A data-driven hierarchical decoupling design method for complex structural systems

Abstract

The present invention discloses a data-driven hierarchical decoupling design method for complex structural systems, including: performing performance demand analysis on the complex structural system and determining performance indicators; performing hierarchical decoupling analysis based on the structure and performance indicators of the complex structural system and establishing a system hierarchical model; completing the hierarchical parameter definition of the hierarchical complex structural system, and forming a hierarchical structural design system with mutual correlation between levels through design variables, performance characterization parameters and transition variables; on this basis, in a "top-down" manner, identifying the specific correlation pattern between levels through data drive, and performing sequential design on each level; then carrying out design synthesis of each level, and evaluating the system performance and reliability to complete the design of the system. The present invention establishes the correlation with the sub-level parameters through data learning at each level, forming a correlation relationship between designs of different levels, with the goal of ensuring overall performance, forming a "top-down" overall efficient design method.

Description

Technical Field

[0001] The present invention belongs to the technical field of complex equipment system design, and in particular relates to a data-driven hierarchical decoupling design method for complex structural systems. Background Art

[0002] Complex structural systems are composed of numerous interconnected and interacting components, such as wind turbine structures, offshore oil and gas platform structures, automobile bodies, ship structures, aircraft structures, and shield machine structures. The design of complex structural systems is crucial to their ultimate performance and reliability. Complex structural systems are often constructed from a large number of components of varying shapes through welding, bolting, bonding, and riveting. These components are strongly coupled and collectively determine the overall structural performance. Through years of accumulated technology and experience, the design of complex structural systems has developed a specific process.

[0003] Taking the automotive body structure system as an example, the same part has different modes and degrees of correlation with different subsystem functions (such as frontal or side impact safety). Here, the correlation mode refers to the mechanism by which different parts interact to determine the subsystem's function. Furthermore, different design parameters of the same part, such as sheet thickness and local curvature, have different modes and degrees of correlation with the specific functions of their subsystem. Here, the correlation mode refers to the mechanism by which different structural or geometric parameters jointly influence the specific performance parameters of the component. Finally, after the component level, further subdivision can be performed depending on the actual situation. For example, after the macro-level thickness and geometric parameters of carbon fiber-reinforced polyurethane panels, the next level can be further subdivided: carbon fiber layup and content. These different levels, from the underlying micro-material design and performance, part geometry design and performance, subsystem design and performance, and system design and performance, interact with each other to jointly determine the performance of the entire system. To meet the different system performance requirements of complex structural systems, the interaction mechanisms at each level vary, including the modes and degrees of correlation between sub-levels, forming the design problem of complex hierarchical structural systems.

[0004] Traditional approaches analyze the performance requirements of complex systems and employ a "manual iteration" of trial and error to design subsystems and components into the entire system. Engineers, based on experience or regulatory requirements, conduct component design, including material selection, component design, integrated design, and performance testing. They then iterate based on component performance test results. Subsystem performance is evaluated to determine if component designs need adjustment; and system performance is evaluated to determine if subsystem designs need adjustment. This process is repeated until system performance meets design or regulatory requirements. However, this multi-level, "bottom-up," "manual iteration" design process, from components to the entire system, is inefficient. Furthermore, the resulting design is sensitive to component parameters, performance, and subsequent manufacturing processes. Due to the uncertainty of these factors, the final product may not meet established performance requirements, resulting in poor robustness. Summary of the Invention

[0005] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the existing technology and provide a data-driven hierarchical decoupling design method for complex structural systems. Through data learning, the correlation between parameters at different levels is established, and the correlation relationship between designs at different levels is formed, with the goal of ensuring overall performance, forming a "top-down" overall design method.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a data-driven hierarchical decoupling design method for complex structural systems, comprising the following steps:

[0008] Perform performance requirement analysis on complex structural systems and determine performance indicators; the performance indicators include performance evaluation indicators and labels. The performance evaluation indicators are used to evaluate system-level performance. The labels refer to the evaluation results of the performance evaluation indicators judged by the evaluation criteria, including feasible solution labels and infeasible solution labels.

[0009] Based on the structure and performance indicators of the complex structural system, a hierarchical decoupling analysis is performed to establish a system hierarchical model. The system hierarchical model includes multiple sub-levels and defines the basic parameters of each sub-level in the model. The basic parameters include design variables, performance characterization parameters, and transition variables. The design variables are used to represent the basic design elements of this level. The performance characterization parameters are used to characterize the performance of each sub-level that affects the performance indicators of this level. The transition variables are used to facilitate the current level to regulate the performance characterization parameters of the sub-level.

[0010] According to the design goal, the hierarchical elements of the non-lowest sub-level are digitally simulated given the design variables and transition variables to obtain the performance evaluation indicators and characterization performance parameters of the current level, and the performance evaluation indicators are judged according to the given evaluation criteria. Different design variables and characterization performance parameters are labeled to obtain a data set, which includes different design variables and characterization performance parameters and corresponding labels. The data label entropy reduction method is used to mine the key design variables and their value ranges that meet the design goals and the key characterization performance parameters and their value ranges. The key design variables and their value ranges are the current level design set; the key characterization performance parameters and their ranges are passed to the next level and used as one of the judgment constraints for the feasible solution or infeasible solution of the next level; the above process is repeated for each non-lowest sub-level to complete the design of each level; the lowest sub-level data is mined to obtain the design variables and their value ranges. The design variables and their value ranges are the current level design set;

[0011] Integrate the design sets obtained at each level to obtain the system design set, construct and connect components according to the design variables in the system design set, obtain the complex structure system design simulation model, and use performance indicators to verify the performance of the complex structure system simulation model.

[0012] As a preferred technical solution, the performance requirement analysis of the complex structural system includes the following steps:

[0013] Conduct qualitative and quantitative analysis of the performance requirements of complex structural systems based on prior knowledge, and determine various performance evaluation indicators and their feasible standards;

[0014] The labels are set based on an indication of whether the feasibility criteria of the evaluation metric are met.

[0015] As a preferred technical solution, the hierarchical decoupling analysis based on the structure and performance indicators of the complex structural system includes the following steps:

[0016] A complex structural system is structurally analyzed according to performance indicators, and the complex structural system is divided into levels from upper to lower levels according to the structural analysis results to construct a system hierarchical model; the first level of the system hierarchical model represents the uppermost complex structural system as a whole, and the bottom level represents the lowest level component or material design, and its specific design scale is determined according to actual conditions.

[0017] As a preferred technical solution, the design objectives include:

[0018] For non-bottom-level levels, the goal of the hierarchical design is to obtain the design subdomains The design subdomain needs to meet the feasibility probability of the performance index (p(ψ(′g′)|γ c )) and the robustness of performance indicators (Φ(ψ(′g′)|γc )) are maximized, where 'g' represents the feasible solution label; set the design variable x and the transition variable x t The constraint interval is as follows:

[0019] x l <x<x u

[0020] x tl <x t <x tu

[0021]

[0022] Among them, x l represents the upper limit of the design variable x, x u represents the lower limit of the design variable x, x tl Represents the transition variable x t The upper limit of x tu Represents the transition variable x t The lower limit of Represents a design constraint.

[0023] As a preferred technical solution, the marking method comprises the following steps:

[0024] According to the set design goals, the system response under the combination of given design variables and performance parameters determined by transition variables is determined. Only when the performance parameter range passed from the previous level is met and other performance evaluations meet the design goals, the feasible solution label 'g' is output, that is, the design variable is a feasible solution. Otherwise, the infeasible solution label 'p' is output, that is, the design variable is an infeasible solution. The data set of design variables, performance parameters and their corresponding labels at this level is obtained.

[0025] As a preferred technical solution, the design objectives include:

[0026] For the lowest sub-level, the goal of hierarchical design is to obtain the design subdomain <x> of this level. The probability (p(ψ(′g′)|x)) and robustness (Φ(ψ(′g′)|x)) of the design subdomain to meet the performance index are maximized, where ′g′ represents the feasible solution label. The constraint interval of the design variable x is set as follows:

[0027] x l <x<x u

[0028]

[0029] Among them, xl represents the upper limit of the design variable x, xu represents the lower limit of the design variable x, Indicates that the performance of the design components at this level meets the requirements of the previous level transferred to this level. Represents a design constraint.

[0030] As a preferred technical solution, the transfer of the performance characterization parameters and their ranges to the next level includes:

[0031] The subset of characterization performance parameters that meet the set design conditions is passed to the next level, and the next level uses the subset of characterization performance parameters as constraints. When the performance index of the hierarchical component meets the evaluation standard ψ('g'), and its characterization performance parameters meet the characterization performance parameters passed down from the level When , the solution marked as 'g' is feasible.

[0032] As a preferred technical solution, the dataset label entropy reduction method identifies the design variables that meet the performance indicators and the respective subdomains that characterize the performance parameters, including the following steps:

[0033] The dataset S obtained by simulation is divided into subsets based on the entropy reduction model. The dataset S includes the characteristic parameters of all samples at the current level, that is, the design variables and performance parameters, as well as the corresponding labels. The dataset is divided by selecting the characteristic parameter A with the maximum information gain GR(A) value, as shown in the following formula:

[0034]

[0035] Among them, G(A) is the information gain, that is, the entropy of the current data set S The entropy after dividing the data set using the characteristic parameter A The difference between the two, SI(A) is the information increment of the data set partition, which means the amount of information generated by partitioning the data set by A, m is the number of labels, p i is the proportion of the number of designs of the i-th label to the total number, N is the number of data subsets generated after the data set is divided, S j is the subset generated after splitting the data set S, For subset S j The information entropy of |·| represents the number of samples in the sample set;

[0036] By following the above indicators, the set S is recursively divided until the set conditions are met and the division is stopped to obtain the recursive division result;

[0037] The recursive partitioning results are analyzed to obtain the subset with the most labels 'g' and the highest purity, and all key design variables and their value ranges and key performance parameters and their value ranges used to obtain the subset are recorded.

[0038] As a preferred technical solution, if the characteristic parameter A is a discrete variable, then N is the number of discrete values; if the characteristic parameter A is a continuous variable, then N is 2.

[0039] As a preferred technical solution, it is characterized in that, in the process of recursively dividing the current set D, the subset S j The division stops when one of the following conditions (1) to (3) is met:

[0040] (1) Subset S j The label purity p i (S j ) reaches a preset value;

[0041] (2)|S j | Reaching the preset minimum value;

[0042] (3) The current dataset S is split a predefined number of times.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] In order to solve the problem that the design process of traditional complex structural system from parts to systems is "bottom-up", which makes the design of components fail to fully consider the overall performance of the system, and the design efficiency caused by the "trial and error" iteration method, a hierarchical design method for complex structural systems has been developed. First, the present invention adopts the "top-down" design idea to establish a systematic hierarchical design method. By characterizing the performance parameters The correlation between different levels is achieved so that the lower level design takes into account the performance requirements of the upper level and establishes a method system with clear system goal orientation. Secondly, a data set partitioning method with decreasing information entropy is proposed, with the goal of maximizing the purity of the sub-data set labels to obtain the design variables x and the performance parameters. Design subdomains to meet the hierarchical performance index ψ(′g ′ ) to maximize the probability of high performance, obtain the value ranges of sub-level key elements and their performance parameters, and improve design efficiency. Through a top-down hierarchical system design methodology, the selection of representation parameters, and design domain reduction, a systematic and efficient hierarchical decoupling design method for complex structural systems is established. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0046] Figure 1This is a flow chart of a data-driven hierarchical decoupling design method for a complex structural system according to an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of a hierarchical design system for a complex structural system according to an embodiment of the present invention;

[0048] Figure 3 For the data set construction, information entropy reduction data set division and data driving γ of the embodiment of the present invention c Schematic diagram of the parameter identification process. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0050] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0051] See also Figure 1 This embodiment provides a data-driven hierarchical decoupling design method for complex structural systems, including the following steps:

[0052] S01. Conduct performance requirement analysis of complex structural systems.

[0053] According to the requirements of experience, test regulations and specifications, a qualitative and quantitative analysis of performance indicators such as stiffness and strength is carried out to determine the main performance indicator system of the complex structural system. In this embodiment, the performance indicator is set as a performance evaluation indicator and its label. The performance evaluation indicator (ψ) is used to characterize the design performance of this level, and the label indicates the evaluation result of the performance evaluation indicator. Each performance evaluation indicator represents a sub-requirement of the target requirement for a complex structural system. For example, based on the need for collision safety, the frontal collision safety requirement for the entire vehicle body structure includes safety sub-requirements such as the peak collision acceleration and the floor intrusion amount. As for the label of the performance evaluation indicator, this embodiment can mark the specific design as 'g' (if satisfied, it is a feasible solution) and 'p' (if not satisfied, it is an infeasible solution) according to the indicator requirements.

[0054] S02. Conduct system-level decoupling analysis.

[0055] like Figure 2 As shown, in this embodiment, a complex structural system is structurally analyzed based on performance indicators. Based on the structural analysis results, the complex structural system is divided into layers from top to bottom to construct a system hierarchical model. The first layer of the system hierarchical model represents the entire complex structural system at the top level, and the bottom layer represents the bottom layer. The top level complex structural system structure, such as the complex structural system itself, and the bottom layer can be a component. Based on actual design requirements, the component can also be further divided into material and process design sub-layers.

[0056] Next, a system hierarchical model is constructed based on the divided hierarchies, where each hierarchical level has its own basic design elements or basic attributes. In this embodiment, these basic design elements or basic attributes are defined as three types of parameters, namely, design variables (x), performance parameters (x), and performance parameters (x). and transition variables (x t ).

[0057] Among them, (1) the design variable x is the parameter directly designed at this level. The relationship it maps may be the connection method, material, shape, thickness, or other basic properties of the complex structural system. Subsequently, the design subdomain is obtained through simulation data driving, and the feasible subdomain of x is obtained based on the characteristics of the design subdomain. The specific steps are shown in steps S03 and S04. It should also be explained that the design variable is not only a continuous variable, such as the thickness of the steel plate, but can also be discrete, such as the connection method between steel plates. The values ​​may include: welding, riveting, bonding, and bolting. This embodiment needs to evaluate the effects or system performance under different values.

[0058] (2) Characterization performance parameters This parameter is used to characterize the performance parameters of each sub-level that affect the performance indicators of this level. It is set to solve the problem that it is difficult to change the basic elements of the underlying levels during the design of this level, and the cost of system simulation evaluation is high. For example, directly designing the shape of the components in the frontal collision simulation model of the whole vehicle is difficult and the evaluation cost is high. The average stiffness of the components has a significant impact on the performance of this level, so it can be used as a performance parameter to characterize the performance. By changing the thickness of the component sheet or the material stiffness, the average stiffness of the component can be easily changed, that is, changing the transition variable x t Through data mining, the average stiffness range requirements of components that meet the safety of the entire vehicle are obtained, which is used as the performance target or constraint of component design. The correlation between the system and the components is established, and the system performance indicator-oriented design is realized at the component level.

[0059] (3) Transition variable x t, are basic transition parameters that are easily parameterized in order to change the performance parameters, as described above. When reaching the part level, there is no further level, and only design variables (such as thickness, shape, and material) are present, without transition variables.

[0060] Through the above decoupling analysis process, the hierarchical division of the complex structural system is completed, and the hierarchical structural design system with mutual correlation between the levels is formed through design variables, performance parameters and transition variables, as shown in the figure. Figure 2 As shown on the left, it provides the basis for the decoupling design of S03.

[0061] S03. Carry out top-down decoupling design at different levels of the data-driven system.

[0062] exist Figure 1 The box diagram on the right of this step shows a data-driven design method for a certain level. Data is obtained through hierarchical digital simulation, and the data is driven to obtain the representation performance requirements of the next level. The correlation between levels is established, and the different levels divided in step S02 are designed.

[0063] This embodiment uses a certain level in the system hierarchy model as a reference, which is recorded as this level, and the level below it is called the next level. For the convenience of description, this embodiment uses "i" to describe the position of a specific level. Figure 1 The data-driven hierarchical design method mentioned in is as follows:

[0064] S03-1 Clarify the parameters of this level, including design variables x and performance parameters Transition variable x t and the evaluation index ψ(′·′).

[0065] S03-2 Establish a mathematical model for hierarchical design and clarify the goals of hierarchical design.

[0066] For non-bottom-level hierarchical design, the goal is to obtain the design subdomain The probability (p(ψ(′g′)|γ c )) and robustness (Φ(ψ(′g′)|γ c )) is maximized. At the same time, the design satisfies the design variables, transition variables and other constraints The requirements are as follows (1):

[0067]

[0068] Among them, x l represents the lower limit of the design variable x, x u represents the upper limit of the design variable x, x tl Represents the transition variable x tThe lower limit of x tu Represents the transition variable x t The upper limit, Represents a design constraint.

[0069] S03-3 carries out simulation evaluation parametric modeling, and performs evaluation calculations of the performance of the representation and hierarchical indicators under different combinations of design variables and transition variables based on the parametric simulation model. Specifically, in this step, parametric modeling needs to establish a model calculation framework for automatic update definition and calculation evaluation of parameters based on different systems and design parameters. Further, random or Latin hypercube sampling is performed within the range of design variables and transition variables to obtain different combinations. The number of sampling combinations depends on the actual model evaluation cost and the scale of the problem; the design variables and transition variable values ​​of different combinations are assigned to the parametric model to obtain the corresponding simulation model, and the corresponding performance evaluation indicators and the response of the representation performance parameters are calculated.

[0070] In addition, the generation of the label is specifically as shown in step S03-4.

[0071] S03-4 generates hierarchical performance simulation data, including input attributes: design variables x and performance parameters and labels. Specifically, based on the hierarchical indicator performance evaluation results of step S03-3 above, labels are defined according to the evaluation criteria: if the performance requirements or the goals of the hierarchical design are met, it is marked as 'g', otherwise it is marked as 'p'. For non-first-level 'g' labels, they must also meet the requirements passed from the previous level. The requirements are met, thereby obtaining a data set with different combinations of design variables and performance parameters (input parameters) and their corresponding performance labels (output parameters).

[0072] S03-5 learns the simulation data and obtains the correlation between the performance of the layer and the next layer, including:

[0073] a. Dataset division. Based on the entropy reduction model, the key parameter value A is selected to subset the dataset S obtained by simulation. The dataset S includes all combinations of design variables and performance parameters at the current level, as well as their corresponding labels. The formula for the parameter selection index (GR(A): information gain rate) is as follows:

[0074]

[0075] Among them, G(A) is the information gain, that is, the entropy of the current data set S The entropy after dividing the data set by variable A The difference in G(A) is the reduction in the entropy of the dataset, so the larger the G(A), the better. It should be explained that the variable A mentioned above includes the design variables and performance parameters at this level. and They can be calculated using formulas (3) and (4) respectively.

[0076]

[0077] Among them, SI(A) is the information increment of data set partition, which means the amount of information generated by partitioning the data set by A, m is the number of labels, and p i is the proportion of the number of designs of the i-th label to the total number, N is the number of data subsets generated after the data set is divided, S j is the subset generated after splitting the data set S, For subset S j The information entropy of |·| represents the number of samples in the sample set.

[0078] It should be explained that, for the number of label types m, in this embodiment, it includes 'g' and 'p', that is, m=2. If there are other classification methods, the number of labels can be set to different numbers according to actual needs; for the number of data subsets N, if the variable A is a discrete variable, then N is the number of discrete values; if A is a continuous variable, all possible values ​​of A can be selected according to ">" and "≤" A to divide the data set into two parts, that is, N=2.

[0079] b. Recursive partitioning of data sets

[0080] In order to obtain a subset with a more pure label, follow step a, recursively divide the data set, and record the attribute A or its corresponding value A and direction (greater than or less than equal to) of each divided data set until a specific condition is met and the division is stopped. Specifically, when the subset S j The division stops when one of the following conditions (1) to (3) is met:

[0081] (1) Subset S j The label purity pi(S j )( or ) reaches a preset value;

[0082] (2)|S j | Reaching the preset minimum value;

[0083] (3) The dataset S is split a predefined number of times.

[0084] In particular, the subset S j The label purity pi(S j ) is generally selected in the interval [0.7, 1.0], and the label purity pi (S j ) value is larger, indicating that it is closer to the design target requirement. The selection should be based on actual needs.

[0085] In order to balance the computational effort and design objectives, the label purity pi (S j ) is preferably set to 0.9.

[0086] S03-6 obtains the hierarchical performance requirements by summarizing the data partitioning process. Design subdomains.

[0087] In step S03-6, all subsets generated by completing the data set division are summarized and the probability of whether the performance index is met (p(ψ(′g′)|γ c )) and the robustness of performance indicators (Φ(ψ(′g′)|γ c ))Maximize the dual objectives to select the subset, and at the same time extract the key A and the partition value from the original data set to the selected subset, and obtain the key variables and their value ranges.

[0088] At this point, all key partitioning variables are identified as key design variables or performance characterization parameters, and the value range of the key variables is the design subdomain, that is, in,<x(ψ(′g′))> As the value subdomain of the design variable at this level; As the performance constraint of the key design elements (components) of the next level, it ensures that the performance of this level satisfies the maximization of p(ψ(′g′)).

[0089] In order to better understand the present invention, a more specific example is given here. Figure 3 As shown in the example. Each sample of the sample original data contains two continuous attributes And the corresponding label (1abel), a total of 210 groups of data. Through data learning, the algorithm realizes three data set divisions (P1, P2, P3), forming four data subsets (R1, R2, R3, R4). It can be seen that: R2 satisfies the maximum p(ψ(′g′)) and the γ of R2 can be obtained. c Contains key parameters and its subdomains

[0090] By iterating the above S03-1 to S03-6 processes, the design of each key element (component) of the sub-level is completed. The design of each sub-level must fully consider the characterization performance requirements of ψ(′g′) and other relevant specification requirements. In this iteration, by evaluating the sub-level x and x t The hierarchical performance of Hierarchical design of 〈x(ψ(′g′))> and the characterization performance requirements for the next sub-level

[0091] For the lowest level component design, as shown in formula (5), there is no next level, so there is no next level to characterize performance parameters and transition variables. Only the model performance evaluation ψ(′g′) designed by the elements of this level is used to meet the phenotypic performance parameter requirements passed from the previous level.

[0092]

[0093] Through the above process, the decoupling design of the complex hierarchical structure system is completed from top to bottom, and the obtained design system with L levels is as follows: Figure 2 As shown. Among them, x L and Represent the design variables and transition variables of the Lth level, respectively, <x L (ψ(′g′))> is the set of design variables at the Lth level that satisfies the maximization of p(ψ(′g′)), To obtain the set of performance parameters that maximize p(ψ('g')) at the Lth level, associate the element designs of the next level with the key elements and performance requirements of the next level design. Finally, a set of designs that satisfy ψ('g') at each level is obtained, which is the subset (subdomain) of design variables.

[0094] S04, matching different levels of design based on system performance, specifically including combining the design subdomains of different levels obtained in step S03 <x 1 (ψ(′g′)), x 2 (ψ(′g′)),…,x L Considering system performance and other configuration requirements, such as the structural adjustments required for individual vehicle accessory installations, design selection or matching is performed within corresponding component subsets (subdomains) to obtain design matching sets at different levels. Furthermore, hierarchical design matching sets are completed and integrated to generate a set containing several subsystem designs.

[0095] S05. Construct a system design simulation model. Use the underlying part design variable values ​​of the final system design set to construct components. Combined with the design of component connection variables, consider system boundaries and load conditions, and integrate the system simulation model construction according to specifications and test requirements.

[0096] S06. Verify the overall system performance of the design: The system design model obtained through the above systematic design method identifies high-performance sub-domains that characterize performance parameters, obtains sub-level key design elements and their performance requirements, and ensures the performance of the hierarchical design. The final system design set can basically guarantee the preset performance requirement ψ(′g′), thereby improving design efficiency.

[0097] It should be noted that, for the sake of convenience, the aforementioned method embodiments are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited to the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously.

[0098] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0099] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A data-driven hierarchical decoupling design method for complex structural systems, characterized by: The steps include: Perform performance requirement analysis on complex structural systems and determine performance indicators; the performance indicators include performance evaluation indicators and labels. The performance evaluation indicators are used to evaluate system-level performance. The labels refer to the evaluation results of the performance evaluation indicators judged by the evaluation criteria, including feasible solution labels and infeasible solution labels. Based on the structure and performance indicators of the complex structural system, a hierarchical decoupling analysis is performed to establish a system hierarchical model. The system hierarchical model includes multiple sub-levels and defines the basic parameters of each sub-level in the model. The basic parameters include design variables, performance characterization parameters, and transition variables. The design variables are used to represent the basic design elements of this level. The performance characterization parameters are used to characterize the performance of each sub-level that affects the performance indicators of this level. The transition variables are used to facilitate the current level to regulate the performance characterization parameters of the sub-level. According to the design goal, the hierarchical elements of the non-lowest sub-level are digitally simulated given the design variables and transition variables, and the performance evaluation indicators and representative performance parameters of the current level are obtained. The performance evaluation indicators are judged according to the given evaluation criteria, and different design variables and representative performance parameters are labeled to obtain a data set. The data set includes different design variables and representative performance parameters and corresponding labels. The data label entropy reduction method is used to mine the key design variables and their value ranges that meet the design goals, and the key representative performance parameters and their value ranges. The key design variables and their value ranges are the current level design set; Pass the key performance parameters and their ranges to the next level and use them as one of the constraints for determining whether a solution is feasible or infeasible at the next level. Repeat the above process for each sub-level to complete the design of each level; mine the data of the lowest sub-level to obtain the design variables and their value ranges, which are the design set of the current level; The method of using the data label entropy reduction method to mine key design variables and their value ranges and key performance parameters and their value ranges that meet the design goals includes the following steps: Data sets obtained from simulation based on entropy reduction model S Subset partitioning, data set S Including the characteristic parameters of all samples in the current level, that is, the design variables and the characterization performance parameters, as well as the corresponding labels, by selecting the one with the maximum information gain The characteristic parameter A of the value divides the data set as follows: in, is the information gain, that is, the current data set S Entropy The entropy after dividing the data set using the characteristic parameter A The difference, The information increment for data set division is the amount of information generated by dividing the data set by feature parameter A, m is the number of labels, For the The proportion of the number of designs of labels to the total number, N The number of data subsets generated after the data set is divided, S j To split the data set S The generated subset, For subset S j The information entropy of |∙| represents the number of samples in the sample set; By following the above indicators Perform recursive partitioning until the set conditions are met and then stop partitioning to obtain the recursive partitioning results; Analyze the recursive partitioning results to obtain feasible solution labels The largest and purest subset, and record all key design variables and their value ranges and key performance parameters and their value ranges used to obtain the subset; Integrate the design sets obtained at each level to obtain the system design set, construct and connect components according to the design variables in the system design set, obtain the complex structure system design simulation model, and use performance indicators to verify the performance of the complex structure system simulation model.

2. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: The performance requirement analysis of the complex structural system includes the following steps: Conduct qualitative and quantitative analysis of the performance requirements of complex structural systems based on prior knowledge, and determine various performance evaluation indicators and their feasible standards; The labels are set based on an indication of whether the feasibility criteria of the evaluation metric are met.

3. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: The hierarchical decoupling analysis based on the structure and performance indicators of the complex structural system includes the following steps: A complex structural system is structurally analyzed according to performance indicators, and the complex structural system is divided into levels from upper to lower levels according to the structural analysis results to construct a system hierarchical model; the first level of the system hierarchical model represents the uppermost complex structural system as a whole, and the bottom level represents the lowest level component or material design, and its specific design scale is determined according to actual conditions.

4. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: The design goals include: For non-bottom-level levels, the goal of hierarchical design is to obtain a design subdomain that maximizes both the feasibility probability of the performance indicator and the robustness of the performance indicator; and to set the constraint intervals of the design variables and transition variables.

5. The data-driven hierarchical decoupling design method for complex structural systems according to claim 4 is characterized in that: The marking method comprises the following steps: According to the set design goals, the system response under the combination of given design variables and transition variables is judged. Only when the range of the characterization performance parameters passed from the previous level is met and other performance evaluations meet the design goals, the feasible solution label is output. , that is, the design variable is a feasible solution, otherwise the output is an infeasible solution label , that is, the design variable is an infeasible solution; obtain the data set of the design variables, performance parameters and their corresponding labels at this level.

6. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: The design goals include: For the lowest sub-level, the goal of hierarchical design is to obtain the design subdomain of this level, which needs to maximize the probability of meeting the performance index and the robustness of the performance index; set the design variables x constraint interval.

7. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: The key characterization performance parameters and their ranges are passed to the next level, including: The subset of characterization performance parameters that meet the set design conditions is passed to the next level, and the next level uses the subset of characterization performance parameters as constraints. When the performance indicators of the hierarchical components meet the evaluation criteria and their characterization performance parameters meet the characterization performance parameters passed down from the level, they are marked as .

8. The data-driven hierarchical decoupling design method for complex structural systems according to claim 1 is characterized in that: If the characteristic parameter If is a discrete variable, is the number of discrete values; if the characteristic parameter If is a continuous variable, is 2.

9. The data-driven hierarchical decoupling design method for complex structural systems according to claim 8, characterized in that: In the process of recursive partitioning of the current data set, the subset The division stops when one of the following conditions (1) to (3) is met: (1) Subset The tag purity reaches the preset value; (2) Reaching a preset minimum value; (3) Current dataset Split a predefined number of times.

Images