A method for reconstructing a circumferential weld geometry model of a cylindrical weld test piece

Abstract

This invention relates to a method for reconstructing the geometric model of the circumferential weld seam of a cylindrical welded specimen: Step 1: Perform 3D scanning and divide the cross-section at the weld seam location; Step 2: Divide the inner and outer diameter data of the cross-section into zones; Step 3: Label the inner and outer diameter values ​​according to their original sequence, and then label them with randomized tags after sorting by value; Step 4: Determine the data distribution function; Step 5: Use the data statistics from Step 2 as the distribution parameter of the data distribution function in Step 4, and generate new random simulation data for the inner and outer diameter data under the data distribution function, and then label them with randomized tags after sorting by value; Step 6: Match the simulation data with randomized tags to the original tags and rearrange them according to the original tag order to obtain the simulation data of the inner and outer diameters of the cross-section; Step 7: Repeat Steps 2 to 6 to reconstruct the three-dimensional model of the weld seam deformation influence zone. This invention can quickly generate a statistically equivalent three-dimensional virtual cylindrical welded specimen.

Description

Technical Field

[0001] This invention relates to a method for reconstructing the geometric model of the circumferential weld seam of a cylindrical welded specimen. Background Technology

[0002] In the manufacturing process of mechanical equipment, welding is the most commonly used method for connecting metal structures. With the increasing maturity of welding technology, almost all mechanical structures designed in the initial design phase can be manufactured. In the early stages of mechanical structure design, finite element method (FEM) simulation analysis has become an important means of verifying the feasibility of structural strength design schemes in order to analyze performance indicators such as structural mechanical strength. Combined with feedback from structural mechanical tests under specified load conditions, it can verify the accuracy of the simulation analysis, creating a mutually reinforcing and virtuous cycle. When conducting mechanical simulations of large welded structures, it is necessary to restore the weld seam structure to its actual structural form as much as possible to effectively improve the accuracy of the analysis of the structural mechanical state. Currently, one common method used domestically and internationally in researching such problems is to perform 3D scanning of the specimen structure to obtain its structural outline, and then use software to process the original data or automatically generate the scanned structural specimen to reconstruct a three-dimensional geometric model for structural mechanical simulation analysis. Another common method for reconstructing the three-dimensional structural model required for mechanical analysis of structures with weld seams is to use an equivalent structure to simulate the weld seam morphology. In this case, the structural shape already differs from the actual specimen; the equivalent model only approximates the mechanical state of the test specimen after mechanical simulation analysis due to the stress effect at the weld seam position. The above-mentioned methods for geometric reconstruction of weld positions are all empirical and experimental verification techniques that need to be verified with actual specimens. However, these methods lack statistical regularity, and the mechanical simulation analysis is uncontrollable for the number of virtual specimens generated by 3D scanning that are actually needed. Based on this method, the number of welding specimens required in reality is large, and the analysis efficiency is low, which increases the design cost. Summary of the Invention

[0003] The purpose of this invention is to provide a method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen. By combining statistics and numerical simulation technology, it can quickly generate a three-dimensional virtual cylindrical welded specimen with statistically equivalent configuration, thereby reducing the dependence on the number of actual welded specimens and significantly improving the efficiency of structural mechanics simulation analysis.

[0004] The objective of this invention is achieved through the following technical solution:

[0005] A method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen includes the following steps:

[0006] Step 1: Perform 3D data scanning on the cylindrical welded structure to determine the weld deformation influence zone and divide the weld deformation influence zone to obtain i cross-sections;

[0007] Step 2: Divide the inner and outer diameter data of the cross section of any weld deformation influence zone into circumferential sections, and perform statistical descriptions on the inner and outer radius data within each section.

[0008] Step 3: According to the actual geometric structure, label each value of the inner and outer diameters of each partition with the original label according to the original sequence, and then sort them according to the size of the value. The sorted values ​​are labeled with random labels, and each random label corresponds to an original label.

[0009] Step 4: Determine the data distribution function based on the data characteristics of each partition;

[0010] Step 5: Use the data statistics results in Step 2 as the distribution parameters of the data distribution function determined in Step 4. Generate new random simulation data for the inner and outer diameter data of each partition of the cross section under the data distribution function. Sort the simulation data within each partition according to the numerical value. The sorted data are then labeled with the disordered labels from Step 2 in circumferential order.

[0011] Step 6: Match the simulation data with disordered labels in each partition of Step 5 with the original labels, rearrange the simulation data in that partition according to the original label order, and then connect the simulation data of the inner and outer diameters in each partition to obtain the reconstructed simulation data of the inner and outer diameters of the cross section.

[0012] Step 7: Repeat steps 2 to 6 i times to obtain simulation reconstruction data of i cross sections, realize the reconstruction of the three-dimensional geometric model of the weld deformation influence zone of the cylindrical welded structure specimen, and perform analysis.

[0013] In step one, one end of the weld deformation influence zone is the starting position for dividing the cross-section, and the other end is the ending position for dividing the cross-section. The number of cross-sections in the weld deformation influence zone is determined based on the mesh structure of the finite element model in the simulation analysis, and the distance between adjacent cross-sections is not greater than the minimum longitudinal finite element mesh size.

[0014] In step one, the distance between adjacent cross sections is equal to the minimum longitudinal finite element mesh size, or the minimum longitudinal finite element mesh size is n times the distance between adjacent cross sections.

[0015] In step one, the selection of cross-sectional radius data points and the node errors on the structural mesh satisfy the requirement that the numerical result error between adjacent mesh densities be less than 5%.

[0016] In step two, the cross-section within the weld deformation influence zone is a non-standard circular structure. The factors to consider in determining the number of circumferential partitions include: 1) the amount of data falling within the corresponding interval limits; 2) the fluctuation of the inner and outer radii of the cross-section; and 3) the number of partitions being even.

[0017] In step four, the data in each partition all conform to the normal distribution function.

[0018] The advantages and positive effects of this invention are as follows:

[0019] 1. The method of this invention acquires a cylindrical welded structure through 3D scanning and divides the weld location into cross-sections. Then, the inner and outer diameters of each cross-section are divided into several data points according to nodes and statistically described. The inner and outer diameter data of each cross-section are then processed by a data distribution function after verification to generate random data. After sequential reconstruction, the geometric dimensions of a cross-section are reproduced. All cross-sections at the weld location are cyclically processed in this way to reconstruct the three-dimensional simulation specimen structure of the cylindrical welded structure. Furthermore, this method can perform data mining and feature extraction on a certain number of real weld data to statistically determine the structural morphological characteristics of the weld joint under this working condition, which is used for the geometric dimension description of this method. This enables the simulation generation of an infinite number of new three-dimensional geometric models of cylindrical welded structure specimens. After obtaining a stable geometric morphological characterization and stable laws, it is not necessary to add new processed welded specimens. The number of actual specimens is controllable, reducing the dependence on the number of actual welded specimens.

[0020] 2. The method of this invention combines statistical simulation, which can quickly reconstruct cylindrical welded structures with welds when the specific parameters of the weld location are unknown, thus significantly improving the efficiency of structural mechanics simulation analysis. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0022] Figure 2 This is a schematic diagram of a cylindrical welded structure with one weld seam.

[0023] Figure 3 for Figure 2 A schematic diagram of the longitudinal section of a medium-sized cylindrical welded structure.

[0024] Figure 4 for Figure 2 Schematic diagram of the transverse cross-section of a medium-sized cylindrical welded structure, showing the locations of welds and non-welds.

[0025] Figure 5 for Figure 2 A schematic diagram of the sequence of inner and outer radii of a random cross-section at the weld location of a medium-sized cylindrical welded structure.

[0026] Figure 6 This is a schematic diagram illustrating the test results of the data distribution function used in the method of this invention.

[0027] Figure 7 for Figure 2A schematic diagram of simulation data of the inner and outer radius sequences of a random cross-section at the weld location of a medium cylindrical welded structure. Detailed Implementation

[0028] The invention will now be described in further detail with reference to the accompanying drawings.

[0029] like Figure 2 As shown, the method of the present invention is illustrated using a cylindrical welded structure with only one weld as an example, because multiple welds are simply a repetition of the reconstruction process of the method of the present invention.

[0030] like Figures 1-7 As shown, the method of the present invention includes the following steps:

[0031] Step 1: Perform 3D data scanning on the cylindrical welded structure to determine the weld deformation influence zone. Divide the weld deformation influence zone into cross-sections based on the minimum finite element mesh size, obtaining i cross-sections. One end of the weld deformation influence zone is the starting position for cross-section division, and the other end is the ending position. Starting from the starting end of the weld deformation influence zone, sequentially select non-standard annular cross-section data. The first selected cross-section is numbered n1, then n2, n3, ..., n... i .

[0032] This invention uses 3D scanning to acquire geometric dimensional data of a cylindrical welded structure, observes the changes in the geometric spatial data of the entire cylindrical welded specimen, and, in addition to organizing the overall structural geometric parameters, focuses on determining the weld deformation-affected zone. Based on the structural dimensional data of the weld deformation-affected zone, it reconstructs the actual specimen's geometric morphology to determine the starting and ending positions of the transverse section.

[0033] like Figures 2-4 As shown, a weld deformation influence zone exists in the longitudinal section at any position of the weld seam of the specimen. To improve the accuracy of the mechanical simulation analysis, several transverse sections are divided along the width direction of the weld deformation influence zone. The number of cross sections in the weld deformation influence zone is determined based on the mesh structure of the finite element model for the actual mechanical simulation analysis. The distance between adjacent cross sections is not greater than the minimum longitudinal finite element mesh size, and the distance between cross sections can be equal to the minimum longitudinal finite element mesh size, or the minimum longitudinal finite element mesh size is n times the distance between adjacent cross sections (n ​​is an integer greater than 1), which does not affect the mechanical simulation analysis. The selection of the cross sections in the weld deformation influence zone corresponds to the node size on the finite element mechanical simulation analysis structure mesh and error control is performed. The selection of cross section radius data points and the node error on the structure mesh satisfy the numerical result error between adjacent mesh densities being less than 5%, so as to meet the requirement of finite element mechanical simulation being insensitive to the mesh.

[0034] Step 2: Divide the inner and outer diameter data of the cross section of any weld deformation influence zone into circumferential regions, and perform statistical descriptions on the data within each region.

[0035] like Figure 4 As shown, the cross-section of a cylindrical welded structure will only have two forms: a standard circular structure at the non-weld location and a non-standard circular structure at the weld location. The method of this invention only targets a single cross-section within the weld deformation influence zone; within this zone, all cross-sections exhibit a non-standard circular structure. Figure 5 As shown, the cross-section includes an inner diameter and an outer diameter that are not standard circles.

[0036] The cross-section of the weld location is divided into circumferential sections. The number of circumferential sections needs to consider the following factors: 1) the amount of data falling within the corresponding section limit; 2) the fluctuation of the inner and outer radii of the non-standard ring; 3) the number of sections should ideally be even, preferably designed according to a geometric sequence such as 2, 4, 6, 8, ... The requirement for the amount of data falling within the section limit is to meet the statistical sample size requirement when performing probability distribution tests on a set of data. A small sample size will lead to a larger statistical bias, affecting the final reconstruction of the three-dimensional geometric data. The degree of fluctuation of the inner and outer radii of the non-standard ring is related to the actual weld location and shape of the specimen. The fewer large fluctuation periods in the data falling within a single limit, the better. The requirement for an even number of sections is to facilitate the initial position of the division nodes.

[0037] The following is based on Figure 5 The illustrated embodiment further illustrates this point.

[0038] Figure 5 The cross-section numbered n1 is shown below. The non-standard annular inner and outer diameter data are shown in Table 1. The inner diameter sampling point is M. 1n =124 data points, with the same M sampling points for the outer diameter. 1w =124 data points. As shown in Table 1 below, the inner and outer diameter data are random variables. Randomly select a point as the starting point for the inner and outer diameter data and plot the sequence data in polar coordinates as follows: Figure 5 As shown, the inner diameter of the standard circular cross-section of the cylindrical specimen is 401.035 mm and the outer diameter is 413.335 mm, which serves as a reference.

[0039]

[0040] Table 1

[0041] like Figure 5As shown in the figure, after observation and analysis, the cross-sectional data of the weld position is divided into four partitions according to the sequence fluctuation, labeled as phase 1, phase 2, phase 3 and phase 4. Each phase contains 32 data points. Statistical descriptions are performed on the data in the four phases. This is a well-known technique in the field. The results are shown in Table 2 below.

[0042]

[0043] Table 2

[0044] Step 3: According to the actual geometric structure, label each value of the inner and outer diameters obtained after partitioning with the original label according to the original sequence, and then sort them according to the size of the value. The sorted values ​​are labeled with random labels, and each random label corresponds to an original label.

[0045] According to the actual geometric structure, each value of the inner and outer diameters obtained after partitioning is labeled with the original label according to the original sequence. Thus, a cross-sectional annulus corresponds to a set of data structured like a dictionary in a computer programming language. The labels need to be designed to be able to sort according to specific rules.

[0046] Taking the first five outer diameter data in Table 1 above as an example, each outer diameter value obtained after partitioning according to the actual geometric structure is labeled with the original label A according to the original sequence. The original sequence is: A1 = 413.44, A2 = 413.49, A3 = 413.57, A4 = 413.72, A5 = 413.94.

[0047] The five sets of data above are sorted according to their numerical values ​​to form a randomized label B, where: B1 = 413.94, B2 = 413.72, B3 = 413.57, B4 = 413.49, B5 = 413.44. Here, randomized label B1 corresponds to the original label A5, randomized label B2 corresponds to the original label A4, and so on.

[0048] All data in Table 1 above are labeled with the original label A in the manner described above, and then sorted by numerical value to obtain the disordered label B. As can be seen from the above, each disordered label corresponds to a unique original label.

[0049] Step 4: Determine the data distribution function based on the data conditions of each partition.

[0050] To test the distribution type of the data in Table 1 above, we first determine a data distribution function, and then calculate the corresponding series statistics. Determining the distribution type of a set of data is very easy; it can be verified using statistical methods or software. This is a fundamental method in statistics and a well-known technique in the field.

[0051] This invention assumes that these data follow a normal distribution. Then, it uses MINITAB statistical analysis software to test the distribution type. If the assumed distribution is accepted, the next step is performed; if the assumed distribution is rejected, a new distribution type is assumed. In this embodiment, after analysis using MINITAB software, if the obtained parameter P-value > 0.05, it indicates that the assumption that this set of data follows a normal distribution is reasonable. Taking the one-phase data of the cross-sectional outer diameter in this embodiment as an example, the verification result is as follows... Figure 6 As shown in Table 2 above, the data in each partition conforms to a normal distribution, but the distribution parameters vary. The reason for using a normal distribution in this invention is that modifying the normal distribution parameters is sufficient to meet the analytical requirements of circular tube morphology, and most data will generally meet the requirements of a normal distribution.

[0052] Step 5: Use the data statistics results from Step 2 (Table 2) as the distribution parameters of the data distribution function determined in Step 4. Generate 32 new random simulation data for each phase of the inner and outer diameters under the new data distribution function, resulting in a total of 256 random simulation data. Then, sort the simulation data within each partition phase according to their numerical values. The sorted data are then labeled with the disordered labels obtained in Step 2 in a circular direction, with the maximum value also being the disordered label B1.

[0053] Step Six: Match the simulation data with disordered labels from Step Five to the original labels, and rearrange the simulation data according to the original labels. For example, if the disordered labels after sorting the simulation data by size in Step Five are B1, B2, B3, B4, and B5, as shown in Step Three, B1 corresponds to A5, B2 to A4, B3 to A3, B4 to A2, and B5 to A1. Therefore, rearrange the five sets of simulation data in the order A1, A2, A3, A4, and A5. This invention rearranges the 256 random simulation data obtained in Step Five according to the original labels in the above manner. Connecting the simulation data within the four phases of the inner and outer diameters together yields the reconstructed simulation data of a random cross-sectional inner and outer diameter sequence within the weld deformation influence zone. The reconstructed data is as follows: Figure 7 As shown.

[0054] Step 7: Repeat steps 2 to 6 i times to obtain simulation reconstruction data for i cross sections, that is, to realize the reconstruction of the three-dimensional geometric model of the weld deformation influence zone of the cylindrical welded structure specimen and perform analysis.

[0055] To analyze the accuracy of simulation results, this invention measures the quality of a model's prediction by calculating the mean absolute percentage error (MAPE) of the simulation data for a random cross-section at the weld location, using the inner and outer radius sequences of the simulated data. This is a well-known technique in the field, and the calculation formula is MAPE = sum(|r* j -r j | / r j ) / N, where N is the sample size, r j This is the actual value of the radius, r* j The values ​​are for radius simulation. Verification shows that the mean relative absolute error (MAPE) of this invention is within 0.5‰. Furthermore, the statistical error between the reconstructed weld geometry model using this method and the finite element mechanical analysis of the 3D scanned geometric model of the structural specimen is less than 5%, which meets the requirements for mechanical simulation analysis.

[0056] This invention utilizes statistical methods to describe the weld structure morphology through geometric dimensional data. Furthermore, by performing data mining and feature extraction on a sufficient number of real weld data points, the structural morphological characteristics of the weld joint under this working condition can be statistically analyzed and used for the geometric dimensional description of this invention. Once a stable geometric morphology characterization and stable patterns are obtained, there is no need to add new processed welding specimens. This reduces the dependence on the number of actual welding specimens, achieving the goal of saving on the number of actual welding samples, while significantly improving the efficiency of structural mechanics simulation analysis. The stability of the geometric morphology reconstruction in this invention stems from the fineness of the three-dimensional mesh. The finer the mesh, the better the stability of the reproduced and reconstructed morphology. Ensuring geometric morphology stability is achieved by refining the mesh exponentially, reducing the mesh size.

[0057] In addition, the method of the present invention can quickly reconstruct a cylindrical welded structure with a weld. The simulation study object's three-dimensional structure is based on the input of structural parameters to realize virtual structure modeling. However, when the specific parameters are unclear (such as the non-standard circle in this embodiment), the modeling speed cannot be improved. The method of the present invention combines statistics to realize the rapid reconstruction of three-dimensional structural size parameters in a statistical sense.

[0058] The method of this invention belongs to the method of equivalent structure simulation of weld morphology, that is, the structural shape is different from the actual specimen, and the obtained equivalent model is close to the mechanical state of the test specimen after mechanical simulation analysis of the stress effect at the weld position.

Claims

1. A method for reconstructing the geometric model of the circumferential weld seam of a cylindrical welded specimen, characterized in that: Includes the following steps: Step 1: Perform 3D data scanning on the cylindrical welded structure to determine the weld deformation influence zone and divide the weld deformation influence zone to obtain i cross-sections; Step 2: Divide the inner and outer diameter data of the cross section of any weld deformation influence zone into circumferential sections, and perform statistical descriptions on the inner and outer radius data within each section. Step 3: According to the actual geometric structure, label each value of the inner and outer diameters of each partition with the original label according to the original sequence, and then sort them according to the size of the value. The sorted values ​​are labeled with random labels, and each random label corresponds to an original label. Step 4: Determine the data distribution function based on the data characteristics of each partition; Step 5: Use the data statistics results in Step 2 as the distribution parameters of the data distribution function determined in Step 4. Generate new random simulation data for the inner and outer diameter data of each partition of the cross section under the data distribution function. Sort the simulation data within each partition according to the numerical value. The sorted data are then labeled with the disordered labels from Step 2 in circumferential order. Step 6: Match the simulation data with disordered labels in each partition of Step 5 with the original labels, rearrange the simulation data in that partition according to the original label order, and then connect the simulation data of the inner and outer diameters in each partition to obtain the reconstructed simulation data of the inner and outer diameters of the cross section. Step 7: Repeat steps 2 to 6 i times to obtain simulation reconstruction data of i cross sections, realize the reconstruction of the three-dimensional geometric model of the weld deformation influence zone of the cylindrical welded structure specimen, and perform analysis.

2. The method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen according to claim 1, characterized in that: In step one, one end of the weld deformation influence zone is the starting position for dividing the cross-section, and the other end is the ending position for dividing the cross-section. The number of cross-sections in the weld deformation influence zone is determined based on the mesh structure of the finite element model in the simulation analysis, and the distance between adjacent cross-sections is not greater than the minimum longitudinal finite element mesh size.

3. The method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen according to claim 2, characterized in that: In step one, the distance between adjacent cross sections is equal to the minimum longitudinal finite element mesh size, or the minimum longitudinal finite element mesh size is n times the distance between adjacent cross sections.

4. The method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen according to claim 2, characterized in that: In step one, the selection of cross-sectional radius data points and the node errors on the structural mesh satisfy the requirement that the numerical result error between adjacent mesh densities be less than 5%.

5. The method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen according to claim 1, characterized in that: In step two, the cross-section within the weld deformation influence zone is a non-standard circular structure. The factors to consider in determining the number of circumferential partitions include: 1) the amount of data falling within the corresponding interval limits; 2) the fluctuation of the inner and outer radii of the cross-section; and 3) the number of partitions being even.

6. The method for reconstructing the geometric model of the circumferential weld of a cylindrical welded specimen according to claim 1, characterized in that: In step four, the data in each partition all conform to the normal distribution function.

Images