A method for quantifying uncertainty analysis of aerodynamic noise in dynamic atmospheric environment

Abstract

A method for quantifying and analyzing the uncertainty of aerodynamic noise in a dynamic atmospheric environment includes the following steps: 1) combining intrinsic orthogonal decomposition with a back-propagation neural network model to establish an order model of the spatial and frequency domain distribution of aerodynamic noise; 2) extracting samples based on the atmospheric parameter dynamic change model, using a reduced-order model to predict the sample aerodynamic noise, and obtaining uncertainty quantification results and parameter sensitivity analysis results. Compared with traditional analysis methods, the method provided by the present invention can greatly improve the analysis efficiency while ensuring accuracy, effectively give the uncertainty change range of aerodynamic noise and the sensitivity of each parameter, reduce the huge amount of calculation in the initial design of the aircraft, shorten the design cycle, and has practical engineering significance.

Description

Technical Field

[0001] The present application relates to the technical field of aerodynamic noise analysis of aircraft, and in particular to an uncertainty quantification analysis method of aerodynamic noise in a dynamic atmospheric environment. Background Art

[0002] Aerodynamic noise is a sound with chaotic amplitude and frequency and statistically irregularity directly generated by airflow. In reality, atmospheric parameters are in constant dynamic change with time and geographical location, causing the actual aerodynamic noise environment faced by aircraft flight to be far more complex than the results predicted by the standard atmospheric model, which will seriously affect the performance stability and flight safety of the aircraft.

[0003] Uncertainty quantification methods can provide uncertain random solutions to the variables of interest, thereby supporting the refined design of the aircraft's shape, structure, and flight control system. Currently, the commonly used aerodynamic noise uncertainty quantification methods include the Monte Carlo method and the polynomial chaos method, but neither of them can meet the requirements of aerodynamic noise uncertainty quantification analysis in a dynamic atmospheric environment. Therefore, it is urgent to develop an uncertainty quantification analysis method for aerodynamic noise in a dynamic atmospheric environment to reduce the huge amount of calculations in the initial design of the aircraft and shorten the design cycle. Summary of the invention

[0004] In order to overcome the defects of the prior art, the purpose of the present invention is to provide a method for quantifying and analyzing the uncertainty of aerodynamic noise in a dynamic atmospheric environment, establish a reduced-order model of aerodynamic noise, and use the reduced-order model to predict sample aerodynamic noise, quickly and accurately give the range of variation of aerodynamic noise uncertainty, and ensure the performance stability and flight safety of the aircraft.

[0005] To achieve the above object, the present invention provides a method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment, comprising the following steps:

[0006] 1) The reduced-order model of aerodynamic noise is established by combining the intrinsic orthogonal decomposition method with the back-propagation neural network model;

[0007] 2) Samples are extracted based on the dynamic change model of atmospheric parameters, and the reduced-order model is used to predict the sample aerodynamic noise to obtain uncertainty quantification results and parameter sensitivity analysis results.

[0008] Furthermore, the step 1) further includes:

[0009] 21) Construct the characteristic matrix of the system according to the response values ​​of the sample points;

[0010] 22) According to the system characteristic matrix, obtain its eigenvalue and eigenvector;

[0011] 23) Selecting the truncated m-dimensional basis vector to fit the sample space characteristics according to the eigenvalue, where m is a positive integer less than n;

[0012] 24) Under the truncated basis vector, using the least squares method, respectively calculate the coefficients of the truncated basis vector under each sample point;

[0013] 25) Using the back propagation neural network model, an approximate fitting relationship between the design sample points and the coefficients under the truncated basis vector is established, and a reduced-order model of aerodynamic noise in a dynamic atmospheric environment is constructed.

[0014] Furthermore, the step 21) further includes:

[0015] The optimized Latin hypercube sampling method is used to obtain sample points in the design space;

[0016] Using computational fluid dynamics numerical simulation to obtain the response value of each sample point;

[0017] The system characteristic matrix is ​​constructed using the sample point response values.

[0018] Furthermore, the system characteristic matrix is: Among them, S is the system characteristic matrix, n is a positive integer, U (i) is the sample point response value, i=1,2,…,n.

[0019] Furthermore, the step 22) further includes:

[0020] Perform singular value decomposition on the system characteristic matrix to obtain its characteristic vector; the singular value decomposition formula is: S T SV=VΛ,Ψ0=SV,

[0021] Where Λ is a diagonal matrix formed by eigenvalues, and V is a matrix with columns S T The matrix consisting of the eigenvectors of S.

[0022] Furthermore, the step 24) further includes:

[0023] Under the truncated basis vectors, the approximate relationship Using the least squares method, the coefficients of the truncated basis vectors at each sample point are calculated respectively. Among them, U (i) is the sample point response value, and Ψ is the truncated basis vector.

[0024] Furthermore, the step 2) further includes:

[0025] A Gaussian distribution dynamic change model of atmospheric parameters is established, sample points are extracted according to the Gaussian distribution within a range of 3σ, the corresponding prediction truncated basis vector coefficients are obtained using the constructed back propagation neural network model, and the predicted response value of the sample point is obtained under the truncated basis vector;

[0026] The distribution characteristics of aerodynamic noise are statistically analyzed to obtain the uncertainty quantification analysis results, and based on the Sobol sensitivity analysis method, the Sobol sensitivity index of atmospheric parameters is obtained.

[0027] To achieve the above objectives, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a program running on the processor, and the processor runs the steps of the above-mentioned method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment.

[0028] To achieve the above object, the present invention also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed, execute the steps of the above-mentioned method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment.

[0029] Compared with the traditional analysis method, the uncertainty quantification analysis method of aerodynamic noise in a dynamic atmospheric environment described in the present invention has the following beneficial effects: it can greatly improve the analysis efficiency while ensuring accuracy, and effectively give the uncertainty variation range of aerodynamic noise and the sensitivity of each parameter, which has practical engineering significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention but do not constitute a limitation of the present invention.

[0031] In the attached picture:

[0032] Figure 1 It is a flow chart of the uncertainty quantification analysis method of aerodynamic noise in a dynamic atmospheric environment according to the present invention;

[0033] Figure 2 A flow chart of establishing a reduced-order model of aerodynamic noise according to the present invention;

[0034] Figure 3 A flow chart of a method for obtaining uncertainty quantification results and parameter sensitivity analysis results according to the present invention;

[0035] Figure 4 is a spatial distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention;

[0036] Figure 5 is a frequency domain distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention;

[0037] Figure 6 It is a probability distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention. DETAILED DESCRIPTION

[0038] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0039] In an embodiment of the present invention, samples are extracted according to a dynamic change model of atmospheric parameters, and a reduced-order model is used to predict the sample aerodynamic noise, so as to obtain uncertainty quantification results and parameter sensitivity analysis results, thereby calculating the spatial and frequency domain distribution characteristics of the aerodynamic noise of the aircraft in a dynamic atmospheric environment, wherein the design space of the aircraft wing model is: flight speed 5-10Ma, flight angle of attack 0-8°, atmospheric static pressure 79.7-5529.3Pa and atmospheric static temperature 216.6-270.6K.

[0040] Figure 1 The following is a flow chart of the method for quantifying and analyzing the uncertainty of aerodynamic noise in a dynamic atmospheric environment according to the present invention. Figure 1 , the uncertainty quantification analysis method of aerodynamic noise in a dynamic atmospheric environment of the present invention is described in detail.

[0041] First, in step 101, a reduced-order model of aerodynamic noise is established by combining an intrinsic orthogonal decomposition method with a back-propagation neural network model.

[0042] Figure 2 For the flow chart of establishing the reduced-order model of aerodynamic noise according to the present invention, the following will refer to Figure 2 , the specific process of establishing the reduced-order model of aerodynamic noise of the present invention is described in detail.

[0043] First, in step 201, a system characteristic matrix is ​​constructed according to the sample point response values.

[0044] In the embodiment of the present invention, the number of samples is n and the number of design variables is s. First, the sample space is divided into s dimensions, and each dimension is equally divided into n small intervals. In this way, the sample space is equally divided into n small intervals. s small squares, and then in n s Select n squares from the squares, and make sure that there is only one square selected in any row and any column, and then randomly select a point from each of the selected n squares to form n sample points, and then use the maximum and minimum distance criterion to generate the final design space sample point I through the element exchange update operation (i)(i = 1, 2, …, n). The aerodynamic noise is calculated by a non - linear acoustic calculation method. The steady - state RANS calculation is carried out using the SST k - ω turbulence model. The spatial discretization uses a second - order upwind scheme with a coupled TVD limiter, and the time discretization uses a second - order implicit scheme. Then, the statistically averaged results obtained from the steady - state RANS calculation are interpolated onto the NLAS calculation grid, and the turbulence is artificially reconstructed. The spatial and time discretization formats are the same as those in the RANS calculation, and the time step Δt = 5×10 -5 s, and the response value U of each sample point is obtained (i) (i = 1, 2, …, n); according to the response values of the sample points, the system characteristic matrix S = {U (i)}| n i＝1 , that is:

[0045] In step 202, the system characteristic matrix is subjected to singular value decomposition to obtain the eigenvalues and eigenvectors of the system characteristic matrix.

[0046] In the embodiment of the present invention, for the system characteristic matrix S, singular value decomposition is performed to obtain the eigenvalues ξ and eigenvectors Ψ0 of the matrix S, S T SV = VΛ, Ψ0 = SV. Where: Λ is the diagonal matrix formed by the eigenvalues ξ; V is the matrix composed of the eigenvectors of S T S.

[0047] In step 203, the basis vectors containing all the main characteristics of the samples are truncated to approximately fit the spatial characteristics of all the samples.

[0048] In the embodiment of the present invention, according to the eigenvalues ξ, a small part of the eigenvectors of the system characteristic matrix that contain all the main characteristics of the samples are selected, that is, the truncated basis vectors Ψ are m - dimensional (m < n), and the spatial characteristics of all the samples are approximately fitted.

[0049] In step 204, an approximate fitting relationship is established between the design sample points and the coefficients of the truncated basis vectors.

[0050] In the embodiment of the present invention, under the truncated basis vectors Ψ, by the approximate relationship, using the least - squares method, the coefficients of the truncated basis vectors are calculated for each sample point. From the corresponding relationship between the sample points and the coefficients of the truncated basis vectors, using a back - propagation neural network model, an approximate fitting relationship between the design sample points and the coefficients under the truncated basis vectors is established to construct a reduced - order model of aerodynamic noise in a dynamic atmospheric environment.

[0051] In step 102, samples are extracted according to the dynamic change model of atmospheric parameters, and the reduced - order model is used to predict the aerodynamic noise of the samples to obtain the uncertainty quantification result and the parameter sensitivity analysis result.

[0052] Figure 3 For a flow chart of the method for obtaining uncertainty quantification results and parameter sensitivity analysis results according to the present invention, reference will be made to Figure 3 , a flow chart of the method for obtaining uncertainty quantification results and parameter sensitivity analysis results of the present invention is described in detail.

[0053] First, in step 301, sample points are extracted.

[0054] In the embodiment of the present invention, a Gaussian distribution dynamic change model of atmospheric parameters is established; thereafter, sample points are extracted within a range of 3σ according to the Gaussian distribution.

[0055] In step 302, the predicted response value of the sample point is obtained.

[0056] In the embodiment of the present invention, after extracting sample points according to the atmospheric parameter dynamic change model, the corresponding prediction truncated basis vector coefficient a is obtained by using the constructed back propagation neural network model. (e) ; Then, under the truncated basis vector, the predicted response value of the sample point is obtained

[0057] In step 303, the uncertainty quantification analysis results and parameter sensitivity analysis results of the aerodynamic noise are obtained.

[0058] In the embodiment of the present invention, the predicted response value U of the sample point is obtained (e) After that, the distribution law is statistically analyzed to obtain the results of the uncertainty quantification analysis of aerodynamic noise; and the Sobol sensitivity index of atmospheric parameters is obtained by using the Sobol sensitivity analysis method.

[0059] Figure 4 is a spatial distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention, such as Figure 4 As shown in the figure, the sound pressure level at the leading edge of the wing is generally greater than that at the trailing edge. The maximum sound pressure level at the wing node is 167.6 dB, located at the leading edge of the wing; the minimum is 116 dB, located at the trailing edge of the wing. The disturbance of the sound pressure level at the leading edge is generally less than that at the trailing edge, with the maximum disturbance of the sound pressure level being 5.5 dB and the minimum being 1.1 dB.

[0060] Figure 5 is a frequency domain distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention, such as Figure 5 As shown in the figure, the energy of aerodynamic noise is mainly concentrated around 100 Hz; the maximum sound pressure level in the frequency band is 140 dB, the minimum sound pressure level is 94 dB; the maximum disturbance of the sound pressure level in the frequency band is 1.2 dB.

[0061] Figure 6 is a probability distribution diagram of aerodynamic noise in a dynamic atmospheric environment according to the present invention, such as Figure 6As shown in the figure, when the disturbances of atmospheric static pressure and atmospheric static temperature satisfy the Gaussian distribution dynamic change model, the disturbance of sound pressure level also approximately presents Gaussian distribution. However, due to the nonlinear relationship between atmospheric static pressure, atmospheric static temperature and aerodynamic noise, the distribution of sound pressure level is not a standard Gaussian distribution.

[0062] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a program that runs on the processor, and when the processor runs the program, the steps of the above-mentioned method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment are executed.

[0063] The present invention also provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed, the steps of the above-mentioned method for quantifying and analyzing the uncertainty of aerodynamic noise in a dynamic atmospheric environment are executed. The method for quantifying and analyzing the uncertainty of aerodynamic noise in a dynamic atmospheric environment is described in the introduction of the aforementioned part and will not be repeated here.

[0064] Those skilled in the art can understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions recorded in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment, characterized in that: The following steps are involved: 1) The reduced-order model of aerodynamic noise is established by combining the intrinsic orthogonal decomposition and the back propagation neural network model; 2) Extract samples based on the dynamic change model of atmospheric parameters, use the reduced-order model to predict the sample aerodynamic noise, and obtain uncertainty quantification results and parameter sensitivity analysis results; The step 1) further comprises: 21) Construct the system characteristic matrix according to the sample point response value; 22) According to the system characteristic matrix, obtain its eigenvalue and eigenvector; 23) Selecting the truncated m-dimensional basis vector to fit the sample space characteristics according to the eigenvalue, where m is a positive integer less than n; 24) Under the truncated basis vector, using the least squares method, respectively calculate the coefficients of the truncated basis vector under each sample point; 25) Using the back propagation neural network model, an approximate fitting relationship between the design sample points and the coefficients under the truncated basis vector is established to construct a reduced-order model of aerodynamic noise in a dynamic atmospheric environment; The step 21) further comprises: Divide the sample space into s dimensions, and divide each dimension into n small intervals, so that the sample space is divided into n s small squares; in n s Select n squares from the squares so that only one square is selected in any row and any column; randomly select a point from each of the n selected squares to form n sample points, and use the maximum and minimum distance criterion to generate the final design space sample point I through element exchange update operation (i) (i=1,2,…,n), where n is the number of samples and s is the design variable; Aerodynamic noise is calculated by nonlinear acoustic calculation method, and steady RANS calculation is performed using SST k-ω turbulence model. The spatial discretization adopts the second-order upwind format coupled with TVD limiter, and the time discretization adopts the second-order implicit format. The statistical average results obtained by the steady-state RANS calculation are interpolated onto the NLAS computational grid, and the turbulence is artificially reconstructed. The spatial and temporal discretization formats are the same as those of the RANS calculation, with a time step of Δt = 5 × 10 -5 s, get the response value U of each sample point (i) (i=1,2,…,n); According to the response values ​​of the sample points, construct the system characteristic matrix S = {U (i) }| n i＝1 .

2. The method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to claim 1 is characterized in that: The system characteristic matrix is: Among them, S is the system characteristic matrix, n is a positive integer, U (i) is the sample point response value, i=1,2,…,n.

3. The method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to claim 2 is characterized in that: The step 22) further includes: Perform singular value decomposition on the system characteristic matrix to obtain its characteristic vector; the singular value decomposition formula is: S T SV=VΛ,Ψ0=SV, Where Λ is a diagonal matrix formed by eigenvalues, and V is a matrix with columns S T The matrix consisting of the eigenvectors of S.

4. The method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to claim 2 is characterized in that: The step 24) further includes: Under the truncated basis vectors, the approximate relationship Using the least squares method, the coefficients of the truncated basis vectors at each sample point are calculated respectively. Among them, U (i) is the sample point response value, and Ψ is the truncated basis vector.

5. The method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to claim 1 is characterized in that: The step 2) further includes: A Gaussian distribution dynamic change model of atmospheric parameters is established, sample points are extracted according to the Gaussian distribution within a range of 3σ, the corresponding prediction truncated basis vector coefficients are obtained using the constructed back propagation neural network model, and the predicted response value of the sample point is obtained under the truncated basis vector; The distribution characteristics of aerodynamic noise are statistically analyzed to obtain the uncertainty quantification analysis results, and based on the Sobol sensitivity analysis method, the Sobol sensitivity index of atmospheric parameters is obtained.

6. An electronic device comprising a memory and a processor, wherein the memory stores a program that runs on the processor, and when the processor runs the program, the steps of the method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to any one of claims 1 to 5 are executed.

7. A computer-readable storage medium having computer instructions stored thereon, wherein when the computer instructions are executed, the steps of the method for quantifying and analyzing aerodynamic noise uncertainty in a dynamic atmospheric environment according to any one of claims 1 to 5 are executed.

Images