Turbine blade surface model generation method based on CT multi-directional slice data fusion

Abstract

The application provides a turbine blade surface model generation method based on CT multidirectional slice data fusion, acquires multidirectional slice point cloud data of turbine blade horizontal scanning direction and vertical scanning direction point cloud data through CT scanning, carries out initial registration on the pretreated turbine blade horizontal and vertical direction scanning slice point cloud data by using a sampling consistency initial registration algorithm, sets convergence precision, applies an iterative closest point algorithm to splice and fuse, and obtains complete turbine blade structure point cloud data; finally, the simplified point cloud is projected to a two-dimensional plane through a normal line, a Delaunay growth algorithm is adopted to establish a triangular mesh for the point cloud, the two-dimensional triangular mesh is mapped back to the point cloud three-dimensional space, new triangles are continuously generated outward, and all triangular meshes are generated. The method fuses the obtained CT multidirectional slice data in the initial registration + fine registration mode, improves the point cloud data precision, and greatly improves the turbine blade surface model generation quality.

Description

Technical Field

[0001] This invention belongs to the field of digital inspection technology for aero-engine turbine blades, specifically relating to a method for generating turbine blade surface models based on CT multi-directional slice data fusion. Background Technology

[0002] Aero engines are the power source of aircraft and a core component of them. The quality of engine blades directly affects engine performance, and the machining accuracy and manufacturing level of the blades are important factors affecting engine performance, safety, and lifespan. Turbine blades, in particular, are characterized by high operating temperatures, complex stresses, and high requirements for manufacturing precision.

[0003] Currently, turbine blades are generally formed using precision investment casting. However, their geometry inevitably differs from the design specifications during the overall forming process, and defects such as cracks, inclusions, or holes may occur. Furthermore, due to the high geometric complexity of the turbine blade surface, machining is difficult, and certain machining errors may exist in the blade profile during processing. To improve the high-temperature resistance of turbine blades during operation, a heat transfer enhancement technology using an impact-and-turbulence-column double-wall cooling structure is employed.

[0004] As mentioned above, the accuracy of the profile and internal structure of hollow turbine blades is crucial. Therefore, a comprehensive inspection of the turbine blade structure is usually performed after machining. Traditional digital inspection of turbine blades typically uses CT tomographic image reconstruction technology to construct the surface structure of aero-engine turbine blades and detect whether there are machining errors or forming defects. Summary of the Invention

[0005] The 3D point cloud data obtained by CT slice image reconstruction technology has extremely high accuracy. However, experimental research has found that when using this technology to reconstruct the surface model of a hollow turbine blade, if only the point cloud data obtained from a single slice direction is used, there will be gaps in the point cloud data in the direction perpendicular to the scanning direction due to the slicing. This results in the loss of key structural feature data in the CT slice data in the scanning direction, and there are a large number of redundant points, which seriously affects the generation effect of the turbine blade surface model and reduces the detection accuracy of the turbine blade.

[0006] To address the aforementioned issues, this invention provides a method for generating turbine blade surface models based on CT multi-directional slice data fusion. Multi-directional slice data is obtained through CT scanning. The data in the horizontal and vertical directions are registered using a sampling consistency initial registration algorithm and an iterative nearest point algorithm to complement the point cloud data, thereby obtaining turbine blade point cloud data containing more comprehensive key structural features and generating a high-precision turbine blade surface model.

[0007] The technical solution of this invention is: a method for generating turbine blade surface models based on CT multi-directional slice data fusion, comprising the following steps:

[0008] Step 1: Obtain multi-directional slice point cloud data of turbine blades through CT scanning. The multi-directional slice point cloud data includes point cloud data in the horizontal scanning direction and point cloud data in the vertical scanning direction.

[0009] Step 2: Preprocess the obtained multi-directional slice point cloud data;

[0010] Step 3: Perform point cloud stitching and fusion on the preprocessed CT multi-directional slice point cloud data to obtain the stitched and fused overall point cloud data;

[0011] Step 3.1: Import the horizontal and vertical scan slice point cloud data of the turbine blade, and perform initial registration on the horizontal and vertical scan slice point cloud data of the turbine blade using the sample consistency initial registration algorithm to obtain the initial transformation matrix;

[0012] Step 3.2: Using the multi-directional slice point cloud data and the initial transformation matrix, obtain the overall point cloud data after initial registration;

[0013] Step 3.3: Using the overall point cloud data and the initial transformation matrix, apply the iterative nearest point algorithm to select corresponding point pairs in the multi-directional slice point cloud data, set the convergence precision, and calculate the optimal rigid body transformation;

[0014] If the convergence accuracy is not met, repeat step 3.3;

[0015] If the convergence accuracy is met, proceed to step 4;

[0016] Step 4: After stitching and merging the overall point cloud data, simplify the point cloud by retaining features and remove redundant points from the turbine blade point cloud model.

[0017] Step 5: Use the simplified overall point cloud data of the turbine blade to generate a surface model of the turbine blade.

[0018] Furthermore, the point cloud data in the horizontal scanning direction is point cloud data parallel to the turbine blade edge, and the point cloud data in the vertical scanning direction is point cloud data perpendicular to the turbine blade edge.

[0019] Furthermore, in step 3.3, the convergence accuracy is the mean square error of the point cloud data iteration, and the range is set to [0.001, 0.5].

[0020] Furthermore, step 4, the method for simplifying the overall point cloud data after stitching and fusion, specifically includes the following steps:

[0021] Step 4.1: Calculate the eigenvalues ​​of the normal and curvature of the fused turbine blade CT multi-directional slice point cloud;

[0022] Step 4.2: Traverse the points in the point cloud, determine whether they belong to feature points based on feature values, and classify the point cloud into feature points and non-feature points;

[0023] Step 4.3: Retain the feature points of the point cloud and simplify the non-feature points.

[0024] Furthermore, in step 5, the specific steps for generating the turbine blade surface model are as follows:

[0025] Step 5.1: Project the simplified point cloud onto a two-dimensional plane using normals, and use the Delaunay growth algorithm to create a triangular mesh for the point cloud;

[0026] Step 5.2: Map the two-dimensional triangular mesh back to the three-dimensional point cloud space using the spatial topological relationships within the neighborhood;

[0027] Step 5.3: Repeat steps 5.1 and 5.2 to continuously generate new triangles outwards until all triangular meshes are generated, and output the turbine blade surface model.

[0028] Furthermore, in step 2, the preprocessing of the point cloud data includes removing noise points and NAN points.

[0029] The effects of this invention are:

[0030] 1. This invention uses CT scans to acquire multi-directional slice point cloud data, employing both horizontal and vertical scan slice point cloud data from turbine blades. This avoids the problem of missing point cloud data in directions perpendicular to the scanning direction caused by using only point cloud data obtained from a single direction of CT slices.

[0031] 2. In this invention, an initial registration + fine registration method is adopted. First, the point cloud data of the horizontal and vertical scan slices of the turbine blade are initially registered using the sample consistency initial registration algorithm. Then, by setting the convergence accuracy, the iterative nearest point algorithm is applied to stitch and fuse the data to obtain the point cloud data of the complete turbine blade structure. This avoids the possibility of getting stuck in local optima when using fine registration directly, and greatly improves the accuracy of the point cloud data.

[0032] 3. The method of the present invention is applicable to high-precision digital inspection of turbine blades, with high reliability, and greatly improves the quality of turbine blade surface model generation. Attached Figure Description

[0033] Figure 1 Flowchart of the method for generating turbine blade surface model of the present invention;

[0034] Figure 2 Figure 1 shows a comparison of the generated surface models based on CT scans. Figure 2 shows the generated surface model in the horizontal scanning direction; Figure 3 shows the generated surface model in the vertical scanning direction; and Figure 4 shows the generated surface model using the CT multi-directional slice fusion method of the present invention. Detailed Implementation

[0035] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0036] Reference Figures 1-2 Taking standard parts including curved surfaces and standard parts containing cylinders and cubes as examples, CT scans are performed to simulate the curved surface features of turbine blades and generate surface models. The specific implementation steps are as follows:

[0037] Step 1: Obtain multi-directional slice point cloud data of turbine blades through CT scanning;

[0038] Turbine blade parts are scanned using an industrial cone-beam CT scanner to obtain point cloud data in both the horizontal and vertical scanning directions.

[0039] The point cloud data in the horizontal scanning direction is the point cloud data parallel to the turbine blade edge, and the point cloud data in the vertical scanning direction is the point cloud data perpendicular to the turbine blade edge.

[0040] Step 2: Perform preprocessing on the multi-directional slice point cloud data of the turbine blades obtained in Step 1;

[0041] Point cloud data preprocessing includes removing noise points and NAN points.

[0042] Step 3: Perform point cloud stitching and fusion on the preprocessed CT multi-directional slice point cloud data from Step 2;

[0043] To avoid getting stuck in local optima when using fine registration directly, this invention uses an initial registration + fine registration approach to register point cloud data.

[0044] Step 3.1: Import the horizontal and vertical scan slice point cloud data of the turbine blade, and perform initial registration on the horizontal and vertical scan slice point cloud data of the turbine blade using the sample consistency initial registration algorithm, and output the transformation matrix corresponding to the minimum registration error as the initial transformation matrix.

[0045] Step 3.2: Using the initial transformation matrix obtained from the initial registration and the multi-directional slice point cloud data of the part, obtain the overall point cloud data of the part after initial registration;

[0046] Step 3.3: Using the initial transformation matrix after initial registration and the overall point cloud data of the part, and selecting the corresponding relationship point pairs of the multi-directional slice point cloud data of the part according to the iterative nearest point algorithm, calculate the optimal rigid body transformation until the convergence accuracy requirement is met, and obtain the overall point cloud data of the turbine blade after splicing and fusion. If the convergence accuracy requirement is not met, repeat the selection of corresponding relationship point pairs and calculate the optimal rigid body transformation.

[0047] The convergence accuracy is the mean square error of the point cloud data iteration, and the range is set to [0.001, 0.5].

[0048] In this embodiment, the convergence accuracy is set to 0.01, which meets the accuracy requirement and obtains the overall point cloud data of the parts after splicing and fusion.

[0049] Step 4: Calculate the feature values ​​of the multi-directional slice point cloud of the turbine blade after fusion. For the overall point cloud data after stitching and fusion, use a feature-preserving point cloud simplification method to remove redundant points in the part point cloud model. Specific steps include:

[0050] Step 4.1: Calculate the eigenvalues ​​of the normal and curvature of the point cloud;

[0051] Step 4.2: Traverse the points in the point cloud, determine whether the point cloud belongs to feature points based on the feature values, and classify the point cloud into feature points and non-feature points;

[0052] Step 4.3: Retain the feature points of the point cloud and simplify the non-feature points.

[0053] The point cloud is determined based on its feature values. If it is a feature point, the feature points are retained; otherwise, the non-feature points are simplified.

[0054] Step 5: Using the simplified overall point cloud data, generate a turbine blade surface model. The specific steps are as follows:

[0055] 5.1 Project the simplified point cloud onto a two-dimensional plane using normals, and use the Delaunay growth algorithm to create a triangular mesh for the point cloud;

[0056] 5.2. Map the two-dimensional triangular mesh back to the three-dimensional space of the point cloud through the spatial topological relationships within the neighborhood;

[0057] 5.3 Repeat steps 5.1 and 5.2 to traverse all point clouds and continuously generate new triangles outwards until all triangular meshes are generated, and output the turbine blade surface model.

[0058] Figure 2This paper compares the surface model generation methods using CT multi-directional slice data fusion. Figure (a) shows the surface model generated in the horizontal CT scan direction, Figure (b) shows the surface model generated in the vertical CT scan direction, and Figure (c) shows the surface model generated by CT multi-directional slice fusion. As can be seen from the figures, Figure (a) uses point cloud data obtained from the horizontal direction of the CT slice, resulting in gaps in point cloud data in the vertical direction perpendicular to the scan direction due to slicing, leading to an incomplete surface model. Figure (b) uses point cloud data obtained from the vertical direction of the CT slice, resulting in gaps in point cloud data in the horizontal direction perpendicular to the scan direction, also leading to an incomplete surface model. Figure (c) uses both horizontal and vertical data for registration, complementing the point cloud data, resulting in a complete and high-quality part surface model.

[0059] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention, and such modifications or substitutions should all be covered within the protection scope of the present invention.

Claims

1. A method for generating turbine blade surface models based on CT multi-directional slice data fusion, characterized in that, Includes the following steps: Step 1: Obtain multi-directional slice point cloud data of turbine blades through CT scanning. The multi-directional slice point cloud data includes point cloud data in the horizontal scanning direction and point cloud data in the vertical scanning direction. Step 2: Preprocess the obtained multi-directional slice point cloud data; Step 3: Perform point cloud stitching and fusion on the preprocessed CT multi-directional slice point cloud data to obtain the stitched and fused overall point cloud data; Step 3.1: Import the horizontal and vertical scan slice point cloud data of the turbine blade, and perform initial registration on the horizontal and vertical scan slice point cloud data of the turbine blade using the sample consistency initial registration algorithm to obtain the initial transformation matrix; Step 3.2: Using the multi-directional slice point cloud data and the initial transformation matrix, obtain the overall point cloud data after initial registration; Step 3.3: For the overall point cloud data and the initial transformation matrix, apply the iterative nearest point algorithm to select the corresponding point pairs of the multi-directional slice point cloud data, set the convergence precision, and calculate the optimal rigid body transformation; If the convergence accuracy is not met, repeat step 3.3; If the convergence accuracy is met, proceed to step 4; Step 4: After stitching and merging the overall point cloud data, simplify the point cloud by retaining features and remove redundant points from the turbine blade point cloud model. Step 5: Generate a turbine blade surface model using the simplified overall point cloud data of the turbine blade.

2. The method for generating a turbine blade surface model based on CT multi-directional slice data fusion according to claim 1, characterized in that, The point cloud data in the horizontal scanning direction is the point cloud data parallel to the turbine blade edge, and the point cloud data in the vertical scanning direction is the point cloud data perpendicular to the turbine blade edge.

3. The method for generating turbine blade surface models based on CT multi-directional slice data fusion according to claim 2, characterized in that, In step 3.3, the convergence accuracy is the mean square error of the point cloud data iteration, and the range is set to [0.001, 0.5].

4. The method for generating a turbine blade surface model based on CT multi-directional slice data fusion according to claim 1, characterized in that, Step 4, the method for simplifying the overall point cloud data after stitching and fusion, specifically includes the following steps: Step 4.1: Calculate the eigenvalues ​​of the normal and curvature of the fused turbine blade CT multi-directional slice point cloud; Step 4.2: Traverse the points in the point cloud, determine whether the point cloud belongs to feature points based on the feature values, and classify the point cloud into feature points and non-feature points; Step 4.3: Retain the feature points of the point cloud and simplify the non-feature points.

5. The method for generating turbine blade surface models based on CT multi-directional slice data fusion according to claim 1, characterized in that: In step 5, the specific steps for generating the turbine blade surface model are as follows: Step 5.1: Project the simplified point cloud onto a two-dimensional plane using normals, and use the Delaunay growth algorithm to create a triangular mesh for the point cloud; Step 5.2: Map the two-dimensional triangular mesh back to the three-dimensional point cloud space using the spatial topological relationships within the neighborhood; Step 5.3: Repeat steps 5.1 and 5.2 to continuously generate new triangles outwards until all triangular meshes are generated, and output the turbine blade surface model.

6. The method for generating turbine blade surface models based on CT multi-directional slice data fusion according to claim 1, characterized in that: In step 2, the preprocessing of the point cloud data includes removing noise points and NAN points.

Images