Independent jacket modeling optimization method based on gradient descent method
By combining the divide-and-conquer method and the gradient descent method, the problem of the reliance on human experience in the jacket structure model was solved, and the rapid and efficient structural optimization was achieved, generating multiple optimized jacket models.
Patent Information
- Application Number
- CN202511793549.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-02
- Publication Date
- 2026-03-17
AI Technical Summary
In the existing technology, the quality of the jacket structure model depends on the experience of the engineers, which leads to long modeling time and low efficiency, making it difficult to establish a relatively optimized structural model in a short period of time.
An autonomous modeling and optimization method combining divide-and-conquer and gradient descent is adopted to achieve autonomous optimization of the jacket structure through geometric topology optimization, model techno-economic evaluation, and iterative optimization design of structural dimensions.
Several optimized jacket structure models were generated within an hour, significantly improving modeling efficiency and quantifying the degree of model optimization.
Smart Images
Figure CN121683221A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of artificial intelligence, and in particular to a multi-scheme, autonomous optimization modeling method for optimizing the foundation structure of offshore wind turbine jackets. Background Technology
[0002] The jacket structure is a common type of foundation structure in marine engineering, used to support the superstructure and transfer its load to the soil. Under the premise of meeting specifications, engineers build a jacket structure model based on their experience, then use structural analysis software to calculate and analyze the model. Based on the results, the topology and dimensions of the model are adjusted, and the model is analyzed again using the same software. This process is iterated until a model with relatively optimal technical and economic efficiency is obtained. The quality of the jacket structure model depends on the experience level of the engineers and the timeframe of the task.
[0003] Gradient descent is used to find the extrema of a function. It is a first-order optimization algorithm that can drive the model to converge iteratively. Summary of the Invention
[0004] Technical problems to be solved
[0005] The technical problem to be solved by this invention is how to establish several optimized jacket models in a short period of time, taking into account the environmental load conditions of the marine engineering site.
[0006] Technical solution
[0007] This invention provides an autonomous modeling and optimization method for jacket structures based on the divide-and-conquer and gradient descent methods. It comprises three aspects: First, geometric topology optimization design. Combining geometric termination conditions, a divide-and-conquer method is employed, recursively calling the divide-and-conquer program to optimize the geometric topology, and introducing a minimum design layer number for the jacket to simplify the optimization computation. Second, design of the model's techno-economic evaluation equations. The target value range is discretized and classified, establishing the model's equivalent techno-economic value. Combining the equivalent techno-economic value with the number of critical failure components, a techno-economic evaluation of the constructed jacket model is performed. Third, autonomous iterative optimization design of structural dimensions. Adhering to the design concept of moderate structural redundancy, targeted optimization strategies are adopted for different types of structures, and the gradient descent method is used to achieve relative convergence of the iterative results.
[0008] In some embodiments of the present invention, the geometric topology optimization design includes:
[0009] First, the geometric constraints are determined, and based on these constraints, the minimum number of design layers is calculated. Layer-by-layer optimization of the geometric topology is then performed to achieve overall optimization. Specifically, a divide-and-conquer approach is adopted, decomposing the optimization objective into the sum of the current layer's optimization value and the remaining n-1 layers' optimization values. The optimization calculation for the remaining n-1 layers recursively calls the ontology function until only 2 layers remain.
[0010] In some embodiments of the present invention, the design of the model evaluation equation includes:
[0011] Secondly, the ratio of structural stress to its bearing capacity is reflected by a techno-economic value. If the techno-economic value is greater than 1, it indicates that the structural components have exceeded their bearing capacity and have failed under stress. Therefore, two indicators are designed to evaluate the model: one is the number of failed components, N. broken The other is the comprehensive average of the representative values of technology and economy, N. normal First, the techno-economic values are discretized into six intervals, representing structural stress levels of no stress, slight stress, moderate stress, normal stress, high stress, and failure due to stress. The midpoint of the techno-economic values for each interval is used to represent the techno-economic value for that interval. The proportion of structural components in each interval is calculated as the frequency. A weighted average of the representative values for each interval is then taken to obtain the comprehensive average N of the representative values. normal This is called the model equivalent techno-economic value. Only models with zero damaged components are considered qualified models. The quality of a model is reflected by its equivalent techno-economic value. A value that is too low indicates that the model has too much structural redundancy and is uneconomical, while a value that is too high indicates that the redundancy is too low and is unsafe.
[0012] In some embodiments of the present invention, the autonomous iterative optimization design of structural dimensions includes:
[0013] Furthermore, considering setting the overall target techno-economic value to 0.85, the structure has suitable redundancy, achieving both economy and safety. The differences between the current and target techno-economic values of each component are analyzed, and the average proportion of these differences is set as the descent gradient. The structural dimensions are adjusted accordingly, and iterative calculations are performed to ultimately obtain several jacket structure models with no structural component failures and excellent overall economic performance.
[0014] Beneficial effects
[0015] This invention provides an autonomous modeling and optimization method for duct stents based on gradient descent and autonomous iterative optimization. Compared with existing technologies, it has at least the following two advantages: First, this invention can establish several relatively optimized duct stent structure models within hours, significantly improving the efficiency of structural modeling and optimization calculations. Second, the model evaluation equation designed in this invention can quantitatively evaluate the optimization degree of the duct stent structure model. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the overall process of the present invention.
[0017] Figure 2 This is a schematic diagram of the guide frame components.
[0018] Figure 3 This is a schematic diagram of the two-layer diagonal bracing structure of the present invention.
[0019] Figure 4 This is a schematic diagram of the divide-and-conquer method for multi-layered diagonal bracing structure calculation of the present invention. Detailed Implementation
[0020] This invention provides an autonomous modeling and optimization method for jacket structures based on divide-and-conquer and gradient descent. Through geometric topology optimization design, the geometric topology of the model is determined. Based on this, an initial model M0 is established by arbitrarily assigning values to the structural dimensions. Stress analysis is performed on the initial model to obtain the technical and economic values of each component. Using a pre-determined discretization and classification method for these technical and economic values, the frequency of each interval is calculated. Then, a weighted average of the number of damaged components and representative values is calculated. For components whose technical and economic values differ significantly from the target values, the dimensions are adjusted using gradient descent, thereby generating a new model. Iterative calculations are performed on the new model to generate a series of new models. Several models without structural component failure are selected, thus realizing the function of autonomous modeling and optimization of jacket structures.
[0021] To make the technology, objectives and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and the accompanying drawings.
[0022] This invention provides an autonomous modeling and optimization method for duct stent structures based on the divide-and-conquer method and gradient descent method. Figure 1 This is a schematic diagram illustrating the implementation process of the present invention. The process mainly includes three stages:
[0023] Phase one involves creating a geometric model of the jacket structure through geometric topology optimization design. The height of the jacket structure is calculated based on environmental data such as water depth. Figure 2The schematic diagram shows the components of the jacket structure. The geometric topology of the columns, transition sections, and tower base is relatively clear and has almost no variability. Therefore, the optimization of the geometric topology focuses on the optimization of the diagonal braces. Thus, the geometric topology optimization problem is simplified to the diagonal brace length optimization problem. Furthermore, considering the symmetry of the diagonal braces on each side of the jacket structure, the total diagonal brace length optimization problem is simplified to the total length optimization problem of the diagonal braces on a certain side of the jacket structure. A divide-and-conquer method is used to calculate the total diagonal brace length, decomposing it into the length of the diagonal brace on the nth layer of the jacket structure and the sum of the lengths of the diagonal braces on the remaining n-1 layers. When there are only 2 layers of diagonal braces, the total length calculated for 2 layers is returned. For the calculation of the diagonal brace length for 2 layers of the jacket structure, a numerical calculation method is used to calculate the total length of the diagonal braces at different layer positions sequentially, thereby solving for the shortest length. Figure 3 This is a schematic diagram of a two-layer diagonal bracing structure. Figure 4 This is a schematic diagram illustrating the divide-and-conquer method for calculating multi-layered braced structures. The calculation formula for multi-layered braced structures using the divide-and-conquer method is:
[0024]
[0025] Phase Two: Creating an initial jacket structure model and conducting model evaluation. Based on the jacket geometric model created in Phase One, arbitrary initial values are assigned to the jacket structure to obtain the initial jacket structure model. Environmental data and load data are input to obtain the initial jacket model M0. Structural strength analysis is performed on the initial jacket model to obtain the analysis results of each component. The analysis results are processed to form technical and economic values. The technical and economic values are discretized into 6 segments: [0,0.1], (0.1,0.3], (0.3,0.75], (0.75,0.95], (0.95,1.0], and (1.0,+∞), representing almost no stress, excessive redundancy, moderate redundancy, normal, low redundancy, and structural failure, respectively. The midpoint value of the technical and economic values of each segment represents the technical and economic value of that segment. The frequency of structural components in each segment is statistically analyzed to obtain the comprehensive average of representative values N. normal This is named the model equivalent techno-economic value. The number of structural members whose techno-economic value lies in the segment (1.0, +∞) is extracted, representing the number of members that fail under stress, N. broken . Use U i Indicates the representative value of the segment, α i Let represent the proportion of each segment, then the model evaluation equation is:
[0026]
[0027] Phase Three involves creating several jacket optimization models through autonomous optimization design of structural dimensions. Using the gradient descent method, a target technical-economic value of 0.85 is set. The current technical-economic value of a component is divided by the arithmetic mean of this value and the target technical-economic value to obtain the structural dimension adjustment gradient for that component in the current iteration. The component dimensions are then adjusted, and different adjustment methods are used for different components, taking into account their differences.
[0028] Table 1. Component Iterative Adjustment Methods
[0029]
[0030] Each adjustment generates a new jacket structure model. Stress analysis is performed on this new model, followed by a technical and economic evaluation. The dimensions of each component are then adjusted to generate another new jacket structure model. This process is iterated repeatedly to obtain a series of new models. Based on the model evaluation criteria described in Stage Two, several optimized models are selected.
[0031] This invention can generate several optimized duct stent structure models within hours. The models are self-optimized without human intervention, which greatly improves the modeling efficiency of duct stent structure models.
[0032] The embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. It should be noted that implementations not illustrated or described in the drawings or the main text of the specification are forms known to those skilled in the art and are not described in detail. Furthermore, the definitions of the elements and methods above are not limited to the various specific structures, shapes, or methods mentioned in the embodiments; those skilled in the art can easily modify or substitute them. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Moreover, unless specifically described or steps must occur sequentially, the order of the above steps is not limited to those listed above and can be varied or rearranged according to the desired design. Furthermore, the above embodiments can be used in combination with each other or with other embodiments based on design and reliability considerations; that is, technical features from different embodiments can be freely combined to form more embodiments. The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. The application provides a kind of self-optimization method of self-optimization of gradient descent method, autonomous iteration for the self-optimization of gradient descent method, in the model of jacket, optimization objective and constraint condition are established, in the optimization process, the method of model parameter self-optimization and model evaluation equation are established;Said optimization objective is simplified, and is divided into two parts of geometric topology optimization and structural size optimization.
2. In the geometric topology optimization of claim 1, the minimum design layer of the jacket is calculated by combining the strut angle constraint condition, further, the divide and conquer method is used to solve the geometric topology optimization solution by recursive call of divide and conquer program;Finally, the geometric topology optimization of claim 1 is applied to the jacket model.
3. In the structural size optimization of claim 1, the model parameter self-optimization objective of claim 1 is established, the target value range is discretized and classified according to the concept of moderate redundancy design, different classification adopts targeted optimization strategy, and gradient descent method is used to realize relative convergence of iteration result;On the model technical and economic evaluation equation of claim 1, the equivalent technical and economic representative value of the model is established, combined with the equivalent technical and economic representative value and the critical failure number of component, the technical and economic evaluation of the built jacket model is carried out, and finally the model of claim 1 is self-optimized.
4. The jacket model established by the optimization method of claim 1, claim 2 and claim 3 is open and diverse, providing users with a variety of choices.