A method for constructing a production line topology reference architecture

Abstract

The present invention discloses a method for constructing a production line topology reference architecture. Based on the big data of the company's historical production line topology, the method uses computers and machine learning algorithms to extract the company's commonly used production line topology reference architectures, forming the company's typical production line topology community. This is conducive to recording the company's typical production line characteristics and production habits, realizing the reuse of production line structure and production line construction knowledge, reducing the workload of production line designers, and improving the efficiency of the company's production line construction. At the same time, the method of the present invention, to a certain extent, eliminates the subjective decision-making interference of designers. The reference architecture extracted by the computer is more objective and has more reference value, and also has higher maturity and stability.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise production line construction, and in particular to a method for constructing a production line topology reference architecture. Background Art

[0002] Currently, traditional production line construction methods involve building a production line from start to finish based on specific production functional requirements, or combining several existing small workstations performing specific production tasks to form a production line based on their functions. These designs are essentially serial, fail to consider the overall production line, and are time-consuming to build. Furthermore, they suffer from a lack of connection and integration, and rely heavily on the designer's experience. This inevitably leads to numerous design flaws. Many complex production lines currently in operation, both domestically and internationally, fail to meet pre-designed specifications due to poor initial planning or errors. While this traditional approach to production line construction may not pose a significant problem for small businesses with a small number of production lines and a simple structure, it undoubtedly multiplies the workload and production risks for large enterprises with a large number of production lines and a complex structure. Furthermore, the inability to reuse production line structures significantly reduces production line construction efficiency and increases R&D costs. Summary of the Invention

[0003] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a method for constructing a production line topology reference architecture. This method is based on the big data of the company's historical production line topology structure, and uses computers and machine learning algorithms to extract the company's commonly used production line topology reference architecture to form the company's typical production line topology community. This is conducive to recording the company's typical production line characteristics and production habits, realizing the reuse of production line structure and production line construction knowledge, reducing the workload of production line designers, and improving the company's production line construction efficiency. At the same time, this method has to some extent eliminated the subjective decision-making interference of designers, and the reference architecture extracted by the computer is more objective and has more reference value, and also has higher maturity and stability.

[0004] To achieve the above objectives, the technical solutions provided by the present invention are:

[0005] A method for constructing a production line topology reference architecture includes the following steps:

[0006] S1. Calculate production line x A and production line x B The comprehensive similarity of

[0007] S2, combined production line x A and production line x B Comprehensive similarity calculation production line x A and production line x B The similarity S A,B;

[0008] S3, using n historical production line cases, calculate the x of any two production lines one by one according to the calculation steps of steps S1 and S2. i and production line x j Similarity between i,j , obtain the production line fuzzy compatibility matrix S containing n similarity values;

[0009] S4. Construct a multi-level granularity quotient space based on the production line fuzzy compatibility matrix S;

[0010] S5. Finding a quotient space with the optimal granularity from the multi-level granularity quotient space;

[0011] S6. Build a production line topology reference architecture based on the business space of optimal granularity.

[0012] Furthermore, calculating the comprehensive similarity between production lines specifically includes:

[0013] By calculating the production line x A Device v in A,i and production line x B Device v in B,j The matching degree and similarity of the four attributes of device v are obtained. A,i and device v B,j The comprehensive similarity of

[0014] The four attributes are raw materials, processing technology, product category, and processing quality;

[0015] According to this method, calculate the production line x one by one A Each equipment and production line in x B The pairwise comprehensive similarity s of each device in act (v A,i ,v B,j ), and then calculate the production line x based on the comprehensive similarity between these devices and the number of devices A and production line x B The comprehensive similarity s act (x A ,x B );

[0016] The calculation formula used is as follows:

[0017]

[0018]

[0019] Where,

[0020] ss fea is the device attribute v A,i and vB,j The similarity of production product categories between ss type is the device attribute v A,i and v B,j The similarity of processing quality between ss mat is the device attribute v A,i and v B,j The raw material matching degree between them is 1 if inclusive, otherwise 0; qua is the device attribute v A,i and v B,j The matching degree of the processing technology between them is 1 if it is inclusive, otherwise it is 0; V A For production line x A The number of devices in V B For production line x B The number of devices in the .

[0021] Furthermore, calculate the production line x A and production line x B The similarity S A,B include:

[0022] Calculate production line x A and production line x B The similarity of the topological structure s seq (x A ,x B ):

[0023]

[0024] Calculate production line x A and production line x B The similarity S A,B :

[0025] s A,B =w act *s act (x A ,x B )+w seq *s seq (x A ,x B ) (1-5)

[0026] Where,

[0027] M A,B For production line x A and x B The number of edge matching between the two relationships; E A For production line x A The sum of the number of edges in E B For production line x B The total number of relationship edges in wact For production line equipment on production line x A and x B The weight of the overall similarity between seq For the production line topology, in production line x A and x B The weight of the overall similarity between them.

[0028] Furthermore, the calculation formula of the production line fuzzy compatibility matrix S is as follows:

[0029]

[0030] The similarity of the same production line is 1, and s i,j =s j,i .

[0031] Furthermore, in step S4, a granular computing algorithm based on the fuzzy compatible quotient space is used to input the production line fuzzy compatible matrix S and output a series of mutually transformable granular layers {X(λ)|0≤λ≤1} with different granularities, i.e., the quotient space, where X is the granular layer and λ is the granularity. The specific calculation process is as follows:

[0032] Input production line fuzzy compatibility matrix S

[0033] Step 1: for loop: perform the first to the mth loop, i.e. for m granularity λ i Calculation is performed to obtain m granular layers. The process includes:

[0034] Step 1.1: Initialize the historical production line case set A = {x1, x2, ..., x n},gather gather

[0035] Step 1.2: Traverse loop I: Traverse any production line x in A j ;

[0036] Step 1.2.1: Set the production line x j Transfer from set A to set B;

[0037] Step 1.2.2: Traverse the loop ||: Traverse any production line x in A k ;

[0038] Step 1.2.2.1: Judgment condition: production line x j 、x k The similarity S(x j ,x k ) is greater than or equal to the granularity λ i , if not satisfied, jump to the next loop of the traversal loop ||, if satisfied, perform the following steps:

[0039] Step 1): Set the production line x k Transfer from set A to set B;

[0040] Step 2): Traverse loop III: Traverse any production line x in A s ;

[0041] Step 3): Judgment condition: Production line x k 、x s Similarity S(x k ,x s ) is greater than or equal to the granularity λ i

[0042] , if satisfied, then the production line x s Transfer from set A to set B. If the conditions are not met, jump to the next loop of traversal loop III.

[0043] Step 1.2.3: Transfer set B to set C as a subset of set C;

[0044] Step 1.2.4: Determine whether the set A is an empty set. If so, return the granularity λ. i The corresponding granular layer X(λ) = C, and the traversal loop I ends; if not, let the set Then skip this cycle and continue to the next cycle;

[0045] Step 1.3: Check the condition: Is i equal to m? If yes, terminate the for loop and the algorithm ends. If not, continue the for loop.

[0046] Finally, output m granularity λ i The corresponding granular layer X(λ)=C.

[0047] Furthermore, finding the quotient space with the optimal granularity from the multi-level granularity quotient space includes:

[0048] The Shannon information entropy concept is used to evaluate the granularity of the quotient space. The quotient space X(λ k The granularity of ) refers to the average amount of information required to completely distinguish all individual production lines in the granular layer, and the formula is as follows:

[0049]

[0050] Where,

[0051] g is the quotient space X(λ k ) The number of granules produced; G i For the quotient space X(λ k ) in the i-th production line; |G i | is the quotient space X(λ k) contains the number of production line individuals in the i-th production line particle; log2(|G i |) is to completely distinguish the production line G i The amount of information required by all production line individuals (assuming that the production line particle G i The probability of the j-th production line individual being classified into one class is equal);

[0052] The information entropy of the coarse-grained quotient space is large; from the coarse-grained quotient space X(λ k-1 ) to the fine-grained quotient space X(λ k The information gain generated during the refinement of ) is calculated as follows:

[0053] IG[X(λ k )]=E[X(λ k-1 )]-E[X(λ k )] (1-8)

[0054] Finally, the information gain and comprehensive similarity are combined to obtain the quotient space of the optimal granularity.

[0055] Furthermore, based on the optimal granularity of the business space, a production line topology reference architecture is constructed, including:

[0056] Based on the quotient space of the optimal granularity, a typical production line topology sequence between the production lines contained in each production line particle is extracted;

[0057] A dynamic programming method improved from the longest common subsequence algorithm is used to calculate the maximum length common subsequence (LCS) between all production lines in each production line particle in the quotient space of the optimal granularity (the number of LCSs belonging to different production line particles is equal to the number of production line particles in the quotient space of the optimal granularity).

[0058] For each LCS obtained above, we continue to use the ontology-based calculation method to abstract its attributes and obtain the lowest superclass of all device attributes in the domain ontology to obtain better universality and representativeness. The calculation formula is as follows:

[0059] vtr j,j∈LCS =C super (vt 1,j ,vt 2,j ,...,vt i,j ) (1-10)

[0060] C super (vt 1,j ,vt 2,j ,...,vt i,j ) represents the attribute abstraction of the j-th matched device node of all i production line individuals;

[0061] The collection of all abstracted production line equipment nodes and production line topology relationship edges is further assembled into a new production line topology reference architecture. The production line topology reference architecture corresponding to each production line particle is:

[0062] PRM i =(Vr i ,Er i ,vtr i ) (1-11)

[0063] Where,

[0064] VR i ={vr i,1 ,vr i,2 ,...,vr i,n} is the set of matching production line device nodes; vtr i It is the lowest superclass of the abstract production line equipment attributes; i ={er i,j,k |er i,j,k =vr i,j *vr i,k ,1≤j,k≤n i} is an abstract production line topology relationship edge set.

[0065] Ultimately, a production line topology reference architecture is extracted from each production line particle in the optimal granularity quotient space, and each production line topology reference architecture is represented by the above-mentioned production line device node set and relationship edge set.

[0066] Furthermore, a dynamic programming method improved from the longest common subsequence algorithm is used to calculate the maximum length of the common subsequence LCS (in terms of line x A and production line x B For example), including:

[0067] 1) Start matching from the first device node in the production line topology;

[0068] 2) Based on the recursive formula shown, match the production line device nodes one by one and stack the successfully matched device nodes to the LCS (i.j) middle:

[0069]

[0070] Where,

[0071] s t The similarity threshold preset by the user to distinguish similar devices from dissimilar devices; C super (v A,i ,vB,j ) is the attribute abstraction between two matching production line equipment nodes, i.e., the superclass;

[0072] 3) Recursively repeat steps 1) and 2) to get the final LCS (i.j) result;

[0073] 4) If a production line particle contains more than two production line individuals, the LCS obtained in the first three steps (i.j) The remaining individuals are then matched one by one, i.e., steps 1) to 3) are recursively repeated until all individuals are matched and compared, and the final LCS corresponding to the particles of the production line is obtained.

[0074] Compared with the existing technology, the principles and advantages of this solution are as follows:

[0075] This solution, based on a company's historical production line topology data, uses computers and machine learning algorithms to extract commonly used production line topology reference architectures, forming a community of typical production line topologies for that company. This helps document the company's typical production line characteristics and production habits, enabling the reuse of production line structure and construction knowledge, reducing the workload of production line designers and improving the efficiency of the company's production line construction. This solution also eliminates the subjective decision-making interference of designers to a certain extent. The reference architecture extracted by computers is more objective and valuable, and also has a high degree of maturity and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the services required for use in the embodiments or the prior art descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0077] Figure 1 This is a principle flow chart of a method for constructing a production line topology reference architecture according to the present invention;

[0078] Figure 2 Schematic diagram of fuzzy compatible quotient space. DETAILED DESCRIPTION

[0079] Before describing the specific embodiments, let’s first define the basic concepts.

[0080] Each production line involves multiple pieces of equipment working together to perform manufacturing. Different equipment has different production capabilities. This paper uses four metrics (raw materials, processing technology, product category, and processing quality) to measure these capabilities. Furthermore, a production line is composed of multiple pieces of equipment with different capabilities arranged in different sequences. The resulting arrangement is called the production line topology.

[0081] Taking a mobile phone production line as an example, commonly used equipment includes chip placement machines, labelers, glue dispensers, soldering machines, glue dispensers, and inkjet printers. Each device has four attributes: a. Raw materials: plastics, electronic components, hardware, and packaging materials; b. Processing technology: solder paste printing, painting, silk screen printing, welding, and assembly; c. Product category: 3C products, home appliances, food, and apparel; and d. Processing quality: roughing, finishing, and precision machining. The topology of a mobile phone production line is the resulting arrangement of these devices according to different production plans and processing sequences. Obviously, different products require different types and quantities of equipment, as well as different equipment arrangements. Examples include linear, U-shaped, tree-shaped, and ring-shaped production lines.

[0082] The present invention will be further described below in conjunction with specific embodiments:

[0083] like Figure 1 As shown, the method for constructing a production line topology reference architecture described in this embodiment includes three stages:

[0084] First, the similarity of production lines is calculated based on device attributes and production line topology. This stage adopts ontology inclusion and semantic similarity calculation techniques. Second, a fuzzy compatibility quotient space is constructed. By comprehensively considering and calculating information gain and similarity, the optimal granularity for clustering production line topology structures is found to obtain more representative and accurate particles. Finally, by analyzing the similarity and matching relationship between the relationship edges of device nodes and production line topology structures, the common subgraph of each particle in the optimal granular layer is identified and merged to form a reference architecture of the production line topology structure.

[0085] The specific steps are as follows:

[0086] S1. Calculate production line x A and production line x B The comprehensive similarity of , the process includes:

[0087] By calculating the production line x A Device v in A,i and production line x B Device v in B,j The matching degree and similarity of the four attributes of device v are obtained. A,i and device vB,j The comprehensive similarity of

[0088] The four attributes are raw materials, processing technology, product category, and processing quality;

[0089] According to this method, calculate the production line x one by one A Each equipment and production line in x B The pairwise comprehensive similarity s of each device in act (v A,i ,v B,j ), and then calculate the production line x based on the comprehensive similarity between these devices and the number of devices A and production line x B The comprehensive similarity s act (x A ,x B );

[0090] The calculation formula used is as follows:

[0091]

[0092]

[0093] Where,

[0094] ss fea is the device attribute v A,i and v B,j The similarity of production product categories between ss type is the device attribute v A,i and v B,j The similarity of processing quality between ss mat is the device attribute v A,i and v B,j The raw material matching degree between them is 1 if inclusive, otherwise 0; qua is the device attribute v A,i and v B,j The matching degree of the processing technology between them is 1 if it is inclusive, otherwise it is 0; V A For production line x A The number of devices in V B For production line x B The number of devices in the .

[0095] S2, combined production line x A and production line x B Comprehensive similarity calculation production line x A and production line x B The similarity S A,B The process includes:

[0096] Calculate production line x Aand production line x B The similarity of the topological structure s seq (x A ,x B ):

[0097]

[0098] Calculate production line x A and production line x B The similarity S A,B :

[0099] s A,B =w act *s act (x A ,x B )+w seq *s seq (x A ,x B ) (1-5)

[0100] Where,

[0101] M A,B For production line x A and x B The number of edge matching between the two relationships; E A For production line x A The sum of the number of edges in E B For production line x B The total number of relationship edges in w ac t is the production line equipment in production line x A and x B The weight of the overall similarity between seq For the production line topology, in production line x A and x B The weight of the overall similarity between them.

[0102] S3, using n historical production line cases, calculate the x of any two production lines one by one according to the calculation steps of steps S1 and S2. i and production line x j Similarity between i,j , we get the production line fuzzy compatibility matrix S containing n similarity values; the calculation formula of the production line fuzzy compatibility matrix S is as follows:

[0103]

[0104] The similarity of the same production line is 1, and s i,j =s j,i .

[0105] S4. Construct a multi-level granularity quotient space based on the production line fuzzy compatibility matrix S. The process is as follows:

[0106] Using the granular computing algorithm based on the fuzzy compatible quotient space, the production line fuzzy compatible matrix S is input and a series of mutually transformable granular layers {X(λ)|0≤λ≤1} with different granularities are output, i.e., the quotient space, where X is the granular layer and λ is the granularity. The specific calculation process is as follows:

[0107] Input production line fuzzy compatibility matrix S

[0108] Step 1: for loop: perform the first to the mth loop, i.e. for m granularity λ i Calculation is performed to obtain m granular layers. The process includes:

[0109] Step 1.1: Initialize the historical production line case set A = {x1, x2, ..., x n},gather gather

[0110] Step 1.2: Traverse loop I: Traverse any production line x in A j ;

[0111] Step 1.2.1: Set the production line x j Transfer from set A to set B;

[0112] Step 1.2.2: Traverse the loop ||: Traverse any production line x in A k ;

[0113] Step 1.2.2.1: Judgment condition: production line x j 、x k The similarity S(x j ,x k ) is greater than or equal to the granularity λ i , if not satisfied, jump to the next loop of the traversal loop ||, if satisfied, perform the following steps:

[0114] Step 1): Set the production line x k Transfer from set A to set B;

[0115] Step 2): Traverse loop III: Traverse any production line x in A s ;

[0116] Step 3): Judgment condition: Production line x k 、x s Similarity S(x k ,x s ) is greater than or equal to the granularity λ i , if satisfied, then the production line x s Transfer from set A to set B. If the conditions are not met, jump to the next loop of traversal loop III.

[0117] Step 1.2.3: Transfer set B to set C as a subset of set C;

[0118] Step 1.2.4: Determine whether the set A is an empty set. If so, return the granularity λ. i The corresponding granular layer X(λ) = C, and the traversal loop I ends; if not, let the set Then skip this cycle and continue to the next cycle;

[0119] Step 1.3: Check the condition: Is i equal to m? If yes, terminate the for loop and the algorithm ends. If not, continue the for loop.

[0120] Finally, output m granularity λ i The corresponding granular layer X(λ)=C.

[0121] The algorithm finally outputs the quotient space S containing all the production line granularities. Figure 2 As shown in Figure 2, different λ values ​​correspond to quotient spaces of different granularities, i.e., granular layers of different granularities. A larger λ value means a finer granularity in the quotient space, more production line particles in its granular layer, and fewer production line individuals in each production line particle.

[0122] S5. In the hierarchical structure of S, although it contains a series of granular layers, not all granular layers can provide as much valuable information as possible to support the construction of the production line topology reference architecture. Since the number of production line particles in the quotient space is proportional to the granularity of the quotient space, although more particles can obtain more reference structures, when the granularity of the quotient space becomes finer (that is, the λ value becomes larger), the similarity threshold between the production line individuals that generate this granular layer is high; this means that the typical production line topology structure cannot be effectively abstracted, which in turn leads to a long sequence of typical production line topology structures, making the extracted typical production line topology structure sequence not universal. On the contrary, a coarser granularity of the quotient space (that is, a smaller λ value) means a lower similarity threshold for generating granular layers, which in turn means fewer production line particles and lower similarity between production line individuals; this situation will cause the typical production line topology structure sequence to be shorter, making the extracted typical production line topology structure sequence not adaptable. Therefore, in order to find a more adaptable and universal typical production line topology reference architecture, this embodiment uses two metrics for measuring granularity (information gain and minimum similarity) to find the optimal granularity layer. The process includes:

[0123] The Shannon information entropy concept is used to evaluate the granularity of the quotient space. The quotient space X(λ k The granularity of ) refers to the average amount of information required to completely distinguish all individual production lines in the granular layer, and the formula is as follows:

[0124]

[0125] Where,

[0126] g is the quotient space X(λ k ) The number of granules produced; G i For the quotient space X(λ k ) in the i-th production line; |G i | is the quotient space X(λ k ) contains the number of production line individuals in the i-th production line particle; log2(|G i |) is to completely distinguish the production line G i The amount of information required by all production line individuals (assuming that the production line particle G i The probability of the j-th production line individual being classified into one class is equal);

[0127] The information entropy of the coarse-grained quotient space is large; from the coarse-grained quotient space X(λ k-1 ) to the fine-grained quotient space X(λ k The information gain generated during the refinement of ) is calculated as follows:

[0128] IG[X(λ k )]=E[X(λ k-1 )]-E[X(λ k )] (1-8)

[0129] Finally, the quotient space with both large information gain and large individual minimum similarity is taken as the quotient space with the optimal granularity.

[0130] S6. Build a production line topology reference architecture based on the optimal granularity of the business space. The process is as follows:

[0131] Based on the quotient space of the optimal granularity, a typical production line topology sequence between the production lines contained in each production line particle is extracted;

[0132] The dynamic programming method improved from the longest common subsequence algorithm is used to calculate the maximum length of the common subsequence LCS (in terms of line x A and production line x B For example), specifically including:

[0133] 1) Start matching from the first device node in the production line topology;

[0134] 2) Based on the recursive formula shown, match the production line device nodes one by one and stack the successfully matched device nodes to the LCS (i.j) middle:

[0135]

[0136] Where,

[0137] s t The similarity threshold preset by the user to distinguish similar devices from dissimilar devices; C super (v A,i ,v B,j ) is the attribute abstraction between two matching production line equipment nodes, that is, the superclass; act (v A,i ,v B,j ) is production line x A Each equipment and production line in x B The pairwise comprehensive similarity of each device in ;

[0138] 3) Recursively repeat steps 1) and 2) to get the final LCS (i.j) result;

[0139] 4) If a production line particle contains more than two production line individuals, the LCS obtained in the first three steps (i.j) Then match the remaining individuals one by one, i.e. recursively repeat steps 1) to 3) until all individuals are matched and compared, and the final LCS corresponding to the particles of the production line is obtained;

[0140] For each LCS obtained above, we continue to use the ontology-based calculation method to abstract its attributes and obtain the lowest superclass of all device attributes in the domain ontology to obtain better universality and representativeness. The calculation formula is as follows:

[0141] vtr j,j∈LCS =C super (vt 1,j ,vt 2,j ,...,vt i,j ) (1-10)

[0142] C super (vt 1,j ,vt 2,j ,...,vt i,j ) represents the attribute abstraction of the j-th matched device node of all i production line individuals;

[0143] The collection of all abstracted production line equipment nodes and production line topology relationship edges is further assembled into a new production line topology reference architecture. The production line topology reference architecture corresponding to each production line particle is:

[0144] PRM i =(Vr i ,Er i ,vtr i ) (1-11)

[0145] Where,

[0146] VR i ={vr i,1 ,vr i,2 ,...,vr i,n} is the set of matching production line device nodes; vtr i It is the lowest superclass of the abstract production line equipment attributes; i ={er i,j,k |er i,j,k =vr i,j *vr i,k ,1≤j,k≤n i} is an abstract production line topology relationship edge set.

[0147] Ultimately, a production line topology reference architecture is extracted from each production line particle in the optimal granularity quotient space, and each production line topology reference architecture is represented by the above-mentioned production line device node set and relationship edge set.

[0148] The embodiments described above are only preferred embodiments of the present invention and are not intended to limit the scope of implementation of the present invention. Therefore, any changes made based on the shape and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing a production line topology reference architecture, characterized in that: The following steps are involved: S1. Calculate production line x A and production line x B The comprehensive similarity of S2, combined production line x A and production line x B Comprehensive similarity calculation production line x A and production line x B The similarity S A,B ; S3, using n historical production line cases, calculate the x of any two production lines one by one according to the calculation steps of steps S1 and S2. i and production line x j Similarity between i,j , obtain the production line fuzzy compatibility matrix S containing n similarity values; S4. Construct a multi-level granularity quotient space based on the production line fuzzy compatibility matrix S; S5. Finding a quotient space with the optimal granularity from the multi-level granularity quotient space; S6. Build a production line topology reference architecture based on the optimal granularity of the business space; Based on the optimal granularity of the business space, a production line topology reference architecture is constructed, including: Based on the quotient space of the optimal granularity, a typical production line topology sequence between the production lines contained in each production line particle is extracted; A dynamic programming method improved from the longest common subsequence algorithm is used to calculate the maximum length common subsequence LCS between all production lines in each production line particle in the quotient space of the optimal granularity; For each LCS obtained, we continue to use the ontology-based calculation method to abstract its attributes and obtain the lowest superclass of all device attributes in the domain ontology to obtain better universality and representativeness. The calculation formula is as follows: to j,j∈LCS =C super (see 1,j ,vt 2,j ,...,vt i,j ) (1-10) C super (vt 1,j ,vt 2,j ,...,vt i,j ) represents the attribute abstraction of the j-th matched device node of all i production line individuals; The collection of all abstracted production line equipment nodes and production line topology relationship edges is further assembled into a new production line topology reference architecture. The production line topology reference architecture corresponding to each production line particle is: PRM i =(Turn i ,Is i ,vtr i ) (1-11) Where, VR i ={vr i,1 ,vr i,2 ,...,vr i,n } is the set of matching production line device nodes; vtr i It is the lowest superclass of the abstract production line equipment attributes; i ={er i,j,k |er i,j,k =vr i,j *vr i,k ,1≤j,k≤n i } is the abstract production line topology relationship edge set; Ultimately, a production line topology reference architecture is extracted from each production line particle in the optimal granularity quotient space. Each production line topology reference architecture is represented by the aforementioned set of production line device nodes and relationship edge sets. A dynamic programming method improved from the longest common subsequence algorithm is used to calculate the maximum length common subsequence LCS between all production lines in each production line particle in the quotient space of the optimal granularity, including: 1) Start matching from the first device node in the production line topology; 2) Based on the recursive formula, match the production line device nodes one by one, and stack the successfully matched device nodes to the LCS (i.j) middle: Where, s t The similarity threshold preset by the user to distinguish similar devices from dissimilar devices; C super (v A,i ,v B,j ) is the attribute abstraction between two matching production line equipment nodes, that is, the superclass; act (v A,i ,v B,j ) is production line x A Each equipment and production line in x B The pairwise comprehensive similarity of each device in ; 3) Recursively repeat steps 1) and 2) to get the final LCS (i.j) result; 4) If a production line particle contains more than two production line individuals, the LCS obtained in the first three steps (i.j) The remaining individuals are then matched one by one, i.e., steps 1) to 3) are recursively repeated until all individuals are matched and compared, and the final LCS corresponding to the particles of the production line is obtained.

2. The method for constructing a production line topology reference architecture according to claim 1, characterized in that: Calculating the comprehensive similarity between production lines specifically includes: By calculating the production line x A Device v in A,i and production line x B Device v in B,j The matching degree and similarity of the four attributes of device v are obtained. A,i and device v B,j The comprehensive similarity of The four attributes are raw materials, processing technology, product category, and processing quality; According to this method, calculate the production line x one by one A Each equipment and production line in x B The pairwise comprehensive similarity s of each device in act (v A,i ,v B,j ), and then calculate the production line x based on the comprehensive similarity between these devices and the number of devices A and production line x B The comprehensive similarity s act (x A ,x B ); The calculation formula used is as follows: Where, ss fea is the device attribute v A,i and v B,j The similarity of production product categories between ss type is the device attribute v A,i and v B,j The similarity of processing quality between ss mat is the device attribute v A,i and v B,j The raw material matching degree between them is 1 if inclusive, otherwise 0; qua is the device attribute v A,i and v B,j The matching degree of the processing technology between them is 1 if it is inclusive, otherwise it is 0; V A For production line x A The number of devices in V B For production line x B The number of devices in the .

3. The method for constructing a production line topology reference architecture according to claim 1, characterized in that: Calculate production line x A and production line x B The similarity S A,B include: Calculate production line x A and production line x B The similarity of the topological structure s seq (x A ,x B ): Calculate production line x A and production line x B The similarity S A,B : s A,B =w act *s act (x A ,x B )+w seq *s seq (x A ,x B ) (1-5) Where, M A,B For production line x A and x B The number of edge matching between the two relationships; E A For production line x A The sum of the number of edges in E B For production line x B The total number of relationship edges in w act For production line equipment on production line x A and x B The weight of the overall similarity between seq For the production line topology, in production line x A and x B The weight of the overall similarity between act (x A ,x B ) is production line x A and production line x B The comprehensive similarity.

4. The method for constructing a production line topology reference architecture according to claim 1, wherein: The calculation formula of the production line fuzzy compatibility matrix S is as follows: The similarity of the same production line is 1, and s i,j =s j,i .

5. The method for constructing a production line topology reference architecture according to claim 1, wherein: In step S4, a granular computing algorithm based on fuzzy compatible quotient space is used to input the production line fuzzy compatible matrix S and output a series of mutually transformable granular layers {X(λ)|0≤λ≤1} with different granularities, i.e., quotient space, where X is the granular layer and λ is the granularity. The specific calculation process is as follows: Input production line fuzzy compatibility matrix S Step 1: for loop: perform the first to the mth loop, i.e. for m granularity λ i Calculation is performed to obtain m granular layers. The process includes: Step 1.1: Initialize the historical production line case set A = {x1, x2, ..., x n },gather gather Step 1.2: Traverse loop I: Traverse any production line x in A j ; Step 1.2.1: Set the production line x j Transfer from set A to set B; Step 1.2.2: Traverse loop II: Traverse any production line x in A k ; Step 1.2.2.1: Judgment condition: production line x j 、x k The similarity S(x j ,x k ) is greater than or equal to the granularity λ i , if not, jump to the next loop of traversal loop II, if satisfied, perform the following steps: Step 1): Set the production line x k Transfer from set A to set B; Step 2): Traverse loop III: Traverse any production line x in A s ; Step 3): Judgment condition: Production line x k 、x s The similarity S(x k ,x s ) is greater than or equal to the granularity λ i If satisfied, then the production line x s Transfer from set A to set B. If the conditions are not met, jump to the next loop of traversal loop III. Step 1.2.3: Transfer set B to set C as a subset of set C; Step 1.2.4: Determine whether the set A is an empty set. If so, return the granularity λ. i The corresponding granular layer X(λ) = C, and the traversal loop I ends; if not, let the set Then skip this cycle and continue to the next cycle; Step 1.3: Check the condition: Is i equal to m? If yes, terminate the for loop and the algorithm ends. If not, continue the for loop. Finally, output m granularity λ i The corresponding granular layer X(λ)=C.

6. The method for constructing a production line topology reference architecture according to claim 1, characterized in that: Finding the quotient space with the optimal granularity from the multi-level granularity quotient space includes: The Shannon information entropy concept is used to evaluate the granularity of the quotient space. The quotient space X(λ k The granularity of ) refers to the average amount of information required to completely distinguish all individual production lines in the granular layer, and the formula is as follows: Where, g is the quotient space X(λ k ) The number of granules produced; G i For the quotient space X(λ k ) in the i-th production line; |G i | is the quotient space X(λ k ) contains the number of production line individuals in the i-th production line particle; log2(|G i |) is to completely distinguish the production line G i The amount of information required for all production line individuals in the The information entropy of the coarse-grained quotient space is large; from the coarse-grained quotient space X(λ k-1 ) to the fine-grained quotient space X(λ k The information gain generated during the refinement of ) is calculated as follows: IG[X(λ k )]=E[X(λ k-1 )]-E[X(λ k )] (1-8) Finally, the information gain and comprehensive similarity are combined to obtain the quotient space of the optimal granularity.

Images