Combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method and device based on multi-source data fusion and PINN

Abstract

The invention discloses a combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method and device based on multi-source data fusion and PINN. Comprising the following steps: acquiring measurement data of a flow field in a combustion chamber and a three-dimensional luminous field structure of a flow field flame under the same space coordinate system by adopting a one-dimensional Raman scattering technology, a two-dimensional plane laser-induced fluorescence technology, a two-dimensional particle image velocity measurement technology and a three-dimensional chemical self-luminous chromatography technology respectively; combining time and space coordinates of the combustion chamber flow field to construct training data; constructing a combustion chamber flow field parameter reconstruction model based on a physical information neural network, and training the combustion chamber flow field parameter reconstruction model by adopting the training data to obtain a trained combustion chamber flow field parameter reconstruction model; and inputting the time and space coordinates of the combustion chamber flow field to be reconstructed into the trained combustion chamber flow field parameter reconstruction model to obtain corresponding flow field parameters and predicted values of luminous intensity, thereby solving the problem that the existing combustion chamber flow field multi-parameter three-dimensional quantitative measurement is difficult.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of combustion chamber flow field analysis, in particular to a combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method and device based on multi-source data fusion and PINN. BACKGROUND

[0002] Due to the significant differences in combustion characteristics between hydrogen fuel and traditional fuels, the flame emission is weak and the spectral signal is weak, which brings great challenges to the measurement and analysis of the combustion process. In order to deeply understand the combustion characteristics of hydrogen fuel in the combustion chamber, optimize the combustion process, improve the combustion efficiency and reduce pollutant emissions, it is necessary to develop advanced measurement technology. Optical measurement technology plays an important role in combustion research and has made significant progress in recent years, such as laser-induced fluorescence (PLIF), spontaneous Raman scattering spectroscopy (Raman), tracer particle velocimetry (PIV), etc. These technologies can achieve high-precision measurement of temperature field, concentration field and velocity field in the combustion process.

[0003] In addition, multi-parameter synchronous test optical testing technology can obtain multiple combustion parameters simultaneously in the same experiment by combining multiple optical measurement methods, thereby more comprehensively analyzing the combustion process. This multi-parameter synchronous measurement technology can effectively overcome the limitations of single measurement method and provide more abundant data support for hydrogen combustion research. However, due to the particularity of hydrogen combustion, further research and optimization of these technologies are still needed to meet the testing needs of hydrogen turbine power combustion chambers.

[0004] At the same time, three-dimensional flow field reconstruction technology as an important tool of modern scientific research can convert complex two-dimensional measurement data into simple and intuitive three-dimensional models, providing strong support for visualization and analysis of the combustion process. Physical information neural network (PINN) is a new type of deep learning model that can embed physical laws into the training process of neural network and can be used to realize the inversion reconstruction of flow field parameters. This neural network combines the powerful function of machine learning and the constraint of physical laws, which can further improve the prediction accuracy and generalization ability of the model. By establishing a three-dimensional reconstruction method of combustion flow field based on laser measurement data results and physical information neural network, the running condition of the combustion flow field can be more comprehensively understood, and more accurate basis can be provided for the design and optimization of hydrogen turbine power combustion chambers.

[0005] Nowadays, due to the limitations of measurement technology resolution and applicable conditions, only single-dimensional and single-type flow field parameters of the combustion chamber can be measured in the traditional combustion chamber test. Although various two-dimensional and three-dimensional test technologies have been proposed in recent years, their application in the combustion chamber still has certain limitations, and the data fusion between different dimensions and different types of flow field parameters still has great deficiencies. The quantitative measurement of multi-parameter three-dimensional data of the hydrogen combustion chamber flow field has great research value and significance for the design and research of the hydrogen fuel combustion chamber in the current aviation field, and this technology still has a great gap in the current research field. SUMMARY

[0006] The purpose of the present application is to propose a combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method and device based on multi-source data fusion and PINN for the above-mentioned technical problems.

[0007] In a first aspect, the present application provides a combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method based on multi-source data fusion and PINN, comprising the following steps:

[0008] The one-dimensional Raman scattering technology, two-dimensional planar laser-induced fluorescence technology, two-dimensional particle image velocimetry technology and three-dimensional chemical autofluorescence tomography technology are respectively used to obtain the measurement data of the concentration field, temperature field and velocity field of the flow field in the combustion chamber and the three-dimensional light emission field structure of the flow field flame in the same spatial coordinate system, and the training data is constructed in combination with the time and spatial coordinates of the combustion chamber flow field;

[0009] A combustion chamber flow field parameter reconstruction model based on a physical information neural network is constructed, and the combustion chamber flow field parameter reconstruction model is trained using the training data to obtain a trained combustion chamber flow field parameter reconstruction model;

[0010] The time and spatial coordinates of the combustion chamber flow field to be reconstructed are input into the trained combustion chamber flow field parameter reconstruction model to obtain the predicted values of the corresponding flow field parameters and light emission intensity, and the flow field parameters include the concentration field, temperature field and velocity field of the main products.

[0011] As a preferred, the one-dimensional Raman scattering technology, two-dimensional planar laser-induced fluorescence technology, two-dimensional particle image velocimetry technology and three-dimensional chemical autofluorescence tomography technology are respectively used to obtain the measurement data of the concentration field, temperature field and velocity field of the flow field in the combustion chamber and the three-dimensional light emission field structure of the flow field flame in the same spatial coordinate system, specifically including:

[0012] On the spatial level, the coordinate system of each device is pre-calibrated using an optical calibration target before the combustion chamber is installed to build a unified spatial coordinate reference system;

[0013] In the time dimension, the measurement data of the one-dimensional Raman scattering technology, the two-dimensional planar laser-induced fluorescence technology, the two-dimensional particle image velocimetry technology and the three-dimensional chemiluminescence tomography technology are synchronously processed in time sequence by using a time-delay signal generator, and the time interval of each two kinds of measurement technologies should be less than 10 μs;

[0014] The one-dimensional Raman data are collected by using the one-dimensional Raman scattering technology, the two-dimensional fluorescence intensity distribution map is collected by using the two-dimensional planar laser-induced fluorescence technology, the two-dimensional displacement vector grid is collected by using the two-dimensional particle image velocimetry technology, and the two-dimensional chemiluminescence images at multiple angles are collected by using the three-dimensional chemiluminescence tomography technology;

[0015] The one-dimensional Raman data, the two-dimensional fluorescence intensity distribution map and the two-dimensional displacement vector grid are aligned to the same time by using a time interpolation method with the collection time of the two-dimensional chemiluminescence image as a reference, to obtain the time-aligned one-dimensional Raman data, the time-aligned two-dimensional fluorescence intensity distribution map and the time-aligned two-dimensional displacement vector grid;

[0016] The time-aligned one-dimensional Raman data are subjected to cubic spline interpolation by using a spatial interpolation method to generate the measurement data of the three-dimensional grid-based concentration field of the main product;

[0017] The time-aligned two-dimensional fluorescence intensity distribution map and the time-aligned two-dimensional displacement vector grid are mapped to the three-dimensional grid coordinates by using a bilinear interpolation method, and the measurement data of the three-dimensional grid-based temperature field and velocity field are respectively converted.

[0018] As preferred, the wavelength of the laser used in the one-dimensional Raman scattering technology is 532 nm, the time resolution should be at least 50 ns, the spatial resolution should be at least 0.2 mm, and the measurement error should be less than 5%;

[0019] The wavelength of the laser used in the two-dimensional planar laser-induced fluorescence technology is 308.520 nm, the time resolution should be at least 100 ns, the spatial resolution should be at least 0.1 mm, and the measurement error should be less than 5%;

[0020] The wavelength of the laser used in the two-dimensional particle image velocimetry technology is 532 nm, the time resolution should be at least 50 μs, the spatial resolution should be at least 0.2 mm, and the measurement error should be less than 5%;

[0021] The time resolution of the three-dimensional chemiluminescence tomography technology is better than 10 μs, the spatial resolution is better than 1 mm, and the measurement error is less than 5%.

[0022] As preferred, the combustion chamber flow field parameter reconstruction model comprises an input layer, a plurality of hidden layers and an output layer, the time and space coordinates of the combustion chamber flow field are input into the input layer, the nonlinear relationship is obtained through mathematical modeling in the plurality of hidden layers, the multi-dimensional features are obtained, and the multi-dimensional features are input into the output layer to generate the subsequent flow field parameters.

[0023] As preferred, the loss function used in the training process of the combustion chamber flow field parameter reconstruction model is:

[0024] ;

[0025] Among them, represents the data fitting term of the concentration field, represents the data fitting term of the temperature field, represents the data fitting term of the velocity field, represents the data fitting term of the luminous field, represents the physical consistency term. The first weight, the second weight, the third weight, the fourth weight and the fifth weight are respectively, the data fitting term of the concentration field is the mean square error between the predicted value of the concentration field and the measured data of the concentration field; the data fitting term of the temperature field is the mean square error between the predicted value of the temperature field and the measured data of the temperature field; the data fitting term of the velocity field is the mean square error between the predicted value of the velocity field and the measured data of the velocity field; the data fitting term of the luminous field is the L1 loss between the predicted value of the luminous intensity and the three-dimensional luminous field structure.

[0026] As preferred, the physical control equation on which the physical consistency term is based comprises continuity equation, momentum equation, energy equation and component equation.

[0027] The square sum of the residuals of the continuity equation, the momentum equation, the energy equation and the component equation is obtained, and the physical consistency term is obtained, as shown in the following formula:

[0028] ;

[0029] Among them, represents the residual of the continuity equation, represents the residual of the momentum equation, represents the residual of the energy equation, represents the residual of the component equation.

[0030] In the second aspect, the application provides a combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction device based on multi-source data fusion and PINN, comprising:

[0031] The data acquisition module is configured to acquire, in the same spatial coordinate system, measurement data of a concentration field, a temperature field and a velocity field of main products of a flow field in the combustion chamber and a three-dimensional light-emitting field structure of a flame of the flow field by using one-dimensional Raman scattering technology, two-dimensional planar laser-induced fluorescence technology, two-dimensional particle image velocimetry technology and three-dimensional chemical autofluorescence tomography technology respectively, and construct training data in combination with time and spatial coordinates of the flow field of the combustion chamber;

[0032] The model construction module is configured to construct a combustion chamber flow field parameter reconstruction model based on a physical information neural network and train the combustion chamber flow field parameter reconstruction model by using the training data to obtain a trained combustion chamber flow field parameter reconstruction model.

[0033] The prediction module is configured to input time and spatial coordinates of a combustion chamber flow field to be reconstructed into the trained combustion chamber flow field parameter reconstruction model to obtain predicted values of corresponding flow field parameters and light-emitting intensities, the flow field parameters including a concentration field, a temperature field and a velocity field of main products.

[0034] In a third aspect, the present application provides an electronic device, comprising one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any of the implementation manners of the first aspect.

[0035] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program, when the computer program is executed by a processor, the method described in any of the implementation manners of the first aspect is implemented.

[0036] In a fifth aspect, the present application provides a computer program product, comprising a computer program, when the computer program is executed by a processor, the method described in any of the implementation manners of the first aspect is implemented.

[0037] Compared with the prior art, the present application has the following beneficial effects:

[0038] (1) The combustion chamber flow field multi-parameter three-dimensional quantitative reconstruction method based on multi-source data fusion and PINN provided by the present application realizes more accurate description of the flow field and light-emitting field in the combustion chamber by three-dimensional multi-source data fusion and modeling of the physical information neural network, and can obtain flow field data such as temperature field, concentration field and velocity field of the combustion chamber under different test conditions, thereby providing strong support for the development of hydrogen combustion chambers.

[0039] (2) The multi-parameter three-dimensional quantitative reconstruction method of combustion chamber flow field based on multi-source data fusion and PINN proposed in this invention combines measurement data from multiple measurement technologies, performs data fusion on multi-source data, builds PINN network to process training data, extracts key features to accurately predict flow field performance, solves the problem of difficulty in multi-parameter three-dimensional quantitative measurement of existing combustion chamber flow field, and provides strong support for combustion chamber operation status diagnosis and combustion chamber improvement design.

[0040] (3) The multi-parameter three-dimensional quantitative reconstruction method of combustion chamber flow field based on multi-source data fusion and PINN proposed in this invention embeds physical constraints in the multi-source three-dimensional inversion reconstruction of combustion chamber flow field. By using physical consistency terms to force the flow field to satisfy the multi-parameter conservation law, it solves the problem of lack of physical law constraints in traditional 3D-CTC technology. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating an embodiment of the multi-parameter three-dimensional quantitative reconstruction method for combustion chamber flow field based on multi-source data fusion and PINN.

[0043] Figure 2 This is a flowchart illustrating an embodiment of the multi-parameter three-dimensional quantitative reconstruction method for combustion chamber flow field based on multi-source data fusion and PINN.

[0044] Figure 3 A schematic diagram of a three-dimensional synchronous measurement platform built according to the multi-source data fusion and PINN-based three-dimensional quantitative reconstruction method for multi-parameter combustion chamber flow field in an embodiment of this application;

[0045] Figure 4 This is a schematic diagram of the framework structure and training process of the PINN network in the multi-parameter three-dimensional quantitative reconstruction method of combustion chamber flow field based on multi-source data fusion and PINN, which is an embodiment of this application.

[0046] Figure 5 This is a schematic diagram of a multi-parameter three-dimensional quantitative reconstruction device for combustion chamber flow field based on multi-source data fusion and PINN, which is an embodiment of this application.

[0047] Figure 6 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0049] Figure 1 This application illustrates an embodiment of a method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN, comprising the following steps:

[0050] S1 uses one-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry, and three-dimensional chemiluminescence tomography to obtain the concentration field, temperature field, and velocity field of the main products in the combustion chamber flow field, as well as the three-dimensional luminescence field structure of the flow field flame, in the same spatial coordinate system. Training data is constructed by combining the time and spatial coordinates of the combustion chamber flow field.

[0051] In specific embodiments, one-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry, and three-dimensional chemiluminescence tomography are used respectively to acquire measurement data of the concentration field, temperature field, and velocity field of the flow field in the combustion chamber, as well as the three-dimensional luminescent field structure of the flow field flame, in the same spatial coordinate system. Specifically, this includes:

[0052] At the spatial level, an optical calibration target is used to pre-calibrate the coordinate system of each device before installation in the combustion chamber, so as to establish a unified spatial coordinate reference system.

[0053] At the time level, a time delay signal generator is used to synchronize the measurement data of one-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry and three-dimensional chemical autoluminescence tomography in time series. The time interval between each two measurement techniques should not exceed 10 μs.

[0054] One-dimensional Raman data was acquired using one-dimensional Raman scattering technology, two-dimensional fluorescence intensity distribution maps were acquired using two-dimensional planar laser-induced fluorescence technology, two-dimensional displacement vector grids were acquired using two-dimensional particle image velocimetry technology, and two-dimensional chemiluminescence images were acquired from multiple angles using three-dimensional chemiluminescence tomography.

[0055] Using the acquisition time of the two-dimensional chemiluminescence image as a reference, the one-dimensional Raman data, the two-dimensional fluorescence intensity distribution map and the two-dimensional displacement vector grid are aligned to the same time using the time interpolation method, resulting in time-aligned one-dimensional Raman data, time-aligned two-dimensional fluorescence intensity distribution map and time-aligned two-dimensional displacement vector grid;

[0056] The time-aligned one-dimensional Raman data were subjected to cubic spline interpolation using spatial interpolation methods to generate three-dimensional meshed concentration field measurement data of the main products.

[0057] The time-aligned two-dimensional fluorescence intensity distribution map and the time-aligned two-dimensional displacement vector grid were mapped to three-dimensional grid coordinates through bilinear interpolation, and the measurement data of the three-dimensional gridded temperature field and velocity field were obtained respectively.

[0058] In a specific embodiment, the laser used in the one-dimensional Raman scattering technique has a wavelength of 532 nm, a temporal resolution of at least 50 ns, a spatial resolution of at least 0.2 mm, and a measurement error of no more than 5%.

[0059] The wavelength of the laser used in two-dimensional planar laser-induced fluorescence technology is 308.520 nm, the temporal resolution should be at least 100 ns, the spatial resolution should be at least 0.1 mm, and the measurement error should not exceed 5%.

[0060] The laser used in two-dimensional particle image velocimetry has a wavelength of 532 nm, a temporal resolution of at least 50 μs, a spatial resolution of at least 0.2 mm, and a measurement error of no more than 5%.

[0061] The temporal resolution of three-dimensional chemiluminescence chromatography is better than 10 μs, the spatial resolution is better than 1 mm, and the measurement error is no more than 5%.

[0062] For details, please refer to Figure 2 The multi-parameter three-dimensional quantitative reconstruction method for combustion chamber flow field based on multi-source data fusion and PINN proposed in this application includes quantitative measurement of multi-source data of combustion chamber flow field and qualitative reconstruction of three-dimensional flame, multi-source data fusion, and construction of a PINN network model. This method is applicable to combustion chambers of various fuels, with the best application effect in pure hydrogen or hydrogen-blended combustion chambers. The embodiments of this application use a hydrogen combustion flame as an example for illustration. Quantitative measurement or qualitative observation of the concentration field, temperature field, velocity field, and luminescence field of hydrogen combustion flames typically involves various advanced measurement technologies. The embodiments of this application employ the following technical methods:

[0063] One-dimensional Raman scattering (1D-Raman) can be used to measure the concentration of the flow field in a hydrogen combustion chamber. By analyzing Raman spectroscopy, the spatial distribution of the concentration of the main products can be obtained, i.e., the concentration field. Two-dimensional planar laser-induced fluorescence (2D-PLIF) can be used to measure the spatial distribution of temperature, i.e., the temperature field, by using a laser of a specific wavelength to excite specific fluorescent molecules (e.g., OH radicals) and detecting the fluorescence they release. Two-dimensional particle image velocimetry (2D-PIV) can be used to measure the velocity field of a hydrogen combustion chamber. By recording the displacement of tracer particles under the illumination of a laser sheet, the velocity distribution of the fluid can be calculated, i.e., the velocity field. Three-dimensional chemiluminescence tomography (3D-CTC) can qualitatively reconstruct the luminescence intensity distribution in the three-dimensional space of the flame flow field, i.e., the luminescence field structure, starting from two-dimensional chemiluminescence images taken from multiple angles. The luminescence field structure in the embodiments of this application is a qualitative parameter, but it can be quantified by luminescence intensity.

[0064] The data and images acquired by the aforementioned measurement techniques need to be processed and analyzed using specialized software to extract the required measurement data for the concentration, temperature, and velocity fields, as well as the three-dimensional luminous field structure of the flow field flame. To facilitate subsequent data fusion, multiple measurement techniques need to be performed simultaneously, ensuring that the testing times of different techniques do not overlap while remaining as close as possible. Therefore, precise synchronization of lasers, cameras, and other measurement equipment is required, ensuring that all measurement techniques operate at the correct time and that the laser emission and signal collection times of multiple techniques do not conflict and are sufficiently close.

[0065] Furthermore, the design of the experimental setup also needs to consider how the various measurement instruments using different technologies can be spatially distributed to avoid mutual interference. It requires precise simulation of the experimental scenario, correct selection of measurement points, and ensuring that the instruments of different technologies can be arranged in a reasonable layout, with the spatial locations of the data sources as close as possible. By combining the above measurement technologies, a three-dimensional synchronous measurement platform for multi-source data measurement of the combustion chamber flow field can be designed, such as... Figure 3 As shown.

[0066] In space, the coordinate systems of each device are pre-calibrated using an optical calibration target before installation in the combustion chamber, establishing a unified spatial coordinate reference system. Subsequently, the incident mode of the quantitative measurement laser is adjusted using optical devices such as reflectors to ensure that multiple measurement planes have overlapping areas in space. The device layout is optimized through simulation to ensure that the three types of lasers can surround the core area of ​​the combustion chamber. The 3D-CTC observation system consists of fiber bundles and high-speed cameras. Each sub-fiber bundle is equipped with a 50mm focal length large aperture lens. To reduce the impact of lens distortion on the experiment and ensure that the emission field structure data and quantitative measurement data are spatially aligned, the fiber bundles and lenses need to measure the combustion flame on the same plane and at the same height. The acquisition of all the above optical measurement technologies does not necessarily require them to be located in the same spatial position. It is only necessary to obtain multi-source data in the same spatial coordinate system. While reducing the difficulty of optical alignment, it ensures that the concentration, temperature, velocity, and emission field data can be fused in the same spatial coordinate system to reduce interpolation errors in subsequent data preprocessing.

[0067] In terms of timing, consistency among the three quantitative measurement techniques is achieved through a delay signal generator. Since the system frequency of 2D-PIV differs from that of 2D-PLIF or 1D-Raman, a dual-frequency, nanosecond-level, six-channel delay signal generator is required. In one example, the 2D-PLIF acquisition exposure time is determined to be 200 ns, followed by 2D-PIV acquisition exposure for 200 μs, and finally 1D-Raman acquisition. The timing of the three lasers relative to the acquisition time is adjusted according to the exposure times of the three measurement techniques to achieve temporal proximity. The self-luminous signal generated by the flame is acquired by sub-fiber bundles at different angles and their corresponding camera lenses during the acquisition process of the three quantitative measurement techniques, and imaged on the main fiber head. Finally, the projection on the main fiber head is acquired by a high-speed camera.

[0068] Data preprocessing was performed on multi-source data from different measurement techniques. Spatial interpolation was used to perform cubic spline interpolation on 1D-Raman data to generate three-dimensional gridded concentration field measurement data. Two-dimensional data from 2D-PLIF and 2D-PIV were mapped to three-dimensional grid coordinates using bilinear interpolation. The qualitative observation luminescence field structure served as a spatial framework to assist in locating the relative positions of quantitative data. Using the acquisition time of the 3D-CTC qualitative observation luminescence field as a reference, temporal interpolation was used to align the quantitative data from 1D-Raman, 2D-PLIF, and 2D-PIV to the same time. Through data preprocessing, the spatiotemporal resolution of datasets from different sources was matched to achieve multi-source data fusion. Training data was constructed from the concentration, temperature, and velocity fields of the main products of the combustion chamber flow field acquired under known time and spatial coordinates, along with the three-dimensional luminescence field structure of the flow field flame. Training and validation sets were then generated in the same time and spatial coordinate system for subsequent construction of a combustion chamber flow field parameter reconstruction model based on a Physical Information Neural Network (PINN).

[0069] S2. Construct a combustion chamber flow field parameter reconstruction model based on a physical information neural network and train the combustion chamber flow field parameter reconstruction model using training data to obtain the trained combustion chamber flow field parameter reconstruction model.

[0070] In a specific embodiment, the combustion chamber flow field parameter reconstruction model includes an input layer, several hidden layers, and an output layer. The time and spatial coordinates of the combustion chamber flow field are input into the input layer. After mathematical modeling in several hidden layers, nonlinear relationships are obtained, resulting in multi-dimensional features. These multi-dimensional features are then input into the output layer to generate the subsequent flow field parameters.

[0071] In a specific embodiment, the loss function used during the training of the combustion chamber flow field parameter reconstruction model is:

[0072] ;

[0073] in, This represents the data fitting term for the concentration field. This represents the data fitting term for the temperature field. The data fitting term represents the velocity field. The data fitting term represents the luminescence field. Indicates a physical consistency term; The weights are the first, second, third, fourth, and fifth, respectively. The data fitting term for the concentration field is the mean square error between the predicted value and the measured value of the concentration field; the data fitting term for the temperature field is the mean square error between the predicted value and the measured value of the temperature field; the data fitting term for the velocity field is the mean square error between the predicted value and the measured value of the velocity field; and the data fitting term for the luminescence field is the L1 loss between the predicted value of the luminescence intensity and the three-dimensional luminescence field structure.

[0074] In specific embodiments, the physical governing equations upon which the physical consistency term is based include the continuity equation, momentum equation, energy equation, and composition equation.

[0075] By summing the squares of the residuals for the continuity equation, momentum equation, energy equation, and composition equation respectively, we obtain the physical consistency term, as shown in the following equation:

[0076] ;

[0077] in, Represents the residuals of the continuity equation. This represents the residual of the momentum equation. Represents the residuals of the energy equation. This represents the residual of the component equation.

[0078] Specifically, the combustion chamber flow field parameter reconstruction model based on Physical Information Neural Network (PINN) constructed in the embodiments of this application utilizes experimentally measured training data, combined with physical information guidance and constraints, to achieve the inversion and reconstruction of flow field parameters. By integrating information from different data sources, it generates reconstructed flow field results that conform to physical laws. A schematic diagram of the combustion chamber flow field parameter reconstruction model based on Physical Information Neural Network (PINN) and the PINN training process constructed in the embodiments of this application is shown below. Figure 4 As shown.

[0079] A combustion chamber flow field parameter reconstruction model based on a Physical Information Neural Network (PINN) was constructed. The PINN network comprises an input layer, several hidden layers, and an output layer. The construction process includes setting parameters such as the number of hidden layers, the number of neurons, the learning rate, and the optimization method. Partial differential terms are added to the PINN network's error function as physical constraints to reduce data dispersion. Multi-source fusion of three-dimensional flow field data (measurements of concentration, temperature, and velocity fields) and a three-dimensional emission field structure are used as labels during PINN network training for subsequent training and validation. The input layer, located within the PINN network, primarily receives time t and spatial coordinates (x, y, z). The PINN network contains multiple hidden layers to extract feature information from different dimensions and to mathematically model nonlinear relationships. The output layer generates the reconstructed flow field parameters.

[0080] The training process of the PINN network is as follows: Training data is fed into the PINN network, the model is initialized to generate relevant weights and corresponding output values, and a loss function is calculated between the output values ​​and the true values. The result of the loss function is optimized using gradient descent, and then the optimized correction values ​​are forward-propagated to the hidden layers to adjust the initial weights. After multiple iterations and weight updates, the training of the PINN network is complete. Essentially, the training process of the PINN network continuously optimizes the parameters of the network model, aiming to form a set of parameters that make the model's output infinitely close to the target sample. Through continuous fitting of the loss function, the final output of the PINN network is the reconstructed 3D flow field parameters and the predicted values ​​of luminescence intensity.

[0081] The loss function of the constructed PINN network includes multiple data fitting terms ( ), physical consistency item ( ) and multiple weights ( ), where the concentration field data fitting term The mean square error (MSE) used to measure the difference between the predicted values ​​(Y) of the concentration field and the measured data of the concentration field; the data fitting term for the temperature field. The mean square error between the predicted temperature field (T) and the measured temperature field data is used to measure the mean square error; the data fitting term for the velocity field is also used. The mean square error used to measure the difference between the predicted velocity field (u) and the measured velocity field data from the PINN network; the data fitting term for the luminescence field. This is used to measure the difference between the relevant outputs of the PINN network (such as luminescence intensity (I)) and the qualitative observation data of 3D-CTC (using L1 loss to adapt to the qualitative data). Taking the concentration field data fitting term as an example, assuming there are measured concentration data from N spatiotemporal points, the calculation formula is as follows:

[0082] ;

[0083] In the formula, This represents the concentration value predicted by the PINN network. This represents the measured data of the concentration field. The data fitting term between the temperature field and the velocity field (...). , The calculation formula for ) is similar. Since the emission field data is qualitative, it needs to be processed using L1 loss (mean absolute error, MAE), and the calculation formula is as follows:

[0084] ;

[0085] In the formula, This represents the luminous intensity predicted by the PINN network. This represents the luminous field structure reconstructed by 3D-CTC, and this data is mainly used as a benchmark to enhance the reliability of multi-source fusion.

[0086] Data fitting term ( The term ) is used to measure the difference between the PINN network output and the training data, while the physical consistency term ( The physical consistency term is used to measure whether the output of the PINN network satisfies the physical governing equations. The calculation is based on the residuals of the physical governing equations. The physical governing equations of the three-dimensional flow field include the continuity equation, momentum equation, energy equation, and composition equation. Taking the x-direction in three dimensions as an example:

[0087] ;

[0088] ;

[0089] ;

[0090] ;

[0091] Where ρ represents fluid density, t represents time, u represents velocity, p represents pressure, τ represents shear stress, and g x Let T represent the gravitational acceleration in the x-direction, T represent the temperature, k represent the thermal conductivity, and c represent the thermal conductivity. p This represents the specific heat capacity at constant pressure. h q Y q D qThese represent the reaction rate, specific enthalpy, mass fraction, and diffusion coefficient of the main product q, respectively.

[0092] For the physical governing equations of a three-dimensional flow field, the residual R is defined as the difference between the left-hand and right-hand sides of the governing equations. For example, the residual of the continuity equation is:

[0093] ;

[0094] The sum of squares of the residuals of all physical governing equations is the physical consistency term. ):

[0095] ;

[0096] For different weights in the loss function expression ( The value of ) requires different weights for different data fitting terms. The weights of all items should be controlled within the range of 1 to 10, while the weight of the physical consistency item ( The weights are controlled within the range of 0.001 to 0.1. The numerical adjustment of the weights needs to be based on the order of magnitude of different loss terms, while also considering the specific combustion chamber flow field and the degree of influence of multi-source data fusion on the PINN network, to ensure a balance between data accuracy and physical rationality. Using the physical control equations as part of the loss function through the PINN network can also be used to achieve multi-source data fusion, which places higher demands on the results.

[0097] Finally, the predicted values ​​of the three-dimensional flow field parameters and luminescence intensity output by the PINN network were compared and verified with the validation set data. Based on the comparison and verification results and error analysis, the parameters of the PINN network were adjusted and updated to continuously optimize the model's prediction performance, providing comprehensive information on the combustion chamber's operating status. This resulted in a trained combustion chamber flow field parameter reconstruction model. The trained PINN network can integrate information from different data sources, learn the intrinsic relationships between data, and generate prediction results that conform to physical laws, ultimately yielding accurately predicted three-dimensional flow field parameters and luminescence intensity of the reconstructed combustion chamber.

[0098] S3. Input the time and space coordinates of the combustion chamber flow field to be reconstructed into the trained combustion chamber flow field parameter reconstruction model to obtain the corresponding flow field parameters and predicted values ​​of luminescence intensity. The flow field parameters include the concentration field, temperature field and velocity field of the main products.

[0099] Specifically, the trained combustion chamber flow field parameter reconstruction model obtained in the embodiments of this application is deployed. Then, the time and spatial coordinates of any combustion chamber flow field to be reconstructed are input into the trained combustion chamber flow field parameter reconstruction model, and the corresponding flow field parameters and luminescence intensity prediction values ​​can be obtained, that is, the predicted values ​​of the concentration field, temperature field, velocity field, and luminescence intensity of the main products are obtained. Finally, the predicted values ​​of the three-dimensional flow field parameters and luminescence intensity output should have a time resolution better than 10 μs, a spatial resolution better than 1 mm, and a reconstruction error of no more than 5%.

[0100] The core innovation of this invention is a combustion chamber flow field parameter reconstruction model based on a physical information neural network. Physical constraints are embedded in the multi-source three-dimensional inversion reconstruction of the combustion chamber flow field. A physical consistency term forces the flow field to satisfy multi-parameter conservation laws, solving the problem of traditional 3D-CTC technology lacking physical constraints. Furthermore, multi-source data fusion achieves multi-dimensional and multi-parameter prediction results. Quantitative measurement data of temperature, concentration, and velocity fields are unified into a single three-dimensional grid through coordinate mapping, achieving cross-dimensional fusion. Then, minimizing the residuals of the physical equations drives the predicted values ​​to conform to the internal operating laws of the combustion chamber flow field, meeting the requirements for flow field prediction in complex and changing environments for hydrogen fuel cell aero-engines.

[0101] Further reference Figure 5 As an implementation of the methods shown in the above figures, this application provides an embodiment of a multi-parameter three-dimensional quantitative reconstruction device for combustion chamber flow field based on multi-source data fusion and PINN. This device embodiment is similar to... Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0102] This application provides a multi-parameter three-dimensional quantitative reconstruction device for combustion chamber flow field based on multi-source data fusion and PINN, including:

[0103] Data acquisition module 1 is configured to acquire the concentration field, temperature field, and velocity field of the main products of the flow field in the combustion chamber, as well as the three-dimensional luminescence field structure of the flow field flame, in the same spatial coordinate system using one-dimensional Raman scattering technology, two-dimensional planar laser-induced fluorescence technology, two-dimensional particle image velocimetry technology, and three-dimensional chemical autoluminescence tomography technology, respectively, and to construct training data by combining the time and spatial coordinates of the combustion chamber flow field.

[0104] Model building module 2 is configured to build a combustion chamber flow field parameter reconstruction model based on a physical information neural network and train the combustion chamber flow field parameter reconstruction model using training data to obtain a trained combustion chamber flow field parameter reconstruction model.

[0105] Prediction module 3 is configured to input the temporal and spatial coordinates of the combustion chamber flow field to be reconstructed into the trained combustion chamber flow field parameter reconstruction model to obtain the corresponding flow field parameters, which include the predicted values ​​of the concentration field, temperature field, velocity field and luminescence intensity of the main products.

[0106] Figure 6 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. For example... Figure 6 As shown, the electronic device of this embodiment includes a processor 601 and a memory 602; wherein the memory 602 is used to store computer execution instructions; and the processor 601 is used to execute the computer execution instructions stored in the memory to implement the various steps performed by the electronic device in the above embodiment. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0107] Alternatively, the memory 602 can be either standalone or integrated with the processor 601.

[0108] When the memory 602 is set up independently, the electronic device also includes a bus 603 for connecting the memory 602 and the processor 601.

[0109] This invention also provides a computer storage medium storing computer execution instructions, which, when executed by processor 601, implement the above method.

[0110] This invention also provides a computer program product, including a computer program that, when executed by a processor 601, implements the above-described method.

[0111] In the embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0112] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0113] Furthermore, the functional modules in the various embodiments of this invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit formed by the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0114] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor 601 to execute some steps of the methods of the various embodiments of this application.

[0115] It should be understood that the processor 601 described above can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor, or the processor 601 can be any conventional processor 601. The steps of the method disclosed in this invention can be directly manifested as the hardware processor 601 executing the steps, or as a combination of hardware and software modules within the processor 601 executing the steps.

[0116] The memory 602 may include high-speed RAM memory, and may also include non-volatile memory NVM, such as at least one disk storage device, and may also be a USB flash drive, portable hard drive, read-only memory, disk or optical disc, etc.

[0117] Bus 603 can be an Industry Standard Architecture (ISA), a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Bus 603 can be divided into address bus, data bus, control bus, etc. For ease of illustration, the bus 603 in the accompanying drawings of this application is not limited to only one bus 603 or one type of bus 603.

[0118] The aforementioned storage medium can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium accessible to general-purpose or special-purpose computers.

[0119] An exemplary storage medium is coupled to a processor 601, enabling the processor 601 to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor 601. The processor 601 and the storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor 601 and the storage medium can exist as discrete components in an electronic device or a host device.

[0120] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0121] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN, characterized in that, Includes the following steps: One-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry, and three-dimensional chemiluminescence tomography were used to obtain the concentration field, temperature field, and velocity field of the main products in the combustion chamber flow field, as well as the three-dimensional luminescence field structure of the flow field flame, in the same spatial coordinate system. Training data were constructed by combining the time and spatial coordinates of the combustion chamber flow field. A combustion chamber flow field parameter reconstruction model based on a physical information neural network is constructed, and the training data is used to train the combustion chamber flow field parameter reconstruction model to obtain a trained combustion chamber flow field parameter reconstruction model. The temporal and spatial coordinates of the combustion chamber flow field to be reconstructed are input into the trained combustion chamber flow field parameter reconstruction model to obtain the corresponding flow field parameters and predicted values ​​of luminescence intensity. The flow field parameters include the concentration field, temperature field and velocity field of the main products.

2. The method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN as described in claim 1, characterized in that, Measurement data of the concentration field, temperature field, and velocity field of the flow field within the combustion chamber, as well as the three-dimensional luminescent field structure of the flame, were obtained in the same spatial coordinate system using one-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry, and three-dimensional chemiluminescence tomography. Specifically, these included: At the spatial level, an optical calibration target is used to pre-calibrate the coordinate system of each device before installation in the combustion chamber, so as to establish a unified spatial coordinate reference system. At the time level, a time delay signal generator is used to synchronize the measurement data of one-dimensional Raman scattering, two-dimensional planar laser-induced fluorescence, two-dimensional particle image velocimetry and three-dimensional chemical autoluminescence tomography in time series. The time interval between each two measurement techniques should not exceed 10 μs. One-dimensional Raman data was acquired using one-dimensional Raman scattering technology, two-dimensional fluorescence intensity distribution maps were acquired using two-dimensional planar laser-induced fluorescence technology, two-dimensional displacement vector grids were acquired using two-dimensional particle image velocimetry technology, and two-dimensional chemiluminescence images were acquired from multiple angles using three-dimensional chemiluminescence tomography. Using the acquisition time of the two-dimensional chemiluminescence image as a reference, the one-dimensional Raman data, the two-dimensional fluorescence intensity distribution map, and the two-dimensional displacement vector grid are aligned to the same time using a time interpolation method, resulting in time-aligned one-dimensional Raman data, time-aligned two-dimensional fluorescence intensity distribution map, and time-aligned two-dimensional displacement vector grid. The time-aligned one-dimensional Raman data is subjected to cubic spline interpolation using a spatial interpolation method to generate three-dimensional gridded concentration field measurement data of the main products. The time-aligned two-dimensional fluorescence intensity distribution map and the time-aligned two-dimensional displacement vector grid are mapped to three-dimensional grid coordinates through bilinear interpolation, and the measurement data of the three-dimensional gridded temperature field and velocity field are obtained respectively.

3. The method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN as described in claim 1, characterized in that, The laser used in the one-dimensional Raman scattering technique has a wavelength of 532 nm, a temporal resolution of at least 50 ns, a spatial resolution of at least 0.2 mm, and a measurement error of no more than 5%. The wavelength of the laser used in the two-dimensional planar laser-induced fluorescence technique is 308.520 nm, the temporal resolution should be at least 100 ns, the spatial resolution should be at least 0.1 mm, and the measurement error should not exceed 5%. The laser used in the two-dimensional particle image velocimetry technology has a wavelength of 532 nm, a temporal resolution of at least 50 μs, a spatial resolution of at least 0.2 mm, and a measurement error of no more than 5%. The three-dimensional chemiluminescence chromatography technique has a temporal resolution better than 10 μs, a spatial resolution better than 1 mm, and a measurement error of no more than 5%.

4. The method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN as described in claim 1, characterized in that, The combustion chamber flow field parameter reconstruction model includes an input layer, several hidden layers, and an output layer. The time and spatial coordinates of the combustion chamber flow field are input into the input layer. After mathematical modeling in several hidden layers, a nonlinear relationship is obtained, resulting in multi-dimensional features. The multi-dimensional features are then input into the output layer to generate the subsequent flow field parameters.

5. The method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN as described in claim 1, characterized in that, The loss function used in the training process of the combustion chamber flow field parameter reconstruction model is: ； in, This represents the data fitting term for the concentration field. This represents the data fitting term for the temperature field. The data fitting term represents the velocity field. The data fitting term represents the luminescence field. Indicates a physical consistency term; The weights are respectively the first weight, the second weight, the third weight, the fourth weight, and the fifth weight. The data fitting term for the concentration field is the mean square error between the predicted value and the measured data of the concentration field; the data fitting term for the temperature field is the mean square error between the predicted value and the measured data of the temperature field; the data fitting term for the velocity field is the mean square error between the predicted value and the measured data of the velocity field; and the data fitting term for the luminescence field is the L1 loss between the predicted value of the luminescence intensity and the three-dimensional luminescence field structure.

6. The method for multi-parameter three-dimensional quantitative reconstruction of combustion chamber flow field based on multi-source data fusion and PINN as described in claim 5, is characterized in that, The physical governing equations upon which the physical consistency term is based include the continuity equation, momentum equation, energy equation, and composition equation. The physical consistency term is obtained by summing the squares of the residuals of the continuity equation, momentum equation, energy equation, and composition equation, respectively, as shown in the following equation: ； in, Represents the residuals of the continuity equation. This represents the residual of the momentum equation. This represents the residual of the energy equation. This represents the residual of the component equation.

7. A three-dimensional quantitative reconstruction device for multi-parameter combustion chamber flow field based on multi-source data fusion and PINN, characterized in that, include: The data acquisition module is configured to acquire the concentration field, temperature field, and velocity field of the main products of the flow field in the combustion chamber, as well as the three-dimensional luminescence field structure of the flow field flame, in the same spatial coordinate system using one-dimensional Raman scattering technology, two-dimensional planar laser-induced fluorescence technology, two-dimensional particle image velocimetry technology, and three-dimensional chemiluminescence tomography technology, respectively. Training data is constructed by combining the time and spatial coordinates of the combustion chamber flow field. The model building module is configured to build a combustion chamber flow field parameter reconstruction model based on a physical information neural network and train the combustion chamber flow field parameter reconstruction model using the training data to obtain a trained combustion chamber flow field parameter reconstruction model. The prediction module is configured to input the temporal and spatial coordinates of the combustion chamber flow field to be reconstructed into the trained combustion chamber flow field parameter reconstruction model to obtain the predicted values ​​of the corresponding flow field parameters and luminescence intensity. The flow field parameters include the concentration field, temperature field, and velocity field of the main products.

8. An electronic device, comprising: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

Images