Method for rapidly determining age forming manufacturing parameters of aluminum alloy integral panel

Abstract

The invention belongs to the technical field of integral stiffened wallboard structure manufacturing, and relates to a method for rapidly determining age forming manufacturing parameters of an aluminum alloy integral wallboard. Comprising the following steps of S1, representing a component as a prediction sample according to a discrete representation method, and performing rebound and strength prediction through a deep learning prediction model; s2, optimization of technological parameters and iterative compensation of mold surface springback are carried out based on a prediction result; and S3, establishing a finite element model of the component according to an iteration result, and checking and correcting the finite element model. According to the method, deep learning and a traditional finite element simulation technology are combined, the key process parameters and the springback compensation molded surface of age forming of the aluminum alloy integral wall plate can be rapidly and accurately determined, and technical auxiliary support is provided for forming and manufacturing of a large complex integral stiffened wall plate.

Description

Technical Field

[0001] This invention belongs to the field of manufacturing technology of integral stiffened wall panel structures, and relates to a method for rapidly determining the manufacturing parameters of aluminum alloy integral wall panel aging forming. Background Technology

[0002] Large integral stiffened panels are crucial load-bearing structures for aircraft wings and fuselages, and aging forming technology is one of the main manufacturing technologies for large integral panels, as well as an important direction for technological development in the aerospace equipment field. Aging forming technology has advantages such as high forming accuracy, high repeatability, and low residual stress; however, the formed components exhibit significant springback, requiring springback compensation to determine the mold surface. Currently, the most commonly used springback compensation method is the numerical iterative trial-and-error method based on finite element simulation, which uses finite element simulation coupled with an accurate material creep deformation constitutive model to perform multiple rounds of "springback prediction, deviation calculation, and mold surface compensation" through numerical trial and error. However, with the increasing size and complexity of the panels to be formed, the time cost of simulation calculations and springback compensation calculations has increased significantly. In some local areas where structural stiffness or curvature changes abruptly, convergence is even difficult, making it impossible to predict the springback compensation situation.

[0003] In recent years, with the development of artificial intelligence, machine learning methods have begun to be applied in the field of engineering manufacturing. Using deep learning technology to build neural networks and applying them to the calculation of springback compensation in the forming of large integral stiffened panels can not only greatly improve the calculation efficiency in the initial stage of springback compensation, but also optimize forming process parameters. Finally, a precise finite element model is used for local correction, thereby achieving rapid and accurate positioning of the manufacturing parameters for the aging forming of integral panels. Summary of the Invention

[0004] Purpose of the invention Based on the above analysis, the purpose of this invention is to provide a method for rapidly determining the manufacturing parameters of aluminum alloy integral wall panels during aging forming. This method utilizes deep learning technology to rapidly predict the springback and strength of integral wall panels during forming, completes the initial mold surface springback compensation and locates the forming process parameter range, effectively reducing computation time and resource consumption. Finally, it combines a precise finite element model for correction to further ensure the accuracy of the results.

[0005] Technical solution A method for rapidly determining the manufacturing parameters of aluminum alloy integral wall panels during aging forming, comprising the following steps: S1: The component is represented as a prediction sample according to the discrete representation method, and the rebound and strength are predicted by the deep learning prediction model. S2: Optimize process parameters and perform iterative compensation for mold surface springback based on prediction results; S3: Based on the iteration results, establish a finite element model of the component for verification and correction.

[0006] Furthermore, step S1 specifically includes: The desired integral wall panel structure is discretized and represented as a set of prediction samples. The discretization method is as follows: the structural and technological features of any point i on the wall panel are used as input parameters for the samples, and the springback difference ΔY is used as the input parameters for the samples. i xz and yield strength value σ i As output parameters, the predicted sample set {D} is formed. i |i=1,...,N},D i It includes the input and output parameters for that point.

[0007] Furthermore, the structural and technological features of any point i are specifically the skin thickness δ. i s Fiber height δ i r Structural variation coefficient θ i x+ θ i x- θ i z+ θ i z- Forming time t, mold radius in the x direction R i x The radius R of the mold in the z-direction i x .

[0008] Furthermore, the structural change coefficient and the springback difference are calculated using the following formulas:

[0009]

[0010]

[0011]

[0012]

[0013] Where δ x+ δ x- δ z+ and δ z- The ratio of the skin thickness or rib height at the structural abrupt change points i in the positive x-direction, negative x-direction, positive z-direction, and negative z-direction to the skin thickness or rib height at the current point. s x+ s x-s z+ and s z- These are the distances from the points of structural abrupt change in the aforementioned directions. x0 and z0 are the coordinates of point i before it takes shape, and y... t Let i be the springback coordinates after point i is formed.

[0014] Furthermore, among the parameters in the predicted sample set, the mold radius R i x Mold radius R i z The Y-axis coordinate of the overall wall panel target surface 目标 The forming time t is calculated based on the boundary values ​​of the aging forming process used. The predicted sample set is input into the trained deep learning prediction model to obtain the springback and strength values ​​predicted by the model.

[0015] Furthermore, step S2 specifically involves: calculating the 10th percentile value σ based on the intensity values ​​of all discrete points predicted by the deep learning prediction model. 0.1 To determine whether the strength of the formed structure meets the strength index requirements and to avoid the influence of extreme predicted values ​​on the judgment.

[0016] Determine if σ 0.1 ≥N1σ 指标 , where σ 指标 The strength index of the structure is determined according to the overall panel forming requirements, N1σ 指标 N1 represents the minimum strength that the structure is required to achieve, and N1 is a value greater than 1 according to project requirements.

[0017] Furthermore, if we judge σ 0.1 ≥N1σ 指标 If no, it indicates that the strength of the component after forming is too low, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N} j+1 =t j -D,t j Let be the forming time after the j-th compensation, and D be the forming time compensation value, which is 0.1 in hours. Other appropriate values ​​can also be used according to project requirements. Then, the deep learning model prediction is performed again.

[0018] If we determine σ 0.1 ≥N1σ 指标 If it is positive, then continue to check if σ is positive. 0.1 ≤N2σ 指标 N2 > N1. If we judge σ 0.1 ≤N2σ 指标 If no, it indicates that the strength of the component is redundant after forming, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N}j+1 =t j +D, then re-predict using the deep learning model. If we determine σ 0.1 ≤N2σ 指标 A positive value indicates that the current forming time is reasonable and that surface springback compensation can be carried out.

[0019] Furthermore, the rebound values ​​ΔY=Y are then calculated for all discrete points predicted by the deep learning prediction model. 预测 -Y 目标 , where Y 预测 Let ΔY be the Y-axis coordinate of a discrete point after the overall panel structure is formed, as predicted by deep learning. ΔY represents the distance between the predicted rebound value and the target value at that discrete point. The determination is made as to whether the predicted distance ΔY is the maximum value. max ≤ Engineering error, the engineering error is determined according to the overall panel forming requirements. If ΔY max If the engineering error judgment is negative, then update the Y-axis coordinate of the updated surface to Y. i (j+1) =Y i j +K i j ΔY i j , where Y i j Y is the Y-axis coordinate of the surface corresponding to the discrete point i of the component after the j-th compensation. i 1=Y 目标 K i j =ΔY i j / Y 目标 Then, the prediction sample set {D} is updated based on the new Y-axis coordinates of the surface. i The mold radius R in |i=1,...,N} i x R i z Re-perform the deep learning model prediction until ΔY max ≤ Engineering error is judged as positive, Y i j Use the Y-axis coordinate of the springback compensation mold surface and proceed to step S3.

[0020] Furthermore, step S3 specifically involves: establishing a finite element model in the software ABAQUS based on the desired overall wall panel structure, and then using the Y obtained in step S2... i j In CATIA software, a smooth surface is drawn as the mold profile and input into the established finite element model. The optimized forming time t is used. jAging forming simulation was conducted, and the forming mold surface and process parameters of the overall wall panel were verified and corrected based on the finite element simulation results, ultimately obtaining forming manufacturing parameters that can be applied in engineering.

[0021] The beneficial effects of this application are as follows: This invention proposes a method for efficiently determining the manufacturing parameters of aluminum alloy integral wall panels through the integration of deep learning-assisted rapid prediction and finite element simulation for precise simulation. In the initial stage, a deep learning model completes over 95% of parameter optimization and surface compensation. In the later stage, a finite element model corrects the optimization results, balancing speed and reliability. The core innovation of this invention lies in replacing the traditional finite element method with a deep learning approach that replaces multiple rounds of trial and error, significantly improving prediction efficiency. It also innovatively proposes a dual-objective intelligent optimization method based on strength and springback, dynamically adjusting the forming time t to ensure strength meets standards while avoiding redundancy, and precisely locking the process parameter window. Simultaneously, an adaptive surface compensation algorithm iteratively updates the mold surface, quickly converging to within engineering error limits, solving the convergence problem in complex areas. Finally, finite element simulation refines the calculation results, significantly improving the efficiency of the initial iterative calculations while ensuring accuracy and reducing computational resource waste. The compensated surface of this invention can be directly generated in CATIA, and the parameter output is adapted to actual manufacturing, providing a complete "rapid prediction-intelligent optimization-precise verification" solution for large aerospace wall panels, significantly reducing R&D costs and time. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 The overall wall panel cross-sectional dimensions (unit: mm) are shown in the embodiments of the method of the present invention. Figure 3 This is a schematic diagram illustrating the method for calculating the structural change coefficient. Figure 4 This is a finite element simulation model of the overall wall panel in an embodiment of the method of the present invention; Figure 5 The strength distribution (unit: MPa) of the integral wall panel after forming in the finite element simulation of the embodiment of the method of the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] The features and illustrative embodiments of various aspects of the present invention will now be described in detail. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without requiring some of these specific details. The following description of embodiments is merely intended to provide a better understanding of the invention by illustrating examples of the invention. The invention is by no means limited to any specific setup and method set forth below, but covers any improvements, substitutions, and modifications to the structures, methods, and devices without departing from the spirit of the invention. In the following description, well-known structures and techniques are not shown to avoid unnecessarily obscuring the invention.

[0025] In the description of this invention, it should be noted that the directions or positional relationships indicated by terms such as "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer" are based on the stated directions or positional relationships and are only for the convenience of describing and simplifying the invention, and should not be construed as limiting the invention. Furthermore, the use of ordinal numbers (e.g., "first and second," etc.) is for distinguishing objects and is not limited to this order, and should not be construed as indicating or implying relative importance.

[0026] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly, encompassing both direct connection and indirect connection via an intermediate medium. Those skilled in the art can understand the specific meaning of these terms in this invention based on the specific circumstances.

[0027] It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other, and the various embodiments can be referenced and cited in each other. The present invention will now be described in detail with reference to and in conjunction with the embodiments.

[0028] Example 1 A method for quickly determining the manufacturing parameters for aging forming of integral aluminum alloy wall panels, including the following steps: S1: The component is represented as a prediction sample according to the discrete representation method, and the rebound and strength are predicted by the deep learning prediction model. S2: Optimize process parameters and perform iterative compensation for mold surface springback based on prediction results; S3: Based on the iteration results, establish a finite element model of the component for verification and correction.

[0029] In one embodiment of the present invention, step S1 specifically comprises: The desired integral wall panel structure is discretized into a set of prediction samples. The discretization method involves using the structural and technological features of any point i on the wall panel as input parameters for the samples, and taking the springback difference (ΔY) as the input parameter. i xz ) and yield strength value (σ i ) are used as output parameters to form the prediction sample set {D}. i |i=1,...,N},D i It includes the input and output parameters for that point.

[0030] In one embodiment of the present invention, the structural and technological features of i are specifically the skin thickness (δ) i s ), rib height (δ) i r ), structural change coefficient (θ) i x+ θ i x- θ i z+ θ i z- ), forming time (t), mold radius in the x direction (R) i x ), mold z-direction radius (R) i x ).

[0031] In one embodiment of the present invention, the structural change coefficient and the springback difference are calculated by the following formulas:

[0032]

[0033]

[0034]

[0035]

[0036] Where δ x+ δ x- δ z+ and δ z- The ratio of the skin thickness or rib height at the structural abrupt change points i in the positive x-direction, negative x-direction, positive z-direction, and negative z-direction to the skin thickness or rib height at the current point, and s x+ s x- s z+ and s z- These are the distances from the points of structural abrupt change in the aforementioned directions. x0 and z0 are the coordinates of point i before it takes shape, and y...t Let i be the springback coordinates after point i is formed.

[0037] In one embodiment of the present invention, among the parameters in the prediction sample set, the mold radius (R) i x R i z The Y-axis coordinate of the overall wall panel target surface is Y 目标 The forming time (t) is calculated and is generally determined based on the boundary of the aging forming process used. The predicted sample set is input into the trained deep learning prediction model to obtain the springback and strength values ​​predicted by the model.

[0038] In one embodiment of the present invention, step S2 specifically involves: calculating the 10th percentile value σ based on the intensity values ​​of all discrete points predicted by the deep learning prediction model. 0.1 This is used to determine whether the strength of the formed structure meets the strength index requirements and to avoid the influence of extreme predicted values ​​on the judgment. It determines whether σ... 0.1 ≥N1σ 指标 , where σ 指标 The strength index of the structure is determined according to the overall panel forming requirements, N1σ 指标 N1 represents the minimum strength that the structure is required to achieve, and N1 is a value greater than 1 according to project requirements.

[0039] In one embodiment of the present invention, if σ is determined 0.1 ≥N1σ 指标 If no, it indicates that the strength of the component after forming is too low, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N} j+1 =t j -D,t j Let be the forming time after the j-th compensation, and D be the forming time compensation value, typically 0.1 in hours. Other appropriate values ​​can be used depending on project requirements. Then, the deep learning model is re-predicted. If we determine σ... 0.1 ≥N1σ 指标 If it is positive, then continue to check if σ is positive. 0.1 ≤N2σ 指标 N2 > N1. If we judge σ 0.1 ≤N2σ 指标 If no, it indicates that the strength of the component is redundant after forming, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N} j+1 =t j +D, then re-predict using the deep learning model. If we determine σ 0.1 ≤N2σ 指标A positive value indicates that the current forming time is reasonable and surface springback compensation can be carried out.

[0040] In one embodiment of the present invention, the rebound value ΔY=Y is then calculated for all discrete points predicted by the deep learning prediction model. 预测 -Y 目标 , where Y 预测 Let ΔY be the Y-axis coordinate of a discrete point after the overall panel structure is formed, as predicted by deep learning. ΔY represents the distance between the predicted rebound value and the target value at that discrete point. The determination is made as to whether the predicted distance ΔY is the maximum value. max ≤ Engineering error, the engineering error is determined according to the overall panel forming requirements. If ΔY max If the engineering error judgment is negative, then update the Y-axis coordinate of the updated surface to Y. i (j+1) =Y i j +K i j ΔY i j , where Y i j Let Y be the Y-axis coordinate of the discrete point i of the component after the j-th compensation. i 1=Y 目标 ), K i j =ΔY i j / Y 目标 Then, the prediction sample set {D} is updated based on the new Y-axis coordinates of the surface. i The mold radius (R) in |i=1,...,N} i x R i z Then, re-perform the deep learning model prediction until ΔY is reached. max ≤ Engineering error is judged as positive, Y i j Use the Y-axis coordinate of the springback compensation mold surface and proceed to step S3.

[0041] In one embodiment of the present invention, step S3 specifically involves: establishing a finite element model in the software ABAQUS based on the desired integral wall panel structure, and determining the Y value obtained in step S2. i j In CATIA software, a smooth surface is drawn as the mold profile and input into the established finite element model. The optimized forming time t is then used. j Aging forming simulation was conducted, and the forming mold surface and process parameters of the overall wall panel were verified and corrected based on the finite element simulation results, ultimately obtaining forming manufacturing parameters that can be applied in engineering.

[0042] Example 2 The flowchart of the method of this invention is shown below. Figure 1 As shown, the following is a detailed description using a specific embodiment as an example: S1: The component is represented as a prediction sample according to the discrete representation method, and the rebound and strength are predicted by the deep learning prediction model. S11: For a single-curvature variable-thickness integral wall panel that needs to be formed, its structural dimensions are as follows: Figure 2 As shown, it is discretized into a set of prediction samples. The discretization method is to represent the structural and technological features of any point i on the panel, namely the skin thickness (δ). i s ), rib height (δ) i r ), structural change coefficient (θ) i x+ θ i x- ), forming time (t), mold radius (R) i ) is used as the input parameter for the sample, and the rebound difference (ΔY) is used as the input parameter. i x ) and yield strength value (σ i ) are used as output parameters to form the prediction sample set {D}. i |i=1,...,N},D i This includes the input and output parameters for that point. The structural variation coefficient and springback difference are calculated using the following formulas:

[0043]

[0044]

[0045] Where δ x+ and δ x- The ratio of the skin thickness or rib height at structural abrupt changes in each direction to the skin thickness or rib height at the current point, and s x+ and s x- These represent the distances from the structural abrupt change points in each direction, such as... Figure 3 As shown. x0 and z0 are the coordinates of point i before it is formed, y t Let i be the springback coordinates after point i is formed.

[0046] S12: The target profile of this integral wall panel is R. 目标 =1400mm, then the mold radius (R) in the predicted sample set parameters iThe value is 1400, and the forming time (t) is set to 3 based on the boundary of the two-stage aging forming process used in the integral wall panel, with the unit being hours. The predicted sample set is input into the trained deep learning prediction model to obtain the springback value and strength value predicted by the model.

[0047] S2: Optimize process parameters and perform iterative compensation for mold surface springback based on prediction results; S21: Calculate the 10th percentile value σ based on the intensity values ​​of all discrete points predicted by the deep learning prediction model. 0.1 =481.1MPa, while the strength index value σ 指标 =510MPa, take N1=1.01, then σ 0.1 <N1σ 指标 This indicates that the strength of the component after forming is too low, and the prediction sample set {D} is updated. i The forming time in |i=1,...,N} is t2=t1-0.1, and then the deep learning model prediction is performed again. After 5 iterations, when the forming time is 2.5h, σ 0.1 =516.2MPa, which meets the strength requirements.

[0048] S22: Based on S21, calculate ΔY=Y based on the rebound values ​​of all discrete points predicted by the deep learning prediction model. 预测 -Y 目标 Where ΔY is the vertical distance between the predicted rebound value at a discrete point and the corresponding point on the target surface, and the Y-axis coordinate of the target surface is Y. 目标 It can be calculated from the target profile radius of 1400mm and the half-width of the wall panel of 290mm. The maximum value ΔY during the first forming is obtained from the prediction results of the deep learning model. max =23.28mm, with an engineering error of 2mm, then ΔY max < Engineering error, update the Y-axis coordinate of the profile to Y i (j+1) =Y i j +K i j ΔY i j , where Y i j Let Y be the Y-axis coordinate of the discrete point i of the component after the j-th compensation. i 1=Y 目标 ), K i j =ΔY i j / Y 目标 The calculated surface radius after the first springback compensation is 922 mm, and the prediction sample set {D} is updated. iThe mold radius (R) in |i=1,...,N} i The deep learning model is then re-performed for prediction. After 10 iterations, ΔY... max =1.98mm < engineering error, which meets the requirements. At this time, the radius of the compensation surface is 435mm, and the radius of the whole wall panel after forming is about 1350mm.

[0049] S3: Establish a finite element model of the component based on the iteration results for verification and correction; S31: A finite element model of the integral panel aging forming process is established in ABAQUS. The mold surface radius is 435mm, and a two-stage aging forming process is adopted, with a forming time of 2.5 hours. The constitutive model of the aluminum alloy two-stage aging forming process is embedded into the finite element model. Simulation calculations show that the component forming radius is approximately 1440mm, and the strength distribution is as follows... Figure 4 As shown, it basically meets the manufacturing requirements.

[0050] In the above embodiments, more than ten prediction calculations were performed using a neural network prediction model in the early stage, with each calculation taking less than 10 seconds. In contrast, a single finite element simulation calculation of the entire wall panel takes about 23 hours. It can be seen that the present invention significantly improves the iterative optimization efficiency of the aging forming process parameters of the entire wall panel and saves computing resources.

[0051] The above detailed embodiments are a description of the present invention. It should not be considered that the specific embodiments of the present invention are limited to these descriptions. For those skilled in the art, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the protection scope of the present invention.

Claims

1. A method for rapidly determining the manufacturing parameters of an integral aluminum alloy wall panel during aging forming, characterized in that, The specific steps include: S1: The component is represented as a prediction sample according to the discrete representation method, and the rebound and strength are predicted by the deep learning prediction model. S2: Optimize process parameters and perform iterative compensation for mold surface springback based on prediction results; S3: Based on the iteration results, establish a finite element model of the component for verification and correction.

2. The method as described in claim 1, characterized in that, The specific steps of S1 are as follows: The desired integral wall panel structure is discretized and represented as a set of prediction samples. The discretization method is as follows: the structural and technological features of any point i on the wall panel are used as input parameters for the samples, and the springback difference ΔY is used as the input parameters for the samples. i xz and yield strength value σ i As output parameters, the predicted sample set {D} is formed. i |i=1,...,N},D i It includes the input and output parameters for that point.

3. The method as described in claim 2, characterized in that, The structural and technological features of any point i are specifically the skin thickness δ. i s Fiber height δ i r Structural variation coefficient θ i x+ θ i x- θ i z+ θ i z- Forming time t, mold radius R in the x direction i x The radius R of the mold in the z-direction i x .

4. The method as described in claim 3, characterized in that, The structural variation coefficient and the springback difference are calculated using the following formulas: Where δ x+ δ x- δ z+ and δ z- The ratio of the skin thickness or rib height at the structural abrupt change points i in the positive x-direction, negative x-direction, positive z-direction, and negative z-direction to the skin thickness or rib height at the current point. s x+ s x- s z+ and s z- These represent the distances from the points of structural abrupt change in the aforementioned directions; x0 and z0 are the coordinates of point i before its formation, and y... t Let i be the springback coordinates after point i is formed.

5. The method as described in claim 4, characterized in that, Among the parameters in the prediction sample set, the mold radius R i x Mold radius R i z The Y-axis coordinate of the overall wall panel target surface 目标 The forming time t is calculated based on the boundary value of the aging forming process adopted; the predicted sample set is input into the trained deep learning prediction model to obtain the springback value and strength value predicted by the model.

6. The method as described in claim 5, characterized in that, Step S2 specifically involves: calculating the 10th percentile value σ based on the intensity values ​​of all discrete points predicted by the deep learning prediction model. 0.1 To determine whether the strength of the formed structure meets the strength index requirements and to avoid the influence of predicted extreme values ​​on the judgment; Determine if σ 0.1 ≥N1σ 指标 , where σ 指标 The strength index of the structure is determined according to the overall panel forming requirements, N1σ 指标 N1 represents the minimum strength that the structure is required to achieve, and N1 is a value greater than 1 according to project requirements.

7. The method as described in claim 6, characterized in that, If we determine σ 0.1 ≥N1σ 指标 If no, it indicates that the strength of the component after forming is too low, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N} j+1 =t j -D,t j Let be the forming time after the j-th compensation, and D be the forming time compensation value, which is 0.1 in hours. Other appropriate values ​​can also be used according to project requirements. Then, the deep learning model is re-predicted. If we determine σ 0.1 ≥N1σ 指标 If it is positive, then continue to check if σ is positive. 0.1 ≤N2σ 指标 N2 > N1; if we judge σ 0.1 ≤N2σ 指标 If no, it indicates that the strength of the component is redundant after forming, and the prediction sample set {D} is updated. i The forming time t in |i=1,...,N} j+1 =t j +D, then re-predict using the deep learning model; if σ is judged 0.1 ≤N2σ 指标 A positive value indicates that the current forming time is reasonable and that surface springback compensation can be carried out.

8. The method as described in claim 7, characterized in that, Subsequently, the rebound value ΔY=Y is calculated for all discrete points predicted by the deep learning prediction model. 预测 -Y 目标 , where Y 预测 The Y-axis coordinates of discrete points after the overall panel structure is formed, as predicted by deep learning, are given; ΔY represents the distance between the predicted rebound value and the target value of the discrete point; it is determined whether the distance ΔY is the maximum value. max ≤ Engineering error, the engineering error is determined according to the overall panel forming requirements; if ΔY max If the engineering error judgment is negative, then update the Y-axis coordinate of the updated surface to Y. i (j+1) =Y i j +K i j ΔY i j , where Y i j Y is the Y-axis coordinate of the surface corresponding to the discrete point i of the component after the j-th compensation. i 1=Y 目标 K i j =ΔY i j / Y 目标 Then, the prediction sample set {D} is updated based on the new Y-axis coordinates of the surface. i The mold radius R in |i=1,...,N} i x R i z Re-perform the deep learning model prediction until ΔY max ≤ Engineering error is judged as positive, Y i j Use the Y-axis coordinate of the springback compensation mold surface and proceed to step S3.

9. The method as described in claim 8, characterized in that, Step S3 specifically involves: establishing a finite element model in the software ABAQUS based on the desired overall wall panel structure, and then using the Y obtained in step S2... i j In CATIA software, a smooth surface is drawn as the mold profile and input into the established finite element model. The optimized forming time t is used. j Aging forming simulation was conducted, and the forming mold surface and process parameters of the overall wall panel were verified and corrected based on the finite element simulation results, ultimately obtaining forming manufacturing parameters that can be applied in engineering.

Images