inlet distortion index, wall equivalent cooling ratio, and operating condition vector.

[0046] This embodiment relies on the conventional structure and operating parameters of the combustion chamber during the design phase, and uses a limited sample from a bench database for model training. Required input parameters include: residual gas coefficient variance, average droplet size, spray cone angle, flow organization parameters (such as main swirl number or mixing intensity coefficient), dilution air distribution coefficient, inlet distortion index, and wall equivalent cooling ratio. All of these parameters can be obtained through design drawings, nozzle / guide component characteristics, empirical correlation, or calculations using existing databases, without the need for additional experimental measurements. Operating condition vector The operating conditions are directly determined by the engine design conditions or the operating conditions set by the control system. The output index is the combustion chamber OTDF predicted value, and the operating condition vector. Used to characterize pressure P 3. Temperature T 3. Equivalent ratio Reynolds number Injection momentum flow rate density ratio These are the influencing factors. All the parameters mentioned above are input quantities available during the design phase or under known operating conditions, and do not rely on bench measurements or high-fidelity numerical simulation results.

[0047] Step 2: Variable selection.

[0048] The data then proceeds to preprocessing and variable selection. The input parameters are normalized, and stable selection, group sparse regression, or sensitivity analysis methods are used to select variables with stable contributions.

[0049] All samples were preprocessed using normalization. Stability selection, group sparse regression, or sensitivity analysis methods were used to screen out parameters and interaction terms that contributed stably under multiple operating conditions, thus avoiding overfitting.

[0050] Specifically, a dynamic, data-driven weighting intermediary layer is introduced to decouple and map fixed operating condition inputs with varying parameter importance.

[0051] Using an exponential weighting function This ensures that the weights are always positive, conforming to the physical intuition that "importance" should be positive. Its internal linear combination of exponential forms captures the nonlinear influence of operating conditions on parameter importance. Regarding the exponential weighting function... Normalization is performed to obtain normalized weights. The normalization method is not limited to a specific mathematical expression and is used to make the weights have a comparable scale and can be directly used to predict the condition dependence coefficients of the model. For the working condition vector, i For the index of the input variable, The first in the working condition vector j Each working condition component j For the index of the operating condition components, Let the dimension be the working condition vector. The original weight of the i-th input variable in the stable parameter set under the operating condition vector r. Let be the bias coefficient of the i-th variable. is the sensitivity coefficient of the i-th input variable to the j-th operating condition component.

[0052] By dynamically adjusting weights, the same model can automatically adapt to various extreme operating conditions, from slow ground operation to high-powered high-altitude operation, significantly improving the model's prediction accuracy and generalization ability under untrained conditions. This changes the traditional model that relies on expert experience to manually set parameter weights, achieving fully data-driven and automated weight allocation, avoiding subjectivity, and discovering complex correlations. The adaptive weight model is trained on massive amounts of multi-condition data and is insensitive to noise and fluctuations in individual data points; therefore, compared to fixed-weight models, its prediction results are more stable and reliable. The output weights reveal which design parameters (such as cooling ratio and swirl number) play a dominant role in performance under specific operating conditions, providing engineers with direct and quantitative decision-making basis for optimizing designs.

[0053] Specifically, the extended feature set is filtered using a stable selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set. This ensures that the generation of the stable parameter set does not rely on a single filtering algorithm, but rather utilizes three filtering methods with different principles deployed in parallel.

[0054] Stable selection (resistance to randomness): By using Bootstrap resampling, the ability of parameters to survive small data perturbations is evaluated, and statistically stable parameters are selected.

[0055] Grouped Sparse Regression (Preserving Physical Structure): Grouped Lasso is used to group parameters under the same physical mechanism (such as spray characteristics) for screening. This avoids selecting features that are statistically correlated but physically repetitive or contradictory, ensuring the physical integrity of the screening results.

[0056] Global sensitivity analysis (quantification of contribution): Using the Sobol index, the contribution of each parameter to the uncertainty of the temperature field is directly quantified from the input-output relationship, and the truly influential parameters are screened out.

[0057] Finally, these parameters are generated through integration (such as intersection or voting). This further ensures that these parameters are not significant by chance under specific operating conditions, but are consistently important under the vast majority of operating conditions.

[0058] Constructing a predictive model using a selected "stable parameter set" ensures the robustness of the model's input features. This is a prerequisite and solid foundation for all subsequent high-performance predictive models (including adaptive weight models) to successfully generalize across different operating conditions. Among a large number of features (including interaction terms), it's easy to select features that have only a spurious correlation with the target variable or are only incidentally related on the training set. The stable parameter set selection method, through multi-faceted verification and stability requirements, effectively filters out these unstable "noise" parameters, greatly reducing the risk of model overfitting. The parameter set selected using this method is not only strong in terms of data but also physically reasonable. This ensures that the internal logic of the final predictive model is consistent with the basic principles of combustion science and fluid mechanics.

[0059] Step 3: Prediction model construction.

[0060] Next, adaptive weight modeling is performed, so that the parameter weights are dynamically adjusted according to the operating conditions.

[0061] A constrained multi-parameter prediction model is established, and its representative mathematical expression is as follows:

[0062] ,in, This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. For bias terms, i, j All are indices of input variables and 1≤i <j≤n , n The number of stable variables. Let be the value of the i-th input variable in the stable variable set. Let i be the second-order interaction feature formed by the i-th input variable and the j-th input variable. r Working condition vector ， For working condition vector r The i Normalized weight coefficients of each input variable. For the first i The input variable and the first j Interaction weights of the input variables, Based on and The first one, constructed and normalized in a similar manner i The input variable and the first j Interaction weights of input variables.

[0063] Embed the following physical constraints in the model: ① Monotonicity constraint: ,in, For partial differential equations, The variance of the head residual gas coefficient. ① Non-uniformity of combustion chamber outlet temperature distribution; ② Boundedness constraint: ensuring the prediction results satisfy the logit or sigmoid function. ③ Energy conservation constraint: The predicted outlet temperature rise is ensured through a penalty term in the loss function. It is consistent with the heat of fuel reaction, cooling and dilution effects. These are not measured inputs, but theoretical parameters derived from design parameters such as fuel reaction heat, cooling and dilution gas flow rates, and specific heat. These constraints ensure that the prediction results conform to fundamental physical laws based on mathematical regression.

[0064] Step 4: Model training and validation.

[0065] Then, physical constraint coupling prediction is introduced, in which physical constraints such as monotonicity, boundedness, and energy conservation are introduced.

[0066] During the training phase, the model is fitted using a limited database of historical test benches or simulation samples. Leave-one-out cross-validation is employed to test generalization performance, and the regularization coefficients and constraint weights are adjusted by comparing the model with the measured OTDF (Optical Time Function Depth) to balance prediction accuracy and physical consistency.

[0067] Step 5: Result Prediction and Output.

[0068] Finally, the results are output, obtaining the predicted OTDF value of the combustion chamber and the contribution ranking of each parameter under different operating conditions, thus achieving accurate performance evaluation under cross-operating conditions.

[0069] During the prediction phase, the model automatically calculates adaptive weights and outputs the corresponding OTDF predicted values ​​when given a set of parameters under any given operating condition, without requiring additional experimental temperature measurements or high-fidelity CFD calculations. It also outputs parameter contribution rankings and confidence intervals for design optimization, anomaly diagnosis, and sensitivity analysis. It can be embedded in combustion performance evaluation software or overall design support systems to achieve rapid decision support.

[0070] In this embodiment, a computer device is provided, such as... Figure 3 As shown, it includes a memory 301, a processor 302, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for predicting the outlet temperature of any of the aero-engine combustion chambers.

[0071] Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0072] In this embodiment, a computer-readable storage medium is provided, which stores a computer program that performs any of the above-described methods for predicting the outlet temperature of an aero-engine combustion chamber.

[0073] Specifically, computer-readable storage media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media do not include transient media, such as modulated data signals and carrier waves.

[0074] Based on the same inventive concept, this invention also provides a device for predicting the outlet temperature of an aero-engine combustor, as described in the following embodiments. Since the principle underlying the problem-solving of the device for predicting the outlet temperature of an aero-engine combustor is similar to that of the method for predicting the outlet temperature of an aero-engine combustor, the implementation of the device can refer to the implementation of the method for predicting the outlet temperature of an aero-engine combustor, and will not be repeated. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0075] Figure 4 This is a structural block diagram of a device for predicting the outlet temperature of an aero-engine combustion chamber according to an embodiment of the present invention, such as... Figure 4 As shown, it includes: a parameter acquisition module 401, a preprocessing and parameter filtering module 402, and OTDF The predicted value output module 403 is described below.

[0076] The parameter acquisition module 401 is used to acquire input parameters related to combustion chamber operation and combustion chamber structure in batches, generate an optimal parameter set, obtain combustion chamber operating conditions, and construct an operating condition vector through the combustion chamber operating conditions, wherein the operating condition vector is used to characterize the working state of the combustion chamber.

[0077] The preprocessing and parameter filtering module 402 is used to construct a stable parameter set that has a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions through the preferred parameter set and the operating condition vector, filter all parameters in the stable parameter set, and adjust the weights corresponding to the parameters in the operating condition vector through an adaptive weight model to generate a preprocessed sample set.

[0078] OTDF The prediction output module 403 is used to construct a physical constraint coupling prediction model, determine physical constraints, train the physical constraint coupling prediction model based on the physical constraints, generate a trained physical constraint coupling prediction model, input the preprocessed sample set into the trained physical constraint coupling prediction model, and output the non-uniformity of the combustion chamber outlet temperature distribution.

[0079] In one embodiment, the preprocessing and parameter filtering module includes:

[0080] A standardization processing unit is used to standardize all parameters in the preferred parameter set and the working condition vector to generate a standardized parameter set.

[0081] An extended feature construction unit is used to construct an extended feature set based on the standardized parameter set, wherein the extended feature set consists of the preferred parameter set, the operating condition vector, and interaction terms between each parameter in the preferred parameter set and each parameter in the operating condition vector;

[0082] The stable parameter set generation unit is used to filter the extended feature set through a stable selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set, wherein the variables in the stable parameter set have a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions.

[0083] An adaptive weighting unit is used to construct an adaptive weighting model, inputting the working condition vector into the adaptive weighting model and outputting the weights corresponding to the parameters in the working condition vector.

[0084] A preprocessed sample set generation unit is used to take the stable parameter set containing weights as the preprocessed sample set.

[0085] In one embodiment, the stable parameter set generation unit is further configured to: calculate the selection frequency of each variable in the extended feature set based on a stable selection method; filter out variables with selection frequencies higher than a first preset threshold to generate a first candidate set; group the variables in the extended feature set according to physical correlation; identify significant variable groups using a group sparse regression method; and save all variables in the significant variable groups to a second candidate set; construct a surrogate model of the combustion chamber outlet temperature field; calculate the total sensitivity index of each variable in the extended feature set based on the sensitivity analysis method using the surrogate model; filter out variables with total sensitivity indices higher than a second preset threshold to generate a third candidate set; generate an initial variable set using the first candidate set, the second candidate set, and the third candidate set by taking the intersection or weighted voting; and perform cross-condition consistency verification on each variable in the initial variable set to filter and generate the stable parameter set.

[0086] In one embodiment, the adaptive weighting unit is further configured to construct a condition-dependent exponential weighting function, and to construct an adaptive weighting model based on the exponential weighting function, wherein the exponential weighting function... ,in, For the working condition vector, i For the index of the input variable, The first in the working condition vector j Each working condition component j For the index of the operating condition components, Let the dimension be the working condition vector. The original weight of the i-th input variable in the stable parameter set under the operating condition vector r. Let be the bias coefficient of the i-th variable. Let be the sensitivity coefficient of the i-th input variable to the j-th working condition component; define a target loss function, and train the adaptive weight model using historical datasets through the target loss function, and determine the bias coefficients. and sensitivity coefficient The optimal value.

[0087] In one embodiment, OTDF The predicted value output module includes:

[0088] Predicted value output unit, used for ,in, This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. For bias terms, i, j All are indices of input variables and 1≤i <j ≤n , n The number of stable variables. Let be the value of the i-th input variable in the stable variable set. Let i be the second-order interaction feature formed by the i-th input variable and the j-th input variable. r Working condition vector ， For working condition vector r The i Normalized weight coefficients of each input variable. For the first i The input variable and the first j Interaction weights of input variables.

[0089] In one embodiment, OTDF The predicted value output module also includes:

[0090] Determine monotonicity constraint elements to define monotonicity constraints so that the prediction results satisfy... ,in, For partial differential equations, The variance of the head residual gas coefficient. This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet.

[0091] Define bounded constraint elements to determine bounded constraints so that the prediction results satisfy... ;

[0092] Determine the energy / dilution conservation constraint unit to ensure the outlet temperature rise. In accordance with the heat of fuel reaction and the cooling / dilution heat capacity, wherein the outlet temperature rise is consistent with the heat of fuel reaction and the cooling / dilution heat capacity. These are theoretical parameters based on the heat of fuel reaction, the flow rate of cooling and dilution gases, and the specific heat generated.

[0093] In one embodiment, the above-described apparatus further includes an evaluation and verification module.

[0094] In one embodiment, the evaluation verification module includes:

[0095] The historical test dataset acquisition unit is used to acquire historical test datasets containing multiple batches and various working conditions;

[0096] The first dataset grouping unit is used to divide all working conditions in the historical test dataset into... K A mutually exclusive set of operating conditions, through the aforementioned K Calculate the cross-condition generalization performance index for each mutually exclusive set of operating conditions. K Perform physical consistency verification on each mutually exclusive working condition group to generate the first physical consistency verification result;

[0097] The second dataset grouping unit is used to select a data subset containing at least two independent test batches from the historical test dataset, calculate the cross-batch stability index through the data subset, perform physical consistency verification on the data subset, and generate a second physical consistency verification result.

[0098] The consistency error calculation unit is used to calculate the physical consistency error using the first physical consistency verification result and the second physical consistency verification result.

[0099] An evaluation unit is used to evaluate the physical constraint coupled prediction model based on the cross-condition generalization performance index, the cross-batch stability index, and the physical consistency error.

[0100] The embodiments of the present invention achieve the following technical effects:

[0101] Unlike single-variable fitting, this invention achieves different results through multi-parameter driving and automatic screening skills. By identifying key factors through variable screening, it avoids model complexity and overfitting problems. An adaptive weighting mechanism constructed through stable selection, group sparse regression, or sensitivity analysis dynamically adjusts parameter weights under different operating conditions, achieving cross-condition adaptation and avoiding the limitations of manual settings. By introducing constraints such as monotonicity, boundedness, and energy conservation, the prediction results conform to physical laws (i.e., maintaining consistency of physical constraints), improving reliability. Trained on multiple operating conditions and batches of data, the prediction model can be applied to different pressures, equivalence ratios, and fuel conditions, enhancing extrapolation capabilities. The parameters required for the outlet temperature prediction method are all conventionally measurable or design-obtainable, with low computational overhead, making it suitable for widespread application in the combustion chamber design stage and experimental verification process.

[0102] Obviously, those skilled in the art should understand that the modules or steps of the above-described embodiments of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any particular hardware and software combination.

[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the outlet temperature of an aero-engine combustion chamber, characterized in that, include: Input parameters related to combustion chamber operation and structure are collected in batches to generate an optimal parameter set, combustion chamber operating conditions are obtained, and an operating condition vector is constructed based on the combustion chamber operating conditions. The operating condition vector is used to characterize the working state of the combustion chamber. All parameters in the preferred parameter set and the operating condition vector are standardized to generate a standardized parameter set. Based on the standardized parameter set, an extended feature set is constructed, wherein the extended feature set consists of the preferred parameter set, the operating condition vector, and interaction terms between each parameter in the preferred parameter set and each parameter in the operating condition vector. The extended feature set is then filtered using a stability selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set, wherein the variables in the stable parameter set have a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions. An adaptive weighting model is constructed, and the operating condition vector is input into the adaptive weighting model to output the weights corresponding to the parameters in the operating condition vector. The stable parameter set containing the weights is used as a preprocessed sample set. Constructing an adaptive weight model includes: A condition-dependent exponential weighting function is constructed, and an adaptive weighting model is built based on the exponential weighting function, wherein the exponential weighting function... ,in, For the working condition vector, i For the index of the input variable, The first in the working condition vector j Each working condition component j For the index of the operating condition components, Let the dimension be the working condition vector. The original weight of the i-th input variable in the stable parameter set under the operating condition vector r. Let be the bias coefficient of the i-th variable. Let be the sensitivity coefficient of the i-th input variable to the j-th working condition component; define a target loss function, and train the adaptive weight model using historical datasets through the target loss function, and determine the bias coefficients. and sensitivity coefficient The optimal value; A physical constraint coupling prediction model is constructed, physical constraints are determined, and the physical constraint coupling prediction model is trained based on the physical constraints to generate a trained physical constraint coupling prediction model. The preprocessed sample set is input into the trained physical constraint coupling prediction model, and the non-uniformity of the combustion chamber outlet temperature distribution is output. Constructing a physical constraint coupled prediction model includes: ,in, This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. For bias terms, i, j All are indices of input variables and 1≤i <j≤n , n The number of stable variables. Let be the value of the i-th input variable in the stable variable set. Let i be the second-order interaction feature formed by the i-th input variable and the j-th input variable. r Working condition vector ， For working condition vector r The i Normalized weight coefficients of each input variable. For the first i The input variable and the first j Interaction weights of input variables.

2. The method for predicting the outlet temperature of an aero-engine combustion chamber as described in claim 1, characterized in that, The extended feature set is filtered using a stability selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set, including: Based on the stable selection method, the selection frequency of each variable in the extended feature set is calculated, and variables with selection frequencies higher than a first preset threshold are selected to generate a first candidate set. The variables in the extended feature set are grouped according to their physical correlation. The significant variable groups are identified by the group sparse regression method, and all variables in the significant variable groups are saved to the second candidate set. A proxy model of the combustion chamber outlet temperature field is constructed. Based on the sensitivity analysis method, the total sensitivity index of each variable in the extended feature set is calculated using the proxy model. Variables with a total sensitivity index higher than a second preset threshold are selected to generate a third candidate set. An initial variable set is generated by taking the intersection or weighted voting from the first candidate set, the second candidate set, and the third candidate set; For each variable in the initial variable set, cross-condition consistency verification is performed, and the stable parameter set is generated by filtering.

3. The method for predicting the outlet temperature of an aero-engine combustion chamber as described in claim 1, characterized in that, Determine physical constraints, including: Determine monotonicity constraints so that the prediction results satisfy... ,in, For partial differential equations, The variance of the head residual gas coefficient. This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. Determine boundedness constraints to ensure that the prediction results satisfy... ; Determine the energy / dilution conservation constraint to ensure the outlet temperature rise. In accordance with the heat of fuel reaction and the cooling / dilution heat capacity, wherein the outlet temperature rise is consistent with the heat of fuel reaction and the cooling / dilution heat capacity. These are theoretical parameters based on the heat of fuel reaction, the flow rate of cooling and dilution gases, and the specific heat generated.

4. The method for predicting the outlet temperature of an aero-engine combustion chamber as described in any one of claims 1 to 3, characterized in that, Also includes: After constructing the physical constraint coupling prediction model, the physical constraint coupling prediction model is evaluated, including: Obtain historical test datasets containing multiple batches and various operating conditions; The historical test dataset is divided into all working conditions. K A mutually exclusive set of operating conditions, through the aforementioned K Calculate the cross-condition generalization performance index for each mutually exclusive set of operating conditions. K Perform physical consistency verification on each mutually exclusive working condition group to generate the first physical consistency verification result; Select a data subset containing at least two independent test batches from the historical test dataset, calculate the cross-batch stability index using the data subset, perform physical consistency verification on the data subset, and generate a second physical consistency verification result; The physical consistency error is calculated using the first physical consistency verification result and the second physical consistency verification result. The physical constraint coupled prediction model is evaluated based on the cross-condition generalization performance index, the cross-batch stability index, and the physical consistency error.

5. A device for predicting the outlet temperature of an aero-engine combustion chamber, characterized in that, include: The parameter acquisition module is used to collect input parameters related to combustion chamber operation and combustion chamber structure in batches, generate an optimal parameter set, obtain combustion chamber operating conditions, and construct an operating condition vector through the combustion chamber operating conditions, wherein the operating condition vector is used to characterize the working state of the combustion chamber; The preprocessing and parameter filtering module is used to standardize all parameters in the preferred parameter set and the operating condition vector to generate a standardized parameter set. Based on the standardized parameter set, an extended feature set is constructed, wherein the extended feature set consists of the preferred parameter set, the operating condition vector, and interaction terms between each parameter in the preferred parameter set and each parameter in the operating condition vector. The extended feature set is filtered using a stability selection method, a group sparse regression method, and a sensitivity analysis method to generate a stable parameter set, wherein the variables in the stable parameter set have a stable contribution to the prediction of the combustion chamber outlet temperature field under different operating conditions. An adaptive weighting model is constructed, the operating condition vector is input into the adaptive weighting model, and the weights corresponding to the parameters in the operating condition vector are output. The stable parameter set containing the weights is used as the preprocessed sample set. The preprocessing and parameter filtering module is also used to construct a working condition-dependent exponential weighting function, and to construct an adaptive weighting model based on the exponential weighting function, wherein the exponential weighting function... ,in, For the working condition vector, i For the index of the input variable, The first in the working condition vector j Each working condition component j For the index of the operating condition components, Let the dimension be the working condition vector. The original weight of the i-th input variable in the stable parameter set under the operating condition vector r. Let be the bias coefficient of the i-th variable. Let be the sensitivity coefficient of the i-th input variable to the j-th working condition component; define a target loss function, and train the adaptive weight model using historical datasets through the target loss function, and determine the bias coefficients. and sensitivity coefficient The optimal value; OTDF The prediction output module is used to construct a physical constraint coupling prediction model, determine physical constraints, train the physical constraint coupling prediction model based on the physical constraints, generate a trained physical constraint coupling prediction model, input the preprocessed sample set into the trained physical constraint coupling prediction model, and output the non-uniformity of the combustion chamber outlet temperature distribution. OTDF The predicted value output module includes: Predicted value output unit, used for ,in, This refers to the non-uniformity of the temperature distribution at the combustion chamber outlet. For bias terms, i, j All are indices of input variables and 1≤i <j≤n , n The number of stable variables. Let be the value of the i-th input variable in the stable variable set. Let i be the second-order interaction feature formed by the i-th input variable and the j-th input variable. r Working condition vector ， For working condition vector r The i Normalized weight coefficients of each input variable. For the first i The input variable and the first j Interaction weights of input variables.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method for predicting the outlet temperature of the aero-engine combustion chamber as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that performs the method for predicting the outlet temperature of the aero-engine combustion chamber according to any one of claims 1 to 4.

Images