A constrained registration method for turbine blade datums

Abstract

The present invention discloses a constrained registration method for turbine blade datum, comprising the following steps: measuring point information on the key profile line of the blade to obtain blade profile data; and obtaining a theoretical model of the blade; designing a fitness function containing a penalty term, and using an accelerated PSO algorithm to register the measured points and the theoretical model. The constrained registration method for turbine blade datum proposed by the present invention firstly obtains blade profile data quickly through on-machine measurement or special measuring tools, registers the blade profile data and the blade theoretical model, and ensures the accuracy, rapidity and reliability of blade clamping in subsequent CNC machining.

Description

Technical Field

[0001] The invention relates to the technical field of precision machining, and in particular to a constrained registration method for turbine blade datums. Background Art

[0002] At present, because turbine blades work in high temperature, complex stress, and harsh environment, nickel-based high-temperature alloys are mainly used at home and abroad to manufacture blade blanks through precision casting process. Due to the complex casting process, precision casting turbines are still mainly realized by manual grinding. However, repeated manual grinding and polishing is not only inefficient, inconsistent, and has low yield, but also cannot meet the processing requirements of aero-engine blades, which is not conducive to the subsequent overall assembly of blades and affects the service performance of the engine. In addition, the labor intensity of workers is high and the working environment is harsh. With the rapid development of aviation technology, the performance requirements and output requirements of aero-engines are getting higher and higher, and the demand for efficient CNC processing of precision casting blades is becoming more and more urgent.

[0003] To solve this problem, the three-coordinate measurement is used to measure the precision-cast blades, the position of the blades relative to the fixture is calibrated, and based on the measured blade surface allowance information, the robot adaptive belt grinding method is used to remove quantitative material from the aero-engine precision-cast blades to ensure their processing accuracy. However, the implementation of this method requires the intervention of three-coordinates, and due to problems such as the robot's own accuracy and the accuracy of the removal model, multiple iterative processing is often required to ensure its accuracy and meet the needs of efficient production of large quantities of precision-cast blades.

[0004] This resulted in an array processing device based on independent development. First, a six-point positioning fixture was used to transfer the reference of the precision-cast blade to the casting box with low-melting-point casting, because the six-point positioning had a certain randomness and the deformation of the low-melting-point alloy casting. After the reference was transferred, the precision-cast blade needed to be measured by three-coordinates and clustered based on the measurement data. Finally, the precision-cast blades in the same group were processed with efficient array grinding and polishing. However, this method still did not get rid of the dependence on three-coordinate equipment, which restricted the further improvement of processing efficiency. At the same time, low-melting-point alloys were used in this method, and low-melting-point alloys had the hidden danger of contaminating the blade matrix, and the casting process was gradually being abolished. In summary, there are two main difficulties in the efficient CNC processing of precision-cast blades. First, a new process route suitable for efficient CNC processing of precision-cast blades needs to be formulated; second, the traditional three-coordinate measurement has high cost and long measurement cycle, and the current problem of rapid measurement of blade surface data needs to be solved.

[0005] In view of these two difficulties, and in accordance with the principle of "benchmark first" in mechanical processing, how to quickly obtain blade surface data through on-machine measurement or special measuring tools, and align blade surface data with blade theoretical models, is the primary technical problem at present. This provides a theoretical basis for preparing a process benchmark on the blade tenon or auxiliary fixture based on the blade body surface, ensuring the accuracy, speed and reliability of blade clamping in subsequent CNC processing, and realizing efficient CNC processing of precision casting blades. Summary of the invention

[0006] Purpose of the invention: The purpose of the present invention is to at least partially solve the above-mentioned technical problems, thereby providing a constrained alignment method for turbine blade datums, which quickly obtains blade profile data through on-machine measurement or special measuring tools, aligns blade profile data and blade theoretical models, and ensures the accuracy, speed and reliability of blade clamping in subsequent CNC machining.

[0007] In order to achieve the above object, the present invention provides a constrained registration method for a turbine blade reference, comprising the following steps:

[0008] S100, measuring point information on the key profile line of the blade to obtain blade profile data; and obtaining a theoretical model of the blade;

[0009] S200, design a fitness function with a penalty term, and use the accelerated PSO algorithm to align the measurement points with the theoretical model.

[0010] A further preferred technical solution of the present invention is that the point information on the key surface line of the blade is measured in step S100 to obtain the blade surface data, specifically: after the blade is fixed by mechanical tightening, the point information on the key surface line of the precision casting blade is quickly obtained by on-machine measurement or special measuring tools.

[0011] Preferably, obtaining point information on the key profile line of the blade based on on-machine measurement with a digital micrometer specifically includes:

[0012] S111, measuring the trajectory planned on the blade body according to the blade surface characteristics, the measuring points are planned near the design section lines of the blade basin and the blade back, the tool axis normal vector at the measuring point is perpendicular to the curved surface, and the total number of measuring points is controlled by setting the trajectory error;

[0013] S112, setting the preload ε, approach speed and safety distance parameters of the digital micrometer during measurement, converting the measurement track generated in the CAM software into NC code through a dedicated post-program, and then performing on-machine measurement;

[0014] S113, extract the measurement data corresponding to each measurement point , combined with the measurement point knife position data , tool axis normal vector corresponding to tool position point , preload And the radius of the micrometer side head , the actual coordinates of the blade surface measurement points are solved by formula (1) :

[0015] (1);

[0016] in, i is the number of the measurement point, .

[0017] Preferably, the point information on the key profile line of the blade is obtained based on the offline measurement of the inductance meter, specifically including:

[0018] S121, calibrating the blades as standard blades using three-coordinates;

[0019] S122, use multiple inductance meters to measure the measuring points simultaneously;

[0020] S123, the measurement data when the inductance meter is measured is , the coordinates of the profile data points of the inductance meter corresponding to the standard blade are The head normal vector is , the lateral head radius is , the actual coordinates of the blade surface measurement points are solved by formula (2) :

[0021] (2);

[0022] in i is the serial number of the inductance meter, .

[0023] Preferably, the point position information on the profile line of a small batch of blades is obtained by on-machine measurement, and the point position information on the profile line of a large batch of blades is obtained by a special measuring tool.

[0024] Preferably, the step S200 of designing a fitness function containing a penalty term comprises:

[0025] S211, performing spatial rotation and translation transformation on the measured blade profile points;

[0026] S212, calculate the distance from the measurement point to the blade model after rotation and translation point by point l i ;

[0027] S213, the sum of the squares of each distance is used as the original fitness function , and at the same time determine whether each distance is within the minimum machining allowance Tolerance band Within the range, the penalty term of the fitness function is calculated using formula (3) :

[0028] (3);

[0029] in, is the penalty factor, The calculation formula is:

[0030] (4);

[0031] in, , , and Corresponding to the measurement points The upper deviation, lower deviation, minimum machining allowance and error coefficient of the blade;

[0032] S214, the final fitness function is designed as , the registration optimization objective is Minimum.

[0033] Preferably, the step S211 performs a spatial rotation and translation transformation on the measured blade profile points, specifically:

[0034] The original measurement point is transformed by rotating and translating around a fixed coordinate system. During the transformation, the origin of the fixed coordinate system and the measurement point Rotate around a fixed axis and then translate;

[0035] Let the transformation parameters be and ,in for Rotate the angle around the X axis, for Rotate the angle around the Y axis, for Rotate the angle around the Z axis, , , for The translation along the X, Y, and Z axes after rotating around the X, Y, and Z axes in sequence; Homogenization and left multiplication of transformation matrices in sequence can obtain the measured points after rotation and translation. .

[0036] Preferably, the step S212 of calculating the distance from the measurement point to the blade model after rotation and translation point by point is specifically as follows:

[0037] The blade model is expressed as a triangular mesh model, and a dynamic retrieval tree is established for the triangular mesh surface. The retrieval tree is used to quickly search for the distance to the point Neighboring triangles within the threshold distance ,in , calculate the points in sequence To the adjacent triangle Distance , and take the one with the smallest absolute value as the final distance.

[0038] Preferably, the PSO algorithm after accelerated processing is used to align the measurement points with the theoretical model in step S200, and the specific method is as follows:

[0039] Optimize transformation parameters through PSO algorithm during registration and , that is, searching for the best parameter value in the variable space of 6 transformation parameters;

[0040] , as well as To set the search space, first specify the search range for the movement amount along each coordinate axis. , , To traverse each measurement point, obtain the farthest distance from the origin in the coordinate component of each measurement point; the search range of the rotation amount along each coordinate axis of the search space , and Set by formula (5):

[0041] (5);

[0042] Set the particle swarm to move the search range around the global optimal value after each generation of updates. , and Multiply by a factor , and recalculate the rotation search range , and Then start a new round of optimization.

[0043] The present invention also provides a turbine blade benchmark processing method, firstly adopting the above-mentioned turbine blade benchmark constrained registration method to align the measurement points with the theoretical model, and then performing the following steps:

[0044] S300, obtaining a transformation matrix between the actual installation position of the blade and the installation position in the theoretical model;

[0045] S400, performing inverse coordinate transformation on the tool position point in the reference machining program generated based on the theoretical model;

[0046] S500, using the transformed processing program to perform benchmark processing on the blade.

[0047] Beneficial effect: The constrained registration method of the turbine blade benchmark proposed in the present invention firstly obtains the blade surface data quickly through on-machine measurement or special measuring tools, and aligns the blade surface data and the blade theoretical model to ensure the accuracy, speed and reliability of blade clamping in subsequent CNC machining.

[0048] The present invention takes into account the different design tolerances of different parts of the blade, and specifies upper and lower deviations and minimum margin requirements according to the different positions of each measuring point; after the overall alignment of the blade, the single-section data is torsionally and translationally aligned in the section plane to obtain the best rotation and translation amounts, and the rotation and translation amounts of the alignment of dozens of turbine blades are counted at a time, the rotation and translation amounts are grouped, and a blade model is established for each group. The later alignment is to align the measuring points with different models in turn, and can support grouping.

[0049] The conventional objective function uses the minimum square distance of each data point as the conventional target value, without considering the difference in the margins between the back and the front and rear edges of the leaf basin. The present invention uses the minimum sum of symmetrical point errors as the target value, and the margins are more uniform after alignment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flow chart of a constrained registration method for a turbine blade benchmark of the present invention;

[0051] Figure 2 This is a comparison chart of the optimal appropriate changes using the traditional algorithm and the method of the present invention in the verification example of the present invention;

[0052] Among them, (a) is the optimal moderate change comparison diagram when the population size is 100, and (b) is the optimal moderate change comparison diagram when the population size is 500;

[0053] Figure 3 This is a diagram of cross-section line detection results using the method of the present invention in a verification example of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments, and they should not be understood as limitations on the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0055] Combine the following Figure 1-Figure 3 The invention describes a turbine blade datum constrained registration method and a turbine blade datum processing method.

[0056] Embodiment 1: This embodiment provides a constrained registration method for a turbine blade reference, comprising the following steps:

[0057] S100, measuring point information on the key profile line of the blade to obtain blade profile data; and obtaining a theoretical model of the blade;

[0058] S200, design a fitness function with a penalty term, and use the accelerated PSO algorithm to align the measurement points with the theoretical model.

[0059] Specific instructions:

[0060] In step S100, the point information on the key profile line of the blade is measured to obtain the blade profile data.

[0061] The point position information on the profile line of small batches of blades is mainly obtained through on-machine measurement, and the point position information on the profile line of large batches of blades is obtained through special measuring tools.

[0062] The method for obtaining the point information on the key profile line of the blade based on the on-machine measurement of the digital micrometer is:

[0063] S111, measuring the trajectory planned on the blade body according to the blade surface characteristics, the measuring points are planned near the design section lines of the blade basin and the blade back, the tool axis normal vector at the measuring point is perpendicular to the curved surface, and the total number of measuring points is controlled by setting the trajectory error;

[0064] S112, setting the preload ε, approach speed and safety distance parameters of the digital micrometer during measurement, converting the measurement track generated in the CAM software into NC code through a dedicated post-program, and then performing on-machine measurement;

[0065] S113, extract the measurement data corresponding to each measurement point , combined with the measurement point knife position data , tool axis normal vector corresponding to tool position point , preload And the radius of the micrometer side head , the actual coordinates of the blade surface measurement points are solved by formula (1) :

[0066] (1);

[0067] in, i is the number of the measurement point, .

[0068] The point information on the key profile line of the blade is obtained by offline measurement based on the inductance meter. The method is:

[0069] S121, calibrating the blades as standard blades using three-coordinates;

[0070] S122, use multiple inductance meters to measure the measuring points simultaneously;

[0071] S123, the measurement data when the inductance meter is measured is , the coordinates of the profile data points of the inductance meter corresponding to the standard blade are , the normal vector of the side head is , the lateral head radius is , the actual coordinates of the blade surface measurement points are solved by formula (2) :

[0072] (2);

[0073] in i is the serial number of the inductance meter, .

[0074] Step S200 includes two parts:

[0075] 1. Design a fitness function with a penalty term, including the following steps:

[0076] S211, performing spatial rotation and translation transformation on the measured blade profile points; the method is:

[0077] The original measurement point is transformed by rotating and translating around a fixed coordinate system. During the transformation, the origin of the fixed coordinate system and the measurement point Rotate around a fixed axis and then translate;

[0078] Let the transformation parameters be and ,in for Rotate the angle around the X axis, for Rotate the angle around the Y axis, for Rotate the angle around the Z axis, , , for The translation along the X, Y, and Z axes after rotating around the X, Y, and Z axes in sequence; Homogenization and left multiplication of transformation matrices in sequence can obtain the measured points after rotation and translation. , the calculation formula is:

[0079] ;

[0080] ;

[0081] ;

[0082] S212, calculate the distance from the measurement point to the blade model after rotation and translation point by point l i ;

[0083] In order to quickly calculate the point The distance to the surface is expressed by the precision casting blade model as a triangular mesh model, and a dynamic retrieval tree is established for the triangular mesh surface. The retrieval tree is used to quickly search for the distance point Neighboring triangles within the threshold distance (in ), calculate the points To the adjacent triangle Distance , take the one with the smallest absolute value as the final distance;

[0084] The retrieval node MBR in the dynamic retrieval tree (R-tree) is improved to MBS, and the minimum circumscribed ball of the triangle patch is used as the data node. The node splitting algorithm is used when the node radius and the number of child nodes are greater than the preset threshold, so as to stably and quickly realize the construction of the retrieval tree of the triangular mesh surface and the retrieval of adjacent triangle patches. For acute triangles, the center and radius of the circumscribed circle of the three vertices are used as the center and radius of the data node MBS; for obtuse and right triangles, the midpoint of the obtuse side is used as the center of the data node MBS, and half of the length of the obtuse side is used as the radius of the data node MBS.

[0085] point To the adjacent triangle S Distance The calculation process is as follows:

[0086] Step 1: Calculate the points To the adjacent triangle S Plane Plane Distance , and get the point In plane Plane Upper projection point , judgment point Is it in the triangle patch? S inside, such as dot In the triangle patch S Inside Valid, order , the program ends, otherwise execute Step2;

[0087] Setp2: Get triangle patches S Three sides , , ,make , execute Step3;

[0088] Step 3: Calculate points arrive Distance of the straight line , and judge arrive Is the projection point on the straight line in If it is within the two endpoints, mark Valid, otherwise mark Invalid, execute Step 4;

[0089] Step 4: ,like , execute Step3, otherwise execute Step5;

[0090] Step 5: Judgement , , If there is a valid distance, execute Step 6; otherwise, execute Step 7.

[0091] Step 6: Obtain by comparison , , The minimum valid value in , according to the triangle patch S The normal vector is determined later After the positive and negative , the program ends;

[0092] Step 7: Calculate points one by one To triangle patch S Distance between three vertices , , , compare and obtain , , Minimum value of , according to the triangle patch S The normal vector is determined later After the positive and negative , the program ends.

[0093] S213, the sum of the squares of each distance is used as the original fitness function , the distance from each point to the leaf is recorded as , k is the number of the point, then:

[0094] ;

[0095] At the same time, determine whether each distance is within the minimum machining allowance Tolerance band Within the range, the penalty term of the fitness function is calculated using formula (3) :

[0096] (3);

[0097] in, is the penalty factor, The calculation formula is:

[0098] (4);

[0099] in, , , Corresponding to the measurement points The upper deviation, lower deviation, minimum machining allowance and error coefficient of the blade;

[0100] S214, the final fitness function is designed as , the registration optimization objective is Minimum.

[0101] 2. Use the accelerated PSO algorithm to align the measurement points with the theoretical model;

[0102] The PSO optimization algorithm is an evolutionary optimization algorithm based on collective wisdom invented by Kennedy in 1995. Compared with genetic algorithms, fish swarm algorithms and taboo search algorithms, the PSO algorithm has a simple structure, good robustness, fast convergence speed, and is easy to implement. It is widely used in various engineering optimization problems. For this reason, the present invention selects the PSO algorithm to achieve the registration of measured data and theoretical blade profiles. However, the PSO optimization algorithm is prone to premature convergence when optimizing large-scale, high-dimensional, and nonlinear discrete targets, and it is difficult to efficiently solve the optimization problem of complex problems. During registration, the PSO algorithm is used to optimize the transformation parameters ( γ,β,α,M_x,M_y,M_z), that is, to seek the best parameter value in the variable space of 6 transformation parameters, and the optimization numerical precision is required to be high, so the standard PSO algorithm needs to be improved in order to achieve higher optimization precision and efficiency. The traditional improvement method is to improve the weight factor or learning factor, which improves the convergence speed and optimization results of the algorithm to a certain extent. However, the improved algorithm does not consider the correlation characteristics between the data of each dimension in the optimization space and the influence of the population density in the search space on the final optimization precision, thus limiting the further improvement of the performance of the PSO optimization algorithm in practical applications.

[0103] Optimize transformation parameters through PSO algorithm during registration , that is, searching for the best parameter value in the variable space of 6 transformation parameters;

[0104] , as well as To set the search space, first specify the search range for the movement amount along each coordinate axis. , , To traverse each measurement point, obtain the farthest distance from the origin in the coordinate component of each measurement point; the search range of the rotation amount along each coordinate axis of the search space , and Set by formula (5):

[0105] (5);

[0106] The population density in the search space has a huge impact on the final optimization accuracy, but a larger population size will reduce the execution efficiency of the algorithm. In order to ensure the search accuracy and algorithm operation efficiency, the present invention sets the particle swarm to update every time during the particle update process. M After the generation, the search range is moved to the global optimal value as the center. , , Multiply by a factor , and recalculate the rotation search range , , Then a new round of optimization begins, and high-precision optimization values ​​are quickly obtained by gradually reducing the search range.

[0107] Embodiment 2: This embodiment provides a turbine blade benchmark processing method.

[0108] In this embodiment, the constrained registration method of the turbine blade benchmark of embodiment 1 is first used to align the measurement points with the theoretical model. In the blade profile error, each interface contour is allowed to have a certain twist and center movement error. After the blade is fully registered, the present invention twists the single section data in the section plane. and translation , Registration, obtain the best rotation and translation, count the registered rotation of dozens of turbine blades at a time and translation , , , for rotation and translation , The samples are grouped and a blade model is built for each group. The subsequent registration is to use different models to register the measurement points in turn.

[0109] Then proceed as follows:

[0110] S300, obtaining a transformation matrix between the actual installation position of the blade and the installation position in the theoretical model;

[0111] S400, performing inverse coordinate transformation on the tool position point in the reference machining program generated based on the theoretical model;

[0112] S500, using the transformed processing program to perform benchmark processing on the blade.

[0113] Based on the above embodiments, the above methods were integrated into the blade processing software developed in the laboratory, and a benchmark preparation software system was developed and verified in two blade experiments and production.

[0114] 1. For the 01 model blade, the blade body tolerance requirement after post-polishing is ±0.075mm, and the inlet and exhaust edge requirement is -0.15~0.10mm. Due to the size limitation of the tenon itself, it is difficult to directly process the tenon part. Therefore, a conversion fixture is designed. The blade is fixed in the conversion fixture by screws. After on-machine measurement and alignment, the six reference planes on the conversion fixture are refined.

[0115] The measurement trajectory was planned and generated in the CAM (UG) software and the on-machine measurement was performed. The lower deviation was set to -0.075mm according to the requirements of the blade body. The upper deviation was set to 1mm according to the characteristics of the later removal and fitness function. Taking into account the later polishing removal (≥0.01mm) and clamping error (≤0.03mm), the minimum machining allowance of the blade was set to 0.05mm. After processing the measured data, the blade was aligned. When optimizing the alignment parameters, the mobile search range was set to ±3, ±3, ±2, and the rotation search range was automatically calculated by the system. The penalty factor was , the inertia weight is 0.9~0.4, the maximum number of iterations of the termination condition is 60, and the traditional algorithm and the improved algorithm of the present invention are respectively used ( M =20, ), the optimal moderate change comparison is performed on the instances with population sizes of 100 and 500. The comparison results are as follows Figure 2 As shown, it is proved that the convergence speed of the fitness value of the improved algorithm of the present invention is significantly improved.

[0116] If each measuring point meets the processing requirements after registration, the reference preparation program is exported to process each reference plane of the auxiliary fixture to complete the reference preparation work. The prepared precision casting blade and auxiliary fixture are fixed in a special fixture, and the fixture coordinate system is established on a three-dimensional coordinate measuring machine (Beijing Hangrui Siwei PONY866, Metrolog XG13 contact measurement system, MPE=(2.5+ 4L / 1000)um, probe diameter 1.5mm) using Renishaw SP25M contact scanning probe. The six cross-section lines of the precision casting blade are tested. The test results are as follows Figure 3 As shown, the minimum deviation of the tested section is -0.03mm. Taking into account the measurement error, processing error and clamping error, the deviation is within the specified range and can meet the processing requirements.

[0117] At the same time, the time for each link of the benchmark preparation is shown in Table 1. According to statistics, the measurement time of 30 surface data points is 3min22s, the measurement point registration calculation time is 2min, and the benchmark processing time is 2min45s. The automatic tool change function was not used in the experiment. The total preparation time is 11min37s. If the measurement and processing are distributed to two machine tools, and the processing is extended by an automatic tool magazine, the benchmark preparation time can be controlled within 5min, which is significantly more efficient than the current manual alignment and clamping and three-coordinate measurement and registration method (on-site statistics are about 8min).

[0118] Table 1 Benchmark preparation time statistics:

[0119]

[0120] 2. After post-process grinding and polishing, the blade body tolerance of the 02 model blade is -0.02~0.06mm, and the intake and exhaust edge requirement is -0.08-0.12mm. The batch size of this blade is large and the tenon size is large, so the inductance meter gauge and tenon reference preparation fixture are designed, and the reference preparation is achieved by milling six reference surfaces on the tenon. The six reference surfaces on the tenon of the successfully aligned blade are processed, and the blades after the reference preparation are array processed to verify the feasibility and stability of the modified process.

[0121] It has been proven that this process can fully meet the mass production of this type of blades. Statistics from the production site show that the average time for blade measurement and alignment is ≤1.5min, and the average time for reference surface milling is ≤3min. By changing the theoretical model within the tolerance range, the blades can be quickly grouped while preparing the blade reference. Compared with the previous manual clamping and three-coordinate measurement grouping, the blade clamping and grouping time is shortened to less than 20% of the original time, greatly improving the production efficiency of the blades.

[0122] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes in form and details may be made without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A constrained registration method for turbine blade datums, characterized in that: The steps include: S100, measuring point information on the key profile line of the blade to obtain blade profile data; and obtaining a theoretical model of the blade; When measuring the point information on the key profile line of the blade and obtaining the blade profile data, the blade is fixed by mechanical clamping, and the point information on the key profile line of the precision casting blade is quickly obtained by in-machine measurement or special measuring tools; Among them, the on-machine measurement based on the digital micrometer obtains the point information on the key profile line of the blade, specifically including: S111, measuring the trajectory planned on the blade body according to the blade surface characteristics, the measuring points are planned near the design section lines of the blade basin and the blade back, the tool axis normal vector at the measuring point is perpendicular to the curved surface, and the total number of measuring points is controlled by setting the trajectory error; S112, setting the preload ε, approach speed and safety distance parameters of the digital micrometer during measurement, converting the measurement track generated in the CAM software into NC code through a dedicated post-program, and then performing on-machine measurement; S113, extract the measurement data d corresponding to each measurement point i , combined with the measurement point knife position data P i , the tool axis normal vector v corresponding to the tool position point i , preload ε and micrometer head radius R, the actual coordinate C of the blade profile measurement point is solved by formula (1) i : C i =P i +(d i -ε+R)·v i (1); Wherein, i is the number of the measuring point, i=0,…,n; The point information on the key profile line of the blade is obtained through offline measurement based on the inductance meter, including: S121, calibrating the blades as standard blades using three-coordinates; S122, use multiple inductance meters to measure the measuring points simultaneously; S123, the measurement data when the inductance meter is measured is d i The coordinates of the profile data point of the inductance meter corresponding to the standard blade are P i , the normal vector of the side head is v i , the radius of the side head is R, and the actual coordinate C of the blade profile measurement point is solved by formula (2): i : C i =P i +(d i +R)·v i (2); Where i is the number of the inductance meter, i = 0,…,n; S200, design a fitness function with a penalty term, and use an accelerated PSO algorithm to align the measurement points with the theoretical model; When designing a fitness function with a penalty term, the upper deviation, lower deviation, minimum machining allowance and error coefficient are specified according to the position of each measurement point; the specific methods include: S211, performing spatial rotation and translation transformation on the measured blade profile points; S212, calculate the distance l from the measurement point to the blade model after rotation and translation point by point i ∈{l0,l1,…,l n }; S213, taking the sum of the squares of each distance as the original fitness function T0, and judging whether each distance is within the tolerance band [E I , E S ], the penalty term T1 of the fitness function is calculated using formula (3): The calculation formula is: Wherein, i is the number of the measuring point, i=0,…,n; δ i and μ i are the upper deviation, lower deviation, minimum machining allowance and error coefficient of the blade at the measuring point respectively; S214. Design the final fitness function as T=T0+T1, and the registration optimization target is to minimize T.

2. The constrained registration method for turbine blade datums according to claim 1, characterized in that: The point position information on the profile line of small batches of blades is obtained by on-machine measurement, and the point position information on the profile line of large batches of blades is obtained by special measuring tools.

3. The constrained registration method for turbine blade datums according to claim 1, characterized in that: In step S211, the measured blade profile points are subjected to spatial rotation and translation transformation, specifically: The coordinate transformation of the original measurement point is carried out by rotating and translating around a fixed coordinate system. During the transformation process, the origin of the fixed coordinate system and the measurement point C i (x i ,y i ,z i ) Rotate around a fixed axis and then translate; Let the transformation parameters be γ, β, α, M x ,M y and M z , where γ is C i Rotation angle around the X axis, β is C i Rotation angle around the Y axis, α is C i Rotation angle around the Z axis, M x 、M y 、M z C i The translation along the X, Y, and Z axes after rotating around the X, Y, and Z axes in sequence; i Homogenization and left multiplication of transformation matrices in sequence can obtain the measured points after rotation and translation.

4. The constrained registration method for turbine blade datums according to claim 1, characterized in that: In step S212, the distance between the measurement point and the blade model after rotation and translation is calculated point by point, specifically: The leaf model is expressed as a triangular mesh model, and a dynamic retrieval tree is established for the triangular mesh surface. The retrieval tree is used to quickly search for adjacent triangular facets S within the threshold distance from point P. j , where j = 0, 1, ..., m, calculate the distance from point P to the adjacent triangle S in sequence j The distance d j , and take the one with the smallest absolute value as the final distance.

5. The constrained registration method for turbine blade datums according to claim 4, characterized in that: In step S200, the PSO algorithm with accelerated processing is used to align the measurement points with the theoretical model. The specific method is as follows: During registration, the transformation parameters γ, β, α, and M are optimized by the PSO algorithm. x ,M y and M z , that is, searching for the best parameter value in the variable space of 6 transformation parameters; ±Range_x, ±Range_y, and ±Range_z are the search ranges for the movement along each coordinate axis when setting the search space. max , Y max , Z max To traverse each measurement point, the farthest distance from the origin in the coordinate component of each measurement point is obtained respectively; the search range of the rotation amount along each coordinate axis of the search space ±Range_γ, ±Range_β and ±Range_α is set by formula (5): The particle swarm is set to be centered on the global optimal value after each update of several generations. The moving search range ±Range_x, ±Range_y and ±Range_z are multiplied by the coefficient k, and the rotating search range ±Range_γ, ±Range_β and ±Range_α are recalculated to start a new round of optimization.

6. A turbine blade benchmark processing method, characterized in that: Firstly, the constrained registration method of the turbine blade benchmark described in claim 1 is used to align the measurement points with the theoretical model; After the overall registration of the blade, the single-section data is torsionally and translationally registered in the section plane to obtain the best rotation and translation. The registered rotation and translation of dozens of turbine blades are counted at a time, and the rotation and translation are grouped. A blade model is established for each group. The later registration is to use different models to register the measurement points in turn; then the following steps are performed: S300, obtaining a transformation matrix between the actual installation position of the blade and the installation position in the theoretical model; S400, performing inverse coordinate transformation on the tool position point in the reference machining program generated based on the theoretical model; S500, using the transformed processing program to perform benchmark processing on the blade.

Images