CN106407502A - Optimum matching-based blade section line profile parameter evaluation method - Google Patents

Abstract

The invention discloses a method for evaluating blade profile profile parameters based on best matching, which uses rough registration of mid-arc lines, effectively eliminates bad points in measurement, and performs profile evaluation through the principle of minimum conditions, and based on tolerance constraints, Under the condition of ensuring that the position degree and torsion degree are not out of tolerance, the contour error can be minimized, and the number of out-of-tolerance points can be minimized to achieve the best match, which is conducive to reducing the false reject rate; through this method, the blade can also be detected quickly and accurately The defective parts of the blade guide the direction of further processing or process improvement of the blade.

Description

A Method for Evaluation of Blade Profile Profile Parameters Based on Best Matching

technical field

The invention belongs to the field of inspection, and in particular relates to a method for evaluating blade cross-section profile parameters based on best matching.

Background technique

Blades are the core components of turbomachinery such as steam turbines and aircraft engines. Evaluation to ensure the blade quality is up to standard.

The blade parameters to be detected can be divided into two categories: one is profile characteristic parameters, including front and rear edge radius, chord length, mid-arc, maximum thickness, etc.; the other is profile profile parameters, including profile profile, stacking point position, torsion, etc. The first type of parameters is only related to the measurement data itself, while the calculation of the second type of parameters must first match the measurement data with the theoretical data, which has a great dependence on the matching method and the evaluation method of the profile parameters.

Most of the existing matching evaluation methods use the three-point method (the center of the front and rear edges and the center of gravity) for rough matching, and then evaluate the contour parameters based on the principle of least squares. However, because the edge of the blade is very thin, the manufacturing accuracy is more difficult to guarantee than the back of the blade pot, and it is often not an ideal arc. Due to the limitations of the current measurement technology, the measurement data of the edge of the blade is often not ideal, and the error of fitting the front and rear edges Larger, the accuracy of the three-point method registration is not high enough.

Enterprises generally use random sampling method for testing. The pass rate and reject rate of the test samples affect the evaluation and judgment of the technicians on the whole batch of blades to a certain extent. Therefore, correct and reasonable parts testing methods and quality evaluation standards It is especially important for the production of enterprises. On the one hand, it can promote the adjustment of the process system to reduce the production scrap rate, and on the other hand, it can reduce the false scrap rate caused by unreasonable inspection and evaluation, so as to ensure the qualified rate of parts.

To sum up, it is particularly important to study a reasonable evaluation method of contour parameters that can be more integrated with engineering practice.

Contents of the invention

The purpose of the present invention is to overcome the above-mentioned deficiencies, to provide a method for evaluating blade profile parameters based on optimal matching, to evaluate blades more comprehensively and intuitively, and to effectively reduce the profile error under the premise of ensuring blade performance , Reduce the number of out-of-tolerance points, thereby reducing the rate of false rejects, and guide the improvement direction of blade processing technology.

In order to achieve the above object, the present invention comprises the following steps:

Step 1, read in the theoretical data and measured data of the profile line of the blade, and fit the profile line respectively;

Step 2, data preprocessing: According to the distribution characteristics of the profile data points, the data points are divided into four parts: leaf basin, leaf back, leading edge and trailing edge and stored separately;

Step 3, using the middle arc as a matching feature, initially match the measured data with the theoretical data, and eliminate the gross error data;

Step 4, using the best matching algorithm to accurately match the measured data with the theoretical data;

Step 5: According to the matching result, calculate the twist angle Ψ and position degree w of the blade through the rotation and translation matrix required by the rigid body transformation, and use the limit contour error values of the blade pot, blade back, leading edge, and trailing edge to obtain the profile line Outline degree;

Step 6, generate an error tolerance zone with the standard shape as the bone line according to the given tolerance parameters;

Step seven, perform parameter evaluation and analysis on the profile of the profile of the blade section according to the tolerance zone of the error.

In said step 3, the preliminary matching uses the middle arc as the matching feature to obtain a rough matching result with high accuracy, thereby effectively eliminating measurement bad points and ensuring the correctness of the evaluation of the contour degree of the profile by the minimum condition method, and Provide a better initial value for the ICP algorithm used in fine matching.

In the fourth step, the improved ICP algorithm is used for accurate matching, and the torsional error and bending error are not out of tolerance as constraints, and the contour of the blade basin, blade back, leading edge, and trailing edge is the smallest, and the total number of contour out-of-poor points is the least As the goal; through the rotation matrix R and the translation matrix T, the torsion angle Ψ and the position degree w of the blade section profile line are obtained; based on the best matching, the evaluation of the blade profile parameters in line with the engineering reality is carried out.

The improved ICP algorithm comprises the following steps:

The first step is to use the measured data that has undergone rigid body transformation after preliminary registration as the initial value of fine registration; obtain the corresponding closest point by finding the vertical foot of each measurement point to the theoretical model line;

In the second step, based on the regional tolerance constraints, the minimum condition method is used to evaluate the contour degree of the profile, that is, to minimize the contour error and the number of contour deviation points under the premise of ensuring that the position degree and torsion error do not exceed the tolerance; then the objective function Expressed as:

f(R,T)=min{max{distance(P _{i_p} ', L)+max{distance(P _{i_b} ', L)+max{distance(P _{i_q} ', L)+max{distance(P _{i_h} ', L)}

&&min{N _{oversize_points} };

In the formula, P _{i_p} ′, P _{i_b} ′, P _{i_q} ′P _{i_h} ′ are points obtained by rigid body transformation through the registration process of the measured points of the blade pot, blade back, leading edge, and trailing edge, respectively; L is the theoretical section line; N _{oversize_points} is the number of oversize points located outside the tolerance zone, that is, the number of points located within the inner boundary or outside the outer boundary of the tolerance zone;

In the third step, the rotation matrix R and translation matrix T determined by the quaternion method are as follows:

Then the torsion angle of the measured data of the profile line of the blade is:

Accumulation point position degree:

When both the position degree w and the torsion error Ψ do not exceed the given tolerance parameters, use the obtained rotation-translation matrix R _k , T _k to update the data point set P _K , so that P _k+1 = R _k *P _k + T _k , repeat the ICP iterative process of fine registration; when the error of position or torsion is out of tolerance, or the sum of the maximum values of contour errors in each area reaches the minimum, the iteration is stopped, and it is considered as the best match at this time.

In said step three, the specific method of preliminary matching is as follows. First, extract the middle arc of theoretical data and measurement data, and use the center of the front and rear edges as the starting point and end point of the middle arc; then, respectively fit the theoretical and measured middle arcs And take points at equal intervals, it is required to obtain the data points of the arc in theory equal to the data points of the arc in the actual measurement, and make the subscript of the starting point of the arc in theory consistent with the subscript of the starting point of the arc in the actual measurement; Set as the corresponding points, use the quaternion method to solve the rotation and translation matrix, and perform rigid body transformation on the measured data.

In the step 3, the method for removing gross error data is as follows:

Measure dead point judgment:

After initial matching, if the distance from the measuring point to the theoretical curve:

d _i <min{d _i-2 ,d _i-1 ,d _i+1 ,d _i+2 } or d _i >max{d _i-2 ,d _i-1 ,d _i+1 ,d _i+2 },

And when |d _{i average} -d _i |>δ ₁ , then p _i is a dead point;

Among them: d _{i average} =(d _i-2 +d _i-1 +d _i+1 +d _i+2 )/4; δ ₁ is a given limit value, take δ ₁ =1.5e, e=max{ Section line profile degree tolerance value}.

Compared with the prior art, the present invention uses the middle arc coarse registration, effectively eliminates bad points in measurement, evaluates the contour degree through the principle of minimum conditions, and based on tolerance constraints, ensures that the position degree and torsion degree are not out of tolerance. Under certain circumstances, the contour error can be minimized, the number of out-of-tolerance points can be minimized, and the best matching can be achieved, which is conducive to reducing the rate of false rejects; through this method, the defective parts of the blade can also be detected quickly and accurately, and the further processing or process of the blade can be guided. Improve direction.

Description of drawings

Fig. 1 is the evaluation flow chart of blade section contour parameter of the present invention;

Fig. 2 is the profile line data of theoretical blade section of the present invention;

Fig. 3 is the data of the profiled line of the measured blade section of the present invention;

Fig. 4 is the arc in theory that the present invention obtains;

Fig. 5 shows the situation that the measurement data falls within the tolerance zone after the best matching is performed in the present invention.

detailed description

The present invention will be further described below in conjunction with the accompanying drawings.

Referring to Fig. 1 to Fig. 5, the present invention comprises the following steps:

Step 1: Realize the process of data region division preprocessing

Since this method is based on regional tolerance constraints, it is very important to accurately preprocess the data by dividing the front and back edges of the leaf basin and the back of the leaf basin.

a. Theoretical data uses the method based on the distance error to divide the data area;

Referring to Figure 2, the characteristics of the theoretical data points are: the data points on the front and rear edges are dense, and the data points on the back of the leaf pot and leaf are sparse, so the distance between points within the range of the leaf pot and leaf back is much larger than that of adjacent points within the range of the front and rear edges. distance between points;

The processing steps are as follows:

First, the data points are roughly divided into two free curves, the leaf pot and the leaf back, and both ends of the curves respectively contain a part of the leading edge data points and the trailing edge data points. Next, determine a distance threshold according to the situation of the leaf measurement data points, and then start from the end points of the two free curves, calculate the distance between the two points, and judge whether the distance is within the range of the distance threshold. If the distance is within the threshold range, it is the data point set of the front and rear edges, otherwise the search is stopped, and the remaining points are the data point set of the leaf pot or the leaf back.

b. The measured data uses a combination of curvature error and distance error to accurately divide the data area;

Referring to Figure 3, the actual measurement point set is measured at a very dense point spacing before the front and rear edges, so it is impossible to clearly judge the boundary point between the front and rear edges and the back of the leaf basin;

The processing steps are as follows:

Firstly, the data points are roughly divided into two free curves of the leaf basin and the leaf back, both ends of which respectively contain a part of the leading edge data points and the trailing edge data points; then, calculate the curvature of each point on the curve, and determine A curvature threshold; starting from the endpoints of the two single-valued curves, search for points whose curvature is greater than the curvature threshold, and put them into the leading edge or trailing edge array; next, use the least squares method to fit the rough extraction point set of the leading and trailing edges to obtain a pseudo Finally, the distance between each data point on the two single-valued curves and the fitting center of the leading and trailing edges is approximately equal to the fitting radius to accurately extract the leading and trailing edges.

Step 2, coarse registration of the middle arc

First, extract the middle arc of theoretical data and measured data, and use the center of the front and rear edges as the starting point and end point of the middle arc; obtain the middle arc as shown in Figure 4; then, respectively fit the theoretical and measured middle arcs and equalize To take points at intervals, it is required to obtain the data points of the arc in theory equal to the data points of the arc in actual measurement, and make the subscript of the starting point of the arc in theory consistent with the subscript of the starting point of the arc in actual measurement; finally, take the point set of the arc as Corresponding points, use the quaternion method to solve the rotation and translation matrix, and perform rigid body transformation on the measurement data.

Step 3, see Figure 5, distinguish between qualified points and out-of-poor points (out-of-poor points are displayed with asterisks), and the unqualified position of the contour size can be intuitively obtained;

This method uses the middle arc for rough registration, so that the accuracy of rough registration is high, so that bad points in measurement can be effectively eliminated; followed by tolerance constraints, to ensure that the degree of torsion and position are not out of tolerance, so that the error of the area contour The minimum and minimum number of out-of-tolerance points is the goal to achieve the best match; the minimum condition method is used to evaluate the profile, which is more in line with the actual requirements of the project; through the method of graphic display, the profile parameters of the blade can be quickly evaluated, and the position degree, torsion degree, Contour degree, number of out-of-tolerance points and defect positions; the matching evaluation is accurate and reliable, which can better guide the further processing of blades.

Example:

1) Read in the theoretical data and measured data of the profile line of the blade, and fit the contour line respectively;

2) Data preprocessing: According to the distribution characteristics of the profile data points, the data points are divided into four parts: leaf basin, leaf back, leading edge and trailing edge and stored separately;

3) Find the middle arc, use the middle arc as the matching feature to perform rough matching on the measured data and the theoretical data, and remove the coarse error data;

Judgment criteria for measuring bad points:

If the distance from the measuring point to the theoretical curve satisfies

d _i <min{d _i-2 ,d _i-1 ,d _i+1 ,d _i+2 } or d _i >max{d _i-2 ,d _i-1 ,d _i+1 ,d _i+2 },

And when |d _{i average} -d _i |>δ ₁ , then p _i point is a dead point.

Among them: d _{i average} =(d _i-2 +d _i-1 +d _i+1 +d _i+2 )/4; δ ₁ is a given limit value, take δ ₁ =1.5e, e=max{ Section line contour tolerance value};

4) Using the best matching algorithm, taking the torsion and position not exceeding the tolerance as the constraint conditions, and aiming at the minimum contour degree and the minimum number of deviation points, the measured data and the theoretical data are accurately matched;

5) According to the matching result, calculate the torsion angle Ψ, the position degree w and the profile degree of the blade;

The rotation matrix R and translation matrix T determined by the quaternion method are as follows:

Then the torsion angle of the measured data of the profile line of the blade is:

Accumulation point position degree:

When both the position degree w and the torsion error Ψ do not exceed the given tolerance parameters, use the obtained rotation-translation matrix R _k , T _k to update the data point set P _K , so that P _k+1 = R _k *P _k + T _k , repeat the ICP iterative process of fine registration; stop the iteration when the error of the position or torsion is out of tolerance, or the sum of the maximum values of the contour errors of each area reaches the minimum, and it is considered to be the best match at this time;

6) Generate an error tolerance zone with the standard shape as the bone line according to the given tolerance parameters;

Translate the data points on the theoretical section contour line inward by e1 or outward by e2 along the normal vector of each point, and then use the reverse engineering method to obtain two smooth curves, and the band-shaped area between the two curves is the tolerance zone; Among them, e1 and e2 are respectively given internal and external allowable error values;

7) Evaluation analysis;

Distinguish and display the actual measurement points and out-of-tolerance points in the tolerance zone, and visually see the position of the blade defect; and comprehensively evaluate whether the blade profile is qualified through the contour degree, position degree, torsion error and contour out-of-tolerance points .

Claims (6)

1. a kind of blade profile molded line profile parameters evaluation methodology based on best match is it is characterised in that comprise the following steps：

Step one, reads in gross data and the measured data of blade profile molded line, and matching contour line respectively；

Step 2, carries out data prediction：According to the characteristic distributions of molded line data point, data point is divided into leaf basin, blade back, leading edge Store with four parts of trailing edge and respectively；

Step 3, using mean camber line as matching characteristic, carries out preliminary matches to measured data and gross data, and rejects thick mistake Difference data；

Step 4, carries out accurately mate using best match algorithm to measured data and gross data；

Step 5, according to matching result, is calculated the torsion angle of blade by the rotation needed for rigid body translation and translation matrix Ψ, position degree w, the limit profile error amount using leaf basin, blade back, leading edge, trailing edge tries to achieve molded line profile tolerance；

Step 6, generates the error allowed band with standard profile for bone line according to given tolerance parameter；

Step 7, carries out parameter evaluation analysis according to error allowed band to blade profile molded line profile.

2. a kind of blade profile molded line profile parameters evaluation methodology based on best match according to claim 1, it is special Levy and be, in described step 3, preliminary matches are using mean camber line as matching characteristic, obtain the higher thick coupling knot of accuracy Really, so as to effectively reject measurement bad point it is ensured that evaluate the correctness of molded line profile tolerance with minimal condition method, and mate institute for essence Using ICP algorithm provide preferable initial value.

3. a kind of blade profile molded line profile parameters evaluation methodology based on best match according to claim 1, it is special Levy and be, in described step 4, accurately mate adopt improved ICP algorithm, with torsional error and bending error not overproof be about Bundle condition, with leaf basin, blade back, leading edge, trailing edge region contour degree is minimum, profile overproof point sum minimum as target；By rotation Torque battle array R and translation matrix T try to achieve the torsion angle Ψ and position degree w of blade profile molded line；To be carried out based on best match Meet the actual vane type line profile parameters evaluation of engineering.

4. a kind of blade profile molded line profile parameters evaluation methodology based on best match according to claim 3, it is special Levy and be, described improved ICP algorithm comprises the following steps：

The first step, using after preliminary registration through rigid body translation measured data as essence registration initial value；By seeking each measurement point Intersection point to theory shaped wire to obtain corresponding closest approach；

Second step, based on region tolerance constraints, evaluates molded line profile tolerance using minimal condition method, that is, ensureing position degree and torsion On the premise of error is not overproof, profile error and the overproof points of profile tolerance are made to reach minimum；Then object function is expressed as：

F (R, T)=min { max { distance (P_{i_p}＇, L)+max { distance (P_{i_b}＇, L)+max { distance (P_{i_q}＇, L)+max{distance(P_{i_h}＇, L) }

&&min{N_{oversize_points}}；

In formula, P_{i_p}＇, P_{i_b}＇, P_{i_q}＇ P_{i_h}＇ is respectively leaf basin, blade back, leading edge, the eyeball of trailing edge through registration process by firm Body converts the point of gained；L is theoretical section line；N_{oversize_points}It is the number of the overproof point outside tolerance range, that is, be located at Points within the inner boundary of tolerance range or beyond external boundary；

3rd step, the spin matrix R being determined by Quaternion Method and translation matrix T are following form：

$R=\left[\begin{array}{ccc}{R}_{11}& R_{12}& R_{13}\\ R_{21}& R_{22}& R_{23}\\ R_{31}& R_{32}& R_{33}\end{array}\right],T=\left[\begin{array}{ccc}\mathrm{\&Delta;}x& \mathrm{\&Delta;}y& \mathrm{\&Delta;}z\end{array}\right];$

The then torsion angle of blade profile molded line measured data is：

$\mathrm{\&Psi;}=arctg(-\frac{R_{21}}{R_{22}})=arctg(-\frac{2(q_1{q}_2+q_0{q}_3)}{q_0^2-q_1^2+q_2^2-q_3^2})$

$\mathrm{\&Psi;}\&Element;(-\frac{\mathrm{\&pi;}}{2},\frac{\mathrm{\&pi;}}{2});$

Long-pending folded point position degree：

When position degree w and torsional error Ψ is all less than given tolerance parameter, with rotation translation matrix R tried to achieve_k、T_kLogarithm Strong point collection P_KIt is updated, make P_k+1=R_k*P_k+T_k, repeat the ICP iterative process of essence registration；When position degree or twist error Overproof, or each region contour degree max value of error sum reach during minimum stop iteration it is believed that now be best match.

5. a kind of blade profile molded line profile parameters evaluation methodology based on best match according to claim 1, it is special Levy and be, in described step 3, the concrete grammar of preliminary matches is as follows, first, extract the middle arc of gross data and measurement data Line, using the front and rear edge center of circle as the beginning and end of mean camber line；Then, fitting theory takes with actual measurement mean camber line and at equal intervals respectively Point is equal to actual measurement mean camber line data points it is desirable to obtain theoretical mean camber line data points, and makes the starting point subscript of theoretical mean camber line Consistent with the subscript of actual measurement mean camber line starting point；Finally, with mean camber line point set as corresponding point, solve rotation peace using Quaternion Method Move matrix, and rigid body translation is carried out to measurement data.

6. a kind of blade profile molded line profile parameters evaluation methodology based on best match according to claim 1, it is special Levy and be, in described step 3, the method rejecting gross error data is as follows：

Measurement bad point judges：

After preliminary matches, if measuring point is to the distance of theoretical curve：

d_i<min{d_i-2,d_i-1,d_i+1,d_i+2Or d_i>max{d_i-2,d_i-1,d_i+1,d_i+2},

And | d_{I is average}-d_i|>δ₁When, then p_iFor bad point；

Wherein：d_{I is average}=(d_i-2+d_i-1+d_i+1+d_i+2)/4；δ₁For given limit value, take δ₁=1.5e, e=max { section line wheel Wide degree tolerance value }.