CN-122021038-A - Automatic generation method and system of architecture optimization model based on decision mode

CN122021038ACN 122021038 ACN122021038 ACN 122021038ACN-122021038-A

Abstract

The invention relates to an automatic generation method and system of an architecture optimization model based on a decision mode, comprising the steps of constructing and maintaining a decision mode library, loading and analyzing the architecture model to be optimized, identifying decision points existing in the architecture model, matching the identified decision points with the decision mode in the decision mode library to generate an optimization model segment, synthesizing a plurality of the optimization model segments into a complete and unified optimization model, calling an optimization solver to calculate the optimization model to obtain an optimal solution, and reversely mapping decision variable assignment results in the optimal solution back to the architecture model to generate an optimized architecture scheme. By constructing a formalized decision mode library and a structured mathematical model template, the system can automatically convert decision problems in a DoDAF and other system architecture models into an optimized model capable of solving, and the complicated and error-prone process of manually writing variables, constraints and objective functions by traditional reliance experts is avoided.

Inventors

FANG ZHEMEI
MIAO TIAN
LI XINXI

Assignees

华中科技大学

Dates

Publication Date: 20260512
Application Date: 20260205

Claims (10)

1. The automatic generation method of the architecture optimization model based on the decision mode is characterized by comprising the following steps of: s1, constructing and maintaining a decision mode library, wherein the decision mode library stores a plurality of predefined decision modes, each decision mode corresponds to a formalized mathematical model template, and the mathematical model templates are described in a structured data format and comprise variable templates, constraint templates and objective function templates; s2, loading and analyzing a system architecture model to be optimized, and identifying decision points existing in the system architecture model; S3, matching the identified decision points with the decision modes in the decision mode library, extracting parameters required by the decision modes from the system architecture model in the step S2 for the successfully matched decision points, and injecting the extracted parameters into mathematical model templates corresponding to the decision modes to generate optimized model fragments; s4, synthesizing a plurality of optimization model segments into a complete and unified optimization model; s5, calling an optimization solver to calculate the optimization model to obtain an optimal solution; s6, reversely mapping the decision variable assignment result in the optimal solution back to the system architecture model in the step S2, and generating an optimized system architecture scheme.
2. The method for automatically generating an architecture optimization model based on a decision mode according to claim 1, wherein the mathematical model template adopts a structural data description of a JSON structure.
3. The method of claim 1, wherein the variable templates define types and index sets, and the constraint templates and objective function templates are defined using parameterized expressions.
4. The method for automatically generating an architecture optimization model based on a decision mode according to claim 1, wherein the decision mode comprises at least one or more of a filtering mode, a partitioning mode, an allocation mode, a ranking mode, and a connection mode.
5. The method for automatically generating an architecture optimization model based on decision mode as claimed in claim 1, wherein the step S2 comprises the following steps: S21, receiving a system architecture model to be optimized; s22, traversing the system architecture model elements, and identifying modeling elements for representing candidate system resources, combat activities to be distributed and system capacity targets; S23, extracting attribute information and semantic relations in the identified modeling elements; s24, identifying decision points of the architecture model to be optimized based on the extracted attribute information and semantic relation.
6. The method of claim 1, wherein the architecture model in S2 comprises at least one of DoDAF, UAF, MODAF or NAF architecture frameworks.
7. The method for automatically generating an architecture optimization model based on decision mode as claimed in claim 1, wherein the step S3 comprises the following steps: s31, matching the identified decision points with decision modes in a decision mode library according to a predefined matching rule; S32, extracting candidate entities and parameter sets required by a decision mode from the architecture model to be optimized in the S2 aiming at successfully matched decision points to form a parameter dictionary; S33, injecting the parameter dictionary into a mathematical model template corresponding to the decision mode, and generating an optimized model segment containing variables, constraints and local objective functions.
8. The method for automatically generating an architecture optimization model based on a decision mode as claimed in claim 1, wherein in step S4, the synthesis mode of the optimization model segments includes a global optimization mode or a focus optimization mode, and a plurality of optimization model segments are synthesized according to the synthesis mode selected by the user; synthesizing all optimization model fragments to seek a global optimal solution; The focusing optimization mode is that only the optimization model segments corresponding to the decision points designated by the user are combined, and states of other decision points are used as fixed constraints.
9. The method for automatically generating an architectural optimization model based on a decision mode according to claim 1, wherein the optimization solver adopts one of Gurobi, CPLEX or SCIP.
10. A decision-mode-based architecture optimization model auto-generation system for use in the method of any one of claims 1 to 9, comprising: the decision mode library module is used for constructing and maintaining a decision mode library; The model conversion engine module is used for loading and analyzing the architecture model to be optimized, identifying decision points, matching the identified decision points with decision modes in the decision mode library, extracting parameters required by the decision modes from the architecture model, and injecting the extracted parameters into mathematical model templates corresponding to the decision modes to generate optimized model fragments; The model synthesis module is used for synthesizing a plurality of optimized model segments into a complete and unified optimized model; The optimization solving and explaining module is used for calling an optimization solver to calculate the optimization model to obtain an optimal solution; And the result mapping and scheme generating module is used for reversely mapping the decision variable assignment result in the optimal solution back to the system architecture model to generate an optimized system architecture scheme.

Description

Automatic generation method and system of architecture optimization model based on decision mode Technical Field The invention belongs to the field of architecture design in architecture engineering, and particularly relates to an automatic generation method and system of an architecture optimization model based on a decision mode. Background An architecture framework is a standardized specification describing the architecture, behavior, and relationships of various dimensions of the architecture, organizing architecture information through multiple view models (e.g., combat view OV, system view SV). In the architecture framework design process, architects are faced with a series of typical decision making problems. Such as how to select an optimal subset among multiple candidate systems to meet budget constraints (screening issues), how to assign functions to appropriate physical systems (assignment issues), how to determine the order in which tasks are performed to optimize response times (ordering issues), etc. These problems essentially belong to the combination optimization or resource allocation problem, and if they can be formed into mathematical optimization models (such as mixed integer linear programming, MILP) and automatically solved by means of modern optimization solvers, the scientificity and optimizing capability of the architecture scheme are significantly improved. However, the method generally adopted in the industry is a manual modeling and optimizing method based on expert experience, and specifically comprises the following steps: step 1, architecture modeling, namely firstly, manually creating a series of descriptive models by using an architecture framework by a system architect according to requirements; Step 2, problem identification, namely identifying key decision points needing to be optimized from the models through manual analysis, step 3, manual modeling, namely manually forming the identified decision problems into mathematical optimization models by relying on deep knowledge and experience of field experts, wherein the process requires the experts to manually define decision variables, manually write constraint conditions according to system constraints and manually construct objective functions according to design targets; And 4, solving and analyzing, namely inputting the manually established optimization model into a special solver for calculation, and explaining a solving result. The existing method has the following defects: 1) The decision points in the descriptive model are manually identified by the architect and manually formalized as a mathematical optimization model. The process is inefficient, modeling quality is strongly related to the mathematical ability of architects and field experience, which results in inconsistent modeling process and easy error, and it is difficult to ensure model accuracy in complex systems. 2) The optimization model is one-time, with variables, constraints tightly bound to specific architectural model elements. When the system demand is changed or the architecture evolves, the optimization model needs a large amount of manual modification and adjustment, so that the flexibility is poor, the reusability is low, and quick architecture iteration and design space exploration are difficult to support. 3) The whole flow from architecture modeling to optimization solving has low automation degree and obvious artificial break points. This severely limits the ability to systematically explore a wide design space, and designers often can only evaluate a limited number of alternatives, potentially missing a more optimal architectural design. Based on this, the present application is hereby proposed. Disclosure of Invention The invention aims to provide an automatic generation method and an automatic generation system for a system architecture optimization model based on a decision mode, which can automatically identify architecture decision semantics, generate the optimization model based on a general decision mode and support reverse mapping of results to an architecture scheme, so as to break through the bottleneck of the conventional manual modeling paradigm, improve the modeling accuracy and efficiency, support rapid reconstruction of the optimization model when the requirements are changed, and promote agile architecture design and iteration. In order to achieve the above purpose, the technical scheme of the invention is as follows: An automatic generation method of an architecture optimization model based on a decision mode comprises the following steps: s1, constructing and maintaining a decision mode library, wherein the decision mode library stores a plurality of predefined decision modes, each decision mode corresponds to a formalized mathematical model template, and the mathematical model templates are described in a structured data format and comprise variable templates, constraint templates and objective function templates; s2, load