Fusion processing method for speed measurement data

Abstract

The invention relates to the technical field of data fusion processing, in particular to a fusion processing method for velocity measurement data, which comprises the following steps: analyzing a PIV velocity field by using POD, decomposing the PIV velocity field into POD modals, inputting data to calculate pulsation velocity, performing cubic spline interpolation on the POD modals to PTV particle coordinates, constructing and solving a linear equation set of a time coefficient, and reconstructing the POD modal time coefficient. According to the method, a pulse velocity field is calculated and reconstructed, and particle velocity precision measurement and calculation are carried out, so that the technical problem that in a complex flow field, large errors can be generated due to the smooth effect of the velocity field is solved, and the overall precision of the velocity field is improved by fusing PTV local fine velocity information on the premise that PIV spatial resolution is kept. And the technical effect of reconstructing the velocity field with higher precision is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field related to data fusion processing, and in particular to a fusion processing method for speed measurement data. Background Art

[0002] With the development of modern science and technology, fluid dynamics research has gradually relied on advanced measurement technology to capture the details of complex flow fields. Particle Image Velocimetry (PIV) and Particle Tracking Velocimetry (PTV) are two widely used optical flow field velocity measurement methods. These two technologies can capture the velocity distribution in the flow field at different times and spaces by tracking the movement of tracer particles, and have been widely used in turbulence research, aerodynamics, fluid machinery and other fields.

[0003] When PIV technology is used to obtain the global velocity distribution of the flow field, PIV can theoretically give the velocity value of each pixel in the image. However, due to the use of query window smoothing processing, it is easy to introduce a large velocity field smoothing error, which increases the error of the velocity output value. Although PTV technology shows high accuracy in capturing the local pulsating velocity of a single particle, it is often unable to provide complete flow field information due to the sparsity of tracer particles.

[0004] In summary, the prior art has a technical problem that in a complex flow field, a large error may be generated due to the smoothing effect of the velocity field. Summary of the invention

[0005] The present application provides a fusion processing method for velocity measurement data, aiming to solve the technical problem in the prior art that in a complex flow field, large errors may be generated due to the smoothing effect of the velocity field.

[0006] In view of the above problems, the technical solution to implement this application is:

[0007] The present application provides a fusion processing method for velocity measurement data, wherein the method comprises: using POD analysis, inputting PIV velocity field data U PIV (x, t), decomposed into a set of discrete PO D modes: Where M is the total number of moments of the PIV velocity field, φ m (x) is the spatial function of the m-th order POD mode, is the time coefficient under POD mode; input PTV velocity field data U PTV (x, t), calculate the pulsation velocity u of the PTV particle PTV ; For each time t, the POD modal space function is interpolated to the PTV particle coordinates using the cubic spline interpolation method; the pulsating velocity u of the PTV particle at the same time is used PTV, the equation relationship that the POD modal space function and the POD modal time coefficient should satisfy, construct a linear equation system about the time coefficient, and reconstruct the POD modal time coefficient using the least squares method Use POD modal space functions and reconstructed time coefficients Calculate the reconstructed pulsating velocity field: Among them, a(t) is used to characterize the corrected POD time coefficient; the reconstructed pulsating velocity field is: Calculate particle velocity accuracy.

[0008] In summary, one or more technical solutions provided in the present application solve the technical problem that in a complex flow field, large errors will be generated due to the smoothing effect of the velocity field. It achieves the technical effect of improving the overall accuracy of the velocity field by fusing the local fine velocity information of PTV while maintaining the spatial resolution of PIV, thereby obtaining a higher-precision reconstructed velocity field. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flow chart of a method for fusion processing of speed measurement data is provided for the present application.

[0010] Figure 2 A DNS velocity field schematic diagram of a fusion processing method for velocity measurement data is provided for the present application.

[0011] Figure 3 A schematic diagram of a cross-frame pair of tracer particle images for a fusion processing method of velocity measurement data is provided for the present application.

[0012] Figure 4 A PIV velocity field schematic diagram of a fusion processing method for velocity measurement data is provided for the present application.

[0013] Figure 5 A PTV velocity schematic diagram of a fusion processing method for velocity measurement data is provided for the present application.

[0014] Figure 6 A u-component error statistical distribution diagram of a fusion processing method for speed measurement data is provided for this application.

[0015] Figure 7 A v-component error statistical distribution diagram of a fusion processing method for speed measurement data is provided for this application.

[0016] Figure 8 A scatter plot of speed error for a fusion processing method of speed measurement data is provided for the present application. DETAILED DESCRIPTION

[0017] Embodiment 1

[0018] The present application is described in detail below with reference to the accompanying drawings. Figure 1 As shown, the present application provides a method for fusion processing of speed measurement data, wherein the method comprises:

[0019] S1: Use POD analysis and input PIV velocity field data U PIV (x, t), decomposed into a set of discrete POD modes: S2: where M is the total number of moments of the PIV velocity field, φ m (x) is the spatial function of the m-th order POD mode, is the time coefficient under POD mode; S3: input PTV velocity field data U PTV (x, t), calculate the pulsation velocity u of the PTV particle PTV .

[0020] Specifically, POD (Proper Orthogonal Decomposition) refers to a method for decomposing complex data into a set of orthogonal modes, which is often used in flow field data analysis. Through POD decomposition, the velocity field of a complex flow field can be decomposed into several spatial modes and time coefficients, while simplifying the data; PIV (particle image velocimetry) is a technology that measures fluid velocity by capturing the average cross-frame displacement of suspended particles in a small area of ​​the fluid. The PIV velocity field data is the particle velocity field distribution obtained by this technology; φ m (x) is the modal function obtained by POD analysis, which represents the velocity field distribution characteristics of each area in the flow field.

[0021] Collect and process PIV velocity field data, define the velocity field data matrix, which contains the velocity values ​​of the flow field at different time and space positions. The total number of velocity field data moments is M. Then, the POD method is applied to decompose these data. POD analysis decomposes these complex flow field data into a set of discrete POD modes, that is, each mode is composed of a spatial function φ m (x) and the time coefficient Here m represents the mode number, and the number of modes depends on the characteristic complexity of the flow field.

[0022] The PTV velocity field estimates the average velocity at all PTV particles Then use Get the pulsation velocity u of the PTV particle PTV , providing data support for subsequent analysis.

[0023] S4: For each time t, the POD modal space function is interpolated to the PTV particle coordinates using the cubic spline interpolation method; S5: The pulsating velocity u of the PTV particle at the same time is used PTV , the equation relationship that the POD modal space function and the POD modal time coefficient should satisfy, construct a linear equation system about the time coefficient, and reconstruct the POD modal time coefficient using the least squares method S6: Using POD modal space functions and reconstructed time coefficients Calculate the reconstructed pulsating velocity field: Where a(t) is used to characterize the corrected POD time coefficient; S7: the reconstructed pulsating velocity field:

[0024] Calculate particle velocity accuracy.

[0025] Specifically, the cubic spline interpolation method is a commonly used interpolation method, which uses a cubic polynomial to smoothly fit a set of discrete data points to generate a continuous function for estimating intermediate values ​​between known data points; the POD modal space function refers to the spatial function obtained by decomposing the flow field through POD analysis, which is used to describe the spatial characteristics of the flow field under different modes; the pulsating velocity u PTV Used to reflect the fluctuation part of fluid velocity; the goal of the least squares method is to solve the system of equations by minimizing the sum of squared errors to obtain the best fitting parameters.

[0026] Preferably, the cubic spline interpolation method is used to interpolate the POD modal space function to the PTV particle coordinates. Compared with linear interpolation, cubic spline interpolation can more accurately reflect the smooth motion characteristics of fluid particles over time, thereby improving the accuracy of reconstructing the flow field. In addition, the cubic spline interpolation method can reduce the numerical error generated during the interpolation process, especially when dealing with more complex velocity fields. It can more accurately estimate the value of the POD mode at the particle position, thereby improving the reconstruction effect of the pulsating velocity field.

[0027] At each time t, the POD modal space function is processed by cubic spline interpolation. The POD modal space function is usually defined at discrete spatial points, but the coordinates of the PTV particles may not coincide with these discrete points. Therefore, it is necessary to interpolate the value of the POD modal space function to the coordinates of each PTV particle. The cubic spline interpolation method is used to generate these continuous POD space function estimates to ensure accurate modal space function values ​​at the PTV particle positions.

[0028] At each time t, the pulsating velocity u of the PTV particle PTVThere is a certain linear relationship between the POD modal space function and the POD modal time coefficient. Based on this equation relationship, a linear equation group about the time coefficient is constructed. This equation group will use the known PTV particle pulsation velocity and the interpolated POD modal space function to solve the POD modal time coefficient at each moment. Since the linear equation group may have errors or noise, the least squares method is used to reconstruct the POD modal time coefficient. By minimizing the sum of squared errors, the least squares method can give the best fitting time coefficient estimate, making the reconstructed velocity field closest to the actual PTV data; Then, compare it with the corresponding POD modal space function φ m (x) is multiplied to calculate the reconstructed pulsating velocity field at each moment. The reconstructed pulsating velocity field is the linear superposition of multiple modes, and the formula is: in, Used to characterize the corrected POD time coefficient, u recon (x, t) represents the reconstructed pulsating velocity field. Finally, the reconstructed pulsating velocity field is calculated and compared with the original PIV velocity field to evaluate the corrected velocity accuracy and confirm the validity of the POD mode.

[0029] Furthermore, the present application method also includes:

[0030] φ m (x) is the spatial function of the m-th order POD mode, and the decomposed set of discrete POD modes satisfies the standard orthogonal basis. The spatial modes and time coefficients are calculated using the SVD decomposition method, and the POD modes are arranged in descending order of the modulus of the singular values.

[0031] Specifically, POD modes are a set of spatial functions used to represent complex flow fields. These functions are obtained by decomposing data. POD modes of various orders constitute a set of orthogonal bases, and the strength of each mode at different times is determined by its time coefficient; an orthogonal basis refers to a set of mutually orthogonal and normalized vectors. In this application, it means that each POD mode is orthogonal to each other and the length of each mode is unit length; SVD (Singular Value Decomposition) decomposition is a matrix decomposition technique, which is often used to extract the main modes in the data. In POD analysis, SVD decomposition is used to decompose velocity field data into spatial modes and time coefficients; the singular values ​​in the SVD decomposition are important parameters reflecting the modal strength. The main modes can be identified from strong to weak by arranging them in descending order according to the modulus of the singular values.

[0032] When performing POD analysis, first input the velocity field data and perform singular value decomposition (SVD) on it. This decomposition decomposes the original velocity field data matrix into three parts: spatial modes, singular values, and normalized time coefficient matrices. Among them, the spatial modes represent the spatial variation characteristics of the velocity field, the time coefficients describe the variation of these modes over time, and the singular values ​​indicate the contribution of each mode to the entire system. Specifically, the velocity field data is constructed into a matrix form. These data come from numerical simulations, such as PTV or PIV data. Each column in the matrix represents the velocity field distribution at a certain moment, and each row represents the velocity field value in space.

[0033] The velocity field data matrix is ​​subjected to singular value decomposition to obtain three matrices: one is the spatial mode matrix (including all POD modes), one is the singular value matrix, and the other is the normalized time coefficient matrix. Among them, the spatial mode is a set of orthogonal functions that represent different flow characteristics. Furthermore, the decomposed POD modes satisfy the properties of the standard orthogonal basis, that is, they are independent and orthogonal. Then, the singular values ​​are sorted according to the size of the modulus. The mode with a large modulus represents the main contribution in the velocity field, while the mode with a small modulus represents the secondary feature. According to the size of the singular value, it is possible to choose how many modes to retain for subsequent analysis. Usually, the first few modes with larger moduli can reconstruct the velocity field well, which not only reduces the computational complexity, but also retains the main characteristics of the flow field. Through the above steps, using SVD decomposition, the POD modes and time coefficients are effectively calculated and sorted. The main modes are arranged in descending order of singular values ​​to reconstruct and analyze the main structure of the velocity field. When processing complex flow field data, it can simplify the analysis and extract the most important flow characteristics.

[0034] Furthermore, Used to characterize the corrected POD time coefficient. For a two-dimensional velocity field, it is composed of 2×N PTV (t) equations, the method of the present application also includes:

[0035] Among them, φ u,m (x i ,y i ) is used to characterize the POD space function of the u component of the PIV velocity field at point (x i ,y i ) at the interpolation point; Among them, φ v,m (x j ,y j ) is used to characterize the POD space function of the v component of the PIV velocity field at point (x i ,y i ) at the interpolation point.

[0036] Specifically, in a two-dimensional velocity field, the u component generally represents the horizontal velocity, while the v component represents the vertical velocity.

[0037] For the POD analysis of the two-dimensional velocity field, the reconstructed POD time coefficients are used to characterize the dynamic characteristics of the velocity field. These time coefficients are expressed in the form of a system of linear equations, which constitute equations for calculating the u and v components of the velocity field. In this process, it is first necessary to interpolate the spatial function of the POD mode to a specific spatial point to form the interpolation results for the u and v components; for the PIV velocity field at each moment, the interpolation results of the POD spatial function of the u component and the v component at a given point are obtained respectively. Thus, a system of linear equations containing multiple equations can be established, in which each equation represents the interpolation relationship at a moment.

[0038] The linear equations are solved by the least square method to obtain the modified POD time coefficient. At this time, the obtained time coefficient reflects the contribution of each mode at a specific moment, thereby achieving an accurate description of the u component and the v component. Finally, by combining these modified POD time coefficients with the corresponding POD space function, the reconstructed pulsating velocity field is calculated, and the obtained u and v components can more accurately describe the motion state of the fluid. The above steps effectively improve the accuracy of PIV measurement, making the analysis of flow characteristics more comprehensive and reliable.

[0039] Furthermore, the present application method also includes:

[0040] Let's write column vector: Convert to matrix form:

[0041] Among them, in [φ1,φ2,……,φ M ] matrix, each vector represents a spatial function φ M (t) Interpolation over all PTV particles at time t.

[0042] Specifically, a column vector is a vertically arranged data structure containing multiple elements, commonly used to represent coordinates, speed, or other physical quantities; a matrix is ​​an array of values ​​arranged in rows and columns that can represent more complex data structures, such as linear transformations or system equations.

[0043] Define column vector It is used to represent the interpolation result of the POD modal space function corresponding to the (x, y) position at time t; the POD modal space function φ m (x) is a row vector POD modal time coefficient is a column vector Organized into a matrix, After sorting, we get Among them, u PTV (x i ,y i ,t)、v PTV (x i ,y i ,t) corresponds to the position of the PTV particle in the PIV velocity field (x i ,y i ), u component and u component of . In short, the structured representation of column vectors facilitates data processing and analysis using linear algebra methods, making calculations more efficient and intuitive.

[0044] The matrix form can be used to conveniently analyze the velocity field as a whole, such as calculating the pulsating velocity and extracting the main flow characteristics. In addition, the introduction of the matrix form also provides a basis for subsequent numerical calculations, making it possible to use linear algebra methods to solve related equations or perform optimization processing, effectively integrating time and space information.

[0045] Furthermore, the present application method also includes:

[0046] The time series average of the PIV velocity field at the PTV particle position using cubic spline interpolation as an estimate of the average velocity at each PTV particle.

[0047] Specifically, cubic spline interpolation is used to smoothly estimate the value of unknown data points; time series average refers to averaging multiple sets of data within a period of time to obtain a representative value within that period of time.

[0048] In order to estimate the average velocity at each PTV particle, it is first necessary to perform cubic spline interpolation on the PIV velocity field at the PTV particle position. Furthermore, the velocity field at each moment is determined using the acquired PIV velocity field data, and a time series is selected. These velocity field data contain u and v components, which represent the velocity in the x and y directions of the flow, respectively. On this basis, for the position of each PTV particle, the cubic spline interpolation method is used to calculate the velocity value of the particle at each moment, ensuring a smooth and continuous velocity estimate at the particle position.

[0049] The interpolation results at all times are averaged in time series to calculate the average velocity of each PTV particle. Furthermore, for each particle, the velocity values ​​interpolated at all times are collected and then averaged to finally obtain the average velocity estimate at each PTV particle. The average velocity value provides reliable flow field information for subsequent analysis, which helps to more accurately describe and understand the motion characteristics of the fluid.

[0050] Furthermore, the present application method also includes:

[0051] The spatially continuous POD modal space function is obtained by using the cubic spline interpolation method, and the value of the continuous function on the PTV particle is used as the estimate of the POD modal space function value corresponding to the PTV particle coordinates.

[0052] Specifically, cubic spline interpolation is a method of connecting known data points through multiple cubic polynomial segments, ensuring the continuity of the function value and its first-order and second-order derivatives at each node, which is used to smoothly estimate the value of unknown data points; POD modal space function is the spatial basis function obtained from POD analysis, which represents the spatial characteristics of different modes in the flow field; PTV particles refer to particles tracked in the fluid, and flow information is obtained by monitoring their motion trajectory.

[0053] In order to obtain the spatially continuous POD modal space function and use it to estimate the POD modal space function value corresponding to the PTV particle coordinates, firstly, based on the existing POD analysis results, the spatial functions of each POD mode are extracted. These modal functions are usually obtained by singular value decomposition (SVD) on the velocity field data at a series of discrete moments; the cubic spline interpolation method is used to interpolate these discrete POD modal space functions to obtain a smooth and continuous function. This process includes selecting interpolation nodes, which are usually the values ​​of the POD modal function at discrete time points, and connecting these nodes using the cubic spline interpolation method to ensure that the first-order and second-order derivatives of the interpolation result at each node are continuous.

[0054] After obtaining the continuous POD modal space function, the next step is to evaluate the function at the position of the PTV particle. For each PTV particle, the corresponding POD modal space function value in space is calculated. This can be achieved by substituting the coordinates of the PTV particle into the interpolated POD modal space function to obtain the corresponding function value, thereby assigning a POD modal value to each PTV particle. Through the above steps, the continuous POD modal space function is effectively applied to the PTV particles, providing important basic data for subsequent velocity field reconstruction and flow analysis.

[0055] Furthermore, the present application method also includes:

[0056] Determine the number of modes M that need to be reconstructed PTV , to adjust the correction effect on the velocity field.

[0057] Specifically, the number of modes refers to the specific number of POD modes extracted from the flow field in POD analysis. These modes capture the main motion characteristics in the flow field and are usually selected according to the size of the singular values ​​to optimize the reconstruction and correction of the velocity field. The correction of the velocity field is in fluid mechanics, by adjusting and optimizing the velocity field data, so that the simulation results better fit the experimental data.

[0058] It is known that the number of selected modes directly affects the accuracy and efficiency of velocity field reconstruction. Furthermore, when reconstructing and correcting the velocity field, it is first necessary to determine the number of POD modes that need to be reconstructed; based on the POD analysis results, the singular values ​​of each mode are calculated. These singular values ​​reflect the importance of the corresponding mode in the entire flow field and are usually arranged in descending order. In this process, an attenuation graph of the singular values ​​can be drawn. This graph helps to intuitively show the contribution of the mode to the flow characteristics; the time coefficients corresponding to the modes that are not selected remain unchanged.

[0059] By observing the attenuation trend of singular values, the appropriate number of modes is selected. Generally speaking, the selection of the number of modes should be based on the contribution rate of singular values, that is, the contribution ratio of a certain mode and all previous modes to the total energy (or total variance). When the cumulative contribution rate reaches a set threshold (such as 80% or 90%), it can be considered that the selected number of modes is sufficient to represent the main characteristics of the flow field; using the selected number of modes, the velocity field is reconstructed. In this process, the spatial function of the selected mode is combined with the corresponding time coefficient to construct a corrected velocity field. By comparing with the original velocity field, the correction effect is evaluated, and the number of modes is further adjusted to optimize the correction result. Through the above steps, the required number of POD modes is effectively determined to achieve accurate correction of the velocity field and improve the credibility of the flow field simulation.

[0060] Furthermore, the present application method also includes:

[0061] Each time the POD modal time coefficient is reconstructed using the least squares method, the time t is fixed and only the POD modal time coefficient at one time is reconstructed.

[0062] Specifically, the least squares method aims to solve the unknown parameters by minimizing the sum of squares of errors. It is widely used in data fitting and reconstruction processes to make the deviation between the model and the data as small as possible; the POD modal time coefficient means that in POD analysis, each mode has its corresponding time coefficient, which describes the change of the mode at different times; the fixed time t means that in the reconstruction process, calculations are only performed for a specific time point each time, rather than all times.

[0063] When using the least squares method to reconstruct the time coefficient of the POD mode, the specific operation steps are carried out step by step, and only the POD modal time coefficient of one time t is processed each time. The goal is to reduce the computational complexity and error accumulation that may be caused by processing all times at one time; a specific time point t is selected, at which the spatial function of the POD mode has been accurately estimated by methods such as cubic spline interpolation. Next, the known POD modal spatial function is combined with the pulsating velocity field at that time (such as the pulsating velocity obtained by PTV measurement) to establish a linear equation system about the POD modal time coefficient. In this linear equation system, the unknown variable is the POD modal time coefficient, and the known quantity is the POD spatial function and the corresponding pulsating velocity field data; the POD modal time coefficient at that time can be solved by the least squares method. The goal of the least squares method is to minimize the difference between the reconstructed velocity field and the actual pulsating velocity field, so that the estimated value of the POD modal time coefficient at that time is the most accurate.

[0064] Only one moment is processed each time, which means that the reconstruction process is carried out hour by hour. The hourly reconstruction method can reduce the calculation pressure and ensure that the time coefficient of each moment is closely related to the actual velocity field data at that moment, thereby improving the accuracy of reconstruction; continue until the POD modal time coefficients of all moments are reconstructed one by one.

[0065] Embodiment 2

[0066] Taking the flow around a cylinder with Reynolds number = 100 as an example, the reconstructed pulsating velocity field is verified by comparing the velocity error of the simulated PIV pulsating velocity field and the pulsating velocity field reconstructed based on this application: Higher accuracy when measuring velocity fields based on particle images.

[0067] In this example, the particle image generated in the flow experiment around a cylinder with Reynolds number = 100 is simulated by simulation method. Then, the PIV and PTV algorithms are used to calculate the PIV velocity field U at each moment. PIV and PTV speed data U PTV, and pre-processing of bad pixels is completed for both. The DNS data under this condition comes from http: / / www.dmdbook.com / . Its dimensionless coordinates are enlarged to an image of 1024×550 pixels, and its dimensionless velocity value is multiplied by 1000 as the pixel displacement velocity (the unit becomes pix / s). Particles with an average particle concentration of about 0.006 / pixel are randomly scattered on the image, and then the pixel displacement of each particle within the cross-frame time dt=0.01s is calculated, and the new position of each particle is calculated based on the displacement, and then a cross-frame simulated particle image is generated. Finally, the PIV and PTV algorithms are used for these simulated particle images to calculate the PIV velocity field U at each moment. PIV and PTV speed data U PTV , and complete the pre-processing of bad pixels for both. Figure 2 The figure shows the DNS velocity field under the typical Reynolds number = 100 cylinder flow condition. Figure 3 The following is a typical cross-frame simulation particle image pair under this condition. Figure 4 The PIV velocity field calculated from the simulated particle image is shown in the figure, and the data at the edge of the image is removed to avoid bad pixels. Figure 5 The PTV velocity information calculated from the simulated particle image is shown, and the data at the edge of the image is removed to avoid bad pixels.

[0068] In the program of the embodiment, the member variables that need to be imported as parameter settings of the structure Options are: the string dir_load of the file path where the velocity field data to be imported is located; the string dir_save of the path where the calculation results are saved; the number of moments Nt contained in the velocity field data to be imported; the string CompareOrN that determines whether it is necessary to display the P IV velocity field, the error between the reconstructed velocity field and the true value of the DNS velocity field ot, if "yes" is selected, the real velocity field data needs to be imported and the output drawing will be displayed for comparison effect. If "no" or other character strings are selected, the drawing will not be displayed for comparison effect; the POD analysis energy cutoff threshold of the PIV velocity field, the unit is "%", after testing, in this embodiment, it is best to be not less than 99.5 and less than 100; the number of modes of the reconstructed time coefficient N_PTVmodal; the starting time t_start and the ending time t_end of the time coefficient to be reconstructed; the maximum relative error tol allowed by the least square method when reconstructing the time coefficient, the relative error is considered to be converged if it is less than this value; the maximum number of iterations calculated by the least square method maxit, its value should make the least squares iteration calculation converge within maxit times; the velocity field diagram is drawn with the result of reconstructing the velocity field at the time_plot; Confidence_Range determines the size of the confidence interval when analyzing the error, and the size of the interval of the proportion of the previous Confidence_Range with the smallest error is used to characterize the size of the overall error; nbins is the number of groups of the error statistical histogram.

[0069] The member variables of the structure Input that need to be imported as the PIV / PTV velocity data input are: the matrix X_PIV of the PIV velocity field grid row coordinates, for any i-row and j-column element X_PIV(i,j), it corresponds to the pixel row coordinates of the i-row and j-column grid point in the particle image, and the number of rows I and the number of columns J should be consistent with the number of PIV grids in the row direction and the number in the column direction of the image, and the unit is pixel (pix); the column matrix Y_PIV of the PIV velocity field grid coordinates, for any i-row and j-column element Y_PIV(i,j), it corresponds to the i-row and j-column grid point in the particle image The pixel column coordinates of the grid points, and the number of rows I and columns J should be consistent with the number of PIV grids in the row direction and the number in the column direction of the image, and the unit is pixel (pix); the three-dimensional array U_PIV of the x-direction (i.e., the row direction of the image) component of the PIV velocity field, the matrix formed by each time profile corresponds to X_PIV and Y_PIV one by one, that is, the element U_PIV(i,j,t) at any i-row and j-column at time t is the component of the velocity in the x-direction at the pixel position of the row coordinate X_PIV(i,j) and column coordinate Y_PIV(i,j) at time t, and the unit is pix / s; PI The three-dimensional array V_PIV of the y-direction (i.e., column direction) component of the V velocity field has a matrix consisting of the profiles at each moment that corresponds one-to-one with X_PIV and Y_PIV, i.e., the element V_PIV(i,j,t) at any i-th row and j-th column at moment t is the component of the velocity in the y-direction at the pixel position with row coordinates X_PIV(i,j) and column coordinates Y_PIV(i,j) at moment t, and the unit is pix / s.

[0070] If the member variable CompareOrNot in the Options structure takes the value of "yes" during parameter setting, the DNS velocity data used as the true value of the flow field in this embodiment must also be input into the following member variables of the GroundTrut h structure: the matrix X_DNS of the DNS velocity field grid row coordinates, for any i-row and j-column element X_DNS(i,j), it corresponds to the pixel row coordinate of the i-row and j-column grid point in the particle image, and the number of rows I and the number of columns J should be consistent with the number of DNS grids in the row direction and the number in the column direction of the image, and the unit is pixel (pix); the matrix Y_DNS of the DNS velocity field grid row coordinates, for any i-row and j-column element Y_DNS(i,j), it corresponds to the pixel row coordinate of the i-row and j-column grid point in the particle image, and the number of rows I and the number of columns J should be consistent with the number of DNS grids in the row direction and the number in the column direction of the image, and the unit is pixel (pix); The number of them in the x-direction is consistent with the number in the column direction, and the unit is pixel (pix); the three-dimensional array U_DNS of the x-direction component of the DNS velocity field (i.e., the row direction of the image), the matrix formed by each time profile corresponds to X_DNS and Y_DNS one by one, that is, the element U_DNS(i,j,t) at any i-row and j-column at time t is the component of the velocity in the x-direction at the pixel position of the row coordinate X_DNS(i,j) and the column coordinate Y_DNS(i,j), and the unit is pix / s; the three-dimensional array V_DNS of the y-direction component of the DNS velocity field (i.e., the column direction of the image), the matrix formed by each time profile corresponds to X_DNS and Y_DNS one by one, that is, the element V_DNS(i,j,t) at any i-row and j-column at time t is the component of the velocity in the y-direction at the pixel position of the row coordinate X_DNS(i,j) and the column coordinate Y_DNS(i,j), and the unit is pix / s.

[0071] The pulsating velocity field reconstructed using this application: The input PIV velocity field and PTV velocity information are used to obtain the reconstructed velocity field u recon (x,t) and v recon (x,t). Figure 6 and Figure 7 The figure shows the distribution histogram of the PIV velocity field error and the reconstructed velocity field error at t = Options.time_plot. The horizontal axis is the reconstruction error of the u component and the v component, in pix / s; the vertical axis is the probability density function value of the error; the thick curve is the probability density value of different error sizes; the thin vertical line is the 3σ (3 times the standard deviation) interval of the two errors. It can be seen that the standard deviation of the error of the traditional PIV velocity field is about 8pix / s, while the pulsating velocity field reconstructed by this application is: The standard deviation of the output reconstructed velocity field is about 5pix / s, which shows that the accuracy of the latter is significantly improved compared with the former.

[0072] Figure 8 The error distribution scatter plot is shown, the horizontal axis is the u component error, and the vertical axis is the v component error. Consider using the probability density function of the normal distribution to fit:

[0073]

[0074] Among them is Yes u 、e v The standard deviation of u 、e v The covariance between them. Then the equal probability density line is the elliptic curve in the figure, which represents the distribution range of the error scatter points.

[0075] The above embodiments show that the reconstruction of the pulsating velocity field based on the flow field tracer particle image has high accuracy and can effectively integrate the PTV velocity information into the PIV velocity field information of the points on the grid. Compared with the traditional PIV method, the error of the reconstructed pulsating velocity field is reduced by nearly 50%.

[0076] In summary, the beneficial effects of the embodiments of the present application are:

[0077] 1. By fusing velocity measurement data from different sources (such as PIV and PTV data), the error of a single measurement method can be significantly reduced, and key features of the flow field, such as vortex structure and streamline distribution, can be more effectively extracted, thereby improving the overall accuracy of velocity measurement, better capturing the details of the flow field, and reducing errors caused by measurement noise.

[0078] 2. Combining data with different temporal and spatial resolutions to achieve velocity field reconstruction with higher temporal and spatial resolution can provide more detailed flow field information.

[0079] 3. Due to the use of the method to determine the number of modes that need to be reconstructed M PTV Through the above steps, the required number of POD modes is effectively determined to achieve accurate correction of the velocity field and improve the credibility of flow field simulation.

[0080] In summary, any step can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor, and no unnecessary restrictions are made here.

[0081] Furthermore, the above-mentioned technical scheme only reflects the preferred technical scheme of the technical scheme of the embodiment of the present application. Some changes that may be made to some parts thereof by technicians in this technical field all reflect the novel principles of the embodiment of the present application. Obviously, technicians in this field can make various changes and modifications to the present application without departing from the scope of the present application.

Claims

1. A fusion processing method for speed measurement data, characterized in that: The method comprises: Use POD analysis to input PIV velocity field data U PIV (x, t), decomposed into a set of discrete PO D modes: Where M is the total number of moments of the PIV velocity field, φ m (x) is the spatial function of the m-th order POD mode, is the time coefficient under POD mode; Input PTV velocity field data U PTV (x, t), calculate the pulsation velocity u of the PTV particle PTV ; For each time t, the POD modal space function is interpolated to the PT V particle coordinates using the cubic spline interpolation method; Using the pulsating velocity u of the PTV particles at the same time PTV , the equation relationship that the POD modal space function and the POD modal time coefficient should satisfy, construct a linear equation system about the time coefficient, and reconstruct the POD modal time coefficient using the least squares method Use POD modal space functions and reconstructed time coefficients Calculate the reconstructed pulsating velocity field: in, Used to characterize the corrected POD time coefficient; With the reconstructed pulsating velocity field: Calculate particle velocity accuracy.

2. The method according to claim 1, characterized in that φ m (x) is the spatial function of the m-th order POD mode, and the set of discrete POD modes obtained by decomposition satisfies the standard orthogonal basis; The spatial modes and time coefficients are calculated using the SVD decomposition method, and the POD modes are arranged in descending order of the modulus of the singular values.

3. The method according to claim 1, characterized in that Used to characterize the corrected POD time coefficient. For a two-dimensional velocity field, it is composed of 2×N PTV (t) equations, including: Among them, φ u，m (x i ,y i ) is used to characterize the POD space function of the u component of the PIV velocity field at point (x i ,y i ) at the interpolation point; Among them, φ v，m (x j ,y j ) is used to characterize the POD space function of the v component of the PIV velocity field at point (x i ,y i ) at the interpolation point.

4. The method according to claim 3, characterized in that The method comprises: Let's write column vector: Convert to matrix form: Among them, in [φ1, φ2, ..., φ M ] matrix, each vector represents a spatial function φ M (t) Interpolation over all PTV particles at time t.

5. The method according to claim 1, characterized in that The time series average of the PIV velocity field at the PTV particle position using cubic spline interpolation as an estimate of the average velocity at each PTV particle.

6. The method according to claim 1, characterized in that The spatially continuous POD modal space function is obtained by using the cubic spline interpolation method, and the value of the continuous function on the PTV particle is used as the estimate of the POD modal space function value corresponding to the PTV particle coordinates.

7. The method according to claim 6, characterized in that Determine the number of modes M that need to be reconstructed PTV , to adjust the correction effect on the velocity field.

8. The method according to claim 7, characterized in that Each time the POD modal time coefficient is reconstructed using the least squares method, the time t is fixed and only the POD modal time coefficient at one time is reconstructed.

Images