CN112284290B

CN112284290B - Autonomous measurement method and system for aero-engine blade robot

Info

Publication number: CN112284290B
Application number: CN202011122728.3A
Authority: CN
Inventors: 王耀南; 唐永鹏; 缪志强; 毛建旭; 朱青; 张辉; 周显恩; 江一鸣; 彭伟星; 刘学兵; 林杰
Original assignee: Hunan University
Current assignee: Hunan University
Priority date: 2020-10-20
Filing date: 2020-10-20
Publication date: 2021-09-28
Anticipated expiration: 2040-10-20
Also published as: CN112284290A

Abstract

The invention discloses an autonomous measuring method and system for an aircraft engine blade robot. The method comprises the following steps: calibrating a servo rotary workbench, a structured light three-dimensional scanner and a robot, arranging measuring points according to an aircraft engine blade design model, planning a path measured by the robot, measuring the three-dimensional appearance of the aircraft engine blade, and processing the measurement data of the aircraft engine blade. The system comprises a hardware system and a software system, wherein the hardware system comprises a servo rotary workbench, a blade measuring clamp, a structured light three-dimensional scanner, a robot control cabinet and an industrial computer; the software system comprises a measuring system calibration module, a measuring point layout module, a robot measuring path planning module, a point cloud data processing and visualization module, a three-dimensional model format conversion module, a human-computer interaction module and a measuring process control module. According to the invention, the measuring points are automatically distributed and the measuring path of the robot is generated according to the product design model, so that the autonomous measurement of the robot is realized.

Description

Autonomous measurement method and system for aero-engine blade robot

Technical Field

The invention relates to the technical field of robot three-dimensional measurement, in particular to an autonomous measurement method and system for an aircraft engine blade robot.

Background

The blade of the aircraft engine is a core part of the aircraft engine, directly determines the aerodynamic performance and the service life of the aircraft engine, and is usually made of titanium alloy, high-temperature alloy and other materials, and the manufacturing amount of the blade accounts for about 30% of the total manufacturing ratio of the aircraft engine. When the blade works in a severe environment with high temperature and high pressure for a long time, the thermal barrier coating is easy to fall off, ablate, crack, corrode and the like, and the blade is broken when the damage is serious. The measurement of the three-dimensional shape of the blade is the key for evaluating and improving the processing quality of the blade, and the acquisition of the damage data of the blade profile is also of great significance for the repair and reprocessing of the blade.

Compared with the measurement of other parts, the blade measurement of the aero-engine has the advantages of high measurement precision requirement, high measurement efficiency requirement, complex processing and analysis of blade measurement data, high comprehensive evaluation difficulty and high measurement reliability requirement.

At present, in the measuring process of the blade of the domestic aero-engine, the traditional standard sample plate measuring means still dominates, the efficiency is low, the development is slow, and the integrated process of design, manufacture and detection is severely restricted. In order to meet the requirement of rapid and efficient detection, three-coordinate measuring machines are commonly adopted in developed western countries to detect blades of aero-engines. The detection method is high in precision, good in repeatability, strong in universality and the like, can finish comprehensive measurement on the blade profile of the aero-engine, is one of the methods with the highest measurement precision of the blade profile of the aero-engine at present, but is low in detection efficiency, complex in measurement path planning and directly influences the measurement result by the radius compensation precision of the measuring head.

In order to ensure the measurement precision of the blade of the aero-engine and simultaneously improve the measurement efficiency and the automation degree of the blade of the aero-engine, the combination of the structured light three-dimensional scanner and the robot technology is an important way for solving the technical problem.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the method and the system for automatically measuring the blade of the aircraft engine overcome the defects in the prior art, and improve the efficiency and the degree of automation of blade measurement of the aircraft engine.

The technical solution of the invention is as follows: the method for automatically measuring the blade robot of the aircraft engine comprises the following steps:

s1: calibrating a servo rotary workbench, a structured light three-dimensional scanner and a robot;

s2: arranging measurement points according to an aircraft engine blade design model;

s3: planning a path measured by the robot;

s4: measuring the three-dimensional appearance of the blade of the aero-engine;

s5: measurement data of the aircraft engine blade is processed.

Further, the calibration servo rotary worktable, the structured light three-dimensional scanner and the robot comprise the following steps:

s11: fixing three precise standard balls on a servo rotary worktable, and ensuring that the three precise standard balls are within the working distance and the visual field range of the structured light three-dimensional scanner;

s12: keeping the robot still, rotating the servo rotary worktable at a stable rotation angle alpha, scanning once by the structured light three-dimensional scanner to obtain measurement data, and calculating an average clustering sphere center P;

s13: continuously keeping the robot still, rotating the servo rotary worktable for N times at a stable rotation angle alpha to ensure that the angle is more than 360 degrees, and recording the average clustering center P_iWhere i 1.. times.n, and fitting the set of points { P using a least squares method_iGet the average cluster center P_iRadius R of circular motion around the axis of rotation of the servo-rotating table₀；

S14: the robot acts to set the terminal attitude xi of the robot_jHerein, thisWhile the servo-rotating table rotates by a stable angle theta_iRotating, once scanning by the structured light three-dimensional scanner, and calculating the average clustering sphere center P of observation_ijWherein i 1, 1.. and N, j 1.. and M, and ensure

And when the tail end of the robot is in xi posture_jThe three precision standard balls are within the working distance and the visual field range of the structured light three-dimensional scanner;

s15: observation of result R according to step S13₀And step S14 observing result { P }_ijI 1, N, j 1, M, and performing hand-eye calibration of the robot and the structured light three-dimensional scanner^handT_cameraAnd determining the attitude transformation relationship between the axis of rotation of the servo rotary table and the robot base coordinate system^baseT_axisWherein { hand } is a robot end coordinate system, { base } is a robot base coordinate system, { camera } is a structured light three-dimensional scanner coordinate system, and { axis } is a coordinate system of a servo rotary table rotation axis.

Further, the method for arranging the measuring points according to the aircraft engine blade design model comprises the following steps:

s21: importing an aircraft engine blade design model, and carrying out Poisson disc sampling to generate a point cloud M ═ finish^bladeP_model,i|i＝1,...,N_mIn which N is_mThe number of points in the point cloud M;

s22: clustering point cloud M to generate Gaussian mixture model

Wherein λ_iK is the weight and the mixing quantity of the Gaussian mixture model, mu_i、∑_iAre respectively Gaussian distribution

A mean and covariance matrix of;

s23: according to μ_i、∑_iGenerating a set of candidate measurement points from the working distance D of the structured light three-dimensional scannerClosed { V_i1., K }, where V ═ i ═ 1_iContaining positions v of candidate measurement points_iAnd direction n_i；

S24: from a set of candidate measurement points { V_iChoose the best set of candidate measurement points { D } from | i ═ 1_i1,. n, where n is equal to or less than K, and ensuring that the robot obtains the integrity of the model by measuring, where n is the number of the finally arranged measuring points;

s25: setting D ═ D_iI 1.. n } depending on the set of selected best candidate measurement points D ═ D ·_iI 1, n, determining the optimal sequence of measurement points

Wherein s is_i＝1,...,n。

Further, the planning of the measured path of the robot includes the following steps:

s31: clamping and fixing the aero-engine blade on a blade measuring clamp, and ensuring that the blade is within the working distance and the visual field range of a structured light three-dimensional scanner;

s32: the structured light three-dimensional scanner scans once to obtain measurement data, removes noise in the measurement data, and segments a point cloud of a blade area into a final arch^cameraP_observed,j|j＝1,...,N_sIn which N is_sThe number of points in the point cloud S;

s33: aligning the point cloud S of the blade area of the aircraft engine with the point cloud M of the blade design model of the aircraft engine through a point cloud registration algorithm, and solving a transformation matrix into^modelT_observedGuarantee the corresponding point pair^cameraP_observed,jAnd^bladeP_model,isatisfy the requirement of

Wherein { model } is an aeroengine blade design model coordinate system, { observed } is an aeroengine blade measurement data coordinate system, { blank } is an aeroengine blade coordinate system to be measured,^cameraP_observed,jto measure points in the data for a structured light three-dimensional scanner,^bladeP_model,idesigning points in the model for the aircraft engine blade;

s34: from matrix equations

Solving the transformation matrix between the coordinate systems { blade } and { axis }:^axisT_blade＝(^baseT_axis)^-1·^baseT_hand·^handT_camera·(^modelT_observed)^-1；

s35: according to the optimal sequence of measurement points D_sequence＝{D₁,...,D_k,...D_n}, target measurement point D_k＝{v_k,n_k}＝{^bladev_k,^bladen_kH, calculating the movement of the robot to D_kThe shortest straight path L;

s36: calculating the position and the direction of a target measuring point in a robot base coordinate system:

wherein the content of the first and second substances,

respectively the position and the direction of a measuring point in a rotating shaft coordinate system of the servo rotating workbench,^axisP_camerafor the current position of the structured light three-dimensional scanner in the rotary axis coordinate system of the servo rotary worktable,^axisT_bladeis a transformation matrix between coordinate systems { blade } and { axis }, r₁＝^axisP_camera-(^axisP_camera ^T·e_z)e_z，

e_z＝[0 0 1]^T，s_k＝1,...,n，R_z(theta) rotation of the rotary table about the axis of rotation zA rotation matrix of the angle is formed,

s37: calculating the translational motion of the structured light three-dimensional scanner from the current attitude motion to the target measurement point

Rotation matrix R of minimum rotational motion cos θ I- (1-cos θ) nn^T+ sin θ n ^ wherein,

θ＝arccos(λ)，n＝^basew/||^basew |, n ^ n ═ n_x n_y n_z]^TThe anti-symmetric matrix of (a) is,

for the position of a target measuring point of a structured light three-dimensional scanner,^baseP_camera、^basen_zrespectively the current position and direction of the structured light three-dimensional scanner;

s38: calculating target pose xi of robot motion^*Calculating

x＝m₃₂-m₂₃，y＝m₁₃-m₃₁，z＝m₂₁-m₁₂And according to the quaternion q being w + xi + yj + zk, the target pose of the robot motion

Wherein the content of the first and second substances,

for the position of the robot end at the target measurement point, R^*Is the attitude of the tail end of the robot at a target measuring point, R is a rotation matrix,^baseP_handis the current position of the robot end-point,^baseR_handthe motion trail of the robot is generated by linear interpolation of a robot controller in a robot control cabinet for the current posture of the tail end of the robot;

s39: calculating the rotation angle of the servo rotary table movement

The target position of the servo rotary table is

Minimum time T of servo rotary table movement_amin＝α/ω_maxWherein e is_zIs [ 001 ]]^T，θ_aFor the current position of the servo rotary table sgn () is a sign function, ω_maxMaximum angular velocity of rotation, r, for steady motion of the servo rotary table₁For measuring the radial vector of a point relative to the axis of rotation, r₂The radial vector of the target measurement point relative to the rotation axis is obtained.

Further, the computing robot moves to D_kComprises the following steps:

s351: calculating the current position of the structured light three-dimensional scanner in a rotating shaft coordinate system of the servo rotating workbench:

s352: calculating the optimal measurement point

Description in { axis } coordinate System

S353: calculating the vertical distance of the structured light three-dimensional scanner from the current position to the target measuring point

Wherein the content of the first and second substances,

for the position of the measuring point in the coordinate system of the rotary axis of the servo rotary table,^axisP_camerathe current position of the structured light three-dimensional scanner in a rotary shaft coordinate system of the servo rotary worktable;

s354: calculating the radial vector r of the measuring point relative to the axis of rotation₁And the radial vector r of the target measurement point relative to the axis of rotation₂：

S355: calculating the shortest straight path of the robot moving to the target measuring point:

the shortest time T of the linear motion of the robot_rmin＝L/v_maxWherein v is_maxThe set maximum speed of the linear motion of the robot is obtained.

Further, the method for measuring the three-dimensional appearance of the blade of the aircraft engine comprises the following steps:

s41: controlling the robot and the servo rotary worktable to move to a target pose according to the planned robot measurement path;

s42: adjusting the motion speed v of the robot to be L/T and the angular speed omega of the servo rotary table to be alpha/T so that the two move synchronously, wherein T is max { T {_amin,T_rminIn which T is_rminIs the shortest time of linear motion of the robot, T_aminIs served asThe shortest time of the rotary worktable movement, wherein alpha is the rotation angle of the servo rotary worktable movement;

s43: the structured light three-dimensional scanner scans once, the measured data is transmitted to an industrial computer for processing, and a target measuring point D is updated_k；

S44: steps S41-S43 are repeated until the robot traverses all target measurement points.

Further, the method for processing the measurement data of the blade of the aircraft engine comprises the following steps:

s51: denoising point cloud data;

s52: simplifying the point cloud of the blades of the aircraft engine;

s53: segmenting point clouds in the blade area of the aircraft engine;

s54: splicing point clouds of blades of the aircraft engine;

s55: reconstructing a curved surface model of the blade of the aircraft engine;

s56: and analyzing and evaluating the dimension error of the profile of the blade of the aircraft engine.

The invention also provides an autonomous measuring system of the blade robot of the aircraft engine, which comprises a hardware system and a software system, wherein:

the hardware system comprises a servo rotary workbench, a blade measuring clamp, a structured light three-dimensional scanner, a robot control cabinet and an industrial computer; wherein:

the servo rotary workbench is connected and communicated with the industrial computer through a field bus and is used for being matched with the robot to measure and executing a motion instruction issued by the industrial computer;

the blade measuring clamp is fixed on the servo rotary workbench through a connecting piece and is used for clamping and fixing the aero-engine blade to be measured;

the structured light three-dimensional scanner is fixed on a tail end flange of the robot through a connecting piece and used for collecting measurement data, and the measurement data are transmitted to an industrial computer through an industrial Ethernet for processing;

one end of the robot control cabinet is communicated with the robot through a field bus, and the other end of the robot control cabinet is communicated with the industrial computer through a network and used for controlling the motion of the robot so that the structured light three-dimensional scanner at the tail end of the robot moves according to a planned measuring path;

the software system includes: the system comprises a measuring system calibration module, a robot measuring point layout module, a robot measuring path planning module, a point cloud data processing and visualization module, a three-dimensional model format conversion module, a measuring system virtual simulation module, a measuring process control module and a human-computer interaction module respectively connected with the measuring system virtual simulation module, wherein:

the measuring system calibration module is used for calibrating the servo rotary workbench, the structured light three-dimensional scanner and the robot, and a calibration result is applied to the robot measuring point layout module;

the robot measuring point layout module generates an optimal measuring point according to the aero-engine blade design model and outputs the generated measuring point data to the robot measuring path planning module;

the robot measurement path planning module searches out a path with the shortest motion of the robot according to the generated optimal measurement point and outputs path data to the measurement process control module;

the measurement process control module is used for controlling the robot to move according to a planned path and controlling the servo rotary worktable and the robot to move to a target posture;

the point cloud data processing and visualization module is used for denoising, simplifying, dividing, splicing and reconstructing point cloud data acquired by the structured light three-dimensional scanner and analyzing and visualizing blade profile errors;

the three-dimensional model format conversion module is used for importing various aero-engine blade design models and exporting various three-dimensional models and can perform three-dimensional model format conversion;

the measurement system virtual simulation module is used for simulating the effect achieved by measurement in a real environment and the motion conditions of the servo rotary worktable and the robot, and improving the reliability of the measurement system and the control accuracy and response speed of the servo rotary worktable and the robot;

the man-machine interaction module monitors input, output and intermediate data of each module, provides a graphical user interface for operators to use, and manages the measurement process.

The invention has the following beneficial effects: the invention independently develops an autonomous measuring method and system of an aircraft engine blade robot, automatically generates measuring points and plans a robot measuring path through an aircraft engine blade design model, and realizes the automation of measurement; the structured light three-dimensional scanner realizes large-scale efficient measurement in a mode that the structured light three-dimensional scanner is fixed on a flange at the tail end of the robot through a connecting piece; the cooperative cooperation of the servo rotary table and the robot also greatly reduces the measurement time. Meanwhile, the system is strong in openness and expansibility, and can integrate functional modules such as network communication, motion control, point cloud data processing and visualization, profile size error analysis, ASC/PLY/IGES output and the like so as to meet the requirements of commercial application and academic research.

Drawings

FIG. 1 is a general block diagram of an autonomous measurement method of an aircraft engine blade robot;

FIG. 2 is a schematic diagram of three precision standard balls fixed on a servo rotary worktable, wherein 1 is the servo rotary worktable, 2 is a blade measuring fixture, 8 is the precision standard ball, 4 is a structured light three-dimensional scanner, and 5 is a robot;

FIG. 3 is a flow chart of calculating an average cluster centroid;

FIG. 4 is a schematic diagram of the basic principle of Hough voting for detecting a cloud of spherical points, wherein (a) is the case when the points are co-spherical and (b) (c) is the case when the points are unlikely to be co-spherical;

FIG. 5 is a schematic diagram of the basic idea of calibration of a robot measurement system, in which (a) is a distribution of observation points corresponding to a calibration result, (b) is a schematic diagram of minimum rotational motion between two coordinate systems, and (c) is a scattered distribution of observation points;

FIG. 6 is a schematic diagram of coordinate system configurations and transformation relationships between coordinate systems in a robotic measurement system, where (a) is the coordinate system configuration in the measurement system and (b) is the determination of whether an edge passes through or approaches a measurement object;

FIG. 7 is a flow chart of placement of measurement points according to an aircraft engine blade design model;

FIG. 8 is a feature of a projected contour of a Gaussian ellipsoid, where (a) is the feature where the projected contour contains 6 vertically-directed polygon edges and (b) is the feature where the projected contour contains 4 vertically-directed polygon edges;

FIG. 9 is a flow chart for calculating the gain of information observed from each candidate measurement point;

FIG. 10 is a flow chart of selecting an optimal set of measurement points from a set of candidate measurement points;

FIG. 11 is a flow chart for determining an optimal sequence of measurement points;

FIG. 12 is a flow chart for planning a measurement path for a robot;

FIG. 13 is a schematic diagram of a hardware system of an autonomous measuring system of an aircraft engine blade robot, wherein 1 is a servo rotary table, 2 is a blade measuring fixture, 3 is an aircraft engine blade, 4 is a structured light three-dimensional scanner, 5 is a robot, 6 is a robot control cabinet, and 7 is an industrial computer;

FIG. 14 is a software system architecture diagram of an aircraft engine blade robot autonomous measurement system.

Detailed Description

In order to make the technical solutions of the present invention more clear and definite, the present invention is further described in detail below with reference to the embodiments and the drawings, it should be noted that the embodiments and features of the embodiments of the present application can be combined with each other without conflict.

Referring to fig. 1, fig. 1 is a method for autonomous measurement of an aircraft engine blade robot provided in this embodiment, and includes the following steps:

s1: calibrating a servo rotary workbench 1, a structured light three-dimensional scanner 4 and a robot 5;

s3: planning the path measured by the robot 5;

s4: measuring the three-dimensional appearance of the blade of the aero-engine;

s5: measurement data of the aircraft engine blade is processed.

Further, referring to fig. 2, the calibrating servo rotating table 1, the structured light three-dimensional scanner 4 and the robot 5 include the following steps:

s11: referring to fig. 2, three precision standard balls 8 are fixed on the servo rotary table 1, and the three precision standard balls 8 are ensured to be within the working distance and the visual field range of the structured light three-dimensional scanner 4;

s12: referring to fig. 3, the robot 5 is kept still, the servo rotary table 1 rotates at a stable rotation angle α, the structured light three-dimensional scanner 4 scans once to obtain measurement data, and an average clustering sphere center P is calculated; ,

calculating an average clustering sphere center P, namely detecting and segmenting point clouds distributed on a standard sphere through Hough voting, clustering the segmented point clouds into 3 classes by using a K-means + + algorithm, wherein each class corresponds to a sphere area of one standard sphere, fitting the sphere by a least square method to obtain the sphere center positions of three precise standard spheres 3, and calculating the average clustering sphere center P to represent an observation result, wherein the specific steps are as follows:

s1201: removing noise points in the measured data by a point cloud bilateral filtering algorithm to obtain a point cloud S ═ last page^cameraP_observed,j|j＝1,...,N_sCalculating the normal direction of each point as N ═ retaining curl^cameran_observed,j|j＝1,...,N_sIn which N is_sDefining an empty set C for the number of points in the point cloud S, and initializing i, r as 1, wherein i, r belongs to {1_s}；

S1202: extraction of point p from S, N_rAnd normal n_rDefining the Hough voting accumulator Acc_r；

S1203: extraction of point p from S, N_iAnd normal n_iAnd calculating point pair characteristics:

wherein d ═ p_i-p_r；

S1204: if C is present₁＞ηr_s ²Or C₂> 0 or C₃< 0 or C₄＞cosε₁Then i ← i +is updated1, return to step S1203, where ε₁Is a very small positive value, eta ∈ (0, 4)]，r_sAs the radius of the standard sphere, judgment condition C₁＞ηr_s ²Is explained as p_iAt p_rRadius of

Out of the neighborhood of (1), condition C₂> 0 or C₃< 0 to be interpreted as p_iAnd p_rSpherical surface sharing is impossible, as shown in FIG. 4(b), condition C₄＞cosε₁Is explained as p_iAnd p_rThe directions are parallel and no common sphere is possible, as shown in fig. 4 (c);

s1205: computing

If | | | r-r_s||＜ε₂，

S₁S₂-C₂C₃＞C₁ cosε₃And C₂C₃C₄+S₁S₂C₄+S₁C₃S₃-C₂S₂S₃＞C₁ cosε₄Then the accumulator Acc_rPlus 1, wherein r_sIs the radius of a standard sphere,. epsilon₂、ε₃、ε₄A very small positive value;

s1206: if p is_iAll points in S are traversed and Acc is finished_r＞thresh_voteA 1 is to p_rAdded to the point set C, where thresh_voteIf the minimum number of voting support is not, updating r ← r +1, i ← 1, and returning to step S1202;

wherein, the Hough voting detects the point cloud of the spherical surface, referring to FIG. 4(a), for the point p belonging to the spherical surface_rA point p in the vicinity thereof_iAnd p_rConstructed point pair characteristics

Wherein

By point pair characteristic F ═ (| | d |, (n)_r,d),∠(n_i,d),∠(n_r,n_i))，

It can be known that the constraint condition F is satisfied₂+F₃N and F₂-F₃＝F₄The relaxation constraint is | F₂+F₃-π|＜ε₃And | F₂-F₃-F₄|＜ε₄In which epsilon₃、ε₄A very small positive value;

use of C₁、C₂、C₃、C₄Describing relaxation constraints as S₁S₂-C₂C₃＞C₁ cosε₃And C₂C₃C₄+S₁S₂C₄+S₁C₃S₃-C₂S₂S₃＞C₁cosε₄In which C is₁＝F₁ ²，C₂＝F₁cosF₂，C₃＝F₁cosF₃，C₄＝cosF₄，

If p is_rNear (radius of

In the neighborhood of) has enough p_iIf the voting is supported, then consider p_rBelongs to a spherical area and passes through | | | r-r_s||＜ε₂Judging whether the sphere belongs to the sphere of the standard sphere, wherein eta can be 2 to ensure enough voting number and reduce the search area, and thresh_voteCan be set to a small value and then adjusted by actual conditions, because a strict condition | | | r-r is set_s||＜ε₂，thresh_voteThe value of (c) can be relaxed;

s1207: randomly selecting a point from the point set C as the center C of the initial cluster₁；

S1208: calculate each point x_iDistance D (x) from the center point of the existing cluster_i) Then, the probability of each point being selected as the next cluster center point is calculated:

then selecting the next clustering center point by a wheel disc method;

the wheel disc method specifically operates as follows: calculating a probability distribution function

P₀Randomly generating a random number between 0 and 1, and determining the interval [ P ] to which the random number belongs_k-1(x),P_k(x)]Then x corresponds to_kIs the next cluster center selected;

repeating the step S1208 until K cluster center points are selected, where K is 3, and when the distance between each point and the plurality of cluster centers is calculated, d (x) takes the minimum distance value;

s1209: calculating the distance between each point and each cluster center, returning each point to the cluster center closest to the point, calculating the mean value of each class midpoint as a new cluster center, and if the class center is not changed any more or reaches the maximum iteration number N_cExecuting step S1210, otherwise, continuing to execute step S1209;

s1210: each class { C_kFitting points in 1,2 and 3 to a spherical surface by using a least square method, wherein the centers of the fitted spheres are P₁，P₂，P₃Radius is respectively r₁、r₂、r₃The fitting error is defined as

Wherein least squares are used to fit C_kThe specific operations of midpoint to sphere (circle in space) are:

computing matrices

Using Ax as b and least square solution x as (A)^TA)^-1b, the center (circle center) of the fitting spherical surface is (a b c)^TRadius of

Wherein n is C_kThe number of midpoints;

s13: continuously keeping the robot 5 still, rotating the servo rotary worktable 1 for N times at a stable rotation angle alpha to ensure that the angle alpha is more than 360 degrees, and recording the average clustering center P_iWhere i 1.. times.n, and fitting the set of points { P using a least squares method_iGet the average cluster center P_iRadius R of circular motion around the axis of rotation of the servo-rotating table₀Increasing R in order to generate enough observation points₀The estimation accuracy of (a) may be 10 °, N may be 36;

s14: the robot 5 operates to set the terminal attitude xi of the robot 5_jAt this time, the servo rotary table 1 rotates by a smooth rotation angle θ_iRotating, the structured light three-dimensional scanner 4 scans once to calculate the average clustering sphere center P of observation_ijWherein i 1, 1.. and N, j 1.. and M, and ensure

And when the tail end of the robot 5 is in xi posture_jWhen the three precision standard balls 3 are in the working distance and the visual field range of the structured light three-dimensional scanner 4;

s15: observation of result R according to step S13₀And step S14 observing result { P }_ijI ═ 1,. N, j ═ 1,. M }, the hand-eye calibration of the robot 5 and the structured light three-dimensional scanner 4 is performed^handT_cameraAnd determining the attitude transformation relationship between the axis of rotation of the servo rotary table 1 and the base coordinate system of the robot 5^baseT_axisWherein { hand } is the terminal coordinate system of robot 5, { base } is the base coordinate system of robot 5, { camera } is the coordinate system of structured light three-dimensional scanner 4, and { axis } is the coordinate system of the rotation axis of servo rotation table 1, and the relationship between the coordinate systems refers to fig. 6;

referring to FIG. 5(a), the implementation^handT_cameraAnd^baseT_axissimultaneous calibration, first of all, one of them solving

Make observed { P_ijI 1, N, j 1, M, with radius R around the axis of rotation of the servo rotary table 1₀According to { P_ijThe circular motion of the servo rotary table can recover the direction n of the rotary shaft of the servo rotary table, and the origin of a coordinate system { axis } can be set as the circle center P of the circular motion₀With the z-axis set to n, the x-axis and y-axis can be determined by the minimum rotational movement of the z-axis to the direction n of the { base } coordinate system, as shown in FIG. 5(b), to obtain another solution

As shown in figure 5(c) of the drawings,

nearby^handT_cameraWill result in { P_ijScattered around the circular motion, if errors are accumulated

The larger the distribution, the more scattered; searching for a numerical optimal solution of a calibration problem through a genetic algorithm and particle swarm algorithm mixed algorithm, and specifically comprising the following steps:

S1501：^baseT_handby reading the terminal attitude xi of the robot 5 in the robot controller_jObtained as (x, y, z, α, β, γ), i.e.:

t_xyz＝[x y z]^T，

s1502: is provided with^handT_camera＝T(t_x,t_y,t_z,r_x,r_y,r_z) From t_x∈[x_min,x_max]、t_y∈[y_min,y_max]、t_z∈[z_min,z_max]、r_x∈[α_min,α_max]、r_y∈[β_min,β_max]、r_z∈[γ_min,γ_max]Initializing a particle swarm with the scale of m in a six-dimensional search space, and setting an initial position x of a kth particle_k＝(x_k1,x_k2,x_k3,x_k4,x_k5,x_k6) And velocity v_k＝(v_k1,v_k2,v_k3,v_k4,v_k5,v_k6) K 1.. m, where x_min＝x-Δx₁，x_max＝x+Δx₂，y_min＝y-Δy₁，y_max＝y+Δy₂，z_min＝z-Δz₁，z_max＝z+Δz₂，α_min＝α-Δα₁，α_max＝α+Δα₂，β_min＝β-Δβ₁，β_max＝β+Δβ₂，γ_min＝γ-Δγ₁，γ_max＝γ+Δγ₂Selecting proper delta x according to actual conditions₁、Δx₂、Δy₁、Δy₂、Δz₁、Δz₂、Δα₁、Δα₂、Δβ₁、Δβ₂、Δγ₁、Δγ₂Limiting the scope of the search space;

s1503: calculating the adaptive value of each particle by first calculating the translation t and rotation matrix R_xyz：t＝[x_k1 x_k2 x_k3]^T，

Then

Computing

Fitting the set of points { P } by least squares_ij ^*|i＝1,...N,jM to a sphere with a sphere center P₀Radius of r₀Then from { P_ij ^*Great drawing point set^sP_l ^*Cir, satisfy the condition | calculation^sP_t ^*-P₀||∈(r₀-σ,r₀+ σ), l ═ 1.. S, where S is ∑ tone^sP_l ^*The number of midpoints, σ, is a very small positive value; then, from^sP_l ^*Extraction point in^sP₁ ^*、^sP₂ ^*Calculating n ═^sP₁ ^*×^sP₂ ^*Calculating an adaptive value for each particle

Finally update axis_k＝(P₀,n,r₀)；

S1504: for each particle, its adapted value and the best position X it has experienced_k*＝(X_k1,X_k2,X_k3,X_k4,X_k5,X_k6) Is compared if e_k＜e_k*Then x is_kAs the best position the current particle has experienced;

s1505: each particle will adapt its value to the best position X that the global has experienced_g＝(X_g1,X_g2,X_g3,X_g4,X_g5,X_g6) Is compared if e_k＜e_gThen x is_kAs the current global best position, if r₀∈(R₀-ξ,R₀+ xi), update axis ═ axis_kWhere ξ is a very small positive value;

s1506: selecting, crossing and mutating the particle swarm, calculating the adaptive value of each particle, and updating X_k*And X_g；

S1507: the velocity and position of each particle is updated by first calculating:

wherein k is 1, m, s is 1, 6, v_ks∈[-v_smax,v_smax]，v_smaxIs the maximum search speed in the s-th dimension,

effectively controls and restrains the flight speed of the particles for the contraction factor, simultaneously enhances the local searching capability of the algorithm,

C＝c₁+c₂and C is more than 4, t is the current iteration number, t_maxFor maximum number of iterations, learning factor c₁、c₂Is a non-negative constant, r₁、r₂Obey [0,1 ] for mutually independent pseudo-random numbers]Then, x is calculated_ks(t+1)＝x_ks(t)+v_ks(t+1)；

S1508: if the optimal solution has stagnated and no longer changes or reaches the maximum number of iterations or has obtained a good enough adaptation value, the iteration is stopped, the optimal solution is output,

otherwise, returning to the step S1503;

s1509: calculating a translation t^*And a rotation matrix

Then

S1510: by axis ═ P₀,n,r₀) Calculating the translation t_axis＝P₀Calculating the rotation w ═ e_z×n，θ＝arccos(e_z ^T·n)，n_wR is calculated from the formula rhoderligs |/| w | | |_axis＝cosθI-(1-cosθ)n_wn_w ^T+sinθn_wWherein e_z＝[0 0 1]^T，n_wIs n_w＝[n_x n_y n_z]^TThe anti-symmetric matrix of (a) is,

s1511: computing

Further, referring to fig. 7, the arranging the measurement points according to the aircraft engine blade design model includes the following steps:

s21: importing a CAD (computer-aided design) design model of an aircraft engine blade, and carrying out Poisson disc sampling to generate a point cloud M ═ last opening^bladeP_model,i|i＝1,...,N_mIn which N is_mThe number of points in the point cloud M;

s22: clustering point cloud M to generate Gaussian mixture model

A mean and covariance matrix of;

the method comprises the following steps that a candidate measuring point set is constructed according to Gaussian mixture model parameters of a point cloud M and the working distance D of a structured light three-dimensional scanner 4, the mixing number K is too small to generate enough candidate measuring points, a cavity area is generated in a measuring result, the mixing number K is too large, the calculating time is long, the storage resource consumption is high, it is very important to select a proper K value according to the size and the surface condition of an actual blade, and a larger K value is needed for measurement of the whole blade and the like; the method comprises the following specific steps:

first, a mixture is definedNumber of components K, initializing parameter of each component pi_j,μ_j,∑_jCalculating a log-likelihood function:

wherein j is 1.. K;

then, E-step (expected calculation procedure) calculates the posterior probability:

wherein the hidden variable z_ikDenotes x_iBelongs to the kth gaussian mixture component, j 1.., K;

next, M-step (maximization procedure) updates:

wherein

j＝1,...,K；

Finally, a log-likelihood function is calculated

j 1, K, checking whether the parameter or lnP converges, and if not, returning to continue executing E-step and M-step;

s23: according to μ_i、∑_iAnd the working distance D of the structured light three-dimensional scanner 4 generates a set of candidate measurement points { V }_i1., K }, where V ═ i ═ 1_iContaining positions v of candidate measurement points_iAnd direction n_i；

First, the normal direction { n ] of each point in the point cloud M is calculated_i|i＝1,...,N_m}；

Then, the distance mu is searched from the point cloud M_iK of_μOne neighboring point { p_ij|||p_ij-μ_j||＜ρ,i＝1,...K_μJ ═ 1.. K }, where ρ isA set neighborhood radius;

recalculation

k is 1,2,3, and S is selected to be_kMaximum k, defining a reference direction

Calculating the direction n of the candidate measuring point_sj＝-u_skWherein sgn () is a sign function, u_kIs sigma_jUnit vector of the feature vector of (a), σ_kIs sigma_jIs a characteristic value of_jIs 0, then u_sObtained by cross multiplication of other two feature vectors;

finally, the position v of the candidate measuring point is calculated_sj＝μ_j-(D-ησ_k)n_sjWhere D is the working distance of the structured light three-dimensional scanner 4, and D is the element (D)_min,D_max)，ησ_k＜D-D_min，σ_kIs in the direction u of reference_skParallel eigenvectors u_kCorresponding characteristic value, eta is regulating factor, according to mu_iThe curvature of the nearby point cloud is used to adjust the position of the measuring point, the sampling rate is increased at the place with large curvature, and D can be selected

Eta is selected such that D_min＜D-ησ_k＜D_max；

S24: from a set of candidate measurement points { V_iChoose the best set of candidate measurement points { D } from | i ═ 1_iI | (1.), n }, wherein n is less than or equal to K, and the integrity of the model obtained by the measurement of the robot 5 is ensured, wherein n is the number of the finally arranged measurement points; the method comprises the following specific steps:

s241: referring to fig. 9, the information gain observed from each candidate measurement point is calculated:

first, at a candidate measurement point { v }_sj,n_sjThe possible observed point cloud distribution is

Satisfies n_sju_sk< 0, wherein u_skIs G_kA corresponding reference direction;

then, P is added_s(x) Medium gaussian mixture component G_kPress mu_kAnd candidate measurement point { v_sj,n_sjDistance d between_k＝(μ_k-ν_sj)n_sjSorting from small to large, k being 1_kMinimum gaussian mixture component G_kAdd to set Ω { (k, p)_k1,p_k2,p_k3,p_k4,p_k5,p_k6) In }, p_k1＝μ_k-2σ_k1u_k1，p_k2＝μ_k+2σ_k1u_k1，p_k3＝μ_k-2σ_k2u_k2，p_k4＝μ_k+2σ_k2u_k2，p_k5＝μ_k-2σ_k3u_k3，p_k6＝μ_k+2σ_k3u_k3Wherein p is_k1,p_k2,p_k3,p_k4,p_k5,p_k6Determine G_kCorresponding to a Gaussian ellipsoid, σ_k1、σ_k2、σ_k3Are respectively sigma_kCharacteristic value of (u)_k1、u_k2、u_k3Is sigma_kUnit vector of the feature vector of (1), if ∑_kIs 0, then u_sObtained by cross multiplication of other two feature vectors;

subsequently, step by step from P_s(x) In which d is taken out_kGreater gaussian mixture component G_kIf the parameter μ is for all gaussian components in the set Ω_ΩAnd d_ΩSatisfies the conditions

d_k＝(μ_k-v_sj)n_sj，d_Ω＝(μ_Ω-v_sj)n_sjThen G will be_kAdding to the set omega, otherwise adding to the set omega^*＝{(k,p_k1,p_k2,p_k3,p_k4,p_k5,p_k6) In (1) }; repeating the steps until P_s(x) The medium Gaussian mixture components are each assigned to the set Ω or the set Ω^*Performing the following steps;

next, the sets Ω and Ω are calculated^*Medium gaussian mixture component G_kThe corresponding projection point of the Gaussian ellipsoid in the field of view of the structured light three-dimensional scanner 4, i.e. for a given one of the elements (k, p)_k1,p_k2,p_k3,p_k4,p_k5,p_k6) Calculating

Then, the sets Ω and Ω are calculated^*Medium gaussian mixture component G_kCorresponding projection profile characteristics of Gaussian ellipsoid, and defining the projection profile characteristic form as

Representing measured points from candidate { v }_sj,n_sjObserved Gaussian mixture component G_kThe projection profile of the corresponding Gaussian ellipsoid in the visual field range of the structured light three-dimensional scanner; wherein the projection profile is a convex polygon generated from projection points, and the defined projection profile is characterized by a projection center

And a vector s pointing perpendicularly from the center of projection to the edge of the convex polygon_vComposition of vector s_vThe modulo length of (a) represents the distance of the projection center to the polygon edge, see fig. 8 (a); in addition, the distribution of the projection points may refer to fig. 8(b), and the constructed convex polygon will surround part of the points, and the specific steps are as follows:

computing

Select | | | s_iS with the largest | |_iAs an initial value, i is added to the list Circle { };

selecting one j from List {1,2,3,4,5,6}, i ≠ j, and calculating

From { s } - }₁,s₂,s₃,s₄,s₅,s₆Select s from_kIf all of s_kSatisfies s_ks_v≤||s_v||²K ≠ i ≠ j, j is removed from List ═ {1,2,3,4,5,6}, i is added to Circle, s is added to Circle_vAdded to the Gaussian mixture component G_kSet of projection profile features S_jkAnd mixing s_jAs an initial value s_i；

If the values of the first element and the last element in the list Circle are the same, stopping iteration, otherwise, returning to continue calculating the s_v；

Calculating a Gaussian mixture component G_kAnd (3) calculating the projection outline area of the corresponding Gaussian ellipsoid by sequentially taking { a, b, c }, { a, b, d }, { a, b, e }, and { a, b, f } from Circle

If only 5 elements are in Circle, only S needs to be calculated₁And S₂，

For the set omega^*Medium gaussian mixture component G_kCalculate G_kAnd (3) setting the obtained Circle as { a, b, c, d, e, f, a } of the corresponding projection grid of the Gaussian ellipsoid, sequentially taking two adjacent elements from the Circle, and calculating

Wherein

Which means that the rounding is made up,

i＝1,..,n_v-1，j＝1,...,n_l-1, from

The set of constructs is denoted as

After the above steps are completed, the set Ω is processed next^*Medium gaussian mixture component G_kBy G_kEvaluating whether the Gaussian ellipsoid is shielded or not and the covered degree of the Gaussian ellipsoid during shielding by using the corresponding points in the projection grid of the Gaussian ellipsoid; when G is_kAll the Gaussian ellipsoid projection points are in G_aOutside the projected outline area of Gaussian ellipsoid and G_aAll the Gaussian ellipsoid projection points are in G_kOutside the projected outline area of Gaussian ellipsoid of G_kGaussian ellipsoid and G_aThe Gaussian ellipsoids do not have shielding problem, and the shielding problem is solved through calculation

In that

Projection of direction, i.e. if present

So that

1.. 6, then the projected points are

At G_kOutside the projected outline area of the Gaussian ellipsoid, otherwise, the projected point

At G_kWithin the region of the Gaussian ellipsoid projection profile of G_bThe projection point of the Gaussian ellipsoid is

When G is_kWhen the Gaussian ellipsoid has shielding problem, G should be dealt with first_kGaussian ellipsoid projection contour region calculation interpolation point

Then make statistics of

The covered degree in the shielding process is judged according to the ratio of the number in a certain Gaussian ellipsoid projection outline region to the total number of the interpolation points, and the specific steps are as follows:

let P_s(x) Neutralization candidate measuring point { v_sj,n_sjD < d_kIs made up of a set F of gaussian mixture components_k＝{G_kf|f＝1,...,N_F}，

Number of points in

The coverage statistic variable cover is 0;

if it is not

Or

Or

Or

Or

Or

Then the cover is not changed, otherwise the cover is added with 1, wherein

f＝1,...,N_F；

If the statistical coverage rate

Wherein eta^*For a set threshold, consider the set Ω^*Medium gaussian mixture component G_kIs completely shielded, then P_s(x) Not counting G_kOtherwise P_s(x) Needs to count G_kThe information gain of (1), the existence of partial occlusion also needs to be counted_kThe information gain of (1);

at the candidate measurement point { v_sj,n_sjThe possible observed point cloud distributions are:

finally, the candidate measurement points { v are calculated_sj,n_sjObserving the obtained information gain, namely calculating:

wherein sigma_kIs G_kAnd a reference direction u_skThe characteristic values corresponding to the parallel characteristic vectors;

s242: after information gain observed by each candidate measuring point is obtained through calculation, an optimal group of measuring points { D ] is selected_iReferring to fig. 10, the specific steps are:

from the set of candidate measurement points { V_jSearch Information gain Information_sjLargest candidate measurement point { v_sj,n_sjIs the initial measurement pointWill { v }_sj,n_sjAdd to the set of measurement points D, let P_s(x)＝P_sj ^*(x)；

From the set of candidate measurement points { V_jSelect the next candidate measurement point { v }_si,n_siEnsure two observations P_s(x) And P_si ^*(x) Having the same Gaussian mixture component, and calculating all the same Gaussian components G_sharedIs the contour area S_sharedAnd

wherein P is_sharedFor a set of all identical Gaussian components, S_sharedTaking the minimum projected contour area in the two observations, if G_sharedIf there is no occlusion problem, η is equal to 0, otherwise η is equal to η_k；

Estimate P_s(x) And P_si ^*(x) Calculating JS divergence between:

the KL divergence between two Gaussian mixture models can be estimated by the following steps:

wherein

d is the dimension of the multivariate gaussian distribution, where d is 3;

selection of S_shared,sum＞S_shared,minAnd D_JS(P_si ^*(x)||P_s(x) Maximum { v) }_si,n_siAs next measurement point, { V } from the set of candidate measurement points_jGet rid of { v }_si,n_siAnd will { v }_si,n_siAdd it to the set of measurement points D, update P_s(x)＝P_s(x)+P_si ^*(x)-P_shared(x) In which S is_shared,minFor a set minimum amount of information, P, shared by two adjacent observations_shared(x) Is P_sharedA mixture distribution of all gaussian components;

if D is_JS(P_s(x) If | P (x) < gamma, outputting the measurement point set, otherwise returning to continue from the candidate measurement point set { V |)_jSelect the next candidate measurement point { v }_si,n_siRepeatedly executing subsequent steps until D_JS(P_s(x) | p (x)) < γ, where γ is a small positive value; wherein the condition D is judged_JS(P_s(x) | p (x) < γ describes the difference between the point cloud distribution observed at all measurement points and the overall point cloud distribution, and if the difference is smaller, the point cloud distribution observed at all measurement points can reflect the overall point cloud distribution;

Wherein s is_i＝1,...,n；

Firstly, generating a graph with a constraint relation by discrete points, then searching the graph by a simulated annealing method to traverse the shortest paths of all measurement points, referring to fig. 11, the specific steps are as follows:

s251: connecting three-dimensional discrete points in the set D into a triangular mesh through Delaunay triangulation, and constructing a map Graph { D ═ D_i,d_ijThat is, vertices (measurement points) of the triangular meshes are vertices of Graph, edges (distances between adjacent measurement points) of the triangular meshes are edges of Graph, and a center point of each triangular mesh is calculated

Wherein D_k1,D_k2,D_k3The vertex of the kth triangular patch;

if the Graph edge crosses or is close to the measurement object, then such edge should be removed to generate a reasonable Graph and the triangle patch to which the edge connects is deleted;

the cylindrical surface is used for surrounding a measuring object, and the spatial area of the measuring object is set as

The center of the bottom surface of the cylindrical surface is

The center of the top surface is

Referring to FIG. 6(b), calculate

d_v1＝min{||v_i-p_c||,||v_j-p_c||}，d_v2＝min{||v_i ^*-p_c ^*||,||v_j ^*-p_c ^*| l }, keeping the condition 'lambda' satisfied_v1＞0，λ_v2> 0 and | | s_v1||＞d_min,||s_v2||＞d_min'or' lambda_v1＞0，λ_v2Less than or equal to 0 and s_v1||＞d_min,d_v2＞d_min'or' lambda_v1≤0，λ_v2> 0 and d_v1＞d_min,||s_v2||＞d_min'or' lambda_v1≤0，λ_v2Not more than 0 and d_v1＞d_min,d_v2＞d_min"wherein v is_i ^*、v_j ^*、p_c ^*Are each v_i、v_j、p_cProjection of points on XY plane, d_minSetting the minimum distance between the edge in Graph and a cylindrical surface surrounding a measuring object;

s252: initial annealing temperature T_kK is 0 and a cooling factor α;

s253: according to Graph ═ D_i,d_ij} generating a path

Let global optimal solution Path_best＝Path_i；

S254: generating a random number j, j 1

And (3) calculating:

wherein

Exchanging the neighborhood structure according to the relationship of the edges in the graph to generate a new path, namely, points on two triangular patches with a common edge can generate a new path by exchanging the sequence of passing through two common vertexes;

s255: if P is present_kSatisfy | | P_k-P₁| P < ε_k-P₂If | < epsilon, where epsilon is a very small positive value, the exchange neighborhood structure generates a new path

Calculating path length variation

Otherwise, returning to the step S254;

s256: if according to the probability

Then Path_best＝Path_jWherein random (0,1) is [0,1 ]]Random numbers within the interval;

s257: annealing operation, T_k+1＝αT_k，k←k+1，Path_i＝Path_jIf the convergence criterion is satisfied (from several adjacent Path)_bestObservation of Path_bestWhether the stagnation is not changed any more), the annealing process is ended, and the optimal solution Path is output_bestOtherwise, returning to the step S254;

s258: search Path_bestThe longest side of the middle

The optimal sequence of measurement points is

Further, referring to fig. 12, the planning of the path measured by the robot 5 includes the following steps:

s31: clamping and fixing an aircraft engine blade 3 on a blade measuring clamp 2, and ensuring that the blade is within the working distance and the visual field range of a structured light three-dimensional scanner 4;

s32: the structured light three-dimensional scanner 4 scans once to obtain the measurement data, removes the noise in the measurement data, and segments the point cloud of the blade area into a final arch^cameraP_observed,j|j＝1,...,N_sIn which N is_sThe number of points in the point cloud S;

s33: aligning the point cloud S of the blade area of the aircraft engine with the point cloud M of the CAD design model of the blade of the aircraft engine through a point cloud registration algorithm, and solving a transformation matrix into^modelT_observedThen corresponding to the point pair^cameraP_observed,jAnd^bladeP_model,isatisfy the following requirements

Wherein { model } is a CAD design model coordinate system of the aero-engine blade, { observed } is a measurement data coordinate system of the aero-engine blade, { blank } is a coordinate system of the aero-engine blade to be measured,^cameraP_observed,jto measure points in the data for a structured light three-dimensional scanner,^bladeP_model,idesigning points in the model for the aircraft engine blade;

s34: from the matrix equation:

s35: according to the optimal sequence of measurement points D_sequence＝{D₁,...,D_k,...D_n}, target measurement points

Computing robot 5 moves to D_kThe shortest straight path L comprises the following specific steps:

s351: calculating the current position of the structured light three-dimensional scanner 4 in a rotating shaft coordinate system of the servo rotating worktable 1:

s352: calculating the optimal measurement point

Description in { axis } coordinate System

Namely:

s353: calculating the vertical distance of the structured light three-dimensional scanner 4 moving from the current position to the target measurement point:

wherein e is_z＝[0 0 1]^T，

S355: calculating the shortest straight path of the robot 5 moving to the target measurement point:

the shortest time T of the linear motion of the robot 5_rmin＝L/v_maxWherein v is_maxThe set maximum linear motion speed of the robot 5;

s36: calculating the position of a target measurement point in a robot 5 base coordinate system

And direction

Wherein the content of the first and second substances,

respectively the position and the direction of a measuring point in a rotating shaft coordinate system of the servo rotating workbench,^axisP_camerato servo the current position of the structured light three-dimensional scanner 4 in the rotary axis coordinate system of the rotary table 1,^axisT_bladeis a transformation matrix between coordinate systems { blade } and { axis }, r₁＝^axisP_camera-(^axisP_camera ^T·e_z)e_z，

e_z＝[0 0 1]^T，s_k＝1,...,n，R_z(theta) is a rotation matrix of the servo rotary table 1 rotated by an angle theta around the direction of the rotation axis z,

s37: calculating the translational motion of the structured light three-dimensional scanner 4 from the current attitude motion to the target measurement point:

rotation matrix of minimum rotational movement: r ═ cos θ I- (1-cos θ) nn^T+ sin θ n ^ wherein,

for the position of a target measuring point of a structured light three-dimensional scanner,^baseP_camera、^basen_zrespectively the current position and direction of the structured light three-dimensional scanner 4;

s38: calculating the target pose xi of the robot 5 motion^*：

Wherein the content of the first and second substances,x＝m₃₂-m₂₃，y＝m₁₃-m₃₁，z＝m₂₁-m₁₂according to the quaternion q ═ w + xi + yj + zk, the target pose of the robot 5 motion

Wherein the content of the first and second substances,

for the position of the end of the robot 5 at the target measuring point, R^*The attitude of the tail end of the robot 5 at a target measuring point, R is a rotation matrix,^baseP_handbeing the current position of the end of the robot 5,^baseR_handthe motion trail of the robot 5 is generated by linear interpolation of a robot controller in a robot control cabinet 6 for the current posture of the tail end of the robot 5;

s39: calculating the rotation angle of the movement of the servo rotary table 1

The target position of the rotation of the servo rotary table 1 is set to

And the shortest time is T_amin＝α/ω_maxWherein e is_zIs [ 001 ]]^T，θ_aFor the current position of the servo rotary table 1 sgn () is a sign function, ω_maxMaximum angular velocity of rotation, r, for steady movement of the servo rotary table 1₁For measuring the radial vector of a point relative to the axis of rotation, r₂The radial vector of the target measurement point relative to the rotation axis is obtained.

Further, the method for measuring the three-dimensional appearance of the blade 3 of the aircraft engine comprises the following steps:

s41: controlling the robot 5 and the servo rotary worktable 1 to move to a target pose according to the planned measuring path of the robot 5;

s42: the movement velocity v of the robot 5 and the angular velocity ω of the servo rotary table 1 are adjusted to L/T and α/T, respectivelyMotion synchronization, where T ═ max { T_amin,T_rminIn which T is_rminIs the shortest time, T, of the linear motion of the robot 5_aminThe shortest movement time of the servo rotary table 1 is defined as alpha, and the rotation angle of the servo rotary table 1 is defined as alpha;

s43: the structured light three-dimensional scanner 4 scans once, the measured data is transmitted to the industrial computer 7 for processing, and then the target measuring point D is updated_k；

S44: steps S41-S43 are repeated until the robot 5 traverses all target measurement points.

s51: denoising point cloud data;

s52: simplifying the point cloud of the blades of the aircraft engine;

s53: segmenting point clouds in the blade area of the aircraft engine;

s54: splicing point clouds of blades of the aircraft engine;

s55: reconstructing a curved surface model of the blade of the aircraft engine;

Referring to fig. 13 and 14, the invention further provides an autonomous measuring system for an aircraft engine blade robot, which comprises a hardware system and a software system, wherein:

the hardware system comprises a servo rotary worktable 1, a blade measuring clamp 2, a structured light three-dimensional scanner 4, a robot 5, a robot control cabinet 6 and an industrial computer 7; the servo rotary workbench 1 is connected and communicated with an industrial computer 7 through a field bus, and is used for being matched with the robot 5 to measure and executing a motion instruction issued by the industrial computer 7; the blade measuring clamp 2 is fixed on the servo rotary worktable 1 through a connecting piece and is used for clamping and fixing an aero-engine blade 3 to be measured; the structured light three-dimensional scanner 4 is fixed on a flange at the tail end of the robot 5 through a connecting piece and is used for collecting measurement data, and the measurement data are transmitted to an industrial computer 7 through an industrial Ethernet for processing; one end of the robot control cabinet 6 is communicated with the robot 5 through a field bus, and the other end of the robot control cabinet is communicated with the industrial computer 7 through a network, so that the robot 5 is controlled to move, and the structured light three-dimensional scanner 4 at the tail end of the robot 5 moves according to a planned measuring path;

further, the robot 5 communicates with the industrial computer 7 through a TCP/IP protocol or a field bus;

further, the structured light three-dimensional scanner 4 is connected with the industrial computer 7 through a GigE interface and a network cable;

further, a servo driver of the servo rotary worktable 1 is communicated with the industrial computer 7 through an EtherCAT field bus or an RS232 serial port or a CANOPEN bus;

the measuring system calibration module is used for calibrating the servo rotary workbench 1, the structured light three-dimensional scanner 4 and the robot 5, and the calibration result is applied to the robot measuring point layout module;

the measurement process control module is used for controlling the robot 5 to move according to a planned path and controlling the servo rotary worktable 1 and the robot 5 to move to a target posture;

the point cloud data processing and visualization module is used for denoising, simplifying, dividing, splicing and reconstructing point cloud data acquired by the structured light three-dimensional scanner 4 and analyzing and visualizing blade profile errors;

the man-machine interaction module monitors input, output and intermediate data of each module, provides a graphical user interface for operators to use, and manages the measurement process;

further, the software system is based on a Windows operating system; the ROS2 is applied, and a DDS publish/subscribe system architecture and a QoS (quality of service) strategy are utilized to ensure that data are distributed efficiently and flexibly in real time, meet the application requirements of distributed real-time communication and realize that a plurality of robots simultaneously perform measurement tasks.

In the description above, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore should not be construed as limiting the scope of the present invention.

In conclusion, although the present invention has been described with reference to the preferred embodiments, it should be noted that, although various changes and modifications may be made by those skilled in the art, they should be included in the scope of the present invention unless they depart from the scope of the present invention.

Claims

1. The autonomous measuring method of the aero-engine blade robot is characterized by comprising the following steps:

s1: the calibration servo rotary worktable, the structured light three-dimensional scanner and the robot specifically comprise the following steps:

S14: the robot acts to set the terminal attitude xi of the robot_jAt this time, the servo rotary table rotates by a smooth rotation angle θ_iRotating, once scanning by the structured light three-dimensional scanner, and calculating the average clustering sphere center P of observation_ijWherein i 1, 1.. and N, j 1.. and M, and ensure

s15: observation of result R according to step S13₀And step S14 observing result { P }_ijI 1.. N, j 1.. M }, and the machine is operated in the same manner as described aboveHand-eye calibration of human and structured light three-dimensional scanner^handT_cameraAnd determining the attitude transformation relationship between the axis of rotation of the servo rotary table and the robot base coordinate system^baseT_axisWherein { hand } is a robot end coordinate system, { base } is a robot base coordinate system, { camera } is a structured light three-dimensional scanner coordinate system, and { axis } is a coordinate system of a servo rotary table rotating shaft;

s3: planning a path measured by the robot;

s4: measuring the three-dimensional appearance of the blade of the aero-engine;

s5: measurement data of the aircraft engine blade is processed.

2. The aircraft engine blade robot autonomous measuring method according to claim 1, wherein the arranging of the measuring points according to the aircraft engine blade design model comprises the following steps:

s22: clustering point cloud M to generate Gaussian mixture model

A mean and covariance matrix of;

s23: according to μ_i、∑_iAnd generating a set of candidate measurement points { V } for the working distance D of the structured light three-dimensional scanner_i1., K }, where V ═ i ═ 1_iContaining positions v of candidate measurement points_iAnd direction n_i；

s25: setting D ═ D_i1., n }, according to the selected set of best candidate measurement points D ═ D ·_iI 1, n, determining the optimal sequence of measurement points

Wherein s is_i＝1,...,n。

3. The aircraft engine blade robot autonomous measuring method according to claim 2, wherein the planning of the robot-measured path comprises the following steps:

Wherein { model } is an aeroengine blade design model coordinate system, { observed } is an aeroengine blade measurement data coordinate system, { blank } is an aeroengine blade coordinate system to be measured,^cameraP_observed,jfor structured light three-dimensional scanner measurementsThe point in the volume data is measured by the point,^bladeP_model,idesigning points in the model for the aircraft engine blade;

s34: from matrix equations

wherein the content of the first and second substances,

R_z(theta) for rotary operationThe stage is rotated around the rotation axis z by a rotation matrix of theta degrees,

s38: calculating target pose xi of robot motion^*Calculating

Wherein the content of the first and second substances,

s39: calculating the rotation angle of the servo rotary table movement

The target position of the servo rotary table is

4. An aircraft engine blade robot autonomous measuring method according to claim 3, characterized in that the computing robot moves to D_kComprises the following steps:

s352: calculating the optimal measurement point

At { axis } coordinateDescription of the invention

Wherein e is_z＝[0 0 1]^T，

r₁＝^axisP_camera-(^axisP_camera ^T·e_z)e_z，

5. The aircraft engine blade robot autonomous measuring method according to claim 4, wherein the measuring of the three-dimensional topography of the aircraft engine blade comprises the following steps:

s42: adjusting the motion speed v of the robot to be L/T and the angular speed omega of the servo rotary table to be alpha/T so that the two move synchronously, wherein T is max { T {_amin,T_rminIn which T is_rminIs the shortest time of linear motion of the robot, T_aminThe shortest time of the movement of the servo rotary worktable is alpha, and the alpha is the rotation angle of the movement of the servo rotary worktable;

6. The aircraft engine blade robot autonomous measuring method according to claim 5, characterized in that the processing of the measurement data of the aircraft engine blade comprises the following steps:

s51: denoising point cloud data;

s52: simplifying the point cloud of the blades of the aircraft engine;

s53: segmenting point clouds in the blade area of the aircraft engine;

s54: splicing point clouds of blades of the aircraft engine;

s55: reconstructing a curved surface model of the blade of the aircraft engine;

7. An autonomous measuring system of an aircraft engine blade robot, characterized in that the system comprises a hardware system and a software system, wherein:

the system is based on the autonomous measuring method of the aeronautical transmitter blade robot in any one of claims 1-6.