A method for extracting blade profile data of gas turbine transient and steady-state experiments

Abstract

A method for extracting the surface data of gas turbine transient and steady-state experimental blade cascades, including the arrangement of experimental blade cascade calibration points, the construction of three-dimensional space information corresponding to the experimental blade cascade, the construction of a projection matrix from three-dimensional space to a two-dimensional pixel plane, the calculation of the pixel plane projection of the three-dimensional blade cascade surface information, the two-dimensional pixel plane region decomposition, the extraction of pixel plane blade cascade projection point data, the three-dimensional reconstruction of the two-dimensional pixel plane, etc. The present invention can simply and effectively obtain the experimental data of the blade tip, suction surface, and pressure surface of the three-dimensional blade cascade according to the experimental requirements under a single infrared camera system, and thus provide accurate experimental results for quantitative analysis of experimental data, numerical verification such as CFD, self-programming, and empirical formulas, and the method is suitable for data extraction of complex blade cascade surfaces. The present invention completes the data extraction of the experimental blade cascade only under a single camera system, has the advantages of low cost, wide application range, and accurate extracted data, and is more in line with the use needs of gas turbine researchers.

Description

Technical Field

[0001] The invention relates to the technical field of gas turbine aerodynamic heat transfer and integrated cooling, and in particular to a method for extracting gas turbine transient and steady-state experimental blade cascade profile data. Background Art

[0002] Gas turbines are important working parts in aircraft engines and heavy-duty gas turbines. Improving the gas turbine inlet temperature can effectively improve their thermal efficiency. In this context, the gas turbine inlet temperature has increased year by year. The turbine inlet temperature of advanced heavy-duty gas turbines has exceeded 1600K, and the turbine inlet gas temperature of military aircraft engines can reach 2000K, which is far higher than the ambient operating temperature (1300K) that the materials currently used to manufacture turbines (such as nickel-based high-temperature alloys) can withstand. In order to ensure the normal operation of gas turbines, it is necessary to combine advanced cooling technology, high-temperature resistant materials, and thermal barrier coatings to design more reliable blade solutions.

[0003] The complex flow and strong mixing inside the turbine blade channel greatly increase the difficulty of turbine blade design. In-depth research on the flow, heat transfer and cooling characteristics of gas turbine blade channels has important industrial significance for the development of modern advanced cooling scheme design. Transient and steady-state experiments are important means to conduct mechanistic research on blade channels, among which infrared thermal imaging (IRT) technology is widely used in various heat transfer and cooling experimental studies at home and abroad. However, due to the inherent geometric characteristics of the blade channel, infrared images often contain data on non-test area and various surfaces of the turbine blade, such as end wall, blade body, blade top, etc. This feature greatly limits the researchers' quantitative analysis of experimental data and extended research based on experimental data.

[0004] In order to solve the above problems, existing technologies often require dual infrared cameras and other auxiliary equipment, which undoubtedly greatly increases the experimental cost and the difficulty of operation for researchers. In addition, the inherent technical characteristics of existing technologies greatly limit their application in turbine blade transient experiments. In addition, modern advanced turbine blades have complex curved surface structures, and existing calculations cannot accurately perform three-dimensional reconstruction and data extraction. Summary of the invention

[0005] In order to overcome the defects of the above technologies, the purpose of the present invention is to provide a method for extracting transient and steady-state experimental blade surface data of a gas turbine. This method completes the data extraction of the experimental blade only under a single-camera system, greatly reducing the cost of experimental equipment and the difficulty of experimental operation, and obtaining more accurate experimental data, which is more in line with the needs of gas turbine experimenters.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for extracting blade cascade profile data of a gas turbine transient and steady-state experiment comprises the following steps;

[0008] Step (1): Arrange multiple groups of calibration points according to the experimental areas of interest (such as suction surface, pressure surface, blade top), and determine the coordinates of the calibration points in three-dimensional space and the coordinates of the two-dimensional pixel plane of the infrared camera, that is, the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration points;

[0009] Step (2): converting the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration point into homogeneous coordinates;

[0010] Step (3): matching the homogeneous coordinates of the three-dimensional space point of the calibration point with the homogeneous coordinates of the two-dimensional image point, thereby constructing a homogeneous coordinate projection relationship between the three-dimensional space point and the two-dimensional pixel point;

[0011] Step (4): solving the projection matrix according to the projection relationship;

[0012] Step (5): Dividing the blade grid based on the experimental blade grid geometry and dividing the blade grid into three-dimensional structured or unstructured grids; the surface types include suction surface, pressure surface, and blade tip;

[0013] Step (6): constructing mesh surface information of the profile based on the blade cascade three-dimensional mesh;

[0014] Step (7): using the projection matrix, projecting the mesh nodes and mesh surface information of the profile onto the two-dimensional pixel plane, obtaining the projection points of the profile mesh on the two-dimensional pixel plane and the distribution of the profile on the two-dimensional pixel plane;

[0015] Step (8): performing image segmentation on the two-dimensional pixel plane according to the profile information, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane;

[0016] Step (9): sequentially traverse the projection points of the profile mesh on the two-dimensional pixel plane and extract pixel information;

[0017] Step (10): Based on the projection relationship between the three-dimensional space information of the profile mesh and the pixel plane, the three-dimensional space reconstruction is performed in reverse order to complete the extraction of the gas turbine transient and steady-state experimental blade surface data.

[0018] Considering the limitations of the infrared window range and viewing angle of the gas turbine heat transfer and cooling test bench, the images taken by the infrared camera often contain information from many areas, and this feature greatly limits the further post-processing and utilization of the experimental results by researchers. Therefore, it is necessary to extract the data of the experimental focus area in order to solve this problem.

[0019] According to the geometric structure of the test piece itself and the shooting conditions of the test bench infrared camera, the test piece is preprocessed before the test, that is, calibration points are arranged on the CAD model and the test piece; the calibration points must meet the requirement of being clearly visible in the image taken by the infrared camera.

[0020] After the gas turbine heat transfer and cooling experiments are completed, it is necessary to extract the pixel coordinates (two-dimensional coordinate system) of the calibration points preset on the test piece based on the images taken by the infrared camera. The coordinate system in the CAD model is regarded as the world coordinate system, and the three-dimensional coordinates of the calibration points are regarded as the real coordinates. Based on the above assumptions, the three-dimensional coordinate values ​​of the calibration points are extracted for camera calibration.

[0021] The step (1) requires that multiple sets of calibration points be arranged in the area of ​​interest of the test piece before the experiment, and that they are clearly visible in the image captured by the infrared camera, and that the coordinates of the calibration points in three-dimensional space and in a two-dimensional pixel plane coordinate system be extracted after the experiment is completed.

[0022] In the step (2), the calibrated coordinate values ​​are augmented to obtain their corresponding homogeneous coordinate values. The specific operation is as follows:

[0023] Assume that the coordinates of each calibration point i in the three-dimensional space and the coordinates in the two-dimensional pixel plane coordinate system are Then the corresponding homogeneous coordinates are

[0024] In the step (3), the three-dimensional space homogeneous coordinate value of the calibration point and the pixel coordinate value (two-dimensional coordinate value) form a corresponding point pair (i.e., a projection point pair); in order to establish a projection relationship from the three-dimensional space to the two-dimensional pixel plane, the present invention adopts a pinhole camera model; the advantage of this model is that all points within the camera depth of field, that is, points within the depth range that the camera can focus on for imaging, can be imaged according to the principle of the pinhole camera; for the gas turbine blade row experiment, the blade row height (blade height) of the test piece is relatively small, that is, the depth range of the focused imaging is within the camera depth of field; and the gas turbine blade row experiment can ignore the influence of perspective and distortion on data extraction. The pinhole camera model can better meet the needs of gas turbine blade row experiment data extraction.

[0025] In the pinhole camera model, in order to construct the projection relationship from three-dimensional space to two-dimensional pixel plane, the camera image plane is first introduced. The distance from the image plane to the camera projection center point O is f, and it is perpendicular to the principal axis (parallel to the XY plane), where f is the focal length of the camera;

[0026] On this image plane, the three-dimensional point p w =(x w ,y w ,z w ) is mapped to its two-dimensional projection point p c =(xc ,y c ) has the following corresponding relationship:

[0027]

[0028] Using homogeneous coordinates, it is expressed as follows:

[0029]

[0030] In the pixel plane, the unit is usually pixel, and there is a situation where the projection center of the camera is not at (0, 0, 0), the main axis is not the Z axis, or the pixel plane is not parallel to the XY plane; for this reason, it is necessary to introduce a transformation matrix to further correct the model. The transformation matrix includes a rotation matrix R and a translation matrix T. Therefore, the method of converting the coordinates in the three-dimensional space into the coordinates in the two-dimensional pixel plane coordinate system is:

[0031]

[0032] Among them, the matrix M is the projection matrix from the three-dimensional space to the pixel plane.

[0033] The step (4) uses a direct linear transformation method to solve the projection matrix M, and the specific solution process is as follows:

[0034]

[0035] Based on this, two linear equations are obtained:

[0036]

[0037] In the formula, m 1 , m 2 , m 3 are the first to third row vectors of the projection matrix M respectively.

[0038] SVD is used to solve the above linear equations, and the obtained singular values ​​are the projection matrix parameters.

[0039] In step (5), in order to fully utilize the known three-dimensional information of the turbine blade grid, the profiles (such as the suction surface, pressure surface, blade tip, etc.) of the three-dimensional model of the turbine blade grid need to be divided, and then divided into three-dimensional structured / unstructured grids. Then, the grid node coordinates and grid line and surface information of each profile are derived.

[0040] In step (7), the projection matrix M solved in step (4) is used to perform projection transformation on the grid node coordinates of each surface in turn, so as to convert them into two-dimensional pixel coordinates, which are hereinafter referred to as grid projection points.

[0041] In the step (8), the image captured in the experiment is segmented based on the line and surface information of the grid projection points, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane.

[0042] In the step (9), the grid projection points are traversed in sequence and the values ​​of the grid projection points on the pixel plane are obtained by using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation, so as to extract the image plane information;

[0043] Assuming that the high-resolution infrared camera used has good processing accuracy, that is, each pixel is represented as a rectangular pixel block; the following relationship exists when searching for values ​​using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation on the two-dimensional pixel plane:

[0044] Nearest neighbor interpolation:

[0045] When the grid projection point falls within a pixel block, assume that the coordinates of the four vertices of the pixel block are Q 11 (x 1 ,y 1 ), Q 21 (x 2 ,y 1 ), Q 12 (x 1 ,y 1 ), Q 22 (x 2 ,y 2 ), calculate the grid projection point P h =(x h ,y h ) and the Cartesian distance between the four vertices of the pixel block. Taking the first vertex as an example, the distance is And assign the value of the nearest pixel block vertex to the grid projection point;

[0046] Bilinear interpolation:

[0047] When the grid projection point falls within a pixel block, linear interpolation is first performed twice in the x direction, and then linear interpolation is performed once in the y direction. The calculation formula is as follows:

[0048]

[0049] Bicubic interpolation:

[0050] This method uses 16 points around the value to be found to perform cubic interpolation, and the calculation formula is as follows:

[0051]

[0052] In step (10), based on the projection relationship between the grid three-dimensional coordinates and the grid projection points, that is, the relationship between the grid points and their grid projection points after projective transformation, the grid projection points are back-projected into the three-dimensional space for three-dimensional reconstruction, and the pixel information (experimental results) at the grid projection points is assigned to the three-dimensional space coordinates; thus, the extraction of the transient and steady-state experimental cascade profile data of the gas turbine can be completed.

[0053] Advantages of the present invention:

[0054] (1) The present invention completes the extraction and three-dimensional reconstruction of experimental data based on a single-camera system. Compared with the existing dual-camera system and even multi-camera systems, in the case where high-precision infrared cameras are currently expensive, the cost of building the experimental platform is significantly reduced.

[0055] (2) Since the calibration points on the surface of the test piece and the profile information of the three-dimensional model are combined, high-precision three-dimensional reconstruction can be achieved, that is, it is applicable to more complex gas turbine cascade profiles (such as non-axisymmetric end walls, shaped convergent end walls, etc.) rather than being limited to the reconstruction of simple planes, providing convenience and possibility for researchers to process high-precision and diverse experimental data.

[0056] (3) It is applicable to gas turbine transient and steady-state cascade experiments. The three-dimensional reconstruction and data of the prior art require multiple adjustments of the camera pose (i.e., multiple shooting angles, multiple focal lengths, etc.) during the experiment, which does not meet the requirements of continuous shooting at a high frame rate of the transient experiment camera. Since the true geometric information of the three-dimensional model is introduced, high-precision three-dimensional reconstruction and data extraction can be achieved while meeting the requirements of continuous shooting (transient experiment). Brief description of the drawings

[0057] Figure 1 is a flowchart of a method for extracting the cascade profile data of a gas turbine transient and steady-state experiment of the present invention.

[0058] Figure 2 is a schematic diagram of a pinhole camera model and its projection principle.

[0059] Figure 3 is a schematic diagram of the infrared image shooting of the test piece in the embodiment.

[0060] Figure 4 is a schematic diagram of the calibration points of the test piece CAD model in the embodiment.

[0061] Figure 5 is a schematic diagram of the cascade profile division in the pixel plane in the embodiment.

[0062] Figure 6 is a schematic diagram of three-dimensional reconstruction in the three-dimensional pixel plane. Detailed implementation manners

[0063] The present invention is further described in detail below in conjunction with embodiments.

[0064] Figure 1 The flowchart of the method for extracting the profile data of the transient and steady-state gas turbine blade cascade is shown in FIG. According to the experimental areas of interest (such as the suction surface, pressure surface, and blade tip), multiple groups of calibration points are arranged, and the coordinate points of the calibration points in the three-dimensional space and the coordinates in the two-dimensional pixel plane of the infrared camera are determined respectively, that is, the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration points. The three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration points are converted into homogeneous coordinates. The corresponding coordinate projection relationship is constructed based on the spatial point coordinates and the image point coordinates of the calibration points. The projection matrix from the test piece to the pixel plane is established based on the pinhole camera model, and the projection matrix is ​​solved by the direct linear change method. The blade profile types (such as the suction surface, pressure surface, blade tip, etc.) are divided based on the experimental blade cascade geometry, and the three-dimensional structured / unstructured grids are divided, and then the mesh surface information of the three-dimensional mesh profile of the blade cascade is constructed. Based on the constructed projection matrix, the profile mesh information is projected to the two-dimensional pixel plane. The two-dimensional pixel plane is image segmented according to the profile information, that is, the suction surface area, pressure surface area, blade tip area, etc. in the pixel plane. The projection points of the profile mesh on the two-dimensional pixel plane are traversed in sequence, and the pixel information is extracted. According to the projection relationship between the three-dimensional spatial information of the profile mesh and the pixel plane, the three-dimensional space is reconstructed in reverse to complete the extraction of the gas turbine transient and steady-state experimental blade surface data.

[0065] In one embodiment of the present invention, a heat transfer cooling transient experiment is carried out using a gas turbine rotor blade, and the test piece is preprocessed and experimental data is extracted based on the present invention.

[0066] A method for extracting blade cascade profile data of a gas turbine transient and steady-state experiment comprises the following steps;

[0067] Step (1): Arrange multiple groups of calibration points according to the experimental areas of interest (such as suction surface, pressure surface, blade top), and determine the coordinates of the calibration points in three-dimensional space and the coordinates of the two-dimensional pixel plane of the infrared camera, that is, the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration points;

[0068] Step (2): converting the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration point into homogeneous coordinates;

[0069] Step (3): matching the homogeneous coordinates of the three-dimensional space point of the calibration point with the homogeneous coordinates of the two-dimensional image point, thereby constructing a homogeneous coordinate projection relationship between the three-dimensional space point and the two-dimensional pixel point;

[0070] Step (4): solving the projection matrix according to the projection relationship;

[0071] Step (5): Dividing the blade grid based on the experimental blade grid geometry and dividing the blade grid into three-dimensional structured or unstructured grids; the surface types include suction surface, pressure surface, and blade tip;

[0072] Step (6): constructing mesh surface information of the profile based on the blade cascade three-dimensional mesh;

[0073] Step (7): using the projection matrix, projecting the mesh nodes and mesh surface information of the profile onto the two-dimensional pixel plane, obtaining the projection points of the profile mesh on the two-dimensional pixel plane and the distribution of the profile on the two-dimensional pixel plane;

[0074] Step (8): performing image segmentation on the two-dimensional pixel plane according to the profile information, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane;

[0075] Step (9): sequentially traverse the projection points of the profile mesh on the two-dimensional pixel plane and extract pixel information;

[0076] Step (10): Based on the projection relationship between the three-dimensional space information of the profile mesh and the pixel plane, the three-dimensional space reconstruction is performed in reverse order to complete the extraction of the gas turbine transient and steady-state experimental blade surface data.

[0077] Considering the limitations of the infrared window range and viewing angle of the gas turbine heat transfer and cooling test bench, the images taken by the infrared camera often contain information from many areas, and this feature greatly limits the further post-processing and utilization of the experimental results by researchers. Therefore, it is necessary to extract the data of the experimental focus area in order to solve this problem.

[0078] According to the geometric structure of the test piece itself and the shooting conditions of the test bench infrared camera, the test piece is preprocessed before the test, that is, calibration points are arranged on the CAD model and the test piece; the calibration points must meet the requirement of being clearly visible in the image taken by the infrared camera.

[0079] After the gas turbine heat transfer and cooling experiments are completed, it is necessary to extract the pixel coordinates (two-dimensional coordinate system) of the calibration points preset on the test piece based on the images taken by the infrared camera. The coordinate system in the CAD model is regarded as the world coordinate system, and the three-dimensional coordinates of the calibration points are regarded as the real coordinates. Based on the above assumptions, the three-dimensional coordinate values ​​of the calibration points are extracted for camera calibration.

[0080] The step (1) requires that multiple sets of calibration points be arranged in the area of ​​interest of the test piece before the experiment, and that they are clearly visible in the image captured by the infrared camera, and that the coordinates of the calibration points in three-dimensional space and in a two-dimensional pixel plane coordinate system be extracted after the experiment is completed.

[0081] In the step (2), the calibrated coordinate values ​​are augmented to obtain their corresponding homogeneous coordinate values. The specific operation is as follows:

[0082] Assume that the coordinates of each calibration point i in the three-dimensional space and the coordinates in the two-dimensional pixel plane coordinate system are Then the corresponding homogeneous coordinates are

[0083] In the step (3), the three-dimensional space homogeneous coordinate value of the calibration point and the pixel coordinate value (two-dimensional coordinate value) form a corresponding point pair (i.e., a projection point pair); in order to establish a projection relationship from the three-dimensional space to the two-dimensional pixel plane, the present invention adopts a pinhole camera model; the advantage of this model is that all points within the camera depth of field, that is, points within the depth range that the camera can focus on for imaging, can be imaged according to the principle of the pinhole camera; for the gas turbine blade row experiment, the blade row height (blade height) of the test piece is relatively small, that is, the depth range of the focused imaging is within the camera depth of field; and the gas turbine blade row experiment can ignore the influence of perspective and distortion on data extraction. The pinhole camera model can better meet the needs of gas turbine blade row experiment data extraction.

[0084] In the pinhole camera model, in order to construct the projection relationship from three-dimensional space to two-dimensional pixel plane, the camera image plane is first introduced. The distance from the image plane to the camera projection center point O is f, and it is perpendicular to the principal axis (parallel to the XY plane), where f is the focal length of the camera;

[0085] On this image plane, the three-dimensional point p w =(x w ,y w ,z w ) is mapped to its two-dimensional projection point p c =(x c ,y c ) has the following corresponding relationship:

[0086]

[0087] Using homogeneous coordinates, it is expressed as follows:

[0088]

[0089] In the pixel plane, the unit is usually pixel, and there is a situation where the projection center of the camera is not at (0, 0, 0), the main axis is not the Z axis, or the pixel plane is not parallel to the XY plane; for this reason, it is necessary to introduce a transformation matrix to further correct the model. The transformation matrix includes a rotation matrix R and a translation matrix T. Therefore, the method of converting the coordinates in the three-dimensional space into the coordinates in the two-dimensional pixel plane coordinate system is:

[0090]

[0091] Among them, the matrix M is the projection matrix from the three-dimensional space to the pixel plane.

[0092] The step (4) uses a direct linear transformation method to solve the projection matrix M, and the specific solution process is as follows:

[0093]

[0094] Based on this, two linear equations are obtained:

[0095]

[0096] In the formula, m 1 , m 2 , m 3 are the first to third row vectors of the projection matrix M respectively.

[0097] SVD is used to solve the above linear equations, and the obtained singular values ​​are the projection matrix parameters.

[0098] In step (5), in order to fully utilize the known three-dimensional information of the turbine blade grid, the profiles (such as the suction surface, pressure surface, blade tip, etc.) of the three-dimensional model of the turbine blade grid need to be divided, and then divided into three-dimensional structured / unstructured grids. Then, the grid node coordinates and grid line and surface information of each profile are derived.

[0099] In step (7), the projection matrix M solved in step (4) is used to perform projection transformation on the grid node coordinates of each surface in turn, so as to convert them into two-dimensional pixel coordinates, which are hereinafter referred to as grid projection points.

[0100] In the step (8), the image captured in the experiment is segmented based on the line and surface information of the grid projection points, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane.

[0101] In the step (9), the grid projection points are traversed in sequence and the values ​​of the grid projection points on the pixel plane are obtained by using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation, so as to extract the image plane information;

[0102] Assuming that the high-resolution infrared camera used has good processing accuracy, that is, each pixel is represented as a rectangular pixel block; the following relationship exists when searching for values ​​using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation on the two-dimensional pixel plane:

[0103] Nearest neighbor interpolation:

[0104] When the grid projection point falls within a pixel block, assume that the coordinates of the four vertices of the pixel block are Q 11 (x 1 ,y 1 ), Q 21 (x 2 ,y 1 ), Q12 (x 1 ,y 1 ), Q 22 (x 2 ,y 2 ), calculate the grid projection point P h =(x h ,y h ) and the Cartesian distance between the four vertices of the pixel block. Taking the first vertex as an example, the distance is And assign the value of the nearest pixel block vertex to the grid projection point;

[0105] Bilinear interpolation:

[0106] When the grid projection point falls within a pixel block, linear interpolation is first performed twice in the x direction, and then linear interpolation is performed once in the y direction. The calculation formula is as follows:

[0107]

[0108] Bicubic interpolation:

[0109] This method uses 16 points around the value to be found to perform cubic interpolation, and the calculation formula is as follows:

[0110]

[0111] In the step (10), based on the projection relationship between the grid three-dimensional coordinates and the grid projection points, that is, the relationship between the grid points and the grid projection points after projection transformation, the grid projection points are back-projected into the three-dimensional space for three-dimensional reconstruction, and the pixel information (experimental results) at the grid projection points are assigned to the three-dimensional space coordinates; thereby, the extraction of the transient and steady-state experimental blade surface data of the gas turbine can be completed. Embodiment:

[0112] Reference Figure 1 This example provides a method for extracting blade profile data of a gas turbine transient and steady-state experiment, which specifically includes the following steps:

[0113] 1. Layout of calibration points for cascade test pieces:

[0114] Figure 3 A schematic diagram of the test piece in the test bench area and the infrared camera shooting angle when using this method is given. Figure 3 As shown, in the embodiment, a single infrared camera is used to photograph the surface of the test piece. Since this method only requires a single infrared camera to complete the extraction of experimental data, it effectively reduces the need for the use of multiple expensive high-resolution infrared cameras in the traditional method, thereby effectively reducing the cost of setting up the test bench.

[0115] In this embodiment, the cascade channel and the cascade blade tip area have a large visible range and are the main areas of concern, so multiple groups of calibration points are arranged in the cascade channel and the cascade blade tip area. Figure 4 As shown in , the calibration points are evenly distributed throughout the entire area of ​​interest, ensuring that the calibration information is sufficient and accurate. Figure 5 As shown in the figure, in the image taken by the infrared camera, multiple groups of calibration points are also clearly visible, which proves the rationality of the layout of the calibration points.

[0116] 2. Construct a projection matrix from three-dimensional space to pixel plane:

[0117] Extract separately Figure 5 and Figure 4 The coordinate values ​​of the calibrated points in the pixel plane and the 3D CAD model shown in , and the calibrated coordinate values ​​are augmented to obtain their corresponding homogeneous coordinate values, the specific operations are as follows:

[0118] Assume that the coordinates of each calibration point i in the three-dimensional space and the coordinates in the two-dimensional pixel plane coordinate system are Then the corresponding homogeneous coordinates are

[0119] The three-dimensional space homogeneous coordinate values ​​of the calibration points and the pixel coordinate values ​​(two-dimensional coordinate values) form corresponding point pairs (i.e., projection point pairs). Based on the pinhole camera model, the projection relationship from the three-dimensional space to the two-dimensional plane is constructed, and the projection matrix is ​​solved using the direct transformation method. The schematic diagram of the pinhole camera projection principle is shown in the figure. Figure 2 shown.

[0120] 3. Turbine blade profile and mesh division:

[0121] In order to make full use of the known three-dimensional information of the turbine blade row, the profile of the three-dimensional model of the turbine blade row needs to be divided. In this embodiment, the turbine blade row is geometrically divided into the blade row channel, the blade body region, the blade tip region, the inlet section, the outlet section, etc. Then, in this embodiment, the three-dimensional structured grid is divided to ensure accurate and full use of the three-dimensional profile information, and the grid nodes, grid lines, and surface information of the important areas (blade row channel, blade body region, blade tip region) are exported.

[0122] 4. Calculate the pixel plane projection of the three-dimensional cascade surface information:

[0123] In this embodiment, the three-dimensional grid node coordinates of the cascade channel, the blade body region, and the blade tip region are projected onto the two-dimensional pixel plane through a projection matrix to obtain grid projection points of the grid nodes on the pixel plane.

[0124] 5. Pixel plane area decomposition:

[0125] In this embodiment, the obtained infrared image area is segmented according to the two-dimensional pixel plane projection results of the blade channel profile, blade root profile, and blade tip profile, namely, the blade tip area, blade cascade area, air inlet section, outlet section, etc.

[0126] 6. Pixel plane projection point data extraction:

[0127] Traverse each grid projection point and use nearest neighbor interpolation, bilinear interpolation, or bicubic interpolation to obtain the pixel value of the projection point in the infrared image.

[0128] 7. 3D reconstruction of 2D pixel plane:

[0129] like Figure 6 As shown, the projection relationship between the grid nodes and the grid projection points of the blade area is used to perform three-dimensional reconstruction from the two-dimensional pixel plane to the three-dimensional space, and the pixel values ​​of the grid projection points are assigned to the corresponding grid nodes of the blade area, thereby realizing the three-dimensional reconstruction of the blade area, and the extraction of the temperature distribution data of the blade area on the surface of the test piece is realized by extracting the pixel values ​​of the blade area in the two-dimensional pixel plane.

Claims

1. A method for extracting blade profile data of gas turbine transient and steady-state experiments, It is characterized in that The steps include: Step (1): Arrange multiple groups of calibration points according to the experimental area of ​​interest, and determine the coordinates of the calibration points in three-dimensional space and the coordinates of the two-dimensional pixel plane of the infrared camera, that is, the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration points; Step (2): converting the three-dimensional space point coordinates and the two-dimensional image point coordinates of the calibration point into homogeneous coordinates; Step (3): matching the homogeneous coordinates of the three-dimensional space point of the calibration point with the homogeneous coordinates of the two-dimensional image point, thereby constructing a homogeneous coordinate projection relationship between the three-dimensional space point and the two-dimensional pixel point; Step (4): solving the projection matrix according to the projection relationship; Step (5): Dividing the blade grid based on the experimental blade grid geometry and dividing the blade grid into three-dimensional structured or unstructured grids; the types of the profiles include suction surface, pressure surface, and blade tip; Step (6): constructing mesh surface information of a profile based on the three-dimensional structured or unstructured mesh; Step (7): using the projection matrix, projecting the mesh nodes and mesh surface information of the profile onto the two-dimensional pixel plane, obtaining the projection points of the profile mesh on the two-dimensional pixel plane and the distribution of the profile on the two-dimensional pixel plane; Step (8): performing image segmentation on the two-dimensional pixel plane according to the profile information, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane; Step (9): sequentially traverse the projection points of the profile mesh on the two-dimensional pixel plane and extract pixel information; Step (10): Based on the projection relationship between the three-dimensional space information of the profile mesh and the pixel plane, the three-dimensional space reconstruction is performed in reverse order to complete the extraction of the gas turbine transient and steady-state experimental blade surface data.

2. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that The step (1) requires that multiple sets of calibration points be arranged in the area of ​​interest of the test piece before the experiment, and that they are clearly visible in the image captured by the infrared camera, and that the coordinates of the calibration points in three-dimensional space and in a two-dimensional pixel plane coordinate system be extracted after the experiment is completed.

3. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In the step (2), the calibrated coordinate values ​​are augmented to obtain their corresponding homogeneous coordinate values. The specific operation is as follows: Assume that the coordinates of each calibration point i in the three-dimensional space and the coordinates in the two-dimensional pixel plane coordinate system are Then the corresponding homogeneous coordinates are 4. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 3, It is characterized in that In the step (3), the three-dimensional spatial homogeneous coordinate values ​​of the calibration points and the pixel coordinate values, i.e., the two-dimensional coordinate values, are used to form corresponding point pairs, i.e., projection point pairs; using a pinhole camera model, all points within the depth range within which the camera can focus on imaging can be imaged according to the principle of a pinhole camera; In the pinhole camera model, in order to construct the projection relationship from the three-dimensional space to the two-dimensional pixel plane, the camera image plane is first introduced. The distance from the image plane to the camera projection center point O is f, and it is perpendicular to the main axis, where f is the focal length of the camera; On this image plane, the three-dimensional point p w =(x w ,y w ,z w ) is mapped to its two-dimensional projection point p c =(x c ,y c ) has the following corresponding relationship: Using homogeneous coordinates, it is expressed as follows: In the pixel plane, the unit is usually pixels, and there are cases where the projection center of the camera is not at (0, 0, 0), and the principal axis is not the Z axis or the pixel plane is not parallel to the XY plane; for this reason, it is necessary to introduce a transformation matrix to further correct the model. The transformation matrix includes a rotation matrix R and a translation matrix T; therefore, the method of converting the coordinates in the three-dimensional space into the coordinates in the two-dimensional pixel plane coordinate system is: Among them, the matrix M is the projection matrix from the three-dimensional space to the pixel plane.

5. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 4, It is characterized in that The step (4) uses a direct linear transformation method to solve the projection matrix M, and the specific solution process is as follows: Based on this, two linear equations are obtained: where m 1 , m 2 , m 3 are the row vectors of the first to the third rows of the projection matrix M, respectively; SVD is used to solve the above linear equations, and the obtained singular values ​​are the projection matrix parameters.

6. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In the step (5), in order to fully utilize the known three-dimensional information of the turbine blade grid, it is necessary to divide the surface of the three-dimensional model of the turbine blade grid, and then divide it into three-dimensional structured / unstructured grids, and then derive the grid node coordinates and grid line and surface information of each surface.

7. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In step (7), the projection matrix M solved in step (4) is used to perform projection transformation on the grid node coordinates of each surface in turn to convert them into two-dimensional pixel coordinates.

8. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In the step (8), the image captured in the experiment is segmented based on the line and surface information of the grid projection points, namely, the suction surface area, the pressure surface area, and the blade tip area in the pixel plane.

9. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In the step (9), the grid projection points are traversed in sequence and the values ​​of the grid projection points on the pixel plane are obtained by using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation, so as to extract the image plane information; Assuming that the high-resolution infrared camera used has good processing accuracy, that is, each pixel is represented as a rectangular pixel block; the following relationship exists when searching for values ​​using nearest neighbor interpolation / bilinear interpolation / bicubic interpolation on the two-dimensional pixel plane: Nearest neighbor interpolation: When the grid projection point falls within a pixel block, assume that the coordinates of the four vertices of the pixel block are Q 11 (x 1 ,y 1 ), Q 21 (x 2 ,y 1 ), Q 12 (x 1 ,y 2 ), Q 22 (x 2 ,y 2 ), calculate the grid projection point P h =(x h ,y h ) and the Cartesian distance between the four vertices of the pixel block. Taking the first vertex as an example, the distance is And assign the value of the nearest pixel block vertex to the grid projection point; Bilinear interpolation: When the grid projection point falls within a pixel block, linear interpolation is first performed twice in the x direction, and then linear interpolation is performed once in the y direction. The calculation formula is as follows: Bicubic interpolation: This method uses 16 points around the value to be found to perform cubic interpolation, and the calculation formula is as follows:

10. A method for extracting blade profile data of a gas turbine transient and steady-state experiment according to claim 1, It is characterized in that In the step (10), based on the projection relationship between the grid three-dimensional coordinates and the grid projection points, that is, the relationship between the grid points and the grid projection points after projection transformation, the grid projection points are back-projected into the three-dimensional space for three-dimensional reconstruction, and the pixel information at the grid projection points is assigned to the three-dimensional space coordinates; This will enable the extraction of gas turbine blade profile data for transient and steady-state experiments to be completed.

Images