Lumped temperature dynamics modeling methods, devices, equipment, and storage media

Abstract

This application relates to the field of data-driven modeling of wind tunnel temperature systems, and particularly to a lumped temperature dynamics modeling method, apparatus, equipment, and storage medium. It involves repeatedly acquiring a sequence of sampled data from a continuous wind tunnel device at a preset sampling frequency within a single test cycle. The sampled data includes temperature state variables, input variables, and operating condition variables. The sampled data sequence is then subjected to manifold-consistent denoising to obtain a denoised data sequence. Adaptive Gaussian kernel state derivatives are calculated for the temperature state variables in the denoised data sequence to obtain a sequence of temperature state variable derivatives. Based on multiple denoised data sequences and their corresponding temperature state variable derivative sequences, multi-condition incremental regularized regression ensemble modeling is performed to obtain a lumped temperature dynamics model for the continuous wind tunnel. This application forms a complete technology chain from data denoising and state derivative calculation to adaptive operating condition model construction, achieving accurate modeling of continuous wind tunnel temperature systems.

Description

Technical Field

[0001] This application relates to the field of data-driven modeling of wind tunnel temperature systems, and includes, but is not limited to, a lumped temperature dynamics modeling method, apparatus, device, and storage medium. Background Technology

[0002] Continuous wind tunnels are essential ground-based testing equipment for fundamental aerodynamic research, aircraft and automotive engineering development, and aerodynamic shape optimization of high-performance fluid equipment. During continuous wind tunnel operation, airflow temperature is one of the core parameters determining flow field quality. Temperature not only affects gas density and viscosity, thus altering key aerodynamic parameters such as Reynolds number and Mach number, but its fluctuations also cause thermal expansion and contraction of the test section structure, leading to decreased measurement system stability and affecting test repeatability. To improve the stability of test section temperature, various control methods have been developed in engineering practice, including PID control, model predictive control, feedforward control, model reference adaptive control, and fuzzy control, achieving good engineering application results. Building upon this foundation, establishing an accurate wind tunnel temperature dynamics model suitable for predictive analysis would provide more reliable theoretical support for control law design, accelerate parameter tuning, expand the applicable operating conditions of control strategies, and provide a higher-quality predictive basis for advanced methods such as model predictive control that rely on explicit dynamic models.

[0003] The temperature system of a continuous wind tunnel is characterized by multiple variables, strong coupling, and large thermal inertia. The increase in airflow temperature mainly stems from the work done by the compressor, while the decrease depends on the heat dissipation process of the airflow in the heat exchanger, the heat dissipation capacity of which is primarily determined by the cooling water flow rate. Simultaneously, the dynamic evolution of temperature is also influenced by a combination of factors, including ambient temperature and cooling water inlet temperature. Further increasing its complexity is that wind tunnel tests are typically conducted in discrete, complete test cycles. Different test objects have different requirements for the target Mach number, and their Mach number enhancement strategies and adjustment processes also vary, thus giving wind tunnels the characteristics of multiple operating conditions in actual operation.

[0004] Existing continuous wind tunnel temperature dynamics modeling methods still face certain limitations in practical applications. On the one hand, the mechanistic modeling of continuous wind tunnel temperature systems is usually based on the fundamental theories of fluid mechanics and heat transfer, constructing control equations describing airflow and heat transfer processes based on the conservation of mass, momentum, and energy. Related research mainly focuses on mechanistic analysis and numerical simulation, typically based on idealized assumptions. While these methods can reflect the basic thermodynamic characteristics of the system, their accuracy and stability are difficult to guarantee under actual operating conditions characterized by strong multivariate coupling, frequent switching of operating conditions, and significant external disturbances. Furthermore, these models are highly dependent on parameter accuracy, while in engineering, differences in equipment characteristics, measurement errors, and time delays are common, making them unsuitable for direct real-time prediction and control. On the other hand, data-driven black-box models, such as neural network models and support vector regression, while possessing strong nonlinear fitting capabilities, are highly dependent on high-quality data. When samples are insufficient or operating conditions vary significantly, they are prone to insufficient generalization ability and decreased stability, and lack explicit physical constraints, making it difficult to explain complex thermal process mechanisms. The combined effects of high thermal inertia and significant hysteresis amplify the aforementioned problems, limiting the model's effectiveness in engineering applications for high-precision temperature modeling and control.

[0005] At the level of specific modeling implementation, the above methods still face several key challenges: the noise processing of the collected data lacks a systematic and automated mechanism, usually relying on experience to select wavelet denoising, Gaussian filtering, or frequency domain filtering methods. The parameter selection is highly subjective, which can easily lead to noise residue or loss of effective dynamic information, thus affecting the model identification accuracy; most existing methods are based on discrete modeling. Even if dynamic characteristics are extracted through numerical difference, the difference process will further amplify the noise if the data is not sufficiently denoised, reducing the reliability of the model in the dynamic stage; most studies are limited to a single operating condition or steady-state operating point, making it difficult to cover the multi-condition characteristics of the entire continuous wind tunnel operation, resulting in insufficient model generalization ability; in addition, some methods model each temperature state variable independently, without fully considering the thermal coupling relationship between the compressor, heat exchanger, and pipeline, and ignoring the physical constraints between variables within the system, making it difficult for the model to accurately characterize the overall dynamic evolution process.

[0006] Based on the above analysis, it is necessary to construct a unified modeling framework that balances physical consistency and computational efficiency between mechanistic modeling and data-driven methods. Building upon this, it is essential to develop temperature dynamics modeling methods capable of automated noise reduction of raw data, stable acquisition of state derivatives, adaptation to multiple operating conditions, and explicit characterization of internal coupling relationships within the system. This will enhance the robustness and generalization ability of the model, providing a reliable model foundation for high-precision temperature prediction and control in continuous wind tunnels. Summary of the Invention

[0007] In view of this, the lumped temperature dynamics modeling method, apparatus, equipment, and storage medium provided in the embodiments of this application can realize high-precision temperature modeling of continuous wind tunnels.

[0008] The lumped temperature dynamics modeling method, apparatus, device, and storage medium provided in this application are implemented as follows: One aspect of this application provides a lumped temperature dynamics modeling method, the method comprising: The sampling data sequence is obtained by repeatedly acquiring the state sampling data of the continuous wind tunnel equipment at a preset sampling frequency within a test cycle. The sampling data sequence includes multiple types of sampling data, and each of the sampling data includes temperature state variables, input variables, and operating condition variables. Each acquired sampled data sequence is subjected to manifold-consistent denoising to obtain the corresponding denoised data sequence; By performing adaptive Gaussian kernel state derivative calculation on the temperature state variables in the noise-reduced data sequence, a temperature state variable derivative sequence is obtained. Based on multiple denoised data sequences and corresponding temperature state variable derivative sequences, multi-condition incremental regular regression ensemble modeling is performed to obtain a continuous wind tunnel lumped temperature dynamics model.

[0009] In one possible implementation, the step of performing manifold-consistent denoising on each acquired sampled data sequence to obtain a corresponding denoised data sequence includes: Each acquired sampling data sequence is input into the denoising processing model for manifold-consistent denoising calculation, and the corresponding denoised data sequence is output.

[0010] In one possible implementation, obtaining the temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the denoised data sequence includes: Obtain the temperature state variable sequence in the noise-reduced data sequence, and the temperature state variable sequence in the sampled data sequence; Calculate the corresponding Gaussian kernel width parameter based on each temperature state variable sequence in the sampled data sequence and each temperature state variable sequence in the denoised data sequence; Based on the Gaussian kernel width parameter, the numerical derivative of each temperature state variable sequence is obtained by performing a Gaussian kernel differentiator.

[0011] In one possible implementation, the multi-condition incremental regularized regression ensemble modeling based on multiple denoised data sequences and corresponding temperature state variable derivative sequences yields a continuous wind tunnel lumped temperature dynamics model, including: By concatenating the operating condition variable sequences from each of the noise-reduced data sequences, a global operating condition variable dataset is obtained. The low-dimensional dominant direction of each type of operating condition variable is extracted from the global operating condition variable dataset to obtain the dominant direction matrix; Based on each of the noise reduction data sequences, the attention weights of the included working condition variables relative to each dominant direction are calculated to obtain the corresponding enhancement feature matrix; Based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix, the parameters of the incremental regular regression model are solved to obtain the continuous wind tunnel lumped temperature dynamics model.

[0012] In one possible implementation, the step of inputting each acquired sampling data sequence into a denoising processing model for denoising calculation and outputting a corresponding denoised data sequence includes: Calculate the graph Laplacian matrix corresponding to each variable sequence in the sampled data sequence; The variable sequence is input into the noise reduction processing model, and the corresponding predicted noise reduction variable sequence is output. The data loss is determined based on the variable sequence and the corresponding predicted noise reduction variable sequence. The manifold regularization loss is determined based on the predicted denoising variable sequence and the graph Laplacian matrix. The flow mapping loss is determined based on each predicted denoising variable in the predicted denoising variable sequence. The parameters of the denoising processing model are adjusted according to the data loss, the manifold regularization loss, and the flow mapping loss until convergence, and the current predicted denoising variable sequence is determined to be the denoising variable sequence corresponding to the variable sequence.

[0013] In one possible implementation, the step of solving the incremental regular regression model parameters based on the derivative sequence of the temperature state variables corresponding to each of the denoised data sequences and the enhanced feature matrix to obtain a continuous wind tunnel lumped temperature dynamics model includes: The first and second model parameters are initialized based on the preset regularization coefficients, identity matrix, and zero matrix. The first model parameters and the second model parameters are iteratively updated and solved based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix. The model parameter solution process ends when the preset conditions are met, and the continuous wind tunnel lumped temperature dynamics model is determined based on the current first model parameters and second model parameters.

[0014] In one possible implementation, the method further includes: In response to obtaining the input variables, operating condition variables, and initial temperature state variables of the continuous wind tunnel equipment, the predicted temperature state variables are obtained by performing multi-step recursive prediction based on the input variables and operating condition variables through the continuous wind tunnel lumped temperature dynamics model.

[0015] Another aspect of the embodiments of this application provides a lumped temperature dynamics modeling device, the device comprising: The data acquisition module is used to acquire multiple sampling data sequences obtained by sampling the state of the continuous wind tunnel equipment at a preset sampling frequency within a test cycle. The sampling data sequence includes multiple types of sampling data, and each of the sampling data includes temperature state variables, input variables, and operating condition variables. The noise reduction module is used to perform manifold-consistent noise reduction on each acquired sampled data sequence to obtain the corresponding noise-reduced data sequence. The derivative calculation module is used to obtain the temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the noise-reduced data sequence; The wind tunnel modeling module is used to perform multi-condition incremental regular regression integrated modeling based on multiple denoised data sequences and corresponding temperature state variable derivative sequences, to obtain a continuous wind tunnel lumped temperature dynamics model.

[0016] In one possible implementation, the noise reduction processing module is further configured to: Each acquired sampling data sequence is input into the denoising processing model for manifold-consistent denoising calculation, and the corresponding denoised data sequence is output.

[0017] In one possible implementation, the derivative calculation module is further configured to: Obtain the temperature state variable sequence in the noise-reduced data sequence, and the temperature state variable sequence in the sampled data sequence; Calculate the corresponding Gaussian kernel width parameter based on each temperature state variable sequence in the sampled data sequence and each temperature state variable sequence in the denoised data sequence; Based on the Gaussian kernel width parameter, the numerical derivative of each temperature state variable sequence is obtained by performing a Gaussian kernel differentiator.

[0018] In one possible implementation, the wind tunnel modeling module is further used for: By concatenating the operating condition variable sequences from each of the noise-reduced data sequences, a global operating condition variable dataset is obtained. The low-dimensional dominant direction of each type of operating condition variable is extracted from the global operating condition variable dataset to obtain the dominant direction matrix; Based on each of the noise reduction data sequences, the attention weights of the included working condition variables relative to each dominant direction are calculated to obtain the corresponding enhancement feature matrix; Based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix, the parameters of the incremental regular regression model are solved to obtain the continuous wind tunnel lumped temperature dynamics model.

[0019] In one possible implementation, the noise reduction processing module is further configured to: Calculate the graph Laplacian matrix corresponding to each variable sequence in the sampled data sequence; The variable sequence is input into the noise reduction processing model, and the corresponding predicted noise reduction variable sequence is output. The data loss is determined based on the variable sequence and the corresponding predicted noise reduction variable sequence. The manifold regularization loss is determined based on the predicted denoising variable sequence and the graph Laplacian matrix. The flow mapping loss is determined based on each predicted denoising variable in the predicted denoising variable sequence. The parameters of the denoising processing model are adjusted according to the data loss, the manifold regularization loss, and the flow mapping loss until convergence, and the current predicted denoising variable sequence is determined to be the denoising variable sequence corresponding to the variable sequence.

[0020] In one possible implementation, the wind tunnel modeling module is further used for: The first and second model parameters are initialized based on the preset regularization coefficients, identity matrix, and zero matrix. The first model parameters and the second model parameters are iteratively updated and solved based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix. The model parameter solution process ends when the preset conditions are met, and the continuous wind tunnel lumped temperature dynamics model is determined based on the current first model parameters and second model parameters.

[0021] In one possible implementation, the device further includes: The temperature prediction module is used to respond to the input variables, operating condition variables and initial temperature state variables of the continuous wind tunnel equipment, and to perform multi-step recursive prediction based on the input variables and operating condition variables through the continuous wind tunnel lumped temperature dynamics model to obtain the predicted temperature state variables.

[0022] The electronic device provided in this application includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the method described in this application.

[0023] The computer-readable storage medium provided in this application embodiment stores a computer program thereon, which, when executed by a processor, implements the method provided in this application embodiment.

[0024] The lumped temperature dynamics modeling method, apparatus, device, and storage medium provided in this application acquire multiple sampling data sequences of a continuous wind tunnel device at a preset sampling frequency within a test cycle. These sampling data sequences include multiple types of data, each containing temperature state variables, input variables, and operating condition variables. Each acquired sampling data sequence undergoes manifold-consistent denoising processing to obtain a corresponding denoised data sequence. Adaptive Gaussian kernel state derivative calculation is performed on the temperature state variables in the denoised data sequence to obtain a temperature state variable derivative sequence. Multi-condition incremental regularized regression ensemble modeling is then performed based on multiple denoised data sequences and their corresponding temperature state variable derivative sequences to obtain a lumped temperature dynamics model for the continuous wind tunnel. This application forms a complete technology chain from data denoising and state derivative calculation to adaptive operating condition model construction, achieving accurate modeling of the continuous wind tunnel temperature system. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in 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.

[0026] Figure 1 A flowchart illustrating a lumped temperature dynamics modeling method according to an embodiment of this application is shown; Figure 2 This diagram illustrates a comparison of sampling data and noise reduction data according to an embodiment of this application. Figure 3 A schematic diagram illustrating the derivative of a temperature state variable according to an embodiment of this application is shown; Figure 4 A schematic diagram showing a comparison between a predicted temperature state variable and an actual temperature state variable according to an embodiment of this application is provided. Figure 5 A schematic diagram of a lumped temperature dynamics modeling apparatus according to an embodiment of this application is shown; Figure 6 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0029] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0030] It should be noted that the terms "first, second, third" used in the embodiments of this application are used to distinguish similar or different objects and do not represent a specific order of objects. It can be understood that "first, second, third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0031] The lumped temperature dynamics modeling method of this application embodiment can be executed by any electronic device, including but not limited to mobile phones, wearable devices (such as smartwatches, smart bracelets, smart glasses, etc.), tablet computers, laptops, in-vehicle terminals, PCs (Personal Computers), etc. The functions implemented by this method can be achieved by a processor in the electronic device calling program code. Of course, the program code can be stored in a computer storage medium. Therefore, the electronic device includes at least a processor and a storage medium.

[0032] The lumped temperature dynamics modeling method of this application can be used in any application scenario that requires prediction of wind tunnel temperature parameters, such as the application scenario of precise temperature control during wind tunnel testing, the application scenario of simulation verification during the wind tunnel design stage, and the application scenario of wind tunnel health management and fault diagnosis.

[0033] The lumped temperature dynamics modeling scheme of this application embodiment will be described in detail below with reference to the accompanying drawings.

[0034] Figure 1 A flowchart illustrating a lumped temperature dynamics modeling method according to an embodiment of this application is shown. Figure 1 As shown, the lumped temperature dynamics modeling method of this application embodiment may include the following steps S10-S40.

[0035] For ease of description, the lumped temperature dynamics modeling method of this application embodiment is described using an electronic device as the execution subject. It should be understood that the execution subject of this application embodiment can also be a processor or chip in an electronic device, and this application embodiment does not impose any limitations.

[0036] Step S10: Acquire the sampling data sequence obtained by performing state sampling on the continuous wind tunnel equipment at a preset sampling frequency within one test cycle multiple times.

[0037] In one possible implementation, when lumped temperature dynamics modeling of the continuous wind tunnel equipment is required, the electronic device can sample state data during multiple tests to obtain corresponding sampled data sequences. For example, the time length and sampling frequency of each sampled data sequence acquisition can be the same, allowing the electronic device to acquire multiple sampled data sequences obtained by sampling the state of the continuous wind tunnel equipment at a preset sampling frequency within a single test cycle.

[0038] Optionally, the temperature evolution in a continuous wind tunnel is essentially a multi-node coupled thermodynamic process driven by the heat generated by the compressor, the active heat dissipation of the heat exchanger, and the passive disturbance of environmental conditions. Based on this, according to the physical path of energy transfer, the embodiments of this application determine that the required sampling data for each sampling process includes temperature state variables, input variables, and operating condition variables, which are used to perform lumped temperature dynamics modeling based on the above variables.

[0039] In some embodiments, the temperature state variables can be selected from five measurement points: compressor inlet temperature (Comp_In_T), compressor outlet temperature (Comp_Out_T), heat exchanger inlet temperature (HX_In_T), heat exchanger outlet temperature (HX_Out_T), and test section temperature (Test_T). These five measurement points are distributed sequentially along the airflow direction, covering the complete heat transfer chain from compressor power generation and temperature rise, airflow heat transfer, heat exchanger cooling, to airflow stabilization in the test section. Significant thermodynamic coupling exists between the temperatures at each measurement point: temperature changes at upstream nodes directly affect the thermal state of downstream nodes through airflow convection, while the downstream back pressure effect and changes in thermal boundary conditions also feed back to upstream nodes. Therefore, describing the temperature distribution state of the entire wind tunnel airflow path with a five-dimensional state vector can characterize the thermodynamic dynamics of the system.

[0040] Input variables may include cooling water supply rate (HX_Q), measured in cubic meters per second. The heat exchanger is the actuator in the wind tunnel system with active temperature control capability. The cooling water supply rate directly determines the heat exchanger's heat exchange power, thereby affecting the dynamic response of the heat exchanger's outlet temperature, and indirectly affecting the overall loop temperature distribution through airflow circulation.

[0041] Operating condition variables can include compressor power (Comp_W), measured in megawatts, ambient temperature (Temp_T), and cooling water supply temperature (HX_Sup_T). Compressor power determines the energy input the airflow receives from the compressor and is the dominant factor in the system's heat source; its fluctuations are the primary source of dynamic temperature changes. Ambient temperature affects the overall heat dissipation boundary conditions through heat exchange along the wind tunnel duct walls. Cooling water supply temperature directly determines the heat sink capacity on the cold side of the heat exchanger, significantly impacting heat exchange efficiency. Incorporating these three quantities into the model's operating condition variables effectively improves the prediction accuracy and generalization ability of the model under different operating conditions.

[0042] Step S20: Perform manifold consistency denoising on each acquired sampled data sequence to obtain the corresponding denoised data sequence.

[0043] In one possible implementation, the electronic device performs manifold-consistent denoising by inputting each acquired sampled data sequence into a denoising model for manifold-consistent denoising calculation and outputting the corresponding denoised data sequence. Specifically, the electronic device inputs each variable sequence from the sampled data sequence into the denoising model for manifold-consistent denoising calculation and outputs the corresponding denoised variable sequence.

[0044] Optionally, the denoising processing model in this application embodiment may include two parts: a data reconstruction network and a stream mapping network. In the process of denoising any variable sequence in the sampled data sequence, the electronic device may first construct the data reconstruction network and the stream mapping network respectively, and then perform joint training on the data reconstruction network and the stream mapping network to achieve variable sequence denoising.

[0045] Optionally, the denoising process may include calculating the graph Laplacian matrix corresponding to each variable sequence in the sampled data sequence. The variable sequence is input into the denoising model, and the corresponding predicted denoised variable sequence is output. The data loss is determined based on the variable sequence and the corresponding predicted denoised variable sequence. The manifold regularization loss is determined based on the predicted denoised variable sequence and the graph Laplacian matrix. The flow mapping loss is determined based on each predicted denoised variable in the predicted denoised variable sequence. The parameters of the denoising model are adjusted based on the data loss, manifold regularization loss, and flow mapping loss until convergence, and the current predicted denoised variable sequence is determined as the denoised variable sequence corresponding to the variable sequence.

[0046] In some embodiments, the data reconstruction network can be constructed using time series... Input data, with noise reduction based on compressor inlet temperature. Fully connected neural network for output The data is used to reconstruct the network model: The number of time series sampling points is: The neural network contains There are 1 hidden layer, each containing 1 hidden layer. To enhance the data reconstruction network's ability to represent complex dynamic systems and overcome the limitations of traditional activation functions in characterizing multi-scale features and nonlinear behavior, this application proposes an adaptive dynamic activation function. This activation function, by embedding learnable parameters, explicitly decouples and combines the long-term trend characteristics, multi-scale oscillation characteristics, and local nonlinear mutation characteristics of the fitting dynamic process, thereby enhancing the neural network's ability to approximate complex dynamics and its robustness against noise.

[0047] Optionally, adaptive dynamic activation function Defined as a weighted sum of three learnable components, its expression can be: ,in, , , Trend items Oscillation term and nonlinear correction terms The globally learnable weight coefficients. To fit the monotonically increasing or decreasing behavior of a dynamic system, a hyperbolic tangent function with a learnable scaling factor is used to prevent gradient vanishing and limit the output range. ,in, The trend scaling factor is a learnable parameter that is automatically optimized during network training and is used to adaptively adjust the steepness of trend changes. To capture periodic motions at multiple frequencies present in the system, unlike traditional sinusoidal activation functions that use a fixed frequency, this step constructs a superposition model of sinusoidal basis functions whose frequency and phase can both be learned: ,in, The preset number of harmonic components is set to [value]. ; For the first The learnable angular frequencies of each component are initialized to cover the frequency range of common differential equations from low to high frequencies, with the interval set as follows: And dynamically adjust during training to match the true oscillation period of the data; For the first Learnable phase offset of each component is used to align the starting position of the waveform; For the first The adaptive mixing weights of each component are processed by Softmax normalization to ensure the stability of the contribution of each frequency component.

[0048] Used to enhance the model's ability to represent local drastic changes, spikes, or non-smooth features, as residual compensation for trend and oscillation terms: When constructing the data reconstruction network, the above... This is applied to each hidden layer of the network. Each hidden layer instantiates an independent activation function module, which contains the frequency parameters of different layers. Phase parameters Furthermore, the weight coefficients are not shared between layers. Through the stacking of deep network layers, shallow activation functions tend to learn high-frequency local fluctuations, while deep activation functions tend to learn low-frequency global trends, thereby achieving full-scale feature extraction of complex dynamics from the micro to the macro level.

[0049] In other embodiments, embodiments of this application can reconstruct the training output of the network based on data. Select its front element and combined with sampling interval Construct a stream mapping network input And through a multi-layer fully connected neural network. Establish mapping relationship: ,in, This is the output predicted by the stream mapping network. The network contains... There are 1 hidden layer, each containing 1 hidden layer. This network has 10 neurons and uses the hyperbolic tangent function as its activation function. This network does not need to fit complex dynamic trends, and its structure can be simplified. , .

[0050] For example, embodiments of this application can first calculate the graph Laplacian matrix corresponding to the sampled data sequence. The calculation of this graph Laplacian matrix can be determined based on the adjacency relationships constructed between the sampled data, using, for example, compressor inlet temperature sampling data. ; Calculate the average adjacent distance based on the sampling point sequence. Used to measure the local scale of data: Calculated .

[0051] Only adjacent sampling points are assigned non-zero weights to construct the adjacency weight matrix. To avoid unreasonable connections between non-local points, a Gaussian kernel function is used for calculation. Elements: Calculated for: .

[0052] in accordance with Construct the degree matrix Its diagonal elements yes No. Sum of row elements: , The matrix form is as follows: Calculated for: .

[0053] in accordance with and Calculate the graph Laplacian matrix This is used to constrain the smoothness of the denoising results on the data manifold, thereby effectively suppressing noise while maintaining local continuity and intrinsic structure. Calculated for: .

[0054] After calculating the graph Laplacian matrix, the embodiments of this application can use the Adam optimizer to jointly optimize all parameters of the data reconstruction network and the flow mapping network. The loss function used in the optimization process can include data loss, flow mapping loss and manifold regularization loss, and the loss function can be a weighted sum of the above three losses.

[0055] For example, data loss Used to measure the mean square error between the reconstructed data and the acquired data, ensuring consistency between the noise reduction results and the acquired data: ; Stream mapping loss This is used to constrain the reconstructed states of adjacent sampling points to satisfy the dynamic evolution relationship predicted by the flow map network, thereby enhancing temporal consistency. ,in, for After Each element; manifold regularization loss Used to utilize the Thulaplac matrix Constrain the smoothness of the reconstructed state along the manifold direction, and penalize drastic changes in the reconstructed state on the manifold: ,in, This represents the matrix trace operation.

[0056] After completing the above three loss calculations, the total loss function can be determined as a weighted sum of the three losses using weighting coefficients: ,in, , , They are respectively , , The loss weights are determined and the network parameters are iteratively updated using a gradient optimization algorithm until the loss function converges. Weight parameters are set. , , The compressor inlet temperature was obtained by training the Adam optimizer 20,000 times with a learning rate of 0.001. . Optionally, in this embodiment, the electronic device can input each variable sequence into the denoising processing model separately for denoising, or input the entire sampled data sequence into the denoising processing model for denoising. It can automatically complete the denoising processing of all variables without requiring manual parameter adjustment.

[0057] Figure 2 This diagram illustrates a sampling data and noise reduction data according to an embodiment of this application, such as... Figure 2 As shown, the sampling data obtained in each sampling in this embodiment may include five temperature state variables: compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, heat exchanger outlet temperature, and test section temperature; as well as input variables including cooling water supply volume; and operating condition variables including compressor power, ambient temperature, and cooling water supply temperature. The sampling data sequence composed of multiple types of sampling data collected in each test process (i.e., within a test cycle) may include variable sequences corresponding to each variable. The original signals of heat exchanger outlet temperature, cooling water supply volume, compressor power, and ambient temperature have significant noise, which is significantly improved after noise reduction. However, the noise in the compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, test section temperature, and cooling water supply temperature is relatively weak. After noise reduction, the results are highly consistent with the original data, indicating that the method has good robustness while maintaining signal characteristics. The data after noise reduction has less fluctuation and is smoother than the sampled data.

[0058] Step S30: Obtain the temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the noise-reduced data sequence.

[0059] In one possible implementation, after the electronic device denoises each sampled data sequence to obtain a denoised data sequence, it can perform adaptive Gaussian kernel state derivative calculation on the temperature state variables in the denoised data sequence to obtain a temperature state variable derivative sequence. This derivative process may include obtaining the temperature state variable sequences in both the denoised and sampled data sequences. A corresponding Gaussian kernel width parameter is calculated based on both the temperature state variable sequences in the sampled and denoised data sequences. Numerical differentiation using a Gaussian kernel differentiator is then performed on each temperature state variable sequence based on the Gaussian kernel width parameter to obtain the corresponding temperature state variable derivative sequence.

[0060] Optionally, the kernel width parameter required to calculate the derivative. The kernel function is used to control the local scope of action, directly affecting the smoothness and fidelity of derivative estimation. To adapt to the smoothness and sampling density of data under different temperature conditions, it is necessary to... Adaptive design is performed. The process of calculating the kernel width parameter in this embodiment may include: After centering the noise-reduced compressor inlet temperature data, its autocorrelation function is calculated: ,in, , The mean: .

[0061] Calculated : Search for the first one that satisfies... The delay position is used as the feature delay of the state data: ,in, The autocorrelation decay threshold, The value of follows these rules: ,in, For relative average and relative jump, The calculation method is as follows: ,in, The average relative jump rate of the original sampled data. This represents the average relative hop rate of the denoised data. (Calculated...) ,because , The value can be: If within the entire delay range Never lower than Then take The final calculation yielded... Among them, the lower realm Used to guarantee Not less than twice the sampling interval, to prevent [the following] Too small a value leads to numerical oscillations in the derivative estimate. The calculated... , .

[0062] In some embodiments, after the electronic device calculates the kernel width parameter for each temperature state variable (compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, heat exchanger outlet temperature, and test section temperature), it can perform high-precision derivative calculations on the denoised data, suppressing the influence of noise on the derivative. This step achieves derivative estimation through Gaussian kernel regression and adaptively determines the kernel bandwidth based on the autocorrelation characteristics of the temperature data, thereby achieving a balance between noise suppression and preservation of dynamic details.

[0063] For example, the electronic device can obtain the noise-reduced compressor inlet temperature. Corresponding sampling point location for Construct the radial basis function kernel matrix The elements therein are calculated as follows: ,in, This is the kernel width parameter, used to control the smoothness. (Setting...) Calculations yielded : .

[0064] right The elements in the middle about Analytical differentiation yields the kernel derivative matrix. Its element calculation is Calculated for: .

[0065] Solving the linear equations using ridge regression Obtain the kernel weight vector ,in, To address the numerical instability caused by an excessively large condition number in the ill-conditioned matrix, a regularization parameter is introduced, which is the identity matrix. . use and The derivative of the matrix product is analytically calculated to obtain the derivative sequence: The derivative sequence of the compressor inlet temperature was calculated. for: [2.34082288e-031.74889412e-031.20995894e-037.24065998e-042.90947566e-04 -8.99707326e-05……-3.80229958e-03 -3.38935059e-03 -2.97840969e-03 -2.57669135e-03-2.19162921e-03 -1.83079973e-03].

[0066] Figure 3 A schematic diagram illustrating the derivative of a temperature state variable according to an embodiment of this application is shown, such as... Figure 3 As shown, in this embodiment of the application, when there are five types of temperature state variables, the electronic device calculates the derivative of each of the five types of temperature state variable sequences in each denoised data sequence to obtain the corresponding temperature state variable derivative sequences. All derivative sequences remain smooth and continuous, without the high-frequency oscillations or glitches commonly seen in traditional numerical differentiation due to noise amplification. This indicates that the derivative calculation method has good robustness in suppressing noise sensitivity and can effectively avoid further amplification of the original noise by differentiation operations.

[0067] Step S40: Based on multiple denoised data sequences and the corresponding temperature state variable derivative sequences, perform multi-condition incremental regular regression ensemble modeling to obtain a continuous wind tunnel lumped temperature dynamics model.

[0068] In one possible implementation, after the electronic device performs data denoising and temperature derivative calculation, it can perform multi-condition incremental regularized regression ensemble modeling based on multiple denoised data sequences and corresponding temperature state variable derivative sequences to obtain a lumped temperature dynamics model for the continuous wind tunnel. This process constructs an adaptive incremental regularized regression model to complete the establishment of the lumped temperature dynamics model for the continuous wind tunnel. This step achieves adaptive enhancement of the operating condition characteristics by extracting the dominant direction of the wind tunnel operating conditions and combining an attention mechanism; simultaneously, incremental normal equations are used for cumulative training to improve the model's generalization ability under multiple operating conditions. Finally, a temperature state-space dynamics model with analytical form is obtained.

[0069] Optionally, the modeling process may include concatenating the operating condition variable sequences from each denoised data sequence to obtain a global operating condition variable dataset. From the global operating condition variable dataset, the low-dimensional dominant direction of each type of operating condition variable is extracted to obtain a dominant direction matrix. Based on each denoised data sequence, the attention weights of the included operating condition variables relative to each dominant direction are calculated to obtain the corresponding enhancement feature matrix. Based on the derivative sequences of the temperature state variables corresponding to each denoised data sequence and the enhancement feature matrix, the incremental regularized regression model parameters are solved to obtain a continuous wind tunnel lumped temperature dynamics model.

[0070] In some embodiments, the sampling data sequence collected in a single test cycle can be divided into datasets of model temperature state variables, model input variables, and model operating condition variables. For the first (e.g., the results of the fourth test) collected... The complete wind tunnel data from this experiment was used, and the noise of the collected sampled data sequence was reduced using step S20. Specifically, the electronic equipment can construct a model temperature state variable dataset based on the compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, heat exchanger outlet temperature, and test section temperature. Cooling water supply volume is used as the model input variable dataset. Compressor power, ambient temperature, and cooling water supply temperature were used to construct the model's operating condition variable dataset. .

[0071] : .

[0072] : .

[0073] : .

[0074] The electronic equipment further performs derivative calculations on the compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, heat exchanger outlet temperature, and test section temperature after noise reduction, constructing a dataset of derivatives of the model state variables. : : .

[0075] The model condition variable dataset from the complete experiment is constructed into a global condition variable dataset. : : .

[0076] Furthermore, after determining the global operating condition variable dataset, the electronic device... Feature orientation extraction is performed to extract the core modes that best characterize the differences in operating conditions, thus providing a stable, low-dimensional, and representative operating condition basis for subsequent multi-operating condition incremental modeling. After performing centralization, a zero-mean matrix is ​​obtained: ,in, The total number of data points across the four trials. Then calculate Covariance matrix: ,right Perform eigenvalue decomposition ,in, , .

[0077] Energy normalization is performed on each dominant direction: , Finally, the dominant direction matrix of the working condition is obtained. Calculations yielded : .

[0078] In some embodiments, to enable the model to adaptively adjust its dynamic characteristics according to the current operating conditions, a parameter-free attention mechanism based on cosine similarity is constructed to dynamically allocate operating condition weights and generate an adaptively enhanced feature matrix. Cosine similarity is calculated as follows: ,in, To prevent division by zero for extremely small numbers, take... Map similarity to attention weights that sum to 1. The element calculation is as follows: Calculations yielded : .

[0079] The model state variables, model input variables, and model operating condition variables datasets are concatenated to form the basic feature matrix. The basic features and attention weights are then fused channel-by-channel to obtain the final enhanced features: Calculations yielded : , in, , , Attention weights The three columns of components, While preserving the original physical variable information, the excitation intensity of the current working condition relative to each dominant direction is explicitly encoded, enabling the subsequent regression model to have working condition awareness.

[0080] After calculating the enhanced feature matrix, the electronic device can initialize the first and second model parameters according to the preset regularization coefficients, identity matrix, and zero matrix. Then, based on the derivative sequence of temperature state variables corresponding to each denoised data sequence and the enhanced feature matrix, the first and second model parameters are iteratively updated and solved. The model parameter solution process ends when preset conditions are met, and the continuous wind tunnel lumped temperature dynamics model is determined based on the current first and second model parameters.

[0081] The training process can be repeatedly computed to obtain... , , , , An incremental normal equation accumulation strategy is adopted to update the model parameters on an experimental cycle, avoiding numerical instability caused by splicing multiple experimental data. L2 regularization is also introduced to prevent overfitting. Thus, the normal equation matrix can be initialized as follows: The first model parameters initialized for: , Initialized second model parameters for: , in, Let be the regularization coefficient, and take . The data from the four experiments were iteratively updated sequentially. After global training is complete, the linear equations are solved in one go to obtain the coefficients of the temperature dynamics model. :

[0082] Based on the calculated parameters Establish a lumped temperature state space dynamics model for a continuous wind tunnel. .

[0083] In other embodiments, embodiments of this application can obtain predicted temperature state variables by performing multi-step recursive prediction based on the input variables and operating conditions of the continuous wind tunnel equipment, using a continuous wind tunnel lumped temperature dynamics model.

[0084] Optionally, the process of multi-step recursive prediction using a model in this embodiment may include obtaining initial values ​​for the compressor inlet temperature, compressor outlet temperature, heat exchanger inlet temperature, heat exchanger outlet temperature, and test section temperature. Cooling water supply Compressor power, ambient temperature, and cooling water supply temperature .

[0085] : .

[0086] : .

[0087] : .

[0088] calculate and Attention weights : .

[0089] For each time step From start to prediction end step The data length for this multi-step prediction experiment is 699, set to... The current predicted state variable. Input variables Operating condition variables Concatenation into basic feature vectors: The basic feature vector is combined with the attention weights Channel-by-channel weighted fusion is used to form an enhanced feature vector: Calculate the state derivative. That is, enhancing feature vectors With parameter matrix The product of: The state prediction value at the next moment can be obtained using Euler's formula. The final multi-step prediction result is as follows: .

[0090] Figure 4 This diagram illustrates a comparison between a predicted temperature state variable and the actual temperature state variable according to an embodiment of this application. As can be seen from the diagram, although multi-step prediction is theoretically challenging, especially in systems with multiple state variables where prediction errors tend to accumulate with step size, this model still manages to track the changing trends of the actual values ​​well. The prediction results are highly consistent with the actual signals, indicating that the model performs excellently in suppressing error accumulation and maintaining stability, effectively achieving high-precision simulation of multi-step prediction tasks. This fully verifies the reliability and applicability of the model in multi-step prediction of complex dynamic systems. It should also be noted that the current model is built based on four sets of experimental data. In practical applications, further expanding the modeling dataset is expected to further improve the model's adaptability to multiple operating conditions and its overall generalization ability.

[0091] Based on the above technical features, the embodiments of this application realize the full-process automation and intelligence of continuous wind tunnel temperature dynamics modeling. The beneficial effects of this scheme are mainly reflected in four aspects: First, through manifold consistency denoising and adaptive Gaussian kernel differentiation, noise in the original data can be automatically eliminated and smooth, oscillating state derivatives can be generated without manual intervention, significantly improving the quality and efficiency of data preprocessing; Second, based on the extraction of the dominant direction of the operating condition and the parameterless attention mechanism, the model can dynamically adjust the feature weights according to the changes in compressor power, ambient temperature, and cooling water supply temperature, thereby maintaining excellent prediction accuracy and generalization ability under multiple test cycles and multiple Mach number strategies; Third, five differential equations that completely describe the dynamic temperature propagation from the compressor inlet to the test section are constructed, fully reflecting the internal thermal coupling relationship of the system, providing a solid model foundation for the overall analysis and control of wind tunnel temperature; Fourth, a general technology chain from data denoising, derivative calculation, adaptive modeling of operating conditions to multi-step prediction is formed, which is not only applicable to continuous wind tunnels, but can also be extended to other multivariable, strongly coupled, and thermally inertial thermal systems, with wide engineering applicability and promotion value.

[0092] It should be understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0093] Based on the foregoing embodiments, this application provides a lumped temperature dynamics modeling device, which includes the included modules and the units included in each module, which can be implemented by a processor; of course, it can also be implemented by specific logic circuits; in the implementation process, the processor can be a central processing unit (CPU), microprocessor (MPU), digital signal processor (DSP) or field programmable gate array (FPGA), etc.

[0094] Figure 5 A schematic diagram of a lumped temperature dynamics modeling apparatus according to an embodiment of this application is shown. Figure 5 As shown, the lumped temperature dynamics modeling device in this application embodiment includes: The data acquisition module 50 is used to acquire multiple sampling data sequences obtained by sampling the state of the continuous wind tunnel equipment at a preset sampling frequency within a test cycle. The sampling data sequence includes multiple types of sampling data, and each of the sampling data includes temperature state variables, input variables, and operating condition variables. The noise reduction processing module 51 is used to perform manifold consistency noise reduction processing on each acquired sampled data sequence to obtain the corresponding noise-reduced data sequence; The derivative calculation module 52 is used to obtain the temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the noise reduction data sequence; The wind tunnel modeling module 53 is used to perform multi-condition incremental regular regression integrated modeling based on multiple noise-reduced data sequences and corresponding temperature state variable derivative sequences, to obtain a continuous wind tunnel lumped temperature dynamics model.

[0095] In one possible implementation, the noise reduction processing module 51 is further configured to: Each acquired sampling data sequence is input into the denoising processing model for manifold-consistent denoising calculation, and the corresponding denoised data sequence is output.

[0096] In one possible implementation, the derivative calculation module 52 is further configured to: Obtain the temperature state variable sequence in the noise-reduced data sequence, and the temperature state variable sequence in the sampled data sequence; Calculate the corresponding Gaussian kernel width parameter based on each temperature state variable sequence in the sampled data sequence and each temperature state variable sequence in the denoised data sequence; Based on the Gaussian kernel width parameter, the numerical derivative of each temperature state variable sequence is obtained by performing a Gaussian kernel differentiator.

[0097] In one possible implementation, the wind tunnel modeling module 53 is further configured to: By concatenating the operating condition variable sequences from each of the noise-reduced data sequences, a global operating condition variable dataset is obtained. The low-dimensional dominant direction of each type of operating condition variable is extracted from the global operating condition variable dataset to obtain the dominant direction matrix; Based on each of the noise reduction data sequences, the attention weights of the included working condition variables relative to each dominant direction are calculated to obtain the corresponding enhancement feature matrix; Based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix, the parameters of the incremental regular regression model are solved to obtain the continuous wind tunnel lumped temperature dynamics model.

[0098] In one possible implementation, the noise reduction processing module 51 is further configured to: Calculate the graph Laplacian matrix corresponding to each variable sequence in the sampled data sequence; The variable sequence is input into the noise reduction processing model, and the corresponding predicted noise reduction variable sequence is output. The data loss is determined based on the variable sequence and the corresponding predicted noise reduction variable sequence. The manifold regularization loss is determined based on the predicted denoising variable sequence and the graph Laplacian matrix. The flow mapping loss is determined based on each predicted denoising variable in the predicted denoising variable sequence. The parameters of the denoising processing model are adjusted according to the data loss, the manifold regularization loss, and the flow mapping loss until convergence, and the current predicted denoising variable sequence is determined to be the denoising variable sequence corresponding to the variable sequence.

[0099] In one possible implementation, the wind tunnel modeling module 53 is further configured to: The first and second model parameters are initialized based on the preset regularization coefficients, identity matrix, and zero matrix. The first model parameters and the second model parameters are iteratively updated and solved based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix. The model parameter solution process ends when the preset conditions are met, and the continuous wind tunnel lumped temperature dynamics model is determined based on the current first model parameters and second model parameters.

[0100] In one possible implementation, the device further includes: The temperature prediction module is used to respond to the input variables, operating condition variables and initial temperature state variables of the continuous wind tunnel equipment, and to perform multi-step recursive prediction based on the input variables and operating condition variables through the continuous wind tunnel lumped temperature dynamics model to obtain the predicted temperature state variables.

[0101] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0102] It should be noted that, in the embodiments of this application... Figure 5 The lumped temperature dynamics modeling device shown is illustrative of the module division, representing only a logical functional division; in actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into a single processing unit, exist as separate physical entities, or be integrated into a single unit. The integrated units described above can be implemented in hardware, as software functional units, or a combination of both.

[0103] It should be noted that, in the embodiments of this application, if the above-described methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0104] Figure 6 A schematic diagram of an electronic device according to an embodiment of this application is shown. For example... Figure 6 As shown in the figure, this application provides an electronic device, which can be a server, and its internal structure diagram can be as follows. Figure 6As shown, the electronic device includes a processor 620, a memory, and a transceiver 640 connected via a system bus 610. The processor 620 provides computing and control capabilities. The memory includes a non-volatile storage medium 631 and internal memory 632. The non-volatile storage medium 631 stores an operating system, computer programs, and a database. The internal memory 632 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium 631. The database stores data. The transceiver 640 communicates with external terminals via a network connection. The computer program, when executed by the processor 620, implements the methods described above.

[0105] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor 620, implements the steps of the method provided in the above embodiments.

[0106] This application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the steps in the method provided in the above-described method embodiments.

[0107] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0108] In one possible implementation, the apparatus provided in this application can be implemented as a computer program, which can be configured as follows: Figure 6 The device operates on the electronic device shown. The memory of the electronic device can store the various program modules that make up the above-described apparatus. The computer program composed of the various program modules causes the processor 620 to execute the steps of the methods in the various embodiments of this application described in this specification.

[0109] It should be noted that the descriptions of the storage medium and device embodiments above are similar to the descriptions of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium, storage medium, and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0110] It should be understood that the phrases "one embodiment," "an embodiment," or "some embodiments" mentioned throughout the specification mean that a specific feature, structure, or characteristic related to an embodiment is included in at least one embodiment of this application. Therefore, phrases such as "in one possible implementation," "in one embodiment," or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments; their similarities or commonalities can be referred to mutually, and for the sake of brevity, they will not be repeated here.

[0111] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three kinds of relationships. For example, object A and / or object B can represent three situations: object A exists alone, object A and object B exist simultaneously, and object B exists alone.

[0112] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or modules can be electrical, mechanical, or other forms.

[0114] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0115] In addition, each functional module in the various embodiments of this application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the integrated modules can be implemented in hardware or in the form of hardware plus software functional units.

[0116] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0117] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0118] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0119] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0120] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0121] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A lumped temperature dynamics modeling method, characterized in that, The method includes: The sampling data sequence is obtained by repeatedly acquiring the state sampling data of the continuous wind tunnel equipment at a preset sampling frequency within a test cycle. The sampling data sequence includes multiple types of sampling data, and each of the sampling data includes temperature state variables, input variables, and operating condition variables. Each acquired sampled data sequence is subjected to manifold-consistent denoising to obtain the corresponding denoised data sequence; By performing adaptive Gaussian kernel state derivative calculation on the temperature state variables in the noise-reduced data sequence, a temperature state variable derivative sequence is obtained. Based on multiple denoised data sequences and corresponding temperature state variable derivative sequences, multi-condition incremental regular regression ensemble modeling is performed to obtain a continuous wind tunnel lumped temperature dynamics model.

2. The method according to claim 1, characterized in that, The step of performing manifold-consistent denoising on each acquired sampled data sequence to obtain the corresponding denoised data sequence includes: Each acquired sampling data sequence is input into the denoising processing model for manifold-consistent denoising calculation, and the corresponding denoised data sequence is output.

3. The method according to claim 1, characterized in that, The step of obtaining a temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the denoised data sequence includes: Obtain the temperature state variable sequence in the noise-reduced data sequence, and the temperature state variable sequence in the sampled data sequence; Calculate the corresponding Gaussian kernel width parameter based on each temperature state variable sequence in the sampled data sequence and each temperature state variable sequence in the denoised data sequence; Based on the Gaussian kernel width parameter, the numerical derivative of each temperature state variable sequence is obtained by performing a Gaussian kernel differentiator.

4. The method according to claim 1, characterized in that, The multi-condition incremental regularized regression ensemble modeling based on multiple denoised data sequences and corresponding temperature state variable derivative sequences yields a continuous wind tunnel lumped temperature dynamics model, including: By concatenating the operating condition variable sequences from each of the noise-reduced data sequences, a global operating condition variable dataset is obtained. The low-dimensional dominant direction of each type of operating condition variable is extracted from the global operating condition variable dataset to obtain the dominant direction matrix; Based on each of the noise reduction data sequences, the attention weights of the included working condition variables relative to each dominant direction are calculated to obtain the corresponding enhancement feature matrix; Based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix, the parameters of the incremental regular regression model are solved to obtain the continuous wind tunnel lumped temperature dynamics model.

5. The method according to claim 2, characterized in that, The step of inputting each acquired sampling data sequence into the denoising processing model for denoising calculation and outputting the corresponding denoised data sequence includes: Calculate the graph Laplacian matrix corresponding to each variable sequence in the sampled data sequence; The variable sequence is input into the noise reduction processing model, and the corresponding predicted noise reduction variable sequence is output. The data loss is determined based on the variable sequence and the corresponding predicted noise reduction variable sequence. The manifold regularization loss is determined based on the predicted denoising variable sequence and the graph Laplacian matrix. The flow mapping loss is determined based on each predicted denoising variable in the predicted denoising variable sequence. The parameters of the denoising processing model are adjusted according to the data loss, the manifold regularization loss, and the flow mapping loss until convergence, and the current predicted denoising variable sequence is determined to be the denoising variable sequence corresponding to the variable sequence.

6. The method according to claim 4, characterized in that, The step of solving the incremental regular regression model parameters based on the temperature state variable derivative sequence corresponding to each of the noise-reduced data sequences and the enhanced feature matrix to obtain a continuous wind tunnel lumped temperature dynamics model includes: The first and second model parameters are initialized based on the preset regularization coefficients, identity matrix, and zero matrix. The first model parameters and the second model parameters are iteratively updated and solved based on the derivative sequence of the temperature state variable corresponding to each of the noise-reduced data sequences and the enhanced feature matrix. The model parameter solution process ends when the preset conditions are met, and the continuous wind tunnel lumped temperature dynamics model is determined based on the current first model parameters and second model parameters.

7. The method according to claim 1, characterized in that, The method further includes: In response to obtaining the input variables, operating condition variables, and initial temperature state variables of the continuous wind tunnel equipment, the predicted temperature state variables are obtained by performing multi-step recursive prediction based on the input variables and operating condition variables through the continuous wind tunnel lumped temperature dynamics model.

8. A lumped temperature dynamics modeling device, characterized in that, The device includes: The data acquisition module is used to acquire multiple sampling data sequences obtained by sampling the state of the continuous wind tunnel equipment at a preset sampling frequency within a test cycle. The sampling data sequence includes multiple types of sampling data, and each of the sampling data includes temperature state variables, input variables, and operating condition variables. The noise reduction processing module is used to perform noise reduction processing on each acquired sampled data sequence to obtain the corresponding noise-reduced data sequence; The derivative calculation module is used to obtain the temperature state variable derivative sequence by performing adaptive Gaussian kernel state derivative calculation on the temperature state variable in the noise-reduced data sequence; The wind tunnel modeling module is used to perform multi-condition incremental regular regression integrated modeling based on multiple denoised data sequences and corresponding temperature state variable derivative sequences, to obtain a continuous wind tunnel lumped temperature dynamics model.

9. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 7.

Images