Compensation and adjustment methods, devices, electronic equipment, and media applied to turbine blades

Abstract

This disclosure presents embodiments of a compensation and adjustment method, apparatus, electronic device, and medium applied to turbine blades. One specific implementation of the method includes: performing adaptive compensation detection on the turbine blade to be tested to obtain a three-dimensional point cloud dataset of the blade; performing point cloud feature registration processing on the three-dimensional point cloud dataset of the blade and a preset three-dimensional model of the turbine blade to obtain a blade point cloud registration transformation matrix; determining a set of normal vector deviation values ​​and a set of tool position normal compensation vectors; performing multi-dimensional direct-axis vector decomposition and rotary axis interpolation adjustment on the tool position normal compensation vector set to obtain a set of decomposed compensation vector values ​​and a set of rotary axis adjustment vector values; generating a cutting tool instruction set; and controlling the cutting tool on the spindle of a CNC machine tool to perform compensation cutting adjustment to obtain a compensated and adjusted turbine blade. This implementation completes the machining of the turbine blade in one step, which can improve the quality and adjustment efficiency of the turbine blade, shorten the machining and adjustment time, reduce equipment wear, and improve performance.

Description

Technical Field

[0001] The embodiments of this disclosure relate to the field of computer technology, and more specifically to a compensation adjustment method, apparatus, electronic device, and medium applied to turbine blades. Background Technology

[0002] As a core component of aero-engines and gas turbines, turbine blades' aerodynamic shape and surface quality directly determine the engine's efficiency and performance. Turbine blades are typically made of difficult-to-machine materials such as high-temperature alloys, exhibiting complex geometric features such as twisting, thin walls, and free-form surfaces. This makes precision machining of turbine blades a key area in the manufacturing industry, and the accuracy of milling using cutting tools is a crucial process for ensuring the quality of turbine blade machining. For compensation adjustments applied to turbine blades, the common method is as follows: The turbine blade to be inspected is unloaded from the machine tool and transferred to an inspection device for testing to obtain blade inspection information. Next, using a six-point alignment method, point cloud registration is performed between the blade inspection information and a pre-set 3D turbine blade model to obtain a point cloud registration transformation matrix. Then, based on the point cloud registration transformation matrix, a set of 3D point position deviation information is determined. Then, based on the 3D point position deviation information set, compensation information on the vertical axis is determined. Finally, the turbine blade to be inspected is re-clamped, and based on the vertical axis compensation information, the cutting tool is controlled to perform compensation cutting adjustments. The machined turbine blade is then unloaded again, and the above steps are repeated until no errors are found.

[0003] However, in practice, it has been found that when using the above method to compensate and adjust turbine blades, the following technical problems often arise: Since machining and inspection are two independent processes, the turbine blades need to be disassembled and clamped multiple times. Furthermore, the coordinate system of the workpiece being inspected is not the same as the coordinate system used for machining, which may lead to some deviation. This makes it difficult to guarantee machining accuracy, resulting in low machining and inspection efficiency, prolonged machining time, and lower turbine blade quality and performance. Simultaneously, the registration method based on the six-point alignment method and its focus solely on point cloud position information and vectors on the vertical axis result in a relatively singular consideration in machining, leading to low accuracy of the compensation vector, lower turbine blade quality, increased performance wear on cutting tools, and reduced performance.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the present disclosure concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure provide methods, apparatuses, electronic devices, and media for compensating and adjusting turbine blades to solve one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a compensation adjustment method for turbine blades, comprising: performing adaptive compensation detection on a turbine blade to be inspected using a machine tool probe mounted on a CNC machine tool spindle to obtain a three-dimensional point cloud dataset of the blade; performing point cloud feature registration processing on the aforementioned three-dimensional point cloud dataset of the blade and a preset three-dimensional turbine blade model to obtain a blade point cloud registration transformation matrix; determining the normal vector deviation value set between the aforementioned three-dimensional point cloud dataset of the blade and the aforementioned preset three-dimensional turbine blade model based on the aforementioned blade point cloud registration transformation matrix; and determining a preset tool... The toolpath normal compensation vector set is obtained for each tool position point on the toolpath; the toolpath normal compensation vector set is decomposed into a multidimensional direct axis vector decomposition process to obtain a decomposed compensation vector value set; the toolpath normal compensation vector set is adjusted by rotational axis interpolation to obtain a rotational axis adjustment vector value set; based on the decomposed compensation vector value set and the rotational axis adjustment vector value set, a cutting tool instruction set for the preset toolpath is generated; based on the cutting tool instruction set, the cutting tool on the CNC machine tool spindle is controlled to perform compensation cutting adjustment on the turbine blade to be tested, to obtain the compensated and adjusted turbine blade.

[0008] Secondly, some embodiments of this disclosure provide a compensation and adjustment device for turbine blades, comprising: an adaptive compensation detection unit configured to perform adaptive compensation detection on a turbine blade to be detected using a machine tool probe mounted on a CNC machine tool spindle, thereby obtaining a three-dimensional point cloud dataset of the blade; a point cloud feature registration unit configured to perform point cloud feature registration processing on the aforementioned three-dimensional point cloud dataset of the blade and a preset three-dimensional turbine blade model, thereby obtaining a blade point cloud registration transformation matrix; a first determining unit configured to determine a set of normal vector deviation values ​​between the aforementioned three-dimensional point cloud dataset of the blade and the aforementioned preset three-dimensional turbine blade model based on the aforementioned blade point cloud registration transformation matrix; and a second determining unit configured to determine a preset... The toolpath includes a tool position normal compensation vector set for each tool position point; a multi-dimensional linear vector decomposition unit configured to perform multi-dimensional linear vector decomposition on the tool position normal compensation vector set to obtain a decomposed compensation vector value set; a rotary axis interpolation adjustment unit configured to perform rotary axis interpolation adjustment on the tool position normal compensation vector set to obtain a rotary axis adjustment vector value set; a generation unit configured to generate a cutting tool instruction set for the preset toolpath based on the decomposed compensation vector value set and the rotary axis adjustment vector value set; and a control unit configured to control the cutting tool on the CNC machine tool spindle according to the cutting tool instruction set to perform compensation cutting adjustment on the turbine blade to be tested, to obtain a compensated turbine blade.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, such that when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any implementation of the first aspect.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method as described in any implementation of the first aspect.

[0011] The above embodiments of this disclosure have the following beneficial effects: The compensation and adjustment method for turbine blades in some embodiments of this disclosure can complete the machining of turbine blades in one go through online detection and adjustment, which can improve the quality and adjustment efficiency of turbine blades, shorten the machining and adjustment time, reduce equipment wear and improve performance. Specifically, the reason why the machining accuracy is difficult to guarantee, resulting in low machining and detection efficiency, prolonged machining time, low turbine blade quality and performance, increased performance wear of cutting tools and reduced performance is that: since machining and detection are two independent processes, the turbine blades need to be disassembled and clamped multiple times, and the workpiece coordinate system for detection is not the same as the machining coordinate system, which may have a certain deviation, making it difficult to guarantee machining accuracy, resulting in low machining and detection efficiency, prolonged machining time, and low turbine blade quality and performance; at the same time, the registration method based on the six-point alignment method and only focusing on the point cloud position information and the vector on the vertical axis makes the machining considerations relatively simple, the accuracy of the compensation vector is low, resulting in low turbine blade quality, increased performance wear of cutting tools and reduced performance. Based on this, some embodiments of the compensation and adjustment method for turbine blades disclosed herein can firstly perform adaptive compensation detection on the turbine blade to be inspected using a machine tool probe mounted on the spindle of a CNC machine tool, obtaining a three-dimensional point cloud dataset of the blade. Here, online detection enables subsequent one-time inspection and processing, improving processing and adjustment efficiency. Furthermore, adaptive compensation can offset systematic errors caused by probe installation, spindle runout, and blank clamping, resulting in a high-precision, fully covered three-dimensional point cloud dataset. Secondly, point cloud feature registration processing is performed on the aforementioned three-dimensional point cloud dataset of the blade and a preset three-dimensional turbine blade model to obtain a blade point cloud registration transformation matrix. Here, global precise matching is performed through point cloud feature registration to obtain a more accurate transformation matrix. Thirdly, based on the aforementioned blade point cloud registration transformation matrix, the normal vector deviation value set between the aforementioned three-dimensional point cloud dataset of the blade and the aforementioned preset three-dimensional turbine blade model is determined. Here, the normal vector deviation value set includes not only the three-dimensional point position deviation but also the normal vector deviation, which can more comprehensively reflect the errors between the turbine blade and the preset three-dimensional turbine blade model, and can accurately capture the local features of the turbine blade. Next, based on the aforementioned set of normal vector deviation values, the tool position normal compensation vector set for each tool position point on the preset toolpath is determined. This ensures precise compensation and smooth, continuous compensation between adjacent tool position points, effectively preventing abrupt changes on the turbine blade surface and improving the accuracy of the tool position normal compensation vector set. Subsequently, the aforementioned tool position normal compensation vector set undergoes multi-dimensional direct-axis vector decomposition to obtain a set of decomposed compensation vector values. Here, the tool position normal compensation vector set is converted into motion components of each direct axis, considering the motion characteristics of each direct axis to improve the accuracy of the decomposed compensation vector value set. Finally, the aforementioned tool position normal compensation vector set is adjusted by rotational axis interpolation to obtain a set of rotational axis adjustment vector values.Here, attitude interpolation and correction are performed on the rotary axes of the five-axis machine tool to avoid overcutting, undercutting, and interference caused by attitude deviations. Interpolation adjustment improves the smoothness and accuracy of the rotary axis motion. Then, based on the above-mentioned decomposed compensation vector value set and the above-mentioned rotary axis adjustment vector value set, a cutting tool instruction set for the above-mentioned preset toolpath is generated. Here, by fusing the decomposed compensation vector value set and the rotary axis adjustment vector value set, accurate and complete instructions are generated for subsequent compensation adjustment. Finally, based on the above-mentioned cutting tool instruction set, the cutting tool on the CNC machine tool spindle is controlled to perform compensation cutting adjustment on the turbine blade to be inspected, resulting in a compensated turbine blade. Here, completing the inspection and machining compensation adjustment in one go can improve the efficiency of machining compensation adjustment, shorten the machining time, thereby further improving the quality of the turbine blade, reducing the performance wear of the cutting tool, and improving performance. Therefore, this compensation adjustment method applied to turbine blades can complete the machining of turbine blades in one go through online inspection and adjustment, which can improve the quality and adjustment efficiency of turbine blades, shorten the machining adjustment time, reduce equipment wear, and improve performance. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of the compensation adjustment method applied to turbine blades according to the present disclosure;

[0014] Figure 2 This is an internal test schematic diagram of a blade 3D point cloud dataset in some embodiments of the compensation adjustment method for turbine blades according to this disclosure;

[0015] Figure 3 This is a schematic diagram of the initial registration fitted point cloud obtained by performing initial point cloud registration between the blade three-dimensional point cloud dataset and the preset turbine blade three-dimensional model in some embodiments of the compensation and adjustment method for turbine blades according to this disclosure.

[0016] Figure 4 This is a schematic diagram of the point cloud after initial registration fitting, obtained by re-registering the three-dimensional point cloud dataset of the blade and the preset three-dimensional model of the turbine blade according to some embodiments of the compensation and adjustment method for turbine blades disclosed herein.

[0017] Figure 5 This is a schematic diagram of a preset tool path for machining a turbine blade to be inspected, according to some embodiments of the compensation and adjustment method for turbine blades disclosed herein.

[0018] Figure 6 This is a schematic diagram of the structure of some embodiments of the compensation and adjustment device applied to turbine blades according to the present disclosure;

[0019] Figure 7 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0020] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0021] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0022] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0023] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0024] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0025] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Figure 1 A flow 100 of some embodiments of a compensation adjustment method for turbine blades according to the present disclosure is shown. The compensation adjustment method for turbine blades includes the following steps:

[0027] Step 101: Using a machine tool probe mounted on the spindle of a CNC machine tool, adaptive compensation detection is performed on the turbine blade to be inspected to obtain a three-dimensional point cloud dataset of the blade.

[0028] In some embodiments, the execution entity (e.g., an electronic device) of the above-described compensation and adjustment method applied to turbine blades can perform adaptive compensation detection on the turbine blade to be inspected using a machine tool probe mounted on a CNC machine tool spindle, obtaining a three-dimensional point cloud dataset of the blade. The CNC machine tool can be a platform that processes and inspects the turbine blade to be inspected, can carry the machine tool probe, and controls its motion trajectory according to a set program to achieve online measurement of complex curved surfaces. For example, the CNC machine tool spindle can be a five-axis linkage machining center. The machine tool probe can be a sensor that performs trigger-type or scanning-type measurements on the turbine blade to be inspected. The turbine blade to be inspected can be a core component of an aero-engine or gas turbine that has been preliminarily processed and is awaiting inspection, and the turbine blade's profile is a complex free-form surface. The three-dimensional point cloud data of the blade in the above-described three-dimensional point cloud dataset can be a data set of three-dimensional points characterizing the external profile of the turbine blade to be inspected. Figure 2 As shown, Figure 2 The diagram illustrates the internal testing of the blade's 3D point cloud dataset. In practice, the aforementioned execution entity can first obtain the geometric information set of the turbine blade to be inspected. Then, using a preset geometric scan parameter mapping table, it determines the adaptive scan parameters of the machine tool probe based on the geometric information set. Finally, through the adaptive scan parameters, the machine tool probe is controlled to perform adaptive compensation inspection of the turbine blade to be inspected, thereby obtaining the blade's 3D point cloud dataset.

[0029] In addressing the technical problems mentioned above, the application scenario—the machining compensation and adjustment of turbine blades for aero-engines and gas turbines—often presents the following challenges: Due to the complex geometric surfaces of turbine blades, the machine tool probe cannot adaptively adjust parameters when extracting the 3D point cloud, leading to point cloud occlusion, sparseness, and errors. This results in low accuracy of the 3D point cloud dataset, consequently lower precision in machining compensation and adjustment, lower blade quality, reduced cutting efficiency, compromised stability and performance of the cutting tool system, prolonged adjustment time, and increased system wear. Considering the following requirements for this application scenario: adaptability to high precision, high-density 3D point clouds, complex turbine blade surfaces, and turbine blade point cloud occlusion, we have decided to adopt the following solution:

[0030] In some optional implementations of certain embodiments, adaptive compensation detection of the turbine blade to be inspected is performed using a machine tool probe mounted on the spindle of a CNC machine tool to obtain a three-dimensional point cloud dataset of the blade. This may include the following steps:

[0031] The first step is to dynamically determine the probe scanning parameter set of the machine tool probe based on the turbine blade to be inspected. This probe scanning parameter set can be a set of parameter values ​​obtained by adjusting the scanning parameters of the machine tool probe according to the shape of each geometric part of the turbine blade. For example, the probe scanning parameter set may include: scanning speed and scanning resolution. For example, the dynamic determination may involve increasing the scanning resolution to 0.006mm and reducing the scanning speed to 1.2-1.8mm / s when scanning the blade tip or narrow flow channel regions (e.g., width ≤ 8mm); and maintaining the resolution at 0.01mm and increasing the speed to 3-5mm / s when scanning the blade body. In practice, the execution entity can first use the moving least squares method to perform local surface fitting on a preset three-dimensional model of the turbine blade to be inspected, obtaining the Gaussian curvature set and average curvature set of each point included in the turbine blade. Secondly, it can determine the sidewall distance information set of the turbine blade to be inspected through distance field analysis. Then, using an adaptive parameter mapping table, the adaptive parameter value set corresponding to the sidewall distance information set, Gaussian curvature set, and average curvature set of the turbine blade to be inspected is determined as the probe scanning parameter set. The aforementioned adaptive parameter mapping table can be a pre-set table mapping the scanning parameters of the machine tool probe to the geometric information (sidewall distance information, Gaussian curvature, and average curvature) of the turbine blade to be inspected.

[0032] The second step is to control the machine tool probe to perform detection based on the probe scanning parameter set mentioned above, and obtain the initial three-dimensional point cloud dataset.

[0033] Third, for each constraint segment in the constraint segment information set corresponding to the initial 3D point cloud dataset, perform the following point cloud surface construction steps:

[0034] Sub-step 1: In response to determining that there is a constraint segment length greater than the radius confidence value in the constraint segment length set included in the constraint segment information set, each constraint segment in the constraint segment set corresponding to at least one constraint segment length greater than the radius confidence value is divided into midpoints to generate a midpoint constraint segment group, thus obtaining a midpoint constraint segment group set. The radius confidence value can be the product of the segment radius included in the constraint segment information and a preset threshold. The preset threshold can be a pre-set value. For example, the preset threshold can be 0.5. The constraint segment information can include, but is not limited to, at least one of the following: the axis of the turbine blade to be detected, and the segment containing constraint information.

[0035] Sub-step 2 involves determining the target constraint segment set by combining the midpoint constraint segment set with each constraint segment information corresponding to each constraint segment whose confidence value is less than or equal to the radius.

[0036] Sub-step 3 involves constructing tetrahedrons from the endpoint point cloud data set of the target constraint line segment set to obtain an initial point cloud tetrahedron set. This initial point cloud tetrahedron set can be a set of non-overlapping tetrahedrons formed from the endpoint point cloud data set. The tetrahedron construction can be performed using the Delaunay tetrahedronization algorithm.

[0037] Sub-step 4: In response to determining that there exists at least one target constraint line segment in the target constraint line segment set that is not included in any initial point cloud tetrahedron set, add point cloud data to the midpoint of each target constraint line segment in the at least one target constraint line segment to obtain the first newly added point cloud dataset.

[0038] Sub-step 5: For each initial point cloud tetrahedron in the initial point cloud tetrahedron set, compare the radius-to-side-length ratio of the corresponding initial point cloud tetrahedron with a preset ratio threshold to obtain a comparison result. The radius-to-side-length ratio can be the ratio of the radius of the circumscribed sphere corresponding to the initial point cloud tetrahedron to the target side length. The target side length can be the minimum side length of the initial point cloud tetrahedron. The preset ratio threshold can be a pre-set maximum value for the ratio. For example, the preset ratio threshold can be 2.

[0039] Sub-step 6: In response to determining that there is at least one comparison result in the obtained comparison result set whose radius-to-side-length ratio is greater than the above-mentioned preset ratio threshold, point cloud data is added to the circumcenter of the at least one initial point cloud tetrahedron corresponding to the at least one comparison result to obtain a second newly added point cloud dataset.

[0040] Sub-step 7 involves constructing a tetrahedron from the target newly added point cloud dataset and the endpoint point cloud dataset to obtain the target point cloud tetrahedron set. The target newly added point cloud dataset can be a first newly added point cloud dataset, a second newly added point cloud dataset, or a point cloud dataset composed of the first and second newly added point cloud datasets.

[0041] Sub-step 8: In response to determining that the target constraint segment set and the target point cloud tetrahedron set satisfy a preset constraint condition set, the aforementioned initial 3D point cloud dataset and the newly added target point cloud data are determined as the blade 3D point cloud dataset. The aforementioned preset constraint condition set may include: each target constraint segment in the aforementioned target constraint segment set is located within any target point cloud tetrahedron in the aforementioned target point cloud tetrahedron set; the radius-to-side-length ratio of the target point cloud tetrahedron is less than the aforementioned preset ratio threshold; and the length of the target constraint segment is less than the radius confidence value.

[0042] The above-described technical solution and its related content, as an inventive point of this disclosure, solve the technical problem of "low accuracy in the machining compensation adjustment of turbine blades, resulting in low turbine blade quality." Factors leading to low accuracy in the machining compensation adjustment of turbine blades and low turbine blade quality are often as follows: Due to the complex geometric surfaces of turbine blades, when extracting the 3D point cloud of the turbine blade, the machine tool probe cannot adaptively adjust parameters, leading to problems such as point cloud occlusion, point cloud sparsity, and errors. This results in low accuracy of the 3D point cloud dataset, consequently leading to low accuracy in the machining compensation adjustment of the turbine blade, low turbine blade quality, reduced cutting adjustment efficiency, decreased stability and performance of the cutting tool system, prolonged adjustment time, and increased system wear. Solving these factors can improve the accuracy of the machining compensation adjustment of turbine blades and improve turbine blade quality. To achieve this effect, this disclosure first dynamically determines the probe scanning parameter set of the machine tool probe and controls the machine tool probe to perform detection, obtaining an initial 3D point cloud dataset that can fit the geometric distribution of the turbine blade to be detected, improving the globality and accuracy of the detected 3D point cloud. Secondly, the constraint line segment information is divided into midpoints to ensure that the entire set of target constraint line segments generated subsequently is included within the generated tetrahedron. Next, tetrahedron construction is performed on the endpoint point cloud data set of the target constraint line segment set, which better describes the spatial distribution and geometric features of the point cloud data, improving the accuracy of the 3D point cloud. Then, determining whether the entire set of target constraint line segments is included within the initial point cloud tetrahedron improves the integrity of the generated tetrahedron, reducing gaps and breaks in the generated tetrahedron set, and enhancing the overall integrity and stability of the tetrahedron. Finally, if the target constraint line segment set is not included in the initial tetrahedron set or the radius-to-side-length ratio of the initial point cloud tetrahedron exceeds a preset threshold, a new point cloud dataset is added. This allows for the extraction of a 3D point cloud set that more closely matches the actual 3D point cloud data of the turbine blade under test, providing a more comprehensive and accurate representation of the turbine blade. Finally, the initial 3D point cloud dataset and the target newly added point cloud data are determined as the blade 3D point cloud dataset. Based on the blade 3D point cloud dataset, the cutting tool on the CNC machine tool spindle is controlled to perform compensatory cutting adjustment on the turbine blade to be tested. The resulting compensated turbine blade can improve the accuracy and efficiency of the cutting tool control, improve the accuracy of the compensatory cutting adjustment, thereby improving the quality and performance of the turbine blade, shortening the adjustment time, and improving the performance and stability of the adjustment system corresponding to the cutting tool.

[0043] Step 102: Perform point cloud feature registration processing on the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix.

[0044] In some embodiments, the execution entity can perform point cloud feature registration processing on the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain a blade point cloud registration transformation matrix. The preset turbine blade 3D model can be a pre-defined, stored in the CNC system, theoretical CAD (Computer-Aided Design) 3D model of the turbine blade to be inspected. The blade point cloud registration transformation matrix can be a rigid body transformation matrix describing how, after rotation and translation, the blade 3D point cloud dataset and the preset turbine blade 3D model's 3D point clouds completely overlap. The blade point cloud registration transformation matrix can be a 4x4 matrix including a rotation matrix and a translation vector. 4. Matrix. For example... Figure 3 As shown, Figure 3 The diagram shows a registered 3D point cloud dataset of a blade obtained by performing initial point cloud registration (e.g., pre-alignment) on a 3D point cloud dataset of a blade and a preset 3D turbine blade model. Figure 4 The diagram shows the blade 3D point cloud dataset after transformation by the blade point cloud registration transformation matrix, obtained by re-registering the blade 3D point cloud dataset and the preset turbine blade 3D model with point cloud (for example, by registering the point cloud with six characteristic points determined by the six-point positioning principle).

[0045] In practice, the aforementioned execution entity can use the ICP (Iterative Closest Point) algorithm to perform point cloud feature registration processing on the aforementioned blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix.

[0046] In some optional implementations of certain embodiments, the above-mentioned point cloud feature registration processing of the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix may include the following steps:

[0047] The first step involves inputting the target point cloud dataset and the blade 3D point cloud dataset, both included in the pre-defined turbine blade 3D model, into the first-level point cloud feature extraction network of the point cloud adaptive registration model. This yields the first blade point cloud feature vector and the first target point cloud feature vector. The point cloud adaptive registration model further includes: a second-level point cloud feature extraction network, a first point cloud feature information interaction network, a third-level point cloud feature extraction network, a fourth-level point cloud feature extraction network, a second point cloud feature information interaction network, and a fifth-level point cloud feature extraction network. The point cloud adaptive registration model can be a deep neural network model that performs point cloud registration on the input target point cloud dataset and the blade 3D point cloud dataset to output a blade point cloud registration transformation matrix. The first-level point cloud feature extraction network can be a deep neural network model that extracts point cloud features from both the input target point cloud dataset and the blade 3D point cloud dataset. This first-level point cloud feature extraction network can be a deep neural network model composed of a point cloud spatial alignment network, two multilayer perceptrons, a feature transformation network, convolutional layers, batch normalization layers, and max pooling layers connected in series. The two multilayer perceptrons mentioned above can be multilayer perceptrons that map the input feature vector from 3D to 64D and from 64D to 128D, respectively. The point cloud spatial alignment network mentioned above can use the target point cloud dataset and the leaf 3D point cloud dataset as N... 3 A single-channel image is sequentially input into three concatenated convolutional layers, then into a max-pooling layer, stacked into N nodes, and finally input into a deep neural network model with two fully connected layers. N can be the number of leaf 3D point cloud data points included in the input leaf 3D point cloud dataset. The three convolutional layers can map the feature dimensions from 3 to 64, from 64 to 128, and from 128 to 1024, respectively. The feature transformation network can have the same network structure as the point cloud spatial alignment network, but the three convolutional layers can map the feature dimensions from 64 to 128, from 128 to 1024, and from 1024 to 64, respectively, and the input and output can be different. The second-level point cloud feature extraction network can be a model composed of one-dimensional convolutional layers, batch normalization, and a ReLU (Rectified Linear Unit) function. The third, fourth, and fifth level point cloud feature extraction networks can each have the same network structure as the second level point cloud feature extraction network, but different inputs, outputs, and network parameters. The one-dimensional convolutional kernels in the second, third, fourth, and fifth level point cloud feature extraction networks are 64, 64, and 64, respectively. 64 1.64 128 1. 128 256 1, 256 512 1. The aforementioned point cloud adaptive registration model can be trained in stages using transfer learning. First, a preliminary model training is performed using general 3D point cloud sample data, followed by fine-tuning training using turbine blade point cloud sample data. The loss function for training the aforementioned point cloud adaptive registration model can be a loss function composed of a point cloud distance loss function, a rotation matrix loss function, and a displacement matrix loss function. The aforementioned point cloud distance loss function can utilize the Chamfer Distance (CD) loss function to determine the global distance error between the target point cloud dataset and the blade 3D point cloud dataset before and after transformation, thus adapting to the global shape consistency requirements of overlapping 3D point clouds. The aforementioned rotation matrix loss function can be a function that determines the product of the difference between the predicted rotation quaternion and the true rotation quaternion and the true rotation weight. The aforementioned preset rotation weight can be a pre-set value used to reduce rotation error, and its value can be 4. The aforementioned displacement matrix loss function can be a function that determines the L2 error between the predicted translation vector and the true translation vector.

[0048] The second step involves inputting the first leaf point cloud feature vector and the first target point cloud feature vector into the second-level point cloud feature extraction network to obtain the second leaf point cloud feature vector and the second target point cloud feature vector.

[0049] The third step involves inputting the feature vectors of the second leaf point cloud and the second target point cloud into the first point cloud feature information interaction network to obtain the feature vectors of the third leaf point cloud and the third target point cloud.

[0050] The fourth step involves inputting the feature vectors of the third leaf point cloud and the third target point cloud into the third-level point cloud feature extraction network to obtain the feature vectors of the fourth leaf point cloud and the fourth target point cloud.

[0051] The fifth step involves inputting the feature vectors of the fourth leaf point cloud and the fourth target point cloud into the fourth-level point cloud feature extraction network to obtain the feature vectors of the fifth leaf point cloud and the fifth target point cloud.

[0052] The sixth step involves inputting the feature vectors of the fifth leaf point cloud and the fifth target point cloud into the second point cloud feature information interaction network to obtain the feature vectors of the sixth leaf point cloud and the sixth target point cloud.

[0053] Step 7: Input the feature vector of the sixth leaf point cloud and the feature vector of the sixth target point cloud into the fifth-level point cloud feature extraction network to obtain the feature vector of the seventh leaf point cloud and the feature vector of the seventh target point cloud.

[0054] Step 8: After feature fusion of the feature vectors of the second leaf point cloud, the fifth leaf point cloud, and the seventh leaf point cloud, max pooling is performed to obtain the global leaf point cloud feature vector. Similarly, after feature fusion of the feature vectors of the second target point cloud, the fifth target point cloud, and the seventh target point cloud, max pooling is performed to obtain the global target point cloud feature vector.

[0055] Step nine involves performing multi-layer convolutions on the aforementioned global leaf point cloud feature vectors and the aforementioned global target point cloud feature vectors to obtain the leaf point cloud registration transformation matrix. This multi-layer convolution process can be a stepwise dimensionality reduction process, where the feature vectors are sequentially input into four one-dimensional convolutional layers followed by two fully connected layers. The four one-dimensional convolutional layers can be convolutional layers that sequentially reduce the dimension of the feature vectors from 2048 to 1024, 512, 256, and 7.

[0056] Optionally, the above-mentioned inputting the second leaf point cloud feature vector and the second target point cloud feature vector into the first point cloud feature information interaction network to obtain the third leaf point cloud feature vector and the third target point cloud feature vector may include the following steps:

[0057] The first step is to perform max pooling on the feature vectors of the second leaf point cloud and the second target point cloud, respectively, to obtain the pooled feature vectors of the leaf point cloud and the target point cloud.

[0058] The second step involves stacking the pooled feature vectors of the blade point cloud and the target point cloud to obtain stacked feature vectors of the blade point cloud and the target point cloud. The number of stacking operations can be determined by the number of blade 3D point cloud data points included in the aforementioned blade 3D point cloud dataset.

[0059] The third step is to perform feature fusion on the stacked feature vector of the leaf point cloud, the feature vector of the second leaf point cloud, and the stacked feature vector of the target point cloud to obtain the fused feature vector of the leaf point cloud, which is used as the feature vector of the third leaf point cloud.

[0060] The fourth step involves fusing the second target point cloud feature vector, the stacked target point cloud feature vector, and the stacked leaf point cloud feature vector to obtain the target point cloud fused feature vector, which serves as the third target point cloud feature vector.

[0061] Step 103: Based on the blade point cloud registration transformation matrix, determine the normal vector deviation value set between the blade 3D point cloud dataset and the preset turbine blade 3D model.

[0062] In some embodiments, the execution entity can determine the set of normal vector deviation values ​​between the blade 3D point cloud dataset and the preset turbine blade 3D model based on the blade point cloud registration transformation matrix. The normal vector deviation value in the set can be the difference between the distance between the multiplied blade 3D point cloud dataset and the blade point cloud registration transformation matrix and the distance to the 3D point cloud included in the preset turbine blade 3D model, and the deviation margin on the normal vector.

[0063] In some optional implementations of certain embodiments, determining the normal vector deviation value set of the blade 3D point cloud dataset and the preset turbine blade 3D model based on the blade point cloud registration transformation matrix may include the following steps:

[0064] The first step is to determine the product of the above-mentioned blade point cloud registration transformation matrix and the above-mentioned blade 3D point cloud dataset, which will be used as the registered blade 3D point cloud dataset.

[0065] The second step is to determine the three-dimensional point cloud set in the preset turbine blade three-dimensional model that corresponds to the registered blade three-dimensional point cloud dataset, and use it as the target three-dimensional point cloud set.

[0066] The third step is to determine the normal vector of each target 3D point cloud in the aforementioned target 3D point cloud set, which is then used as the target normal vector to obtain the target normal vector set. In practice, the aforementioned execution entity can use a local plane fitting method to determine the locally optimal plane for each target 3D point cloud, obtaining a set of locally optimal planes. Then, the least squares method is used to solve the equations of the tangent planes corresponding to the aforementioned locally optimal plane sets to obtain the target normal vector set.

[0067] The fourth step is to determine the distance between each registered blade 3D point cloud data in the above-mentioned registered blade 3D point cloud dataset and the corresponding target normal vector in the above-mentioned target normal vector set, and use this distance as the point cloud distance to obtain the point cloud distance set. The point cloud distance can be the Hausdoff distance.

[0068] The fifth step is to determine the projection components of the point cloud distance set onto the target normal vector set, thereby obtaining the normal vector deviation value set.

[0069] Step 104: Determine the tool position normal compensation vector set for each tool position point on the preset tool path based on the normal vector deviation value set.

[0070] In some embodiments, the execution entity can determine the tool position normal compensation vector set for each tool position point on the preset toolpath based on the normal vector deviation value set. The preset toolpath can be a pre-defined cutting trajectory used to cut the turbine blade to be tested to better conform to the preset turbine blade 3D model. The tool position points can be the geometric points where the cutting tool on the CNC machine tool spindle cuts the turbine blade to be tested. For example, the tool position point can be the center point (tool tip) of the bottom of a milling cutter or drill bit. The tool position normal compensation vector in the tool position normal compensation vector set can be the cutting direction and thickness of the cutting tool at the tool position point to eliminate errors with the preset turbine blade 3D model. Figure 5 As shown, Figure 5 The diagram shows the preset path of the turbine blade at each tool point on the preset toolpath (for example, the tool point is an arc with a radius of 236.891 mm, corresponding to the points of 67.577 mm and 95.318 mm, with a left chord angle of 0.73 degrees and a right chord angle of 3.704 degrees, and the preset route can be a path that starts from the points corresponding to 67.577 mm and 95.318 mm, moves from the left chord angle to the right chord angle).

[0071] In some optional implementations of certain embodiments, determining the tool position normal compensation vector set for each tool position point on the preset toolpath based on the aforementioned normal vector deviation value set may include the following steps:

[0072] The first step is to map the above-mentioned normal vector deviation value set to the above-mentioned preset turbine blade 3D model to obtain the deviation turbine blade 3D model. This deviation turbine blade 3D model can represent the deviation distribution of the blade's 3D point cloud dataset.

[0073] The second step is to perform smoothing and denoising on the aforementioned three-dimensional model of the deviated turbine blade to obtain a denoised three-dimensional model of the deviated turbine blade. This smoothing and denoising process can be performed using a Gaussian filter.

[0074] The third step involves determining the semi-variant spatial correlation value set and the inter-point cloud distance set for adjacent deviation 3D point cloud sets within the denoised deviation turbine blade 3D point cloud set. Specifically, adjacent deviation 3D point clouds in the aforementioned set can be pairs of deviation 3D point clouds consisting of each deviation 3D point cloud and its adjacent counterpart. The semi-variant spatial correlation values ​​in the aforementioned set can be obtained by inputting the aforementioned deviation 3D point clouds into a semi-variant function, characterizing the spatial correlation of deviation 3D point cloud pairs. The inter-point cloud distances in the aforementioned inter-point cloud distance set can be the relative distances, i.e., lag distances, between deviation 3D point cloud pairs.

[0075] The fourth step involves generating a fitted point cloud deviation model based on the aforementioned semi-variable spatial correlation value set and the aforementioned point cloud distance set. This fitted point cloud deviation model can be a mathematical function that transforms the discrete semi-variable spatial correlation value set and the aforementioned point cloud distance set into a continuous, smooth mathematical function that describes the global error distribution. In practice, the executing entity can input the aforementioned semi-variable spatial correlation value set and the aforementioned point cloud distance set into a preset fitting function to obtain the fitted point cloud deviation model. This preset fitting function can be one of the following: a spherical model, an exponential model, or a Gaussian model.

[0076] The fifth step involves determining the neighborhood deviation 3D point cloud set relative to each tool position on the preset toolpath, based on the aforementioned fitted point cloud deviation model. The neighborhood deviation 3D point clouds in this set can be deviation 3D point clouds within a range of distance from the tool position. This range can be a numerical value used to quantify the neighborhood range. The range can also be a value obtained by solving the fitted point cloud deviation model using the least squares method. In practice, the executing entity can first solve the fitted point cloud deviation model using the least squares method to obtain the range value. Then, it can determine multiple deviation 3D point clouds within a radius of twice the range value, centered on each tool position, to obtain the neighborhood deviation 3D point cloud set.

[0077] Step 6: Based on the aforementioned neighborhood deviation 3D point cloud set, generate a set of point cloud interpolation equations for each of the aforementioned tool points. These point cloud interpolation equations can be a set of functions used to determine the weights of each neighborhood deviation 3D point cloud in the aforementioned neighborhood deviation 3D point cloud set. For example, the aforementioned point cloud interpolation equations can be a Kriging interpolation function. In practice, the executing entity can first input the aforementioned neighborhood deviation 3D point cloud set into the aforementioned semi-variogram function to obtain the target variogram function. Lagrange multipliers are then introduced into the target variogram function to construct the data variance function of the deviation 3D point to be interpolated and the neighborhood deviation 3D point set, thus obtaining the point cloud interpolation equation set.

[0078] Step 7: Solve the constraints of the above point cloud interpolation equations to obtain a set of normal vector deviation estimates. The normal vector deviation estimates in this set represent the weights, or importance, of the normal vector deviation values. The constraint solution can be performed using matrix operations.

[0079] Step 8: Determine the product of each normal vector deviation estimate in the above normal vector deviation estimate set and the normal vector of the corresponding tool position in each tool position as the tool position normal compensation vector, and obtain the tool position normal compensation vector set.

[0080] In addressing the aforementioned technical problems in the application scenario—industrial-grade machining compensation and adjustment of turbine blades—the following technical issues often arise: Due to the complex geometric surfaces of turbine blades, determining the toolpath normal compensation vector set requires measuring the deviation values ​​of adjacent 3D point clouds, necessitating parameter determination. Traditional sparrow search algorithms for this determination are prone to getting trapped in local optima, resulting in low parameter accuracy. This leads to a low accuracy rate of the toolpath normal compensation vector determined by the parameters, resulting in lower turbine blade quality and performance. This reduces the precision of machining compensation and adjustment, further lowering blade quality, reducing cutting efficiency, impacting the stability and performance of the cutting tool system, extending adjustment time, and increasing system wear. Considering the following requirements for this application scenario: adaptability to high precision, adaptability to complex turbine blade geometry, adaptability to drastically changing edges and torsional deformation of the blades, and adaptability to the anisotropy of the turbine blade axis, we have decided to adopt the following solution:

[0081] In some optional implementations of certain embodiments, the process of determining the tool position normal compensation vector set for each tool position on the preset toolpath based on the above-mentioned normal vector deviation value set, and controlling the cutting tool on the CNC machine tool spindle to perform compensating cutting adjustment on the turbine blade to be tested based on the above-mentioned tool position normal compensation vector, may include the following steps:

[0082] The first step involves generating a minimum interpolation prediction fitness function and an initial interpolation prediction population for the aforementioned normal vector deviation value set. The minimum interpolation prediction fitness can be the mean square error of the interpolated 3D point cloud and blade 3D point cloud data generated from the point cloud interpolation equation set. Individuals in the initial interpolation prediction population can be a set of model parameters corresponding to the point cloud interpolation equation. For example, if the point cloud interpolation equation is a Kriging interpolation function, then the individuals in the initial interpolation prediction population can be the function parameters of the semi-variogram function to be optimized, namely, nugget value, sill value, and range. The initial interpolation prediction population can also include a set of interpolation parameter values. This set of interpolation parameter values ​​can include: a preset population size, a preset execution threshold, a preset proportion of population discoverers, a preset proportion of population scouts, the function dimension of the minimum interpolation prediction fitness function, and individual boundary values. The individual boundary values ​​can be boundary values ​​composed of the maximum and minimum values ​​of individuals. The preset proportion of population discoverers can be the proportion of individuals guiding the population towards potentially high-quality solutions, typically taking values ​​of (0.1, 0.2). The aforementioned preset population scout ratio represents the proportion of individuals that guide the population to avoid risks and increase population diversity; a typical value is (0.1, 0.2). The initial interpolation prediction population can be obtained using the tent map algorithm. The aforementioned preset population size can be a pre-set number of individuals in the initial interpolation prediction population. The aforementioned preset execution threshold can be a pre-set maximum number of iterations.

[0083] The second step is to determine the inverse solution of the initial interpolation prediction population, thus obtaining the initial inverse interpolation prediction population. This determination can be performed using a back-learning strategy.

[0084] The third step involves filtering the initial interpolation prediction population and the initial reverse interpolation prediction population based on the aforementioned minimum interpolation prediction fitness function, resulting in a filtered interpolation prediction population. In practice, the executing entity can first input the initial interpolation prediction population and the initial reverse interpolation prediction population into the minimum interpolation prediction fitness function to obtain a first fitness function value set and a second fitness function value set. There is a one-to-one correspondence between the first fitness function value set and the second fitness function value set. Then, the individual with the largest corresponding value is selected from the first fitness function value set and the second fitness function value set to obtain the filtered interpolation prediction population.

[0085] Fourth, based on the population prediction after screening and interpolation, perform the following population update steps:

[0086] Sub-step 1: Based on the pre-defined proportion of discoverers in the initial interpolation prediction population, randomly select a discoverer interpolation population and a follower interpolation population from the filtered interpolation prediction population. The follower interpolation population can be the population remaining after removing the discoverer interpolation population from the filtered interpolation prediction population. The discoverer interpolation population can be a population composed of randomly selected individuals from the filtered interpolation prediction population whose proportion in the filtered interpolation prediction population is equal to the pre-defined proportion of discoverers. In practice, the executing entity can first determine the population size of the discoverer interpolation population by multiplying the pre-defined population size by the pre-defined proportion of discoverers. Then, randomly select the pre-defined population size of individuals from the filtered interpolation prediction population to obtain the discoverer interpolation population. Finally, the population obtained by removing the discoverer interpolation population from the filtered interpolation prediction population is determined as the follower interpolation population.

[0087] Sub-step 2 involves filtering the follower interpolation population and the filtered interpolation prediction population based on the minimized interpolation prediction fitness function to obtain target follower individuals, first target initial individuals, and second target initial individuals. In practice, the executing entity can first input the filtered interpolation prediction population into the minimized interpolation prediction fitness function to obtain a set of filtered fitness function values. Secondly, it can determine the filtered fitness function values ​​corresponding to the follower interpolation population from the set of filtered fitness function values, using these as the target fitness function value set. Then, it can determine the follower individual corresponding to the target fitness function value with the largest value in the set of target fitness function values ​​as the target follower individual. Finally, it can determine the filtered individuals corresponding to the smallest and largest filtered fitness values ​​in the set of filtered fitness function values ​​as the first target initial individuals and the second target initial individuals.

[0088] Sub-step 3: Based on the target follower individuals and the initial first target individual, perform local individual update processing on the follower interpolation population to obtain the updated follower interpolation population. Specifically, follower individuals in the updated follower interpolation population located in the first half of the filtered interpolation prediction population are updated in position according to the follower update strategy in the sparrow search algorithm. Follower individuals located in the second half of the filtered population are updated in position according to the flight step size determined by the Levy flight strategy. The updated follower interpolation individuals in the updated follower interpolation population can be represented as follows:

[0089] .

[0090] in, Indicates the first Each follower interpolated individual in The first iteration in the next loop The position of the dimension, i.e., the follower interpolated individual after the update. It is represented as a random number that follows a standard normal distribution. Indicates the first The follower interpolated individual with the smallest initial fitness value in the next iteration is the first target initial individual. Indicates the first Each follower interpolated individual in The first iteration in the next loop The position of the dimension. This indicates the number of individuals included in the interpolated predicted population after screening. This indicates the position of the target follower interpolated individual. This refers to Levi's flight strategy. This represents the Lévy flight constant, with a value of 1.5. This represents a random number whose value is in the range [0, 1]. express , It is a one-dimensional matrix, and each element in the matrix is ​​randomly selected from 1 or -1. This represents the transpose of a one-dimensional matrix. This represents a one-dimensional matrix where each element is 1. This represents the gamma function.

[0091] Sub-step 4: Based on the initial individuals of the second target, perform a nonlinear sinusoidal position update on the discoverer interpolation population to obtain the updated discoverer interpolation population. The updated discoverer interpolation individuals in the updated discoverer interpolation population can be represented as:

[0092] .

[0093] in, This represents the nonlinear sinusoidal learning factor. This represents the minimum value of the nonlinear sinusoidal learning factor. This represents the maximum value of the nonlinear sinusoidal learning factor. It represents a random number within the range of [0, 2π]. It represents a random number within the range of [0, 2]. This indicates the position of the initial individual in the second target. This indicates the preset execution threshold. This represents the warning value within the range of [0, 1]. This represents a safe value within the range of [0.5, 1].

[0094] Sub-step 5: Based on the aforementioned minimum interpolation prediction fitness function, update the movement positions of the updated discoverer interpolation population and the updated follower interpolation population to obtain the updated scout interpolation population. In practice, firstly, a scout interpolation population is randomly selected from the updated follower interpolation population and the updated discoverer interpolation population. The ratio of the scout interpolation population to the selected interpolation prediction population can be a preset proportion of scouts. Then, based on the fitness value set obtained after inputting the updated follower interpolation population and the updated discoverer interpolation population into the minimum interpolation prediction fitness function, the scout interpolation population is updated to obtain the updated scout interpolation population. The updated scout interpolation individuals in the updated scout interpolation population can be represented as:

[0095] .

[0096] in, The step size control parameter is a normally distributed random number with a mean of 0 and a variance of 1. Indicates the first The fitness function value of each scout interpolated individual. This represents the fitness function value with the largest fitness function value in the scout interpolation population. The direction of movement of the interpolated individual scout is represented by a uniform random number within the range of [-1, 1]. This represents the fitness function value with the smallest fitness function value in the scout interpolation population. This represents the smallest constant, preventing the denominator from being 0.

[0097] Sub-step 6 involves performing Gaussian-Cauchy updates on the updated discoverer interpolation population, the updated follower interpolation population, and the updated scout interpolation population to obtain a Gaussian-updated interpolation population. The Gaussian-updated interpolated individuals in this population can be individuals obtained by updating the updated discoverer interpolation population, the updated follower interpolation population, and the updated scout interpolation population using Cauchy mutation and Gaussian mutation.

[0098] Sub-step 7: In response to determining that the number of times the above population update step has been executed is greater than or equal to a preset execution threshold, a point cloud interpolation equation set is generated based on the minimized interpolation prediction fitness function and the Gaussian updated interpolation population. The point cloud interpolation equation set is then solved to obtain the tool position normal compensation vector. Based on this tool position normal compensation vector, the cutting tool on the CNC machine tool spindle is controlled to perform compensatory cutting adjustment on the turbine blade to be detected. In practice, the execution entity can first input the Gaussian updated interpolation population into the minimized interpolation prediction fitness function to obtain the target fitness function value set. Secondly, the parameter value set corresponding to the Gaussian updated interpolation individual with the largest target fitness function value is selected from the target fitness function value set and determined as the parameter set corresponding to the point cloud interpolation equation set, serving as the target interpolation equation set. Then, the target interpolation equation set is solved matrixwise to obtain the tool position normal compensation vector. Based on this tool position normal compensation vector, the cutting tool on the CNC machine tool spindle is controlled to perform compensatory cutting adjustment on the turbine blade to be detected.

[0099] Fifth step: In response to determining that the number of times the above execution has been performed is less than the preset execution threshold, the Gaussian update interpolation population is determined as the filtered interpolation prediction population, and the sum of the number of times the execution has been performed and the preset threshold is determined as the number of times the above population update step is performed again.

[0100] The above-described technical solution and its related content, as an inventive point of this disclosure, solve the technical problem of "low accuracy of the tool position normal compensation vector determined by parameters and low quality and performance of turbine blades". The factors contributing to the low accuracy of the tool position normal compensation vector determined by parameters and the low quality and performance of turbine blades are often as follows: Due to the complex geometric surfaces of turbine blades, determining the tool position normal compensation vector set requires measuring the deviation values ​​of adjacent 3D point clouds, necessitating parameter determination. Using traditional sparrow search algorithms for this purpose can easily lead to getting trapped in local optima, resulting in low parameter accuracy. This, in turn, leads to low accuracy of the tool position normal compensation vector determined by parameters, resulting in low quality and performance of the turbine blades. Furthermore, it reduces the precision of machining compensation adjustments for turbine blades, further lowering their quality, reducing cutting adjustment efficiency, impacting the stability and performance of the cutting tool system, prolonging adjustment time, and increasing system wear. Solving these factors can improve the accuracy of the tool position normal compensation vector and improve the quality and performance of the turbine blades. To achieve this effect, this disclosure first minimizes the interpolation prediction fitness function and the initial interpolation prediction population, facilitating subsequent population selection. Second, through a mapping algorithm and a direction learning strategy, the initial interpolation prediction population is selected, resulting in a more uniform distribution and improved quality, accuracy, and diversity, thus enhancing the population's global and stable search capabilities. Third, the follower interpolation population is updated using a Lévy flight strategy, combining short-range and long-range searches to improve global search capabilities and population position diversity, avoiding getting trapped in local optima. Finally, the discoverer interpolation population is updated using a fusion sine and cosine algorithm. By introducing a nonlinear sine learning factor, a larger value in the early stages of iteration enhances global search, while a gradually decreasing value in the later stages improves local search capabilities and increases population position accuracy. Then, the updated discoverer and follower interpolation populations are moved and updated to obtain the updated scout interpolation population. Gaussian-Cauchy updates are then performed on all three populations. Gaussian-Cauchy updates can change rapidly in the early stages of iteration and gradually stabilize in the later stages, reducing the risk of getting trapped in local optima. Gaussian-Cauchy updates can search locally on individual nodes, quickly and accurately locating the global optimum and accelerating convergence. Finally, a set of point cloud interpolation equations is generated and solved to obtain the tool position normal compensation vector. Based on this vector, the cutting tool on the CNC machine tool spindle is controlled to perform compensation cutting adjustments. This improves the accuracy and efficiency of compensation cutting adjustments, enhances the quality and performance of turbine blades, increases the efficiency of compensation adjustments, shortens adjustment time, improves system performance and stability, and reduces wear on various system components.

[0101] Step 105: Perform multidimensional direct-axis vector decomposition on the tool position normal compensation vector set to obtain the decomposed compensation vector value set.

[0102] In some embodiments, the execution entity performs multidimensional linear vector decomposition on the tool position normal compensation vector set to obtain a decomposed compensation vector value set. The decomposed compensation vector value set can be a component of a tool position normal compensation vector on the linear motion axes of the horizontal (x-axis), vertical (y-axis), and vertical (z-axis), i.e., controlling the movement distance of the cutting tool in the three spatial linear directions (horizontal, longitudinal, and vertical), thereby achieving linear positioning of the tool in space.

[0103] Step 106: Perform rotation axis interpolation adjustment on the tool position normal compensation vector set to obtain the rotation axis adjustment vector value set.

[0104] In some embodiments, the execution entity can perform rotational axis interpolation adjustment on the tool position normal compensation vector set to obtain a rotational axis adjustment vector value set. The rotational axis adjustment vector value set can be the components of the tool position normal compensation vector on the first and second rotational axes, i.e., realizing the flipping and rotational movements of the cutting tool to adjust the tool's attitude angle and cutting direction. The first rotational axis can be an axis rotating around the longitudinal axis. The second rotational axis can be an axis rotating around the vertical axis.

[0105] In some optional implementations of certain embodiments, the above-mentioned rotation axis interpolation adjustment of the tool position normal compensation vector set to obtain a rotation axis adjustment vector value set may include the following steps:

[0106] The first step is to determine the dot product of the above tool position normal compensation vector set and the tool position normal vector set corresponding to each tool position point, which is used as the set of cosine values ​​of the tool position point rotation angle.

[0107] The second step involves performing inverse cosine processing on the aforementioned set of tool position rotation angle cosine values ​​to obtain the tool position rotation angle set. Here, the tool position rotation angle in the aforementioned tool position rotation angle set can be the angle between the aforementioned tool position normal compensation vector and the corresponding tool position normal vector, representing the total angle required for the cutting tool to adjust from the current attitude to the target attitude when performing error compensation. The aforementioned inverse cosine processing can be performed by inputting the aforementioned tool position rotation cosine values ​​into an inverse cosine function for calculation.

[0108] The third step involves performing spherical linear interpolation on the tool position normal compensation vector set, the tool position normal vector set, and the tool position rotation angle set, to obtain the interpolated tool position attitude information set. In practice, the executing entity can input the tool position normal compensation vector set, the tool position normal vector set, and the tool position rotation angle set into the spherical linear interpolation algorithm to obtain the interpolated tool position attitude information set.

[0109] The fourth step is to perform interpolation segmentation on the above-mentioned interpolated tool position posture information set to obtain a tool position posture grouping information set. The tool position posture grouping information in this set can be a group of interpolated tool position posture information obtained through segmentation, to facilitate a smooth transition of the rotation axis adjustment vector value group and avoid tool jitter during posture rotation.

[0110] The fifth step involves determining the horizontal direction angle and pitch angle of each tool position attitude group in the aforementioned tool position attitude grouping information set, resulting in a horizontal direction angle set and a pitch angle set, which serve as the set of rotation axis adjustment vector values. The horizontal direction angle set can be the azimuth angle (the angle between the tool position normal compensation vector included in each tool position attitude grouping information and the horizontal axis after mapping onto the horizontal plane). Alternatively, the horizontal direction angle can be the direction angle obtained by inputting the tool position normal compensation vector included in each tool position attitude grouping information into a two-dimensional arctangent function. The pitch angle set can be the angle between the cutting tool axis and the horizontal plane (XOY plane), i.e., the arcsine function value of the vertical component of the tool position normal compensation vector included in each tool position attitude grouping information.

[0111] Step 107: Generate a cutting tool instruction set for the preset toolpath based on the decomposed compensation vector value set and the rotation axis adjustment vector value set.

[0112] In some embodiments, the executing entity can generate a cutting tool instruction set for the preset toolpath based on the decomposed compensation vector value set and the rotary axis adjustment vector value set. The cutting tool instructions in the cutting tool instruction set can be machining program code that the CNC system can recognize and execute, used to control the cutting tool. For example, the cutting tool instructions can be NC (Numerical Control Code) code. In practice, the executing entity can input the decomposed compensation vector value set and the rotary axis adjustment vector value set into NC template code to obtain the cutting tool instruction set. The NC template code can be a pre-set NC code with compensation variables on each axis, used to control the cutting tool for machining adjustments.

[0113] Step 108: According to the cutting tool instruction set, control the cutting tool on the CNC machine tool spindle to perform compensating cutting adjustment on the turbine blade to be tested, and obtain the compensated and adjusted turbine blade.

[0114] In some embodiments, the executing entity can control the cutting tool on the spindle of a CNC machine tool according to the cutting tool instruction set to perform compensatory cutting adjustment on the turbine blade to be tested, thereby obtaining a compensated and adjusted turbine blade. The cutting tool can be a cutting tool used to perform cutting processing on the turbine blade to be tested. For example, the cutting tool can be, but is not limited to, at least one of the following: end mill, ball end mill, turning tool, and drill bit. In practice, the executing entity can input the cutting tool instruction set into the CNC system corresponding to the CNC machine tool for execution, thereby driving the cutting tool on the spindle of the CNC machine tool to perform compensatory cutting adjustment on the turbine blade to be tested, thereby obtaining a compensated and adjusted turbine blade.

[0115] Further reference Figure 6 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a compensation adjustment device applied to turbine blades. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, this compensation adjustment device applied to turbine blades can be specifically applied to various electronic devices.

[0116] like Figure 6As shown, a compensation and adjustment device 600 for turbine blades includes: an adaptive compensation detection unit 601, a point cloud feature registration unit 602, a first determination unit 603, a second determination unit 604, a multidimensional direct-axis vector decomposition unit 605, a rotary axis interpolation adjustment unit 606, a generation unit 607, and a control unit 608. The adaptive compensation detection unit 601 is configured to: perform adaptive compensation detection on the turbine blade to be detected using a machine tool probe mounted on the spindle of a CNC machine tool, obtaining a three-dimensional point cloud dataset of the blade. The point cloud feature registration unit 602 is configured to: perform point cloud feature registration processing on the aforementioned three-dimensional point cloud dataset of the blade and a preset three-dimensional turbine blade model, obtaining a blade point cloud registration transformation matrix. The first determination unit 603 is configured to: determine the normal vector deviation value set of the aforementioned three-dimensional point cloud dataset of the blade and the preset three-dimensional turbine blade model based on the aforementioned blade point cloud registration transformation matrix. The second determination unit 604 is configured to: determine the tool position normal compensation vector set of each tool position point on a preset toolpath based on the aforementioned normal vector deviation value set. The multidimensional direct-axis vector decomposition unit 605 is configured to perform multidimensional direct-axis vector decomposition on the tool position normal compensation vector set to obtain a set of decomposed compensation vector values. The rotary axis interpolation adjustment unit 606 is configured to perform rotary axis interpolation adjustment on the tool position normal compensation vector set to obtain a set of rotary axis adjustment vector values. The generation unit 607 is configured to generate a set of cutting tool instructions for the preset toolpath based on the set of decomposed compensation vector values ​​and the set of rotary axis adjustment vector values. The control unit 608 is configured to control the cutting tool on the CNC machine tool spindle according to the set of cutting tool instructions to perform compensation cutting adjustment on the turbine blade to be tested, to obtain the compensated turbine blade.

[0117] It is understandable that the units described in the compensation and adjustment device 600 applied to turbine blades are similar to those in the reference device. Figure 1 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the compensation adjustment device 600 applied to turbine blades and the units contained therein, and will not be repeated here.

[0118] The following is for reference. Figure 7 It shows a schematic diagram of the structure of an electronic device 700 suitable for implementing some embodiments of the present disclosure. Figure 7 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0119] like Figure 7As shown, the electronic device 700 may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the electronic device 700. The processing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0120] Typically, the following devices can be connected to I / O interface 705: input devices 706 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 707 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 708 including, for example, magnetic tapes, hard disks, etc.; and communication devices 709. Communication device 709 allows electronic device 700 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 7 An electronic device 700 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 7 Each box shown can represent a device or multiple devices as needed.

[0121] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 709, or installed from storage device 708, or installed from ROM 702. When the computer program is executed by processing device 701, it performs the functions defined in the methods of some embodiments of this disclosure.

[0122] It should be noted that, in some embodiments of this disclosure, the computer-readable medium described above may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0123] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0124] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: perform adaptive compensation detection on the turbine blade to be inspected via a machine tool probe mounted on the spindle of a CNC machine tool, obtaining a three-dimensional point cloud dataset of the blade; perform point cloud feature registration processing on the aforementioned three-dimensional point cloud dataset of the blade and a preset three-dimensional turbine blade model, obtaining a blade point cloud registration transformation matrix; determine the normal vector deviation value set between the aforementioned three-dimensional point cloud dataset of the blade and the preset three-dimensional turbine blade model based on the aforementioned blade point cloud registration transformation matrix; and determine the normal vector deviation value set based on the aforementioned normal vector deviation value set. The following steps are performed: First, determine the tool position normal compensation vector set for each tool position point on the preset tool path. Then, perform multi-dimensional direct-axis vector decomposition on the tool position normal compensation vector set to obtain a set of decomposed compensation vector values. Next, perform rotary axis interpolation adjustment on the tool position normal compensation vector set to obtain a set of rotary axis adjustment vector values. Based on the decomposed compensation vector value set and the rotary axis adjustment vector value set, generate a set of cutting tool instructions for the preset tool path. Finally, control the cutting tool on the CNC machine tool spindle according to the cutting tool instruction set to perform compensating cutting adjustment on the turbine blade to be tested, obtaining the compensated turbine blade.

[0125] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including an adaptive compensation detection unit, a point cloud feature registration unit, a first determination unit, a second determination unit, a multidimensional direct-axis vector decomposition unit, a rotation axis interpolation adjustment unit, a generation unit, and a control unit. The names of these units do not necessarily limit the specific unit; for example, the adaptive compensation detection unit may also be described as "a unit that performs adaptive compensation detection on a turbine blade to be detected using a machine tool probe mounted on the spindle of a CNC machine tool to obtain a three-dimensional point cloud dataset of the blade."

[0128] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0129] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A compensation adjustment method applied to turbine blades, comprising: The turbine blade under test is adaptively compensated and detected by a machine tool probe installed on the spindle of a CNC machine tool, and a three-dimensional point cloud dataset of the blade is obtained. Point cloud feature registration processing is performed on the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix; Based on the blade point cloud registration transformation matrix, determine the set of normal vector deviation values ​​between the blade 3D point cloud dataset and the preset turbine blade 3D model; Based on the set of normal vector deviation values, determine the set of tool position normal compensation vectors for each tool position on the preset tool path, including: mapping the set of normal vector deviation values ​​to the preset turbine blade three-dimensional model to obtain the deviation turbine blade three-dimensional model; The three-dimensional model of the deviated turbine blade is smoothed and denoised to obtain a denoised three-dimensional model of the deviated turbine blade. The semi-variable spatial correlation value set and the distance set between adjacent deviated three-dimensional point cloud sets in the deviated three-dimensional point cloud set included in the denoised three-dimensional model of the deviated turbine blade are determined. Based on the semi-variable spatial correlation value set and the distance set between point clouds, a fitted point cloud deviation function model is generated. Based on the fitted point cloud deviation function model, a neighborhood deviated three-dimensional point cloud set is determined. Based on the neighborhood deviated three-dimensional point cloud set, a set of point cloud interpolation equations for each tool position is generated. The point cloud interpolation equations are constrained and solved to obtain a set of normal vector deviation estimates. The product of each normal vector deviation estimate in the set of normal vector deviation estimates and the normal vector of the corresponding tool position in each tool position is determined as the tool position normal compensation vector, thus obtaining a set of tool position normal compensation vectors. The tool position normal compensation vector set is subjected to multidimensional direct axis vector decomposition to obtain a set of decomposed compensation vector values; The tool position normal compensation vector set is adjusted by rotation axis interpolation to obtain a set of rotation axis adjustment vector values; Based on the decomposed compensation vector value set and the rotation axis adjustment vector value set, a cutting tool instruction set for the preset toolpath is generated; According to the cutting tool instruction set, the cutting tool on the spindle of the CNC machine tool is controlled to perform compensating cutting adjustment on the turbine blade to be tested, so as to obtain the compensated and adjusted turbine blade.

2. The method of claim 1, wherein, The step of performing point cloud feature registration processing on the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix includes: The target point cloud dataset and the blade 3D point cloud dataset included in the preset turbine blade 3D model are input into the first-level point cloud feature extraction network included in the point cloud adaptive registration model to obtain the first blade point cloud feature vector and the first target point cloud feature vector. The point cloud adaptive registration model further includes: a second-level point cloud feature extraction network, a first point cloud feature information interaction network, a third-level point cloud feature extraction network, a fourth-level point cloud feature extraction network, a second point cloud feature information interaction network, and a fifth-level point cloud feature extraction network. The first leaf point cloud feature vector and the first target point cloud feature vector are respectively input into the second-level point cloud feature extraction network to obtain the second leaf point cloud feature vector and the second target point cloud feature vector. The second leaf point cloud feature vector and the second target point cloud feature vector are respectively input into the first point cloud feature information interaction network to obtain the third leaf point cloud feature vector and the third target point cloud feature vector. The point cloud feature vector of the third leaf and the point cloud feature vector of the third target are respectively input into the third-level point cloud feature extraction network to obtain the point cloud feature vector of the fourth leaf and the point cloud feature vector of the fourth target. The point cloud feature vector of the fourth leaf and the point cloud feature vector of the fourth target are respectively input into the fourth-level point cloud feature extraction network to obtain the point cloud feature vector of the fifth leaf and the point cloud feature vector of the fifth target. The feature vectors of the fifth leaf point cloud and the fifth target point cloud are respectively input into the second point cloud feature information interaction network to obtain the feature vectors of the sixth leaf point cloud and the sixth target point cloud. The point cloud feature vector of the sixth leaf and the point cloud feature vector of the sixth target are respectively input into the fifth-level point cloud feature extraction network to obtain the point cloud feature vector of the seventh leaf and the point cloud feature vector of the seventh target. After feature fusion of the feature vectors of the second leaf point cloud, the fifth leaf point cloud, and the seventh leaf point cloud, max pooling is performed to obtain the global leaf point cloud feature vector. Similarly, after feature fusion of the feature vectors of the second target point cloud, the fifth target point cloud, and the seventh target point cloud, max pooling is performed to obtain the global target point cloud feature vector. The global blade point cloud feature vector and the global target point cloud feature vector are subjected to multi-layer convolution processing to obtain the blade point cloud registration transformation matrix.

3. The method of claim 2, wherein, The step of inputting the second leaf point cloud feature vector and the second target point cloud feature vector into the first point cloud feature information interaction network to obtain the third leaf point cloud feature vector and the third target point cloud feature vector includes: Max pooling is performed on the feature vectors of the second leaf point cloud and the second target point cloud respectively to obtain the pooled feature vectors of the leaf point cloud and the target point cloud. The pooled feature vectors of the blade point cloud and the pooled feature vectors of the target point cloud are stacked to obtain stacked feature vectors of the blade point cloud and the target point cloud. The stacked feature vectors of the blade point cloud, the second blade point cloud feature vector, and the stacked feature vector of the target point cloud are fused to obtain the fused feature vector of the blade point cloud, which is used as the third blade point cloud feature vector. The second target point cloud feature vector, the target point cloud stacked feature vector, and the leaf point cloud stacked feature vector are fused to obtain the target point cloud fused feature vector, which is used as the third target point cloud feature vector.

4. The method of claim 1, wherein, The step of determining the set of normal vector deviation values ​​between the blade 3D point cloud dataset and the preset turbine blade 3D model based on the blade point cloud registration transformation matrix includes: The product of the blade point cloud registration transformation matrix and the blade 3D point cloud dataset is determined as the registered blade 3D point cloud dataset. The three-dimensional point cloud set corresponding to the registered three-dimensional point cloud dataset of the turbine blade in the preset three-dimensional model is determined as the target three-dimensional point cloud set. Determine the normal vector of each target 3D point cloud in the target 3D point cloud set, and use it as the target normal vector to obtain the target normal vector set; The distance between each registered blade 3D point cloud data in the registered blade 3D point cloud dataset and the corresponding target normal vector in the target normal vector set is determined as the point cloud distance, thus obtaining the point cloud distance set; The projection components of the point cloud distance set onto the target normal vector set are determined to obtain the normal vector deviation value set.

5. The method of claim 1, wherein, The step of performing rotational axis interpolation adjustment on the tool position normal compensation vector set to obtain a rotational axis adjustment vector value set includes: The dot product of the tool position normal compensation vector set and the tool position normal vector set corresponding to each tool position point is determined as the set of cosine values ​​of the tool position point rotation angle; The set of cosine values ​​of the tool position rotation angle is processed by inverse cosine to obtain the set of tool position rotation angles; Based on the tool position normal vector set and the tool position rotation angle set, spherical linear interpolation is performed on the tool position normal compensation vector set to obtain the interpolated tool position attitude information set. The interpolated tool position posture information set is segmented by interpolation to obtain a tool position posture grouping information set; Determine the horizontal axis direction angle and pitch angle of each tool position attitude group division information in the tool position attitude group division information set to obtain the horizontal axis direction angle group and pitch angle group, which are used as the rotation axis adjustment vector value group set.

6. A compensation and adjustment device for turbine blades, comprising: The adaptive compensation detection unit is configured to perform adaptive compensation detection on the turbine blade to be detected by a machine tool probe mounted on the spindle of a CNC machine tool, and obtain a three-dimensional point cloud dataset of the blade. The point cloud feature registration unit is configured to perform point cloud feature registration processing on the blade 3D point cloud dataset and the preset turbine blade 3D model to obtain the blade point cloud registration transformation matrix. The first determining unit is configured to determine the set of normal vector deviation values ​​between the blade three-dimensional point cloud dataset and the preset turbine blade three-dimensional model based on the blade point cloud registration transformation matrix. The second determining unit is configured to determine the tool position normal compensation vector set for each tool position on a preset tool path based on the normal vector deviation value set, including: mapping the normal vector deviation value set to the preset turbine blade 3D model to obtain a deviation turbine blade 3D model; performing smoothing and denoising processing on the deviation turbine blade 3D model to obtain a denoised deviation turbine blade 3D model; determining the semi-variant spatial correlation value set and the inter-point distance set of adjacent deviation 3D point cloud sets in the deviation 3D point cloud set included in the denoised deviation turbine blade 3D model; and determining the semi-variant spatial correlation value set and the inter-point cloud distance set based on the semi-variant value set. A fitting point cloud deviation function model is generated using the spatial correlation value set and the distance set between the point clouds. Based on the fitting point cloud deviation function model, a neighborhood deviation 3D point cloud set is determined. Based on the neighborhood deviation 3D point cloud set, a set of point cloud interpolation equations for each tool position is generated. The point cloud interpolation equations are constrained and solved to obtain a set of normal vector deviation estimates. The product of each normal vector deviation estimate in the set and the normal vector of the corresponding tool position in each tool position is determined as the tool position normal compensation vector, thus obtaining the tool position normal compensation vector set. A multidimensional direct-axis vector decomposition unit is configured to perform multidimensional direct-axis vector decomposition on the tool position normal compensation vector set to obtain a set of decomposed compensation vector values. A rotary axis interpolation adjustment unit is configured to perform rotary axis interpolation adjustment on the tool position normal compensation vector set to obtain a rotary axis adjustment vector value set; The generation unit is configured to generate a set of cutting tool instructions for the preset toolpath based on the set of decomposed compensation vector values ​​and the set of rotation axis adjustment vector values. The control unit is configured to control the cutting tool on the spindle of the CNC machine tool according to the cutting tool instruction set, so as to perform compensating cutting adjustment on the turbine blade to be tested, and obtain the compensated turbine blade.

7. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-5.

8. A computer readable medium having stored thereon a computer program, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-5.

Images