A large flange assembly alignment method based on multiple local images

Abstract

The present invention discloses a large flange assembly alignment method based on multiple local images, belonging to the technical field of manufacturing and assembly alignment. The method comprises the following steps: photographing the flange mating surface at different positions to obtain multiple local images of two flanges to be assembled; extracting the straight line information and circle information in each image pair; calculating the flange gap vector, the sign and the corresponding relationship between the gap vector and the translation adjustment amount, constructing an optimization model to calculate the flange translation adjustment amount; correcting the flange bolt hole position by the calculated flange translation adjustment amount, and modeling and calculating the flange rotation adjustment amount. The present invention can implement assembly alignment of large flange parts by using multiple local field of view high-precision images.

Description

Technical Field

[0001] The invention belongs to the technical field of manufacturing and assembly alignment, and in particular relates to an assembly alignment method for a large flange with multiple local images. Background Art

[0002] The assembly of large flange parts is mainly done manually. During the assembly process, relative offset and rotation between flanges often occur, resulting in large assembly alignment errors. Local collision damage will occur during installation, affecting assembly accuracy and reliability. Therefore, it is necessary to calculate the translation adjustment and rotation adjustment of the flange to be assembled to correct the flange assembly posture. The alignment method based on machine vision to directly obtain the overall image often requires taking a single high-resolution image or using a low-resolution image geometric laser distance sensor to measure the parts with high precision, which has high hardware cost and low efficiency. Summary of the invention

[0003] In view of this, the present invention provides a large flange assembly alignment method based on multiple local images, which can perform assembly alignment on large flange parts by using multiple local field of view high-precision images.

[0004] A large flange assembly alignment method based on multiple local images, the steps of implementing the method are as follows:

[0005] Step 1: photograph the flange mating surface at different positions to obtain multiple partial images of two flanges to be assembled;

[0006] Step 2: Extract the straight line information and circle information in each image pair;

[0007] Step 3: Construct an optimization model to calculate the flange translation adjustment;

[0008] Step 4: Correct the flange bolt hole position through the flange translation adjustment, and calculate the flange rotation adjustment by modeling.

[0009] Furthermore, each image in step 1 includes the edge of the flange and the bolt hole to be paired, wherein two flange images acquired at the same shooting position are combined to generate an image pair, and a total of multiple groups of image pairs are generated.

[0010] Furthermore, the process of extracting straight line information and circle information in step 2 includes: first, performing image filtering and edge extraction on each image pair, performing circle detection in the image through Hough transform on the edge points extracted from each image, and fitting the circle using the least squares method to extract the bolt hole position of the flange; eliminating points near the circle, and performing straight line fitting on the remaining points using the least squares method to obtain the edge position and direction of the flange.

[0011] Furthermore, the process of constructing the optimization model in step 3 to calculate the flange translation adjustment amount is as follows:

[0012] By fitting the linear equation obtained in the i-th local field of view to characterize the flange edge position and direction, the flange edge assembly gap size Cd in the flange local field of view is calculated for the image centering. (i) and direction Generate gap vector By calculating the i-th local field of view direction v FOVd (i) With gap direction The angle between them determines the gap symbol Where Sgn(x) is the sign function, T represents the transpose of the vector (matrix), and the gap will be adjusted with the flange translation Δt = (Δx, Δy) T The adjusted gap is:

[0013]

[0014] The gaps between the obtained N image pairs are calculated to generate an image gap sequence Cns{Cn (1) ,Cn (2) ,...,Cn (N)} Taking the best gap consistency as the optimization goal and the adjustment amount as the optimization variable, an optimization model is established:

[0015]

[0016] Where avg(Cns) is the mean of the set Cns;

[0017] By solving the model, the flange translation adjustment Δt is obtained opt =(Δx opt ,Δy opt ) T .

[0018] Furthermore, the process of modeling and calculating the flange rotation adjustment amount in step 4 is as follows: opt =(Δx opt ,Δy opt ) T The flange bolt hole positions that need to be adjusted have been extracted from each image pair Make corrections The bolt hole positions to be matched are The optimization goal is to minimize the position difference of the bolt holes to be paired after correction, and the flange rotation adjustment amount and the bolt hole distribution radius are the optimization variables to establish the optimization model:

[0019]

[0020] The model is solved and the flange rotation adjustment is calculated.

[0021] Furthermore, in step three, the gradient descent method is used to calculate the translation adjustment amount.

[0022] Furthermore, in step 4, a genetic algorithm is used to calculate the rotation adjustment amount.

[0023] Beneficial effects:

[0024] 1. The present invention provides a large flange assembly alignment method based on multiple local images. By using multiple local field of view high-precision images to assemble large flange parts, the hardware cost of the assembly system is reduced and the efficiency is improved.

[0025] 2. The present invention extracts multi-image gaps and uses image gap consistency as an indicator to evaluate the translation error of large flange parts. It does not need to detect the overall center position of the parts, and only requires a small amount of edge information for calculation, which reduces the complexity of the data to be processed. The rotation error is evaluated using the bolt hole center distance error as an indicator, which is closer to the actual process requirements of flange assembly.

[0026] 3. The evaluation index of gap consistency in the present invention is determined according to the specific assembly task requirements and can usually be measured using the variance of the gap sequence.

[0027] 4. The present invention establishes a translation adjustment optimization model and a rotation adjustment optimization model to calculate the translation adjustment and rotation adjustment of flange assembly. The optimization model uses the designed error evaluation index as the objective function, and the translation adjustment and rotation adjustment as the optimization variables. The translation adjustment and rotation adjustment are calculated separately through model solving, avoiding the large deviation that may be caused by estimating the overall center and direction of the flange through local image information using traditional methods, thereby improving the system assembly alignment accuracy.

[0028] 5. The method described in the present invention firstly performs an optimization calculation of the translation adjustment amount through a translation adjustment amount optimization model, and uses the translation adjustment amount calculation result to correct the rotation adjustment amount optimization model, thereby eliminating the influence of the translation error on the rotation error during the assembly and alignment process, which is more in line with the actual situation of assembly and alignment of large flange parts.

[0029] 6. The calculation method of the translation adjustment amount optimization model in the present invention is selected according to the specific application needs, and usually a gradient descent-based method can be used for fast calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 . Flow chart of the alignment method of the present invention;

[0031] Figure 2 .Schematic diagram of flange local field image acquisition;

[0032] Figure 3 . Schematic diagram of local field of view edge detection and line and circle extraction;

[0033] Figure 4 . Schematic diagram of local field of view gap;

[0034] Figure 5 . Schematic diagram of local field of view direction;

[0035] Figure 6 .Schematic diagram of bolt hole position correction;

[0036] Figure 7 . Schematic diagram of image, line and circle detection in the embodiment. DETAILED DESCRIPTION

[0037] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0038] The purpose of the present invention is to provide an assembly alignment method based on multiple local images for a large flange. According to the method, the overall adjustment amount of the flange can be estimated through several local fields of view, thereby achieving assembly alignment. Figure 1 The flowchart of the method of the present invention is shown. The method acquires local images of multiple locations of the flange to generate image pairs; performs image processing on each image pair to extract straight lines and circles therein, and separately represents the edge and bolt hole of the flange; calculates the gap in the local field of view of the flange through the extracted straight line information, and establishes a translation adjustment amount optimization model to calculate the translation adjustment amount; corrects the extracted bolt hole position through the calculation result of the translation adjustment amount, and establishes a rotation adjustment amount optimization model to calculate the rotation adjustment amount.

[0039] like Figure 2 As shown, the method first obtains multiple sets of local images with bolt holes to be paired at the flange assembly surface. The local field of view images captured at the same position are combined into an image pair. Multiple sets of local field of view image pairs are generated by capturing at multiple locations. The flange with the larger diameter of the two flanges is defined as flange 1, and the flange with the smaller diameter is defined as flange 2.

[0040] like Figure 3 As shown in the figure, each image obtained contains the edge of the flange and the bolt hole to be matched. The edge of the flange is a feature that is approximately a straight line, and the bolt hole is a feature that is approximately a circle. In order to extract the edge information and bolt hole information in the image, each image is first filtered and denoised, and then the image edge is extracted using the edge extraction operator. The straight line and circle in the image are fitted using the straight line and circle fitting algorithm to extract the straight line and circle information in the image. Repeat this operation to extract the straight line and circle information for all the obtained image pairs.

[0041] like Figure 4 As shown in the figure, the corresponding straight lines of flange 1 and flange 2 extracted from an image pair are L1 and L2, and their equations are A1 (i) x+B1 (i) y+C1 (i) =0,A2 (i) x+B2 (i) y+C2 (i) = 0. Define the flange gap as the vector from the edge of flange 1 to the edge of flange 2.

[0042] The reference direction of the gap is:

[0043]

[0044] The intercept vectors of the two lines are:

[0045]

[0046] Gap v cd (i) is the projection of the intercept vector in the reference direction:

[0047]

[0048] v cd (i) Normalized to the gap direction vector The size of the gap is Cd (i) . This generates the gap vector Repeat this operation to extract the gaps in all image pairs.

[0049] like Figure 5 As shown, the direction of the i-th local field of view relative to the flange center is defined as the field of view direction v FOVd (i) , in a suitable assembly, the angle between the gap direction and the field of view direction should be acute, and the gaps in all fields of view should meet this requirement. Define the gap symbol in the local field of view Where Sgn(x) is the sign function. It is easy to know that flange 1 is adjusted according to the translation amount Δt=(Δx,Δy) T After the movement occurs, the changed gap vector is:

[0050]

[0051] Generate a new gap size from the gap sign and the gap vector:

[0052]

[0053] Under appropriate assembly, among the N sets of image pairs obtained, the gaps between each image pair form a sequence Cns{Cn (1) ,Cn (2) ,...,Cn (N)} has a high consistency. The sequence consistency is evaluated by variance, and the consistency index is constructed:

[0054]

[0055] Taking the minimization of consistency index as the optimization goal and the translation adjustment amount as the optimization variable, the optimization model is constructed:

[0056]

[0057] The model is solved to obtain the flange translation adjustment Δt opt =(Δx opt ,Δy opt ) T .

[0058] like Figure 6 As shown in the figure, the center error of the bolt holes of the two flanges to be matched is composed of translation error and rotation error. First, the bolt hole position of flange 1 is corrected by solving the flange translation adjustment. In the i-th image pair, the bolt hole position of flange 1 is Flange 2 bolt hole position is The bolt hole position of flange 1 after correction is At this time, the error between the bolt holes of flange 1 and flange 2 after correction is mainly caused by rotation error. The rotation error is distributed through the bolt hole radius R (i) The rotation adjustment θ is corrected, and the optimization model is constructed with the minimum error of the center distance of the bolt holes in each image as the optimization goal:

[0059]

[0060] Bolt hole distribution radius R (i) The range of is known, and the model is solved to obtain the rotation adjustment amount θ opt .

[0061] The above optimization model can be solved by different optimization algorithms.

[0062] The invention discloses a large flange assembly alignment method based on the coaxial alignment principle. The method estimates the overall assembly error of the flange by using multiple local fields of view, thereby reducing the hardware cost of the assembly system.

[0063] Example:

[0064] The present method is used to assemble and align two flanges (flange 1 and flange 2) to be assembled as an example. Flange 1 has a diameter of 400.2 mm and flange 2 has a diameter of 400 mm. Eight bolt holes with a diameter of 4 mm and a distribution radius of 195 mm are evenly distributed on each flange to form a bolt hole pair. The two flanges have a translation position error of 0.11 mm and 0.15 mm in the x and y directions, respectively, and a 1° direction error (counterclockwise as the positive direction).

[0065] Step 1: Acquire images in 8 local fields of view at the flange mating location, acquire 8 groups of local field of view images, and form image pairs.

[0066] Step 2: Extract the circle information and line information in each image pair to obtain the flange edge position and direction and the bolt hole position and direction, such as Figure 7 As shown, the arrows in each figure indicate straight line extraction, and the crosses indicate circle extraction.

[0067] Step 3: Extract the gap size, direction, field of view direction, and gap sign in each image pair, as shown in Table 1.

[0068] Table 1

[0069]

[0070] Through the above extraction parameters, the optimization model is constructed

[0071]

[0072] Solve the model to obtain the translation adjustment Δt opt =(-30.34,-37.15), the corresponding actual physical adjustment is Δx opt =-111.25um, Δy opt =-136.23um, and the standard error (Δx opt =-110um, Δy opt =-150um) is 13.83um.

[0073] Step 4: Correct the bolt hole position using the calculated translation adjustment amount, as shown in Table 2.

[0074] Table 2

[0075]

[0076] Building an optimization model

[0077]

[0078] In the model, the bolt hole distribution radius R (i)is in a known range (195±0.5mm), and the above model is solved to obtain the rotation adjustment θ opt The value is 1.0037°, which is 0.0037° away from the standard error of 1°. The above calculation results prove the effectiveness of this method.

[0079] In summary, the above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A large flange assembly alignment method based on multiple local images, characterized in that: The steps to implement this method are as follows: Step 1: photograph the flange mating surface at different positions to obtain multiple partial images of two flanges to be assembled; Step 2: Extract the straight line information and circle information in each image pair; Step 3: Construct an optimization model to calculate the flange translation adjustment; Step 4: Correct the flange bolt hole position through the flange translation adjustment amount, and calculate the flange rotation adjustment amount by modeling; In step 1, each image includes the edge of the flange and the bolt hole to be paired, wherein two flange images acquired at the same shooting position are combined to generate an image pair, and a total of multiple groups of image pairs are generated; The process of constructing the optimization model in step 3 to calculate the flange translation adjustment amount is as follows: By fitting the linear equation obtained in the i-th local field of view to characterize the flange edge position and direction, the flange edge assembly gap size Cd in the flange local field of view is calculated for the image centering. (i) and direction Generate gap vector By calculating the i-th local field of view direction v FOVd (i) With gap direction The angle between them determines the gap symbol Where Sgn(x) is the sign function, T represents the transpose of the vector, and the gap will change with the flange translation adjustment amount Δt=(Δx,Δy) T The adjusted gap is: The gaps between the obtained N image pairs are calculated to generate an image gap sequence Cns{Cn (1) ,Cn (2) ,...,Cn (N) } Taking the best gap consistency as the optimization goal and the adjustment amount as the optimization variable, an optimization model is established: Where avg(Cns) is the mean of the set Cns; By solving the model, the flange translation adjustment Δt is obtained opt =(Δx opt ,Δy opt ) T ; The process of modeling and calculating the flange rotation adjustment amount in step 4 is as follows: opt =(Δx opt ,Δy opt ) T The flange bolt hole positions that need to be adjusted have been extracted from each image pair Make corrections The bolt hole positions to be matched are The optimization goal is to minimize the position difference of the bolt holes to be matched after correction. The flange rotation adjustment θ and the bolt hole distribution radius R (i) To optimize variables, an optimization model is established: The model is solved and the flange rotation adjustment is calculated.

2. The large flange assembly alignment method based on multiple local images as claimed in claim 1, characterized in that: The process of extracting straight line information and circle information in step 2 includes: first, performing image filtering and edge extraction on each image pair, performing circle detection in the image by Hough transform on the edge points extracted from each image, and fitting the circle by least squares method to extract the bolt hole position of the flange; eliminating points near the circle, and fitting the remaining points by least squares method to obtain the edge position and direction of the flange.

Images