Steel bridge deck welding inherent deformation prediction method based on digital image correlation technology

Abstract

The invention discloses a method for predicting welding inherent deformation of a steel bridge deck based on a digital image correlation technology. The prediction method comprises the steps that high-temperature-resistant random speckles are manufactured in a welding seam and a near-welding-seam area of a steel bridge deck; digital image related equipment and welding equipment are adjusted, and welding of the steel bridge deck is completed; meanwhile, digital speckle images before and after the test piece is welded are collected and transmitted to a computer through a WiFi network for image pretreatment; and importing the processed digital image into an image analysis processing system, and calculating the welding inherent deformation by using a digital image related algorithm. Different from the conventional method, the digital image related technology directly acquires and analyzes the images before and after welding to obtain the inherent deformation, so that the welding inherent deformation prediction efficiency and precision of the steel bridge deck are greatly improved, and a more efficient means is provided for the welding deformation prediction of a large complex welding structural member.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge construction, and in particular relates to a method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, which is particularly suitable for predicting inherent deformation of steel bridge deck welding. Background Art

[0002] In recent years, orthotropic steel bridge decks have been widely used in large and medium span steel bridges due to their advantages such as less consumables, light weight and high strength, fast construction, and high degree of industrialization. However, due to the complex structure of orthotropic steel bridge decks and the large number of welds, it is also easy to produce welding deformation during the welding process. However, the previous method of calculating inherent deformation is too complicated and time-consuming. For this, we use digital image correlation technology to calculate the inherent deformation and thus predict its welding deformation.

[0003] As a typical representative of non-interference optical measurement methods, the digital image correlation method is different from the traditional interference optical mechanics method. It directly obtains deformation information based on the grayscale change of the digital image before and after the deformation of the material and structure surface. Digital image correlation technology has the advantages of full-field and local deformation measurement, non-contact measurement, strong adaptability, simple equipment, high degree of automation, and wide measurement range. It has been widely used as a common optical measurement and effective means in many fields such as materials science, aerospace, and civil engineering. Therefore, the present invention proposes a method for predicting the inherent deformation of steel bridge deck welding based on digital image correlation technology. Summary of the invention

[0004] In view of the above background technology, the present invention aims to propose a method for predicting the inherent deformation of steel bridge deck welding based on digital image correlation technology to predict the inherent deformation of welding. A new high-temperature resistant speckle production method is used to obtain random speckles that can be stably used at high welding temperatures and have deformation characteristics similar to those of the specimen. Then, the image acquisition system of digital image correlation technology is used to collect digital images of the specimen before and after welding. Finally, the sub-pixel displacement is solved by the relevant algorithm of the image processing system to obtain the inherent deformation of the steel bridge deck. The inherent deformation prediction can provide reference data for the performance of the welded structure and reduce the unsafe factors of the welded structure during the engineering service stage.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, the steps of the method are as follows:

[0007] Step 1: Pre-treat the surface of the steel bridge deck. First, use coarse sandpaper to grind the predicted area to improve the surface roughness of the specimen. After removing the surface oxides and foreign matter, wash it with clean water and wipe it dry to improve the adhesion of the coating on the surface of the specimen.

[0008] Step 2: Create high temperature resistant random speckles on the surface of the steel bridge deck weld and near weld area to provide a carrier for welding deformation information;

[0009] Step 3: Install and adjust digital image-related equipment and welding equipment, set relevant parameters, and use welding robots to complete the welding process of steel bridge decks;

[0010] Step 4: During the welding process, the image acquisition system using digital image correlation technology obtains the speckle images before and after welding, and transmits them to the computer in real time through the WiFi network. A series of sampling points are selected near the weld in the image along the weld direction, and then pre-processed to obtain an image that can be used for direct analysis;

[0011] Step 5: Import the processed digital image into the image analysis and processing system, and use the Fourier transform-based cross-correlation algorithm to calculate the integer pixel displacement to provide the initial displacement value for sub-pixel matching;

[0012] Step 6: Use the inverse combined Gauss-Newton algorithm based on the second-order shape function to calculate the sub-pixel displacement, and then calculate the inherent deformation of the steel bridge deck welding based on the displacement, and the whole process is completed.

[0013] As a preferred technical solution, the production process of high temperature resistant speckle in step 2 is: use a mixed liquid of black cobalt oxide and high temperature paint that can withstand high temperatures of 1900°C as black and white paint to produce speckle, and the formed speckle will not change color or fall off under high welding temperatures, thereby ensuring the quality of the speckle. In order to enhance the randomness of the speckle, first use a spray gun to spray the paint evenly and randomly on the surface of the steel bridge deck, then dry the specimen at room temperature for two hours, then place it in a vacuum drying oven at about 100°C for two hours, and finally take it out and cool it naturally to enhance the adhesion between the paint and the steel bridge deck.

[0014] As the preferred technical solution, the hardware part of the DIC image acquisition system in step 4 is mainly composed of high-precision industrial cameras, ultraviolet light sources, optical lenses, camera brackets, high-temperature infrared thermometers, industrial routers, computers, etc. The image acquisition system, combined with the infrared thermometer, can quickly obtain stable high-quality images while monitoring the welding temperature, and then perform image pre-processing, including image noise reduction, image segmentation and edge detection, to improve the quality of digital images and reduce prediction errors.

[0015] As a preferred technical solution, the image analysis and processing system in step 5 adopts a Fourier transform-based cross-correlation algorithm to achieve high robustness and fast whole-pixel displacement search, providing accurate initial displacement values ​​for the subsequent sub-pixel matching algorithm.

[0016] The deformation sub-area is set to 41*41~61*61 pixels, and the search range is half of the subset. Based on the Fourier cross-correlation algorithm, the overlapping area of ​​the application subset t(x, y) and the current matching subset g(x, y) is used to perform inverse Fourier transform on the mutual power in the frequency domain to obtain the impulse function. By detecting the peak position of the function, the detection of two-dimensional displacement of the image in the spatial domain is realized.

[0017] First, assume that the two subsets are the same, and g(x, y) is the position of t(x, y) after displacement by u and v in the x and y directions respectively. Its expression is as follows:

[0018] g(x,y)=t(xu,yv)

[0019] The relationship between the two-dimensional Fourier transform G and T corresponding to g and t is as follows:

[0020] G(ε,η)=e -j2π(εu+ηv) T(ε,η)

[0021] The cross power spectrum of the two is further calculated as follows:

[0022]

[0023] Among them G * represents the complex conjugate of G, and the peak values ​​of the obtained impulse function correspond to the relative translations u and v of the subset in the spatial domain.

[0024] As a preferred technical solution, the DIC image analysis and processing system in step 6 uses an interpolation method to reconstruct the grayscale image of the sub-pixel area to make it an approximately continuous image, obtain the grayscale information of the sub-pixel position points, and at the same time convert the discrete grayscale distribution into a function of continuous grayscale distribution through interpolation, providing a calculation basis for solving the grayscale gradient in the subsequent sub-pixel algorithm.

[0025] Let S(x) be y = f(x) at the nodes x0, x1, ..., x n The cubic spline interpolation function on , then:

[0026]

[0027] The matrix is ​​changed to solve all first-order derivatives by the pursuit method, and then A0, A1, A2, and A3 are calculated using the following formula.

[0028]

[0029] In the formula, Δx k =x k+1 -x k , Δy k =y k+1 -yk By performing cubic spline interpolation in the X and Y directions respectively, the grayscale and grayscale gradient of bicubic spline interpolation can be obtained.

[0030] As a preferred technical solution, the DIC image analysis and processing system in step 6 uses the reverse combination Gauss-Newton algorithm based on the second-order shape function to calculate the sub-pixel displacement. In each iteration of the Hessian matrix, the increment W(x, y; Δp) is first applied to the reference subset instead of the target subset, and then reversely combined with the current estimate W(x, y; p) to update the target subset.

[0031] Performing a first-order Taylor expansion on Δp and removing high-order small quantities, the zero-mean normalized minimum square distance correlation function can be written as:

[0032]

[0033] Where M represents the half width of the sub-area, Δp represents the change of p, f and g represent the grayscale values ​​of the reference image and the deformed image at the point (x, y). and Represent the average grayscale values ​​corresponding to f and g respectively; is the gradient of the reference subset.

[0034] Find C ZNSSD After finding the local extreme value of (Δp), we can solve:

[0035]

[0036] The solution to the least squares problem is:

[0037]

[0038]

[0039] Where H is the Hessian matrix, which depends only on the reference subset, so it remains unchanged in each iteration. Hessian matrix and grayscale gradient term All calculations are completed before the iteration begins. The second-order form of the shape function is as follows:

[0040]

[0041] Among them, S1…S 18 is a function of the parameter vector p.

[0042]

[0043] After calculation, Δp is obtained and substituted into the Warp function to obtain the incremental Warp function W(x, y; Δp). Then, based on W(x, y; Δp), the original W(x, y; p) is updated. The relationship is as follows:

[0044] W(x,y;p′)=W(x,y;p)W -1 (x, y; Δp)

[0045] Substitute p′ in W(x, y; p′) into the solution formula of Δp to start a new iteration, and repeat the iteration until the preset convergence condition is met. Set the convergence condition to ‖Δp‖=‖p n+1 -p n |‖≤10 -3 After the above iterative process, the reverse combined Gauss-Newton algorithm based on the second-order shape function can converge to the local optimal solution, namely C ZNNSSD (Δp) reaches the local minimum and obtains the sub-pixel level shape parameter vector p.

[0046] p=(u,u x ,u y ,u xx ,u xy ,u yy ,v,v x , v y , v xx , v yy , v yy ) T

[0047] The advantages and positive effects of the present invention are:

[0048] The present invention predicts the inherent deformation of steel bridge deck welding based on digital image correlation technology. By making random high-temperature resistant speckles on the surface of the steel bridge deck and obtaining the speckle images before and after the welding of the test piece, a series of sampling point displacements are extracted based on an efficient image processing and analysis algorithm, thereby obtaining the inherent deformation of welding. The invention has simple equipment and only requires the use of a camera and subsequent image processing and analysis algorithms to complete the rapid and accurate prediction of inherent deformation of welding. The operation process is simple and easy. The non-contact optical measurement has good safety performance and strong adaptability. It can be used to predict the deformation of large and complex welded structures under a large field of view, provide reference data for the performance of welded structures, and reduce unsafe factors of welded structures during the engineering service stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a system diagram of the present invention;

[0050] Figure 2 Schematic diagram of the inherent deformation prediction area of ​​the U-rib-cover plate joint;

[0051] Figure 3 is a step diagram of the present invention;

[0052] Figure 4 It is a flow chart of the present invention.

[0053] In the figure, 1. U-rib-cover plate joint; 2. High temperature resistant random speckle; 3. Welding robot; 4. DIC image acquisition system; 5. DIC image analysis and processing system; DETAILED DESCRIPTION

[0054] The following will be explained in conjunction with the drawings in the embodiments of the present invention; the technical solutions in the embodiments of the present invention will be described in detail and completely; of course, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0055] like Figure 1-3 As shown, the present invention provides a method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, including the following steps:

[0056] Step 1: Pre-treat the surface of the steel bridge deck. First, use coarse sandpaper to grind the predicted area to improve the surface roughness of the specimen, then remove the surface oxides and foreign matter, wash with clean water and wipe dry to improve the adhesion of the coating on the surface of the specimen;

[0057] Step 2: Create high temperature resistant random speckles on the surface of the steel bridge deck weld and near weld area to provide a carrier for welding deformation information;

[0058] Step 3: Install and adjust digital image-related equipment and welding equipment, set relevant parameters, and use welding robots to complete the welding process of steel bridge decks;

[0059] Step 4: During the welding process, the image acquisition system using digital image correlation technology obtains the speckle images before and after welding, and transmits them to the computer in real time through the WiFi network. A series of sampling points are selected near the weld in the image along the weld direction, and then pre-processed to obtain an image that can be used for direct analysis;

[0060] Step 5: Import the processed digital image into the image analysis and processing system, calculate the integer pixel displacement based on the Fourier transform cross-correlation algorithm, and provide the initial displacement value for sub-pixel matching;

[0061] Step 6: Use the inverse combined Gauss-Newton algorithm based on the second-order shape function to calculate the sub-pixel displacement, and then use the obtained sub-pixel displacement to calculate the inherent deformation of the steel bridge deck welding, and the whole process is completed.

[0062] The embodiments of the present invention are described in detail above, but the contents are only preferred embodiments of the present invention and cannot be considered to limit the scope of implementation of the present invention. Any similar changes and improvements made according to the scope of application of the present invention should be regarded as within the concept of the present invention and the protection scope of the claims.

Claims

1. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: Includes the following steps: Step 1: Pre-treat the surface of the steel bridge deck. First, use coarse sandpaper to grind the predicted area to improve the surface roughness of the specimen. After removing the surface oxides and foreign matter, wash it with clean water and wipe it dry to improve the adhesion of the coating on the surface of the specimen. Step 2: Create high temperature resistant random speckles on the surface of the steel bridge deck weld and near weld area to provide a carrier for welding deformation information; Step 3: Install and adjust digital image-related equipment and welding equipment, set relevant parameters, and use welding robots to complete the welding process of steel bridge decks; Step 4: During the welding process, the image acquisition system using digital image correlation technology obtains the speckle images before and after welding, and transmits them to the computer in real time through the WiFi network. A series of sampling points are selected near the weld in the image along the weld direction, and then pre-processed to obtain an image that can be used for direct analysis; Step 5: Import the processed digital image into the image analysis and processing system, and use the Fourier transform-based cross-correlation algorithm to calculate the integer pixel displacement to provide the initial displacement value for sub-pixel matching; Step 6: Use the inverse combined Gauss-Newton algorithm based on the second-order shape function to calculate the sub-pixel displacement, and then calculate the inherent deformation of the steel bridge deck welding based on the obtained sub-pixel displacement, and the whole process is completed.

2. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: The production process of high temperature resistant speckle in step2 is: use a mixed liquid of high temperature resistant black cobalt oxide and high temperature paint as black and white paint to produce high temperature resistant speckle. First, use a spray gun to spray black and white paint evenly and randomly on the surface of the welding area of ​​the steel bridge deck, then dry the specimen at room temperature for two hours, then place it in a vacuum drying oven at about 100℃ for two hours, and finally take out the specimen and cool it naturally to obtain high temperature resistant random speckle.

3. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: The DIC image analysis and processing system in step6 uses a second-order shape function to approximate the actual deformation during welding, which can well describe the uniform and non-uniform deformation of the sub-area. The expression is as follows: Among them, the center point of the reference sub-area before deformation is P(x0, y0), the sub-area of ​​(2M+1)×(2M+1) pixels is established with P as the center, Q(x, y) is an arbitrary point in the reference sub-area, and satisfies x=x0+dx, y=y0+dy, after deformation, point Q moves to Q′(x′, y′), the displacement vector is (u, v), and x′=x+u, y′=y+v; and is the partial derivative of the horizontal displacement with respect to x and y, and is the partial derivative of the vertical displacement with respect to x and y; and is the two-dimensional displacement gradient of the image sub-region.

4. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: In step 6, the DIC image analysis and processing system uses the zero-mean normalized difference square sum formula to measure the similarity of the image sub-areas before and after welding deformation. There is a change in light intensity during the welding process, and this correlation function is insensitive to the overall light intensity change. Its expression is as follows: Where F(x,y) is the grayscale function of the reference sub-region, G(x′,y′) is the grayscale function of the target sub-region, and Represent the average grayscale of the reference sub-area and the target sub-area respectively.

5. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: The DIC image analysis and processing system in step6 uses a Fourier transform-based cross-correlation algorithm to achieve highly robust and fast whole-pixel displacement search, providing accurate initial displacement values ​​for the subsequent sub-pixel matching algorithm. The deformation sub-area is set to 41*41~61*61 pixels, and the search range is half of the subset. Based on the Fourier cross-correlation algorithm, the overlapping area of ​​the application subset t(x, y) and the current matching subset g(x, y) is used to perform inverse Fourier transform on the mutual power in the frequency domain to obtain the impulse function. By detecting the peak position of the function, the detection of two-dimensional displacement of the image in the spatial domain is realized. First, assume that the two subsets are the same, and g(x, y) is the position of t(x, y) after displacement by u and v in the x and y directions respectively. Its expression is as follows: g(x,y)=t(xu,yv) The relationship between the two-dimensional Fourier transform G and T corresponding to g and t is as follows: G(e, h)=e -j2π(εu+ηv) T(e,h) The cross power spectrum of the two is further calculated as follows: Among them G * represents the complex conjugate of G, and the peak values ​​of the obtained impulse function correspond to the relative translations u and v of the subset in the spatial domain.

6. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: In step 6, the DIC image analysis and processing system needs to obtain the grayscale information of non-integer pixel positions. Since the point G (x′, y′) in the deformed image may fall at a sub-pixel position, it is necessary to use the interpolation method to reconstruct the grayscale image of the sub-pixel area to make it an approximately continuous image and obtain the grayscale information of the sub-pixel position. At the same time, the discrete grayscale distribution is converted into a function of continuous grayscale distribution through interpolation, which provides a calculation basis for solving the grayscale gradient in the subsequent sub-pixel algorithm. Let S(x) be y = f(x) at the nodes x0, x1, ..., x n The cubic spline interpolation function on , then: The matrix is ​​changed to solve all first-order derivatives by the pursuit method, and then A0, A1, A2, and A3 are calculated using the following formula. In the formula, Δx k =x k+1 -x k , Δy k =y k+1 -y k By performing cubic spline interpolation in the X and Y directions respectively, the grayscale and grayscale gradient of bicubic spline interpolation can be obtained.

7. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: In step 6, the DIC image analysis and processing system uses the inverse combined Gauss-Newton algorithm based on the second-order shape function to calculate the sub-pixel displacement. In each iteration of the Hessian matrix, the increment W(x, y) is first applied to the reference subset instead of the target subset; Δp), and then compared with the current estimate W(x, y; p) Perform reverse combination to update the target subset. Performing a first-order Taylor expansion on Δp and removing high-order small quantities, the zero-mean normalized minimum square distance correlation function can be written as: Where M represents the half width of the sub-area, Δp represents the change of p, f and g represent the grayscale values ​​of the reference image and the deformed image at the point (x, y). and Represent the average grayscale values ​​corresponding to f and g respectively; is the gradient of the reference subset. Find C ZNSSD After finding the local extreme value of (Δp), we can solve: The solution to the least squares problem is: Where H is the Hessian matrix, which depends only on the reference subset, so it remains unchanged in each iteration. Hessian matrix and grayscale gradient term All calculations are completed before the iteration begins. The second-order form of the shape function is as follows: Among them, S1…S 18 is a function of the parameter vector p. After calculation, Δp is obtained and substituted into the Warp function to obtain the incremental Warp function W(x, y; Δp). Then, based on W(x, y; Δp), the original W(x, y; p) is updated. The relationship is as follows: W(x,y;p′)=W(x,y;p)W -1 (x,y;Δp) Substitute p′ in W(x, y; p′) into the solution formula of Δp to start a new iteration, and repeat the iteration until the preset convergence condition is met. Set the convergence condition to ||Δp||=||p n+1 -p n ||≤10 -3 After the above iterative process, the reverse combined Gauss-Newton algorithm based on the second-order shape function can converge to the local optimal solution, namely C ZNSSD (Δp) reaches the local minimum and obtains the sub-pixel level shape parameter vector p. p=(u,u x ,u y ,u xx ,u xy ,u yy ,v,v x ,v y ,v xx ,v xy ,v yy ) T 。 8. A method for predicting inherent deformation of steel bridge deck welding based on digital image correlation technology, characterized in that: In step 6, the DIC image analysis and processing system uses the sub-pixel displacement obtained in step 5 to calculate the intrinsic strain. According to the measured longitudinal displacement, a curve is fitted and the slope of the curve is calculated to obtain the longitudinal inherent strain. and the longitudinal natural strain ε L The relationship is: By using this relationship, the longitudinal inherent deformation can be obtained, and the lateral inherent deformation can also be obtained accordingly.

Images