A parametric modeling and stiffening optimization method for curved stiffened cylindrical shells under local axial compression

Abstract

This invention discloses a parametric modeling and stiffening optimization method for curved stiffened cylindrical shells under local axial compression, relating to the field of engineering thin-walled stiffened structure design. The method includes the following steps: determining basic and main parameters; designing curved stiffening paths; generating multi-channel structural images of different resolutions based on the stiffening path distribution and stiffening section geometry; establishing a finite element analysis model to calculate the critical buckling load and stiffener mass; building a convolutional neural network model to predict the critical buckling load and stiffener mass; and using the established convolutional neural network model to achieve adaptive optimization design of the curved stiffened structural layout under local axial compression. This invention expands the design space of curved stiffening paths, explores the diversity of curve distributions, and the entire process is simple, easy to operate, and highly efficient, providing a framework for the design of curved stiffened cylindrical shells under local axial compression.

Description

Technical Field

[0001] This invention relates to the field of engineering thin-walled reinforced structure design, specifically to a parametric modeling and reinforcement optimization method for a curved reinforced cylindrical shell under local axial compression. Background Technology

[0002] Integral curved stiffened panels, also known as variable stiffness curved stiffened structures, are a promising new type of load-bearing structure. Curved stiffening design allows for control over the direction, spacing, and position of infinitesimally small straight stiffeners, thus providing greater design flexibility. However, the design characteristics of curved stiffened structures are more complex than those of conventional uniform straight stiffened structures. Considering the distribution and direction of each stiffener leads to an explosive increase in design variables, significantly increasing the workload of optimization calculations. Conversely, considering the distribution of multiple stiffeners simultaneously restricts the freedom of design space. Finding a design method that balances the degree of freedom and complexity in stiffening design is a challenging problem. Summary of the Invention

[0003] To address the issues of overly complex design of curved stiffeners in large design spaces, and the time-consuming and error-prone nature of traditional manual modeling processes, this invention provides a parametric modeling and stiffening optimization method for curved stiffener cylindrical shells under local axial compression. By building a deep learning framework, structural features are extracted, and the layout of curved stiffeners and cross-sectional geometry are optimized, thereby improving the buckling load performance of stiffener cylindrical shells under local axial compression while satisfying stiffener mass constraints.

[0004] The objective of this invention is achieved through the following technical solution: a parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression, the method comprising the following steps: S1. Determine the geometric parameters and structural material parameters of the stiffened shell skin as the basic parameters, and the load boundary condition parameters, the path parameters of the curved stiffening, and the geometric parameters of the stiffened section as the main parameters; S2. Curve stiffening forms are divided into two types: circumferential and axial. Each stiffener is determined only by three control points: the start point, the midpoint, and the end point of the curve. Furthermore, the same control points of the same stiffener curve lie on a straight line, and the distribution of the same control points on a straight line is controlled by an exponential distribution function. S3. With all control points determined, a quadratic B-spline curve is used to draw the stiffening path. Based on the main parameters, a three-channel image containing stiffening path information and stiffening section geometry information is generated. The first two channels contain stiffening path images, and their grayscale values ​​correspond linearly to the stiffening thickness and stiffening height, respectively. The larger the size value, the darker the channel color. The third channel contains only a blank image. The initially formed three-channel image is downsampled to obtain two three-channel images with different resolutions, which together constitute a multi-channel structural image of the stiffened curve. S4. Based on the basic and main parameters, a finite element analysis model of a locally axially compressed thin-walled stiffened cylindrical shell is established using the mesh curling method. The critical buckling load and stiffener mass of the structure are obtained through analysis and used as two labels for the corresponding multi-channel structural images. S5. Using labeled multi-channel structural images as a sample set, two convolutional neural network models are built to predict the critical buckling load and stiffener mass of the stiffened structure, respectively. S6. Based on the established convolutional neural network model, implement the optimization design of the layout of the curve-stiffened structure under local axial compression. The optimization goal is to find the stiffening design scheme with the maximum critical buckling load of the stiffened cylindrical shell under the constraint of stiffener mass. During the optimization process, a dynamic penalty factor that increases with the number of iterations is set.

[0005] Furthermore, in step S1, for a cylindrical shell with a locally distributed axial compressive load that is equidistantly symmetrically distributed, the geometric parameters of the stiffened shell skin include the inner diameter of the cylindrical shell. R、 high H、 Wall thickness t Structural material parameters include material density. ρ Elastic modulus E Poisson's ratio ν Yield strength R eL Load boundary condition parameters include the number of load regions. N Angle of a single load area θ The path parameters for curved reinforcement include the number of circumferential reinforcements in the design area. n 1. Number of axial reinforcement bars in the design area n 2 and six sets of exponential distribution function parameters a , b The geometric parameters of the reinforced section include the height of the stiffeners. h rib thickness t s The exponential distribution function of the reinforced curve path is as follows: in a , b The parameter is a function parameter, and the sampling range is set in the interval [-1, 1], with the difference between the two parameters being less than or equal to 0.5.

[0006] Furthermore, in step S2, the curve stiffening form is divided into two types: circumferential and axial; based on the symmetry of the local axial compression structure, the side surface of the cylindrical shell skin is unfolded and the number of load areas is increased by 2. NThe design area is divided into equal parts to form a single rectangular design area. Within the rectangular design area, the starting point, ending point, and midpoint of the circumferential curve are distributed on the left and right boundaries and the vertical axis of symmetry, respectively, while the starting point, ending point, and midpoint of the axial curve are distributed on the upper and lower boundaries and the horizontal axis of symmetry, respectively. The distribution of the same type of control points on any straight line is controlled by an exponential distribution function, which maps the function to the distribution of points. When there are multiple rectangular design areas n When there are multiple identical curved bars, the distribution of the same control points (one of the start point, end point, and midpoint) of the same curved bars can be obtained according to the following formula: in L i It is the first of the same type of control point i The and the first i -1 The proportion of the distance between control points to the total design length.

[0007] Furthermore, in step S3, after the start point, midpoint, and end point of a curve are determined, a quadratic B-spline curve is selected for fitting. Especially for circumferential curves, in order to make the curves more accurate, based on the symmetry of the local axial compression structure, the rectangular design area is extended along the circumference. Specifically, a similar rectangular design area is added on each of the left and right sides, increasing the number of control points from three to seven. The control point positions of the added left rectangular design area and the original rectangular design area are symmetrical about the left boundary of the original rectangular design area, and the control point positions of the added right rectangular design area and the original rectangular design area are symmetrical about the right boundary of the original rectangular design area.

[0008] Furthermore, in step S3, during the imaging process of the stiffened structure, an RGB three-channel image is used. The initially formed three-channel image is a high-resolution image, which is used to capture the fine geometric features of the stiffeners. The high-resolution image is downsampled to obtain a low-resolution image, which is used to extract the overall layout of the structure. The two three-channel images with different resolutions together constitute a multi-channel structural image of the curve stiffener.

[0009] Further, in step S6, a meta-heuristic algorithm is used for optimization. The optimization goal is to find the stiffened design scheme that maximizes the critical buckling load of the stiffened cylindrical shell under the constraint of stiffener mass. By setting a dynamic penalty factor that increases with the number of iterations, the optimization process is divided into two stages: In the early stage of optimization, the penalty factor is small, and the main optimization goal is to maximize the buckling load, exploring design regions that slightly exceed the mass limit but have great load-bearing potential, and avoiding premature constraint; In the later stage of optimization, the penalty factor is large, forcing the optimization results to converge to the feasible region that satisfies the mass constraint, and performing a refined search.

[0010] Furthermore, in step S6, the formulas for fitness and dynamic penalty factor during the optimization process are as follows: in, Fitness For the fitness function, N To predict the critical buckling load of a structure using a convolutional neural network model, λ As a dynamic penalty factor, P This is a quality penalty item. λ min To minimize the penalty factor, λ max The maximum value of the penalty factor. k To optimize the rounds, k max To maximize the number of optimization attempts, m To predict rib quality using a convolutional neural network model, M For quality constraints.

[0011] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above-described parametric modeling and reinforcement optimization method.

[0012] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the above-described parametric modeling and stiffening optimization method.

[0013] Compared with existing methods, the beneficial effects of the present invention are as follows: 1. This invention uses an exponential distribution function and a quadratic B-spline curve to depict the distribution of ribs throughout the entire design area with fewer variables, which can significantly improve modeling efficiency and support the rapid and batch generation of different reinforcement path schemes.

[0014] 2. This invention characterizes different stiffening information under local axial compression in the form of images, including the distribution of curved paths and the geometric cross-section of the stiffeners, and uses multi-channel images of different resolutions for representation. This image can be used as input to achieve a unified representation of multiple variables and scales, and a convolutional neural network model can be used to predict the load-bearing capacity of the structure, providing a framework for constructing a mapping between stiffeners and structural load-bearing performance.

[0015] 3. This invention addresses the buckling load optimization problem under mass constraints by employing a penalty weight adaptive adjustment mechanism to cleverly achieve a balance between global exploration and local convergence, increasing the probability of finding a better global solution while making the entire optimization process smoother and more stable.

[0016] In summary, the parametric modeling and stiffening optimization method for curved stiffened cylindrical shells under local axial compression proposed in this invention provides ideas for the design of curved stiffened cylindrical shells under local axial compression. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the implementation of the parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression, as provided in this embodiment of the invention. Figure 2 A schematic diagram illustrating the manifestation of curved stiffening on a 1 / 8 cylindrical shell according to an embodiment of the present invention; Figure 3 A schematic diagram illustrating the correspondence between the distribution function and the node layout provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the curved path distribution design on a cylindrical shell provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of a graphical representation method for the structural features of a reinforced cylindrical shell provided in an embodiment of the present invention; Figure 6 A schematic diagram of the load and boundary conditions of the finite element model for buckling analysis of a locally axially compressed cylindrical shell provided in an embodiment of the present invention; Figure 7 A schematic diagram of a buckling load prediction model for a stiffened cylindrical shell based on a convolutional neural network, provided in an embodiment of the present invention; Figure 8 This diagram illustrates the process of optimizing the reinforcement layout of the cylindrical shell according to an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. The accompanying drawings are only schematic representations related to specific embodiments and do not represent the entirety of the invention.

[0020] Figure 1 The flowchart illustrates the implementation of the parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression, as provided in this embodiment of the invention.

[0021] Example 1: The parametric modeling and stiffening optimization method for a curve-stiffened cylindrical shell under local axial compression provided in this invention includes the following steps: Step 1: Determine the stiffened shell skin geometric parameters and structural material parameters as basic parameters, and the load boundary condition parameters, the path parameters of the curved stiffening, and the geometric parameters of the stiffened section as main parameters; in this embodiment, the stiffened shell skin geometric parameters include material density. ρ =2680kg / m 3 Inner diameter of cylindrical shell R =100mm 、 high H =100mm 、 Wall thickness t= 1mm; Structural material parameters include elastic modulus E =70GPa, Poisson's ratio ν =0.3, yield strength R eL =260MPa; Load boundary condition parameters include the number of load regions. N= 4. Angle of a single load area θ =30°; the path parameters for curve reinforcement include the number of circumferential reinforcements in the design area. n 1. Number of axial reinforcement bars in the design area n 2. Six sets of exponential distribution function parameters a , b The geometric parameters of the reinforced section include the height of the stiffeners. h rib thickness t s The exponential distribution function of the reinforced curve path is as follows: in a , b The parameter is a function parameter, and the sampling range is set in the interval [-1, 1], with a maximum difference of 0.5 between them.

[0022] This embodiment mainly optimizes the path parameters and geometric parameters of the stiffened section for curved reinforcement to obtain better structural buckling bearing capacity. The value range of the design parameters for the reinforcement scheme is shown in the table below: Step 2: Curve stiffening is divided into two types: circumferential and axial. Each stiffener is determined by only three control points: the start point, the midpoint, and the end point of the curve. The same control points of the same stiffener curve are on a straight line. The distribution of the same control points on a straight line is controlled by an exponential distribution function. Based on the symmetry of the local axial compression structure, the side surface of the cylindrical shell skin is unfolded and divided into eight similar rectangular design regions according to the location of the local axial compression. Within each rectangular design region, the path definition for the curved stiffening is as follows: Figure 2 As shown. On a 1 / 8 cylindrical shell, the curve stiffening methods are divided into circumferential and axial types. Each stiffener is controlled by only three pathway points: the start point, midpoint, and end point of the curve, anchoring the basic direction and curvature of the curve. The distribution of the same type of control points on any straight line is controlled by an exponential distribution function. Mapping the function to the distribution of points, it follows a monotonically increasing function. f ( x For example, when there are a total of 1,000 people in the design area n When there are multiple identical curved bars, the distribution of the same control points (one of the start point, end point, and midpoint) of the same curved bars can be obtained according to the following formula: in L i It is the first of the same type of control point i The and the first i -1 The proportion of the distance between control points to the total design length, such as Figure 3 As shown.

[0023] Step 3: With all control points determined, a quadratic B-spline curve is used to draw the stiffening path. Based on the main parameters, a three-channel image containing the stiffening path information and the geometric information of the stiffening section is formed. The first two channels contain the stiffening path image, and their gray values ​​correspond linearly to the stiffening thickness and stiffening height, respectively. The larger the size value, the darker the channel color. The third channel contains only a blank image. The initially formed three-channel image is downsampled to obtain two three-channel images with different resolutions, which together constitute a multi-channel structural image of the stiffened curve. When generating a stiffened cylindrical shell example, two types of curves are used: axial and circumferential. For each curve, three sets of nodes (start, midpoint, and end) are required, totaling six sets of nodes. After determining the start, midpoint, and end point of a curve, a quadratic B-spline curve is used for fitting. Particularly for the circumferential curve, to improve accuracy, based on the symmetry of the local axially compressed structure, the rectangular design area is extended along the circumference. Specifically, a similar rectangular design area is added on both the left and right sides, increasing the number of control points from three to seven. The control point positions of the added left rectangular design area are symmetrical about the left boundary of the original rectangular design area, and the control point positions of the added right rectangular design area are symmetrical about the right boundary of the original rectangular design area. For example, when the number of circumferential stiffeners... n 1=10, number of axial reinforcing bars n 2=11, circumferential stiffeners from left to right, each group of nodes a , b The values ​​are [-0.17, -0.28], [0, 0], and [-0.27, 0.14], respectively, representing the axial stiffeners from top to bottom at each group of nodes. a , b When the values ​​are [-0.02, -0.43], [0.1, -0.4], and [0.74, 0.77], the curve group distribution design is as follows: Figure 4 As shown.

[0024] In the visualization of the stiffened structure, RGB three-channel images were used. The two different resolution images had pixel sizes of 256*256*3 and 128*128*3, respectively, with the latter obtained by downsampling the former. The image shows a 1 / 8 cylindrical shell region, with the stiffening paths scaled proportionally according to their distribution. The image information includes curved stiffening path information and curved stiffening section geometry information. The first two channels of the image contain the curved stiffening path image, and their grayscale values ​​are respectively related to the stiffener thickness. t s and rib height h The values ​​of [value] and [value] correspond linearly. The range of rib height and thickness corresponds to the range of line opacity [0.2, 1.0], respectively. The larger the value, the darker the color. The third channel contains only a blank image. Two three-channel images with different resolutions together constitute the structural image of the ribbed curve, such as [image of the ribbed curve]. Figure 5 As shown.

[0025] Step 4: Based on the basic and main parameters, a finite element analysis model of a locally axially compressed thin-walled stiffened cylindrical shell is established using the mesh curling method. The critical buckling load and stiffener mass of the structure are obtained through analysis and used as two labels for the corresponding multi-channel structural images. Based on geometric parameters and structural images, stiffening ribs were first modeled on the design area plane using Abaqus software. The Wrapmesh plugin was then used to implement a mesh rollup method, rolling the ribs into a cylindrical shape. The stiffened structure was further simplified by generating the remaining structure through symmetry, arraying, and copying of the design area, thus improving modeling efficiency. During modeling, the relationship between the ribs and the cylindrical shell was set as a binding constraint, defined through a node set. A reference point RP was set at the top of the cylindrical shell and coupled to the axial compression loading area. Axial compression was applied by setting the axial displacement of the reference point. Furthermore, three translational degrees of freedom (U1, U2, U3) and three rotational degrees of freedom (UR1, UR2, UR3) at the bottom boundary of the cylindrical shell were constrained, as were the other degrees of freedom in the loading area besides the axial translational degree of freedom. Figure 6 As shown. The mass of the stiffeners was obtained using Abaqu software, and the critical buckling load of the structure was calculated using the ABAQUS static general analysis module. The Newton-Raphson iterative algorithm was used to solve the nonlinear buckling problem of the cylindrical shell, and a damping factor of 1e-6 was applied. The model mesh was selected using S4R shell elements, and the element size was adjustable.

[0026] Step 5: Using labeled multi-channel structural images as a sample set, build two convolutional neural network models based on the encoder part of the autoencoder model to predict the critical buckling load and stiffener mass of the stiffened structure, respectively. For a convolutional neural network model used to predict the critical buckling load / stiffening mass of a stiffened structure, two multi-channel images of different resolutions are convolved and pooled respectively, then unfolded into two sets of 1-dimensional long vectors, which are then concatenated. The concatenated long vectors are then input into a fully connected layer to obtain the prediction result.

[0027] The structure and hyperparameters of the convolutional neural network model can be adjusted according to requirements. The model structure established in this embodiment is as follows: Figure 7 As shown, the buckling dataset is divided into training and test sets with a batch size of 64. The loss is calculated using mean squared error (MSE), and the model parameters are updated using the Adam optimizer with a learning rate of 0.001.

[0028] Step 6: Combining the established convolutional neural network model and optimization algorithm, realize the optimized design of the curve stiffening structure layout under local axial compression; A meta-heuristic algorithm is used for optimization, and a genetic algorithm optimization framework is built. In this embodiment, the optimization objective is to find the stiffening design scheme that maximizes the critical buckling load of the stiffened cylindrical shell under the constraint of stiffener mass. The mass constraint is set to 270g, and the fitness and dynamic penalty factor during the optimization process can be obtained from the following formulas: in, Fitness For the fitness function, N To predict the critical buckling load of a structure using a convolutional neural network model, λ As a dynamic penalty factor, P This is a quality penalty item. λ min This is the minimum penalty factor (10 in this example). λ max This is the maximum value of the penalty factor (10000 in this example). k To optimize the rounds, k max This is the maximum number of optimization attempts (200 in this example). m To predict rib quality using a convolutional neural network model, M The quality constraint is 270g in this example. The genetic algorithm uses a crossover rate of 0.5 and a mutation rate of 0.2. Each generation generates a population of 200 individuals, iterates 200 times, and records the best individual in each generation. The optimization process for the reinforced layout is as follows: Figure 8 As shown, as the number of optimization rounds increases, the buckling load approaches convergence after 30 rounds, indicating that the optimization method has a relatively fast convergence speed.

[0029] This invention provides a parametric modeling and stiffening optimization method for curved stiffened cylindrical shells under local axial compression. Addressing the complex design space of curved stiffening in cylindrical shells, it establishes the characteristic equation of the stiffening curve using fewer stiffening path design variables, thus completing the parametric modeling of the cylindrical shell and curved stiffening. This significantly improves efficiency, saves the cost of manual model adjustments, and avoids oversights in manual modeling. Different stiffening information under local axial compression, including the distribution of the curve path and the geometric cross-section of the stiffeners, is represented in RGB image form. Using this image as input, a unified multi-variable, multi-scale representation is achieved. A convolutional neural network model is used to predict the load-bearing capacity of the structure. The training and test sets R of the constructed convolutional neural network model respond to stiffener mass and buckling characteristic values ​​are provided. 2 All values ​​are above 0.97. The optimal buckling load for the curve stiffener is 213.97 kN, which is about 15% higher than the best performance in the initial sample set.

[0030] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is coupled to the processor; the memory is used to store program data, and the processor is used to execute the program data to implement the above-described parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the above-described parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression.

[0032] The above is a further detailed description of the present invention in conjunction with preferred embodiments, and is not intended to limit the present invention. It should be noted that for those skilled in the art, any simple deductions and optimizations made to the present invention based on the core ideas of the present invention should be considered within the protection scope of the present invention.

Claims

1. A parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression, characterized in that, Includes the following steps: S1. Determine the geometric parameters and structural material parameters of the stiffened shell skin as the basic parameters, and the load boundary condition parameters, the path parameters of the curved stiffening, and the geometric parameters of the stiffened section as the main parameters; S2. Curve stiffening forms are divided into two types: circumferential and axial. Each stiffener is determined only by three control points: the start point, the midpoint, and the end point of the curve. Furthermore, the same control points of the same stiffener curve lie on a straight line, and the distribution of the same control points on a straight line is controlled by an exponential distribution function. S3. With all control points determined, a quadratic B-spline curve is used to draw the stiffening path. Based on the main parameters, a three-channel image containing the stiffening path information and the geometric information of the stiffening section is formed. The first two channels contain the stiffening path image, and their gray values ​​correspond linearly to the stiffening thickness and stiffening height, respectively. The larger the size value, the darker the channel color. The third channel contains only a blank image. The initially formed three-channel image is downsampled to obtain two three-channel images with different resolutions, which together constitute a multi-channel structural image of the stiffened curve. S4. Based on the basic and main parameters, a finite element analysis model of a locally axially compressed thin-walled stiffened cylindrical shell is established using the mesh curling method. The critical buckling load and stiffener mass of the structure are obtained through analysis and used as two labels for the corresponding multi-channel structural images. S5. Using labeled multi-channel structural images as a sample set, two convolutional neural network models are built to predict the critical buckling load and stiffener mass of the stiffened structure, respectively. S6. Based on the established convolutional neural network model, implement the optimization design of the layout of the curve-stiffened structure under local axial compression. The optimization goal is to find the stiffening design scheme with the maximum critical buckling load of the stiffened cylindrical shell under the constraint of stiffener mass. During the optimization process, a dynamic penalty factor that increases with the number of iterations is set.

2. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 1, characterized in that, In step S1, for a cylindrical shell with a locally distributed axial compressive load that is equidistantly symmetrical, the geometric parameters of the stiffened shell skin include the inner diameter of the cylindrical shell. R、 high H、 Wall thickness t Structural material parameters include material density. ρ Elastic modulus E Poisson's ratio ν Yield strength R eL Load boundary condition parameters include the number of load regions. N Angle of a single load area θ ; The path parameters for curved reinforcement include the number of circumferential reinforcements in the design area. n 1. Number of axial reinforcement bars in the design area n 2 and six sets of exponential distribution function parameters a , b The geometric parameters of the reinforced section include the height of the stiffeners. h rib thickness t s The exponential distribution function of the reinforced curve path is as follows: in a , b The parameter is a function parameter, and the sampling range is set in the interval [-1, 1], with the difference between the two parameters being less than or equal to 0.

5.

3. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 2, characterized in that, In step S2, based on the symmetry of the local axial compression structure, the side of the cylindrical shell skin is unfolded and the load area is increased by 2 times. N The design area is divided into equal parts to form a single rectangular design area. Within the rectangular design area, the starting point, ending point, and midpoint of the circumferential curve are distributed on the left and right boundaries and the vertical axis of symmetry, respectively, while the starting point, ending point, and midpoint of the axial curve are distributed on the upper and lower boundaries and the horizontal axis of symmetry, respectively.

4. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 3, characterized in that, When there are multiple rectangular design areas n When there are multiple identical curved bars, the distribution of the same control points for the same type of curved bars is obtained according to the following formula: in L i It is the first of the same type of control point i The and the first i -1 The proportion of the distance between control points to the total design length.

5. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 1, characterized in that, In step S3, for the circumferential curve, based on the symmetry of the local axial compression structure, the rectangular design area is extended along the circumference. Specifically, a rectangular design area is added on each of the left and right sides, increasing the number of control points from three to seven. The control point positions of the added left rectangular design area and the original rectangular design area are symmetrical about the left boundary of the original rectangular design area, and the control point positions of the added right rectangular design area and the original rectangular design area are symmetrical about the right boundary of the original rectangular design area.

6. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 1, characterized in that, In step S3, during the visualization of the stiffened structure, an RGB three-channel image is used. The initially formed three-channel image is a high-resolution image, which is used to capture the fine geometric features of the stiffeners. The high-resolution image is downsampled to obtain a low-resolution image, which is used to extract the overall layout of the structure. The two three-channel images with different resolutions together constitute a multi-channel structural image of the curve stiffener.

7. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 1, characterized in that, In step S6, a meta-heuristic algorithm is used for optimization. The optimization goal is to find the stiffened design scheme that maximizes the critical buckling load of the stiffened cylindrical shell under the constraint of stiffener mass. By setting a dynamic penalty factor that increases with the number of iterations, the optimization process is divided into two stages: In the early stage of optimization, the main optimization goal is to maximize the buckling load and explore design regions that slightly exceed the mass limit but have great load-bearing potential, so as to avoid being constrained too early; In the later stage of optimization, the optimization results are forced to converge to the feasible region that satisfies the mass constraint, and a refined search is performed.

8. The parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression according to claim 7, characterized in that, In step S6, the formulas for fitness and dynamic penalty factor during the optimization process are as follows: in, Fitness For the fitness function, N To predict the critical buckling load of a structure using a convolutional neural network model, λ As a dynamic penalty factor, P This is a quality penalty item. λ min To minimize the penalty factor, λ max The maximum value of the penalty factor. k To optimize the rounds, k max To maximize the number of optimization attempts, m To predict rib quality using a convolutional neural network model, M For quality constraints.

9. An electronic device comprising a memory and a processor, characterized in that, The memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the parametric modeling and stiffening optimization method for a curved stiffened cylindrical shell under local axial compression as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the parametric modeling and stiffening optimization method for the curve-stiffened cylindrical shell under local axial compression as described in any one of claims 1-8.

Images