A Functional Module Partitioning Method Based on Fuzzy Clustering and Relevance

Abstract

This invention discloses a functional module partitioning method based on fuzzy clustering and relevance, comprising: decomposing functional modules based on the relevance of function, physics, and information using a relevance matrix to obtain different basic units; dividing functional modules into functional subsets and physical subsets, and using a search algorithm to select merging subsets of functional and physical components to form a module set; and based on the information relevance of functional subsets, physical subsets, and module set, combined with information flow, using fuzzy clustering analysis to achieve logical hierarchical aggregation of modules, thereby realizing functional module identification and integration. This invention achieves hierarchical aggregation of modules with similar functions, realizes modularization and generalization of specific functional applications, identifies common functional modules in the system, coordinates system resource settings, and provides a scientific and reasonable methodological reference for the functional decomposition of ship information systems, with good application prospects.

Description

Technical Field

[0001] This invention relates to the field of system integration design technology, and in particular to a method for dividing functional modules based on fuzzy clustering and relevance. Background Technology

[0002] Currently, in the process of integrating shipboard information systems, the integration is carried out on a subsystem basis. Each subsystem is self-contained, with tightly coupled devices that serve specific subsystems, making cross-system sharing difficult. Furthermore, the lack of integrated design results in devices with redundant or overlapping functions. This architectural structure limits the level of integration optimization and hinders the improvement of overall efficiency.

[0003] To address the aforementioned issues and achieve better resource allocation for ship information systems, as well as enhance system resource sharing, expansion, and upgrade capabilities, it is necessary to optimize the structure and functions of ship information systems and shift from the current integrated design based on subsystems to an integrated design based on functions.

[0004] Therefore, it is necessary to decompose the functions of the ship information system, establish an integrated system hierarchical structure, and design a method from the perspective of service-oriented architecture to horizontally layer the system task channels, extract hierarchical modules with the same nature and similar functions between channels, classify, standardize and generalize devices with the same or similar functions, modularize specific functional applications, plan the system's resource settings in a coordinated manner, form common functional modules of the system, solve the problem of functional module reuse, and provide an integration foundation for the integration of ship information system functions.

[0005] Based on the analysis of relevant requirements and current technical status through the division of system functional modules, the technical problems that need to be solved can be summarized as follows:

[0006] (1) Functional module subdivision and iterative decomposition

[0007] When dividing a system into functional modules, the smaller the granularity of the modules, the more functions can be integrated, but this also increases the workload and cost of integration. Therefore, it is necessary to find a reasonable balance between integration granularity and integration cost. This requires analyzing the classification of system-related objects, researching module division methods, and applying these methods to the modular design of the system, thus laying the foundation for the integrated design of ship information systems.

[0008] (2) Identification and integration of general functional modules

[0009] General function module identification primarily involves grouping system components with common attributes or characteristics together and distinguishing those without. Integration, on the other hand, requires the use of specific rules to recombine them. For a system to accomplish a task, various modules need to cooperate, necessitating information exchange between them. Therefore, the identification and integration of general function modules also require scientific paradigms and methods. Summary of the Invention

[0010] The technical problem this invention aims to solve is to address the shortcomings of existing technologies by providing a functional module partitioning method based on fuzzy clustering and relevance, thereby resolving the issues of low sharing and difficulty in reuse of functional modules in current information systems, and providing a scientific and reasonable methodological reference for the functional decomposition of ship information systems.

[0011] The technical solution adopted by this invention to solve its technical problem is:

[0012] This invention provides a method for functional module partitioning based on fuzzy clustering and relevance, the method comprising the following steps:

[0013] Step 1: For functional modules, establish dictionary identifiers and construct a functional, physical, and information correlation matrix. Decompose the functional modules based on the correlation matrix in terms of functional, physical, and information correlation to obtain different basic units.

[0014] Step 2: Using the disjoint-set data structure, different basic units are divided into functional subsets according to function and physical subsets according to physics. A search algorithm is then used to select functional subsets and physical subsets to form mergeable subsets, which are then merged into a module set. Based on the information correlation between functional subsets, physical subsets, and module sets, and combined with the information flow, fuzzy clustering analysis is used to achieve the logical hierarchical aggregation of modules, thereby realizing the identification and integration of functional modules.

[0015] Furthermore, the representation and characteristics of functional relevance in step 1 of the present invention are specifically as follows:

[0016] The following 0-1 matrix is ​​used to represent the functional similarity between units:

[0017]

[0018] Where n is the number of units, f ij This indicates the functional correlation between unit i and unit j, where 1 indicates that they are related and 0 indicates that they are not related.

[0019] The functional correlation matrix satisfies transitivity, meaning that if there is functional correlation between unit i and unit j, and between unit j and unit k, then there is also functional correlation between unit i and unit k.

[0020] Furthermore, the representation and characteristics of physical correlation in step 1 of the present invention are specifically as follows:

[0021] The following 0-1 matrix is ​​used to represent the physical dependencies between elements:

[0022]

[0023] Where n is the number of units, p ij This indicates the physical correlation between unit i and unit j, where 1 indicates that they have similar physical forms and 0 indicates that they have different physical forms.

[0024] The physical correlation matrix satisfies the transitivity relation, that is, if there is a physical correlation between unit i and unit j, and a physical correlation between unit j and unit k, then there is also a physical correlation between unit i and unit k.

[0025] Furthermore, the representation and characteristics of information relevance in step 1 of the present invention are specifically as follows:

[0026] The following 0-1 matrix is ​​used to represent whether there is information exchange between units, indicating their correlation:

[0027]

[0028] Where n is the number of units, i ij This indicates that there is information exchange between unit i and unit j, and the information flows from unit i to unit j. That is, unit i is the service provider and unit j is the service consumer.

[0029] The information relevance matrix satisfies the transitivity relationship, that is, if information flows from unit i to unit j, and from unit j to unit k, then information can flow from unit i to unit k.

[0030] Further, the method of step 2 of the present invention includes:

[0031] Step 21, Functional Partitioning: Functions must satisfy transitivity; a disjoint-set data structure (Union-Find) is used to integrate modules; the entire element set E = {e1, e2, ..., e...} is then divided into two parts. n} is divided into several subsets {F1, F2, ..., F} according to the principle of functional relevance. f}, e n Represents the basic unit, subset F i With F j They do not overlap with each other, and each subset represents a functionally related unit;

[0032] Step 22, Physical Partitioning: Since the physical form satisfies transitivity, a disjoint-set data structure (Union-Find) is used to integrate the modules; the entire element set E = {e1, e2, ..., e} is divided into two parts. n} is divided into several subsets {P1, P2, ..., P} according to the principle of physical relevance. l}, subset P i With P j They do not overlap with each other, and each subset represents a physically related unit;

[0033] Step 23: Consider functional and physical aggregation: Select subsets that can be merged in terms of both function and physical aspects, and use a search algorithm based on the functional subset {F1, F2, ..., F...} f} and physical subsets {P1, P2, ..., P} l This forms a new set {FP1,FP2,…,FP} that simultaneously satisfies the similarity between the two sets. d}, merge each subset into a module set M = {m1, m2, ..., m} d};;

[0034] Step 24, Hierarchical Aggregation: First, calculate the set M = {m1, m2, ..., m} based on the information relevance of the modules in the subset. d The information relevance of the M set is considered comprehensively, taking into account the functions, physical aspects, and the information flow of the M set. Fuzzy clustering analysis is used to divide and merge the layers.

[0035] Furthermore, the disjoint-set data structure method of the present invention is specifically as follows: the disjoint-set data structure is represented by a tree, all elements belonging to the same set belong to the same tree, and a set is represented by a root. To find the set to which a certain element belongs, the root of the tree to which the element belongs can be found.

[0036] Furthermore, the fuzzy clustering analysis method in step 24 of the present invention is specifically as follows:

[0037] Step a: First, based on the functional and physical module division results and the information correlation matrix, obtain the module characteristic index matrix {x}. ij} n×m ;

[0038] Step b: Establish the module characteristic index matrix {x ij} n×m fuzzy similarity matrix {r ij} n×m ;

[0039] Step c, put r ij Sort the data from largest to smallest, and use this as the basis for selecting the threshold λ.

[0040] Step d: Select the maximum value λ1, which will satisfy r ij =λ1 of xi and x j As a category;

[0041] Step e, take r ij The second largest value λ2 in the range is from {r ij} n×m Find the pairs of elements (x) whose similarity is equal to λ² directly. i ,x j ), accordingly, the equivalence class [x i ] t(R) With [x i ] t(R) Merge;

[0042] Step f continues as in step e until a termination condition is met. Based on information relevance, the termination conditions include: ① meeting a pre-given threshold λ; ② if a contradictory decision rule is generated; ③ the similarity of a certain attribute is lower than a certain value.

[0043] Step g: Undo the last merge, merge adjacent layers with similar functions to form the final classification.

[0044] The beneficial effects of this invention are:

[0045] 1. This method addresses the problems of low resource sharing in information systems and the difficulty in sharing and reusing redundant functions within subsystems. It proposes a functional module partitioning method based on fuzzy clustering and relevance. Targeting the task requirements of the application system, it performs correlation analysis from functional, physical, and informational perspectives to achieve functional module subdivision and iterative decomposition. A module hierarchical aggregation algorithm is proposed to achieve hierarchical aggregation of modules with similar functions, realizing the modularization and generalization of specific functional applications. It identifies common functional modules in the system, coordinates system resource allocation, and provides a scientific and reasonable methodological reference for the functional decomposition of ship information systems, demonstrating promising application prospects.

[0046] 2. Functional module partitioning can be divided into top-down functional independence decomposition and bottom-up functional correlation clustering. However, for complex functional systems, simply performing functional decomposition results in a poorly flexible functional tree, strong subjective bias in the partitioning, and may lead to the separation of underlying functions from the structure. Simply performing functional clustering relies on the bottom-up cohesive relationships between individual units, but the number of units increases dramatically with the complexity of the system (equipment), resulting in an astonishing amount of computational burden and making it difficult to control computational complexity. Therefore, neither functional independence decomposition nor functional correlation clustering alone can effectively achieve the rational partitioning of modules in the modular design of complex functional systems.

[0047] This invention integrates two approaches and proposes a functional module partitioning method based on fuzzy clustering and relevance. This method addresses the functional requirements of application systems by performing correlation analysis from functional, physical, and informational perspectives to achieve functional module subdivision and iterative decomposition. It also proposes a hierarchical module aggregation algorithm to achieve hierarchical aggregation of modules with similar functions, identify common modules, and achieve a scientific partitioning of functional modules. Attached Figure Description

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings:

[0049] Figure 1 This is a functional module partitioning method based on fuzzy clustering and relevance in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram showing the main functional modules of the sensor device according to an embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the module function division result of an embodiment of the present invention;

[0052] Figure 4 This is a schematic diagram of the module physical partitioning results according to an embodiment of the present invention;

[0053] Figure 5 This is a schematic diagram showing the result of comprehensively considering the functional and physical module division in an embodiment of the present invention;

[0054] Figure 6 This is a schematic diagram of the automatic module integration and system layering results according to an embodiment of the present invention;

[0055] Figure 7 This is a general functional block diagram of the sensor device according to an embodiment of the present invention. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0057] Example 1

[0058] The functional module partitioning method based on fuzzy clustering and relevance in this invention includes:

[0059] (1) Functional module subdivision and iterative decomposition based on correlation analysis

[0060] First, the functional modules are decomposed from the perspectives of function, physics, and information to obtain different basic units.

[0061] 1) Functional relevance representation and its characteristics

[0062] Functional relevance refers to the relationship between basic elements. Taking a sensor device as an example, the receiving and transmitting components both function as signal input and output devices. The signal processing and data processing modules both process signals and information, making their functions similar. However, the antenna interface unit and the display and control terminal unit have significantly different functions; the former is an input and output device, while the latter is a display device. Therefore, we use the following 0-1 matrix to represent the functional similarity between elements:

[0063]

[0064] Where n is the number of units, f ij This indicates the functional correlation between unit i and unit j, where 1 indicates that they are related and 0 indicates that they are not related.

[0065] The functional relevance matrix satisfies transitivity, meaning that if there is functional relevance between unit i and unit j, and between unit j and unit k, then there is also functional relevance between unit i and unit k. Utilizing this property, the automatic functional aggregation algorithm is a polynomial-level algorithm, meaning that disjoint-set data structures can be used to achieve module aggregation.

[0066] 2) Physical correlation representation and characteristics

[0067] Physical correlation primarily refers to the physical relationship between basic elements. The connotation of the physical dimension focuses on the physical form of the specific system (equipment) assigned to a function. Some functions may have different physical forms but similar functions; for example, sensor antennas can have array antennas and reflector antennas. Conversely, some functions may have similar physical forms but different functions; for example, receiver cabinets and signal processing cabinets are similar in physical form but different in function. When considering physical correlation for functional module partitioning, it is crucial that only functionally and physically related modules can be integrated into a single module subset. Physical correlation and functional correlation must be analyzed together. Similarly, we use the following 0-1 matrix to represent the physical correlation between elements:

[0068]

[0069] Where n is the number of units, p ij This indicates the physical correlation between unit i and unit j, where 1 indicates that the two have similar physical forms and 0 indicates that the two have different physical forms.

[0070] The physical correlation matrix also satisfies transitivity, meaning that if there is a physical correlation between unit i and unit j, and a physical correlation between unit j and unit k, then there is also a physical correlation between unit i and unit k. Utilizing this property, the subsequent function-based automatic aggregation algorithm is a polynomial-level algorithm, meaning that a disjoint-set data structure can be used to achieve module aggregation.

[0071] 3) Information relevance representation and its characteristics

[0072] Information correlation refers to whether there is information exchange between basic elements. Information exchange can take two forms: unidirectional and bidirectional. For example, there is bidirectional information exchange between the receiver switch matrix and the antenna aperture in a sensor device, but unidirectional information exchange between the receiver unit and the frequency synthesis unit. Similarly, we use the following 0-1 matrix to represent whether there is information exchange between elements, indicating their correlation:

[0073]

[0074] Where n is the number of units, i ij This indicates that there is information exchange between unit i and unit j, and the information flows from unit i to unit j. That is, unit i is the service provider and unit j is the service consumer.

[0075] The information relevance matrix also satisfies the transitivity relationship, that is, if information flows from unit i to unit j, and from unit j to unit k, then information can flow from unit i to unit k.

[0076] (2) By combining information flow and using fuzzy clustering analysis, the logical hierarchy of modules is aggregated to achieve the identification and integration of general functional modules.

[0077] Based on the analysis in the first step, functional module integration and system layering mainly consider factors such as function, physical form, and information flow among the various components. In the module aggregation stage, both functional and physical factors are considered; only functionally and physically related modules can be integrated into subsets. System layering refers to using fuzzy clustering analysis to aggregate functionally similar layers within the integrated subsets, ensuring that the layers satisfy the relationship between information providers and consumers. The module hierarchical aggregation algorithm consists of five steps, each implementing a different function. Executing these steps sequentially completes module aggregation and system layering. Its main idea is as follows:

[0078] 1) Functional Partitioning: Because functions satisfy transitivity, a disjoint-set data structure (Union-Set Union) is used to integrate the modules. This mainly involves partitioning the entire element set E = {e1, e2, ..., e...}. n} is divided into several subsets {F1, F2, ..., F} according to the principle of functional relevance. f}, subset Fi With F j They do not overlap with each other, and each subset represents a functionally related unit.

[0079] 2) Physical partitioning: Because the physical form satisfies transitivity, a disjoint-set data structure (Union-Find) is used to integrate the modules. This mainly involves partitioning the entire element set E = {e1, e2, ..., e...}. n} is divided into several subsets {P1, P2, ..., P} according to the principle of physical relevance. l}, subset P i With P j They do not overlap with each other, and each subset represents a physically related unit.

[0080] 3) Considering functional and physical aggregation: Select subsets that can be merged in terms of both function and physical aspects, and employ a search algorithm. Based on the functional subset {F1, F2, ..., F...} f} and physical subsets {P1, P2, ..., P} l This forms a new set {FP1,FP2,…,FP} that simultaneously satisfies the similarity between the two sets. d}. Merge each subset into a module set M = {m1, m2, ..., m}. d}

[0081] 4) Hierarchical aggregation: First, based on the information relevance of the modules in the subset, calculate the set M = {m1, m2, ..., m}. d The information relevance of the M set is considered comprehensively, taking into account the functions, physical aspects, and the information flow of the M set. Fuzzy clustering analysis is used to divide and merge the layers.

[0082] In steps (1) and (2), the disjoint-set data structure is used, which is represented by a tree. All elements belonging to the same set belong to the same tree. Thus, we can use the root of a tree to represent a set. To find which set an element belongs to, we only need to find the root of the tree containing that element.

[0083] In step 4), the fuzzy clustering analysis method is as follows:

[0084] a) First, based on the functional and physical module division results and the information correlation matrix, obtain the module characteristic index matrix {x}. ij} n×m ;

[0085] b) Establish the module characteristic index matrix {x ij} n×m fuzzy similarity matrix {r ij} n×m ;

[0086] c) r ij Sort the data from largest to smallest, and use this as the basis for selecting the threshold λ.

[0087] d) Select the maximum value λ1, which will satisfy r oj =λ1 of x i and x j As a category;

[0088] e) Take r ij The second largest value λ2 in the range is from {r ij} n×m Find the pairs of elements with a similarity equal to λ² directly.

[0089] (x i ,x j ), accordingly, the equivalence class [x i ] t(R) With [x j ] t(R) Merge;

[0090] f) Continue as in step (e) until the termination condition is met. Depending on the relevance of the information, the termination condition can be one of several options: ① meeting a pre-defined threshold λ; ② if a contradictory decision rule is generated; ③ the similarity of a certain attribute is below a certain value. The specific choice depends on the actual situation.

[0091] g) Undo the last merge, merge adjacent layers with similar functions to form the final classification.

[0092] Example 2

[0093] Taking a system sensor device as an example, this paper decomposes the sensor device into a series of basic functional units based on the principles and main components of various sensor devices. It proposes a description dimension based on three relevance dimensions: function, physics, and information. Relevance analysis is used to analyze the degree of association and mutual influence between units, serving as the basis for module clustering and combination. Aggregation is performed according to the factors influencing relevance, ultimately resulting in functional modules with similar characteristics that facilitate hierarchical aggregation. This provides support for the functional module division of sensor devices. Specific methods are as follows: Figure 1 As shown.

[0094] Based on the module partitioning method based on fuzzy clustering and correlation, and combined with the working principles and composition of three types of sensor devices, their main functional structures are summarized, such as... Figure 2 As shown.

[0095] 1. Establish dictionary identifiers for the original modules, as shown in Table 1.

[0096] Table 1 Original Module Dictionary

[0097]

[0098]

[0099] 2. Construct a correlation matrix

[0100] (1) Define the functional correlation matrix F, as shown in Table 2.

[0101] Table 2. Correlation Matrix F Data Content

[0102]

[0103] (2) Define the physical correlation matrix P, as shown in Table 3.

[0104] Table 3. Physical Correlation Matrix P Data Content

[0105]

[0106] (3) Define the information relevance matrix I, as shown in Table 4.

[0107] Table 4 Information Relevance Matrix I Data Content

[0108]

[0109] 3. Divide functions

[0110] After functional partitioning using the disjoint-set data structure algorithm, the module grouping results are as follows: Figure 3 As shown.

[0111] 4. Perform physical division

[0112] When performing physical partitioning, the disjoint-set data structure algorithm is also used. The original modules are partitioned using matrix P, and the partitioning result is as follows: Figure 4 As shown.

[0113] 5. Taking into account both functional and physical divisions

[0114] By using two nested loops, groups in the F-Group and P-Group sets are traversed to generate groups that satisfy both functional and physical grouping conditions. In this example, the elements in F-Group 1 / P-Group 1, F-Group 2 / P-Group 2, F-Group 3 / P-Group 3, and F-Group 4 / P-Group 4 remain consistent, so the generated FP-Group 1, FP-Group 2, FP-Group 3, and FP-Group 4 remain unchanged. For F-Group 5 / P-Group 5, since F-Group 5 contains modules 7, 14, and 21, while P-Group 5 only contains modules 7 and 21, and module 14 does not simultaneously satisfy both functional and physical grouping conditions, the elements in FP-Group 5, considering both functional and physical grouping, are modules 7 and 21. On the other hand, looking at P-Group 6, the element is module 14. Since there is no corresponding module added in F-Group, P-Group 6 will be retained as FP-Group 6.

[0115] Taking into account both functional and physical partitioning, each subset is a module, as shown in the partitioning results. Figure 5 As shown.

[0116] 6. Layering and Merging

[0117] When layering, first calculate the characteristic index {x} corresponding to each module based on the information relevance matrix I. ij} n×m For example, if unit 5 has information flowing to unit 4, and unit 5 is assigned to module 3 while unit 4 is assigned to module 2, then X 32 =1, when there are multiple information flows between modules, the module degree is accumulated. If unit 12 has an information flow to unit 11, then X 32 =1, for example, the degree matrix {x} obtained from this. ij} n×m As shown in Table 5.

[0118] Table 5 Module Characteristic Index Matrix {x ij} n×m Data content

[0119]

[0120] Establish module characteristic index matrix {x ij} n×m fuzzy similarity matrix {r ij} n×m The minimax method is used here. Establish the fuzzy similarity matrix R as shown in Table 6.

[0121] Table 6 Module Characteristic Index Matrix {r ij} n×m Data content

[0122]

[0123] Sort rij from largest to smallest and use this as the basis for selecting the threshold λ: 0.4>0.33>0.25>0.2>0.14>0.125;

[0124] Choosing the maximum value λ = 0.4 will satisfy r ij x1 and x2 of λ1 are classified as a class; the fuzzy similarity matrix R is established as shown in Table 7.

[0125] Table 7. Fuzzy similarity matrix {r} when λ = 0.4 ij} n×m

[0126]

[0127] Choose the second largest value λ = 0.33, from {r ij} n×m Find the equivalence classes {{x1,x2}, x3, {x4,x5,x6}} with a similarity of 0.33. Considering that the transmit / receive conversion function and the antenna aperture belong to different function sets, this classification violates the decision principle.

[0128] Table 8. Fuzzy similarity matrix {r} when λ = 0.33 ij} n×m

[0129]

[0130] The final merge is cancelled, adjacent layers with similar functions are merged, and the information flow relationships between modules are combined to obtain the final automatic module integration and system layering results, such as... Figure 6 As shown.

[0131] Therefore, based on module clustering and hierarchical algorithms, the main functional modules of each sensor device can be clustered into five major functional sets: antenna aperture function set, transceiver function set, preprocessing function set, computational processing function set (signal processing, data processing), and display and control function set, such as... Figure 7 As shown.

[0132] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0133] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for dividing functional modules based on fuzzy clustering and relevance, characterized in that, The method includes the following steps: Step 1: For functional modules, establish dictionary identifiers and construct a functional, physical, and information correlation matrix. Decompose the functional modules based on the correlation matrix in terms of functional, physical, and information correlation to obtain different basic units. Step 2: Using the disjoint-set data structure, different basic units are divided into functional subsets based on function and physical subsets based on physics. A search algorithm is then used to select functional and physical subsets to form mergeable subsets, which are then merged into a module set. Based on the information correlation between the functional subsets, physical subsets, and module set, and combined with the information flow, fuzzy clustering analysis is used to achieve the logical hierarchical aggregation of modules, thereby realizing the identification and integration of functional modules. The method in step 2 includes: Step 21, Functional Division: Functions must satisfy transitivity; a disjoint-set data structure (Union-Find) is used to integrate modules; the entire element set is then... Divide into several subsets according to the principle of functional relevance. e n Represents basic units and subsets and They do not overlap with each other, and each subset represents a functionally related unit; Step 22, Physical Partitioning: Since the physical form satisfies transitivity, a disjoint-set data structure (Union-Find) is used to integrate the modules; the entire element set is then partitioned. Divided into several subsets according to the principle of physical relevance. subset and They do not overlap with each other, and each subset represents a physically related unit; Step 23: Consider functional and physical aggregation: Select subsets that can be merged in terms of both function and physical aspects, and use a search algorithm based on the functional subsets. and physical subset This forms a new set that simultaneously satisfies the similarity between the two. Merge each subset into a module set. ; Step 24, Hierarchical Aggregation: First, calculate the set based on the information relevance of the modules in the subset. Based on the information relevance, and taking into account the functions, physical aspects, and information flow of the M set, fuzzy clustering analysis is used to perform hierarchical division and merging.

2. The functional module partitioning method based on fuzzy clustering and relevance as described in claim 1, characterized in that, The representation and characteristics of functional relevance in step 1 are as follows: The following 0-1 matrix is ​​used to represent the functional similarity between units: Where n is the number of units, This indicates the functional correlation between unit i and unit j, where 1 indicates that they are related and 0 indicates that they are not related. The functional correlation matrix satisfies transitivity, meaning that if there is functional correlation between unit i and unit j, and between unit j and unit k, then there is also functional correlation between unit i and unit k.

3. The functional module partitioning method based on fuzzy clustering and relevance as described in claim 1, characterized in that, The representation and characteristics of physical correlation in step 1 are as follows: The following 0-1 matrix is ​​used to represent the physical dependencies between elements: Where n is the number of units, This indicates the physical correlation between unit i and unit j, where 1 indicates that they have similar physical forms and 0 indicates that they have different physical forms. The physical correlation matrix satisfies the transitivity relation, that is, if there is a physical correlation between unit i and unit j, and a physical correlation between unit j and unit k, then there is also a physical correlation between unit i and unit k.

4. The functional module partitioning method based on fuzzy clustering and relevance as described in claim 1, characterized in that, The representation and characteristics of information relevance in step 1 are as follows: The following 0-1 matrix is ​​used to represent whether there is information exchange between units, indicating their correlation: Where n is the number of units, This indicates that there is information exchange between unit i and unit j, and the information flows from unit i to unit j. That is, unit i is the service provider and unit j is the service consumer. The information relevance matrix satisfies the transitivity relationship, that is, if information flows from unit i to unit j, and from unit j to unit k, then information can flow from unit i to unit k.

5. The functional module partitioning method based on fuzzy clustering and relevance as described in claim 1, characterized in that, The disjoint-set data structure method is as follows: the disjoint-set data structure uses a tree to represent all elements belonging to the same set. A set is represented by a root. To find the set to which a certain element belongs, you can find the root of the tree containing that element.

6. The functional module partitioning method based on fuzzy clustering and relevance as described in claim 1, characterized in that, The fuzzy clustering analysis method in step 24 is as follows: Step a: First, based on the functional and physical module division results and the information correlation matrix, obtain the module characteristic index matrix. ; Step b: Establish the module characteristic index matrix Fuzzy similarity matrix ; Step c, put r ij Sort by size from largest to smallest, and use as the threshold. Selection criteria; Step d: Select the maximum value. , will satisfy x i and x j As a category; Step e, take r ij The second largest value in ,from Find the similarity directly in the middle. element pairs Accordingly, the equivalence class and Merge; Step f continues as in step e until a termination condition is met. Based on information relevance, the termination condition includes: ① meeting a pre-defined threshold. ; ② Decision rules that produce contradictions; ③ The similarity of a certain attribute is lower than a certain value; Step g: Undo the last merge, merge adjacent layers with similar functions to form the final classification.

Images