Method and system for constructing manufacturing industry space graph based on artificial intelligence graph reasoning

Abstract

This application provides a method and system for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning, relating to the field of electronic digital data processing technology. It includes: S1, extracting the supply chain topology network of logistics-sensitive manufacturing industries and constructing a directed product graph; S2, defining and mapping the supply chain relationship attenuation coefficient as it decreases with increasing shortest path step length, based on the characteristics of logistics-sensitive manufacturing industries; S3, calculating the city's industrial distribution advantage; S4, calculating the strength of directed associations between cities based on the city's industrial distribution advantage and hierarchical attenuation; and S5, constructing a directed spatial map of the city's industries. This method combines the physical objective law of logistics transportation costs decreasing with processing levels and uses artificial intelligence relationship reasoning to achieve electronic digital data processing of supply chain nodes and upstream and downstream relationships, enabling a close integration of AI-driven data processing with the specific physical characteristics of industrial spatial layout.

Description

Technical Field

[0001] This application relates to the field of electronic digital data processing technology, specifically to a method and system for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning. Background Technology

[0002] With the acceleration of regional economic integration and the continuous deepening of digital transformation of supply chains, quantitative analysis of the spatial layout of industrial chains has become a core technical support for regional industrial planning, cross-regional production capacity collaboration, and macroeconomic decision-making.

[0003] Current mainstream analytical approaches to supply chain relationships construct industry maps centered on industry topology networks. These approaches focus on the upstream and downstream business logic and production support relationships between products, building a purely business-oriented supply chain topology structure. However, these technical solutions are not deeply integrated with the physical and geographical context of industry implementation, completely lacking the geospatial attributes of each link in the supply chain. They fail to reflect the distribution characteristics and spatial relationships of specific industry links in different geographical units, making it difficult to support refined industrial planning decisions for regional implementation.

[0004] On the other hand, existing technologies for analyzing regional economic spatial linkages mostly employ traditional spatial economic linkage models, such as the gravity model. These models typically quantify the overall economic linkage strength between cities based on macroeconomic statistical indicators such as total GDP and population size. However, the granularity of the analysis can only remain at the macroeconomic level of cities, and cannot be extended to the specific production links of a particular industrial chain, let alone accurately depict the spatial linkage characteristics of different processing levels of the industrial chain.

[0005] Especially for logistics-sensitive process manufacturing industries such as petrochemicals, ethylene / olefins, and metallurgy, their industrial chains are characterized by strong reliance on physical transportation of upstream and downstream products and continuous processing. Therefore, the spatial layout of these industries is highly sensitive to logistics costs and production continuity. Within the industrial chain of such industries, the closer the processing levels of upstream and downstream products are (e.g., a direct upstream-downstream complementary relationship), the higher the spatial logistics coordination needs and dependence in the production process. If products span multiple processing stages, the spatial agglomeration effect and the strength of direct logistics linkages exhibit a significant step-like attenuation characteristic.

[0006] However, existing technologies do not yet have a quantitative analysis algorithm that can deeply integrate the product hierarchy correlation attenuation characteristics of the industrial chain, the distribution of production capacity of subdivided products between cities, and artificial intelligence relationship recognition and graph reasoning models. Therefore, it is impossible to accurately quantify the strength of targeted collaborative correlation between different cities in specific links of the industrial chain for the aforementioned logistics-sensitive process manufacturing industry. Consequently, it is difficult to meet the actual decision-making needs of precise layout of regional industrial chains and cross-city production capacity collaborative optimization. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this application aims to provide a method and system for constructing a manufacturing industry spatial map based on artificial intelligence graph reasoning. By performing AI entity recognition, upstream and downstream relationship extraction, node importance ranking, and candidate edge confidence assessment on industry chain text data, product description data, and enterprise statistical data, the core product nodes and their upstream and downstream relationships in the target industry chain are identified. Furthermore, the graph structure is validated by combining industry chain knowledge graphs, chemical or manufacturing processing logic, and rule bases. Based on this, a product hierarchy attenuation coefficient is defined for a specific industry chain, and combined with the city's industrial distribution advantage, the strength of directional spatial associations between cities caused by physical processing levels and logistics costs is accurately calculated, generating a directed weighted urban industry spatial map. This provides underlying algorithmic support for regional layout optimization, logistics cost accounting, and strengthening and supplementing the supply chain in heavy chemical industries and process manufacturing industries.

[0008] Specifically, on the one hand, this application provides a method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning, which includes the following steps: S1. Collect supply chain text data, product description data, enterprise statistical data and rule base data of logistics-sensitive manufacturing industries, combine them with supply chain knowledge graphs, use artificial intelligence relationship reasoning to identify the core product nodes of the target supply chain and their upstream and downstream relationships, generate candidate nodes and candidate edges, and complete the verification by combining chemical or manufacturing processing logic, construct a product directed graph, and calculate the shortest path step length between any two product nodes. S2. Define and map the supply chain relationship attenuation coefficient based on the characteristics of logistics-sensitive manufacturing industries. Decrease as the shortest path step size increases; S3. Calculate the city's industrial distribution advantage. Extract the production capacity or number of enterprises of each city at each product node, and calculate the relative advantage of city i in product p as the city's industrial distribution advantage. The calculation formula is: ; in, This represents the production capacity or number of enterprises in city i that produce product p. This represents the total production capacity or number of enterprises producing product p in all cities, where j refers to the j-th city among all cities. S4. Calculate the strength of directed inter-city associations based on the city's industrial distribution dominance and hierarchical decay. For any two cities i and j, and adjacent product pairs in the industry chain. Calculate the directed correlation strength The calculation formula is: ; in, For City i in Products The city's industrial distribution advantage For city j in products The degree of urban industrial distribution advantage; S5. Construct a directed graph of urban industrial space: Traverse all product pairs and accumulate the total logistics space edge weight matrix from city i to city j. matrix elements , It is an asymmetric matrix, based on the normalized total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries.

[0009] Preferably, the artificial intelligence relationship reasoning in S1 refers to: performing text preprocessing, entity recognition, relationship extraction, node importance ranking, and confidence assessment on industrial chain text data, product description data, and enterprise statistical data, so as to output candidate nodes and candidate edges of the product directed graph.

[0010] Preferably, S1 specifically includes the following sub-steps: S11. Extract core product nodes: Based on the chemical or manufacturing processing logic of the target industry chain, collect industry chain text data, product description data and enterprise statistics, clean the data, normalize the terminology and standardize the product names; use artificial intelligence entity recognition model to identify entities of raw materials, intermediate products, finished products and process links, and combine word frequency, production capacity, enterprise coverage and node connectivity to analyze the importance of nodes, and select the core products corresponding to each link as nodes of the graph; S12. Based on the material flow relationship between products, and combined with the results of artificial intelligence relationship reasoning, directed edges are constructed. The specific process is as follows: An artificial intelligence relationship extraction model is used to identify the raw material-product relationship, upstream-downstream relationship, or processing and transformation relationship between products, generating candidate edges and corresponding confidence levels; then, the candidate edges are verified according to the rule base. If product P... A For product P B For the direct production raw materials, a route is established from node P. A Pointing to node P B The directed edges; S13. Calculate the shortest path step length: Use the shortest path algorithm in graph theory to calculate the shortest path step length between any two product nodes. The shortest path step length is defined as the minimum number of directed edges required to traverse from the starting product node to the target product node. S14. No-path processing: Based on the directed product graph obtained in S11 and S12 and verified by the rule base, perform reachability retrieval for any two product nodes; if there is no directed path from the starting product node to the target product node, or if the corresponding candidate relationship fails the confidence threshold screening and rule verification, it is determined that the two nodes have no effective upstream and downstream association, and the shortest path step size is recorded as infinity or directly determined as no association.

[0011] Preferably, in S2, the industrial chain relationship attenuation coefficient is 1.0 when the shortest path step size is 1, 0.6 when the shortest path step size is 2, 0.3 when the shortest path step size is 3, 0.1 when the shortest path step size is 4, and 0 when the shortest path step size is greater than or equal to 5 or when there are no two related product nodes.

[0012] Preferably, the value range of the urban industrial distribution advantage degree in S3 is [0,1].

[0013] Preferably, the total logistics space edge weight matrix from city i to city j in S5 matrix elements The sum of the correlation strengths for all products: .

[0014] Preferably, the total logistics space edge weight matrix from city i to city j For an N×N square matrix, the specific expression is as follows: ; Wherein, the total logistics space edge weight matrix from city i to city j matrix elements The specific expression is as follows: ; Wherein, the diagonal matrix elements W(i,i)=0 and W(i,j)≠W(j,i), and P is the product set.

[0015] Preferably, step S5 specifically includes the following sub-steps: S51. Matrix initialization: Set an NxN zero matrix, where N is the number of cities, and all elements are initialized to 0. S52. Traversal and Accumulation: Traverse all ordered product pairs (p1, p2) in the product set P, calculate the directed association strength from city i to city j for each product pair, and accumulate them sequentially into the total logistics space edge weight matrix from city i to city j. matrix elements middle; S53. Normalization Process: Weighting the total logistics space edge matrix from city i to city j Min-Max normalization is performed, mapping all elements to the interval [0,1]. The calculation formula is as follows: ; in, These are the normalized matrix elements. The maximum value among all elements of the matrix; S54. Remove low-weight edge connections: Set a minimum threshold θ, and apply the following to the elements of the normalized matrix: Perform the following operations: .

[0016] Preferably, minimum threshold The range of values ​​is .

[0017] On the other hand, this application provides an industrial spatial map construction system based on an artificial intelligence graph reasoning method for constructing manufacturing industrial spatial maps, which includes a shortest path step length calculation unit, an industrial chain relationship attenuation coefficient definition unit, an urban industrial distribution advantage calculation unit, a directed association strength calculation unit, and a directed graph construction unit. The shortest path step length calculation unit extracts the supply chain topology network of logistics-sensitive manufacturing industries, combines supply chain knowledge graphs and artificial intelligence relationship reasoning to identify the core product nodes of the target supply chain and their upstream and downstream relationships, constructs a product directed graph, and calculates the shortest path step length between any two product nodes. The definition unit for the supply chain relationship attenuation coefficient is based on the characteristics of logistics-sensitive manufacturing industries and defines and maps the supply chain relationship attenuation coefficient. Decrease as the shortest path step size increases; The urban industrial distribution advantage calculation unit extracts the production capacity or number of enterprises of each city at each product node, and calculates the relative advantage of city i on product p as the urban industrial distribution advantage. ; The directed correlation strength calculation unit is used for any two cities i and j, and product pairs in the industrial chain. Calculate the directed correlation strength ; The directed graph construction unit traverses all product pairs and accumulates them to obtain the total logistics space edge weight matrix from city i to city j. matrix elements middle, It is an asymmetric matrix, based on the total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries.

[0018] Compared with the prior art, the beneficial effects of this application are as follows: (1) The manufacturing industry spatial map construction method based on artificial intelligence graph reasoning in this application deeply integrates the physical objective law that logistics transportation costs decrease with processing level in logistics-sensitive / process manufacturing industries. It closely combines artificial intelligence-driven digital data processing with the physical characteristics of specific industrial spatial layout, and can generate a directed weighted urban industrial spatial map, providing underlying algorithm support for regional layout optimization, logistics cost accounting and strengthening and supplementing the supply chain in heavy chemical industry and process manufacturing industry.

[0019] (2) The manufacturing industry spatial map construction method based on artificial intelligence graph reasoning in this application introduces a hierarchical attenuation coefficient, which truly restores the spatial decoupling effect of heavy industry in cross-processing. It quantifies the processing level distance between products in the industrial chain into a calculable attenuation coefficient, which is a key innovation connecting industrial topology and spatial layout. At the same time, the method in this application integrates the city's industrial distribution advantage with the hierarchical attenuation coefficient, realizing the first quantitative combination of industrial level distance and city production capacity scale.

[0020] (3) The manufacturing industry spatial map construction method based on artificial intelligence graph reasoning in this application generates an asymmetric matrix based on a specific computational model, which can reflect the upstream and downstream directional relationships between cities. It is different from the traditional undirected or symmetric spatial association model and has application innovation value. The spatial map generated based on the asymmetric matrix is ​​an asymmetric directed graph that can accurately identify upstream resource-based supply hubs and downstream processing and consumption hubs. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall method flow for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning in this application; Figure 2 This is a mapping diagram of the hierarchical span and relationship attenuation coefficient of the logistics-sensitive industrial chain in this application; Figure 3 This is a schematic diagram illustrating the principle of weight calculation and attenuation effect of directed edges between cities in this application. Figure 4 This application provides a schematic diagram of the system structure for constructing the industrial spatial map. Figure 5 This is a schematic diagram of the complex urban spatial network of the national olefin industry chain in this application embodiment. Detailed Implementation

[0022] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings.

[0023] Specifically, on the one hand, this application provides a method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning, such as... Figures 1-3 As shown, it includes the following steps: S1 involves collecting supply chain text data, product description data, enterprise statistics data, and rule base data from logistics-sensitive manufacturing industries. Combining this with a supply chain knowledge graph, it utilizes artificial intelligence (AI) relational reasoning to identify core product nodes and their upstream and downstream relationships within the target supply chain. This generates candidate nodes and edges, which are then validated using chemical or manufacturing processing logic. A directed product graph is constructed, and the shortest path length between any two product nodes is calculated. The AI ​​relational reasoning in S1 refers to: preprocessing the supply chain text data, product description data, and enterprise statistics data, performing text preprocessing, entity recognition, relation extraction, node importance ranking, and confidence assessment to output candidate nodes and edges for the directed product graph. Specifically: First, the text data, product description data, and enterprise statistics of the industrial chain are processed by word segmentation, noise reduction, terminology normalization, and structuring. Second, an artificial intelligence entity recognition model is used to extract product entities, process entities, and enterprise entities. Third, an artificial intelligence relation extraction model is used to identify the raw material-product relationships, processing and transformation relationships, and upstream and downstream relationships between entities, and a confidence level is assigned to each candidate relationship. Then, the importance of nodes is ranked based on entity frequency, enterprise coverage, production capacity, and network connectivity. Finally, the candidate nodes and candidate edges are checked for consistency with the manufacturing and processing rule base, and candidate relationships that do not conform to the process logic or have a confidence level below the threshold are deleted, outputting a directed graph of products.

[0024] S1 specifically includes the following sub-steps: S11. Extracting Core Product Nodes: Based on the chemical or manufacturing processing logic of the target industry chain, collect textual data, product description data, and enterprise statistical data of the industry chain. Clean the data, normalize the terminology, and standardize product names. Use an artificial intelligence entity recognition model to identify entities in raw materials, intermediate products, finished products, and process links. Combine word frequency, production capacity, enterprise coverage, and node connectivity to analyze node importance and select the core products corresponding to each link as nodes in the graph. For example, in the ethylene industry chain, core product nodes include: crude oil, naphtha, ethylene, polyethylene, ethylene glycol, etc.

[0025] S12. Based on the material flow relationship between products, and combined with the results of artificial intelligence relationship reasoning, directed edges are constructed. The specific process is as follows: An artificial intelligence relationship extraction model is used to identify the raw material-product relationship, upstream-downstream relationship, or processing and transformation relationship between products, generating candidate edges and corresponding confidence levels; then, the candidate edges are verified according to the rule base. If product P... A For product P B For the direct production raw materials, a route is established from node P. A Pointing to node P BThe directed edges of the product graph. For example: crude oil → naphtha, naphtha → ethylene, ethylene → polyethylene. All product nodes and directed edges together constitute a directed product graph.

[0026] S13. Calculate the shortest path step length: Using the shortest path algorithm in graph theory, calculate the shortest path step length between any two product nodes. The shortest path step length is defined as the minimum number of directed edges required to traverse from the starting product node to the target product node.

[0027] S14. No-Path Handling: Based on the directed product graph obtained in S11 and S12 and validated by the rule base, a reachability search is performed on any two product nodes. If there is no directed path from the starting product node to the target product node, or if the corresponding candidate relationship fails the confidence threshold screening and rule validation, the two nodes are determined to have no effective upstream and downstream association, and the shortest path step size is recorded as infinity or directly determined to be unrelated. In this specific embodiment, the relationships between the product nodes are as follows: Crude oil → naphtha → ethylene → polyethylene; Crude oil → naphtha → ethylene → ethylene glycol; Naphtha → Propylene → Polypropylene; The shortest path step lengths for each product pair are as follows: Crude oil and naphtha: Step 1 means that the two are directly upstream and downstream; Crude oil and ethylene: 2 steps means that the two are transferred through naphtha; Crude oil and polyethylene: 3 steps represent the process of transferring between the two via naphtha and ethylene; Ethylene and polyethylene: Step 1 means that the two are directly upstream and downstream; Naphtha and Polypropylene: 2 steps represent the process of transferring between the two via propylene. Crude oil and polypropylene: 3 steps represent the process of transferring between the two via naphtha and propylene; Polyethylene and polypropylene: There is no direct path between the two, so they are considered unrelated.

[0028] S2. Define and map the supply chain relationship attenuation coefficient based on the characteristics of logistics-sensitive manufacturing industries. It decreases as the shortest path step size increases. The supply chain relationship attenuation coefficient is calculated when the shortest path step size in S2 is 1. The supply chain relationship attenuation coefficient is 1.0 when the shortest path step size is 2. The supply chain relationship attenuation coefficient is 0.6 when the shortest path step size is 3. The supply chain relationship attenuation coefficient is 0.3 when the shortest path step size is 4. The supply chain relationship attenuation coefficient is 0.1 when the shortest path step size is greater than or equal to 5 or when there are no two related product nodes. It is 0.

[0029] S3. Calculate the city's industrial distribution advantage. Extract the production capacity or number of enterprises of each city at each product node, and calculate the relative advantage of city i in product p as the city's industrial distribution advantage. The calculation formula is: ; in, This represents the production capacity or number of enterprises in city i that produce product p. This represents the total production capacity or number of enterprises producing product p in all cities, where j refers to the j-th city among all cities. In a specific embodiment, S3 represents the city's industrial distribution dominance. The value range is [0,1].

[0030] S4. Calculate the strength of directed inter-city associations based on the city's industrial distribution dominance and hierarchical decay. For any two cities i and j, and adjacent product pairs in the industry chain. Calculate the directed correlation strength The calculation formula is: ; in, For City i in Products The city's industrial distribution advantage For city j in products The degree of advantage in the distribution of urban industries.

[0031] S5. Construct a directed graph of urban industrial space: Traverse all product pairs and accumulate the total logistics space edge weight matrix from city i to city j. matrix elements , It is an asymmetric matrix, based on the normalized total logistics space edge weight matrix from city i to city j. Generate a visualized complex network of industrial urban space to obtain a spatial map of logistics-sensitive manufacturing industries.

[0032] Among them, the total logistics space edge weight matrix from city i to city j in S5 matrix elements The sum of the correlation strengths for all products: .

[0033] Total logistics space edge weight matrix from city i to city j The specific expression for an N×N square matrix is ​​as follows: ; Wherein, the total logistics space edge weight matrix from city i to city j matrix elements The specific expression is as follows: ; Among them, the diagonal matrix elements W(i,i)=0 and W(i,j)≠W(j,i).

[0034] In a specific embodiment, S5 constructs the total logistics spatial edge weight matrix from city i to city j. The steps are as follows: S51, Matrix Initialization: Set an NxN zero matrix. , where N is the number of cities, and all elements are initialized to 0.

[0035] S52. Traversal and Accumulation: Traverse all ordered product pairs (p1, p2) in the product set P, calculate the directed association strength from city i to city j for each product pair, and successively accumulate this value to the total logistics space edge weight matrix from city i to city j. matrix elements middle.

[0036] S53. Normalization Process: Weighting the total logistics space edge matrix from city i to city j Min-Max normalization is performed, mapping all elements to the interval [0,1]. The calculation formula is as follows: ; in, These are the normalized matrix elements. This represents the maximum value of all elements in the matrix. The purpose of normalization is to eliminate dimensional differences arising from different industry chains or data volumes, thus making the threshold more universal.

[0037] S54. Remove low-weight edge connections: Set a minimum threshold θ, and apply the following to the elements of the normalized matrix: Perform the following operations: .

[0038] Preferably, minimum threshold The range of values ​​is In a specific implementation, The size determines the sparsity of the map: The larger the value, the stricter the filtering, the fewer connections are retained, the sparser the graph, and only the strongest associations are retained; The smaller the value, the looser the filter, the more connections are retained, and the denser the graph. In practical applications, this is generally preferred. =0.01, meaning that edges with weights in the top 99% after normalization are retained, while weak connections with weights less than 1% of the maximum value are filtered out. This value can also be adjusted according to actual needs.

[0039] On the other hand, this application provides an industrial spatial map construction system based on an artificial intelligence graph reasoning method for constructing manufacturing industrial spatial maps, such as... Figure 4 As shown, it includes a shortest path step length calculation unit 1, an industrial chain relationship attenuation coefficient definition unit 2, an urban industrial distribution advantage calculation unit 3, a directed correlation strength calculation unit 4, and a directed graph construction unit 5.

[0040] The shortest path step length calculation unit 1 extracts the supply chain topology network of logistics-sensitive manufacturing industries, combines supply chain knowledge graphs and artificial intelligence relationship reasoning to identify the core product nodes of the target supply chain and their upstream and downstream relationships, constructs a product directed graph, and calculates the shortest path step length between any two product nodes.

[0041] Unit 2 defines and maps the supply chain relationship attenuation coefficient based on the characteristics of logistics-sensitive manufacturing industries. It decreases as the shortest path step size increases.

[0042] Unit 3 for calculating the city's industrial distribution advantage extracts the production capacity or number of enterprises of each city at each product node, and calculates the relative advantage of city i in product p as the city's industrial distribution advantage. .

[0043] Directed association strength calculation unit 4 is used for any two cities i and j, and product pairs in the industrial chain. Calculate the directed correlation strength .

[0044] The directed graph construction unit 5 traverses all product pairs and accumulates the total logistics space edge weight matrix from city i to city j. matrix elements , It is an asymmetric matrix, based on the total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries. Specific implementation examples: This embodiment, combined with a practical case study of constructing an urban spatial map for the ethylene / olefins industry, provides a detailed explanation of this application.

[0046] In the actual ethylene industry chain, due to the lack of real micro-supply chain data between enterprises, this embodiment makes approximate calculations based on the following core assumptions: Industrial linkage hypothesis: The strength of industrial linkages between cities is positively correlated with the density of their enterprises in the industrial chain.

[0047] Distance decay hypothesis: The strength of the relationship between products in the industrial chain decreases as the distance between upstream and downstream levels increases, that is, the longer the processing link, the weaker the direct logistics collaboration.

[0048] Scale effect hypothesis: The proportion of enterprises reflects the city's industrial advantage in this product field.

[0049] Specifically, the method for constructing an ethylene industry spatial map based on artificial intelligence graph reasoning and hierarchical decay characteristics in this embodiment includes the following steps: S1. Extract the topological network of the ethylene industry chain, combine the industry chain knowledge graph and artificial intelligence relationship reasoning to identify the core product nodes of the ethylene industry chain and their upstream and downstream relationships, construct a directed graph of products, and calculate the shortest path step length between any two product nodes.

[0050] S2. Define and map the attenuation coefficient of the industrial chain relationship based on industry characteristics. The value decreases as the shortest path step size increases. When the products have a direct upstream / downstream relationship (1 step), =1.0; When the relationship between products is indirect (2 steps, such as one step apart), =0.6; When the relationship between products is indirect (3 steps), =0.3; When the relationship between products is indirect (4 steps), =0.1; When there is no relationship between products, similar products, or 5 or more steps... =0.0.

[0051] S3. Calculate the city's industrial distribution advantage. Extract the production capacity or number of enterprises of each city at each product node, and calculate the relative advantage of city i in product p as the city's industrial distribution advantage. .

[0052] In this embodiment, the number of enterprises distributed across N cities nationwide for these products is obtained. The industry advantage of city i for product p is also determined. The calculation formula is: ; in, This represents the production capacity or number of enterprises in city i that produce product p. This represents the total production capacity or number of enterprises producing product p in all cities, where j refers to the j-th city among all cities. The value range is [0,1]. In the formula, city i in the numerator is a fixed input variable, representing the city i currently being calculated; j in the denominator is the traversal index in the summation operation, used to traverse all cities in the country for accumulation.

[0053] S4. Calculate the strength of directed inter-city associations based on the city's industrial distribution dominance and hierarchical decay. For any two cities i and j, and adjacent product pairs in the industry chain. Calculate the directed correlation strength The calculation formula is: ; in, For City i in Products The city's industrial distribution advantage For city j in products The degree of advantage in the distribution of urban industries.

[0054] S5. Construct a directed graph of urban industrial space: Traverse all product pairs and accumulate the total logistics space edge weight matrix from city i to city j. matrix elements W is an asymmetric matrix, based on the total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries.

[0055] because The directionality is such that W(i,j)≠W(j,i). This reflects the differences in the upstream / downstream positions of different cities in the industrial chain, forming an asymmetric directed graph matrix. In this embodiment, the generated edge weights are normalized and a threshold is set. =0.01, filtering weak connections and retaining the top 99% of core connections.

[0056] like Figure 5 As shown, this embodiment applies the method to the national olefin industry chain and successfully outputs a "complex spatial network map of cities in the national olefin industry chain". This map intuitively and accurately identifies the core hub cities of the ethylene industry and discovers the spatial distribution pattern of industrial clusters, fully verifying the scientific nature and practicality of the algorithm in this application.

[0057] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made by those skilled in the art to the technical solutions of this application without departing from the spirit of this application should fall within the protection scope defined by the claims of this application.

Claims

1. A method for constructing a spatial map of a manufacturing industry based on artificial intelligence graph reasoning, characterized in that: It includes the following steps: S1. Collect supply chain text data, product description data, enterprise statistical data and rule base data of logistics-sensitive manufacturing industries, combine them with supply chain knowledge graphs, use artificial intelligence relationship reasoning to identify the core product nodes of the target supply chain and their upstream and downstream relationships, generate candidate nodes and candidate edges, and complete the verification by combining chemical or manufacturing processing logic, construct a product directed graph, and calculate the shortest path step length between any two product nodes. S2. Define and map the supply chain relationship attenuation coefficient based on the characteristics of logistics-sensitive manufacturing industries. Decrease as the shortest path step size increases; S3. Calculate the city's industrial distribution advantage. Extract the production capacity or number of enterprises of each city at each product node, and calculate the relative advantage of city i in product p as the city's industrial distribution advantage. The calculation formula is: ； in, This represents the production capacity or number of enterprises in city i that produce product p. This represents the total production capacity or number of enterprises producing product p in all cities, where j refers to the j-th city among all cities. S4. Calculate the strength of directed inter-city associations based on the city's industrial distribution dominance and hierarchical decay. For any two cities i and j, and adjacent product pairs in the industry chain. Calculate the directed correlation strength The calculation formula is: ； in, For City i in Products The city's industrial distribution advantage For city j in products The degree of urban industrial distribution advantage; S5. Construct a directed graph of urban industrial space: Traverse all product pairs and accumulate the total logistics space edge weight matrix from city i to city j. matrix elements , It is an asymmetric matrix, based on the normalized total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries.

2. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 1, characterized in that: In S1, AI-powered relational reasoning refers to preprocessing text data, product description data, and enterprise statistical data, performing entity recognition, relation extraction, node importance ranking, and confidence assessment to output candidate nodes and candidate edges of a directed graph of products.

3. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 1, characterized in that: S1 specifically includes the following sub-steps: S11. Extract core product nodes: Based on the chemical or manufacturing processing logic of the target industry chain, collect industry chain text data, product description data and enterprise statistics, clean the data, normalize the terminology and standardize the product names; use artificial intelligence entity recognition model to identify entities of raw materials, intermediate products, finished products and process links, and combine word frequency, production capacity, enterprise coverage and node connectivity to analyze the importance of nodes, and select the core products corresponding to each link as nodes of the graph; S12. Based on the material flow relationship between products, and combined with the results of artificial intelligence relationship reasoning, directed edges are constructed. The specific process is as follows: An artificial intelligence relationship extraction model is used to identify the raw material-product relationship, upstream-downstream relationship, or processing and transformation relationship between products, generating candidate edges and corresponding confidence levels; then, the candidate edges are verified according to the rule base. If product P... A For product P B For the direct production raw materials, a route is established from node P. A Pointing to node P B The directed edges; S13. Calculate the shortest path step length: Use the shortest path algorithm in graph theory to calculate the shortest path step length between any two product nodes. The shortest path step length is defined as the minimum number of directed edges required to traverse from the starting product node to the target product node. S14. No-path processing: Based on the directed product graph obtained in S11 and S12 and verified by the rule base, perform reachability retrieval for any two product nodes; if there is no directed path from the starting product node to the target product node, or if the corresponding candidate relationship fails the confidence threshold screening and rule verification, it is determined that the two nodes have no effective upstream and downstream association, and the shortest path step size is recorded as infinity or directly determined as no association.

4. The method for constructing a manufacturing industry spatial map based on artificial intelligence graph reasoning according to claim 1, characterized in that: In S2, the industry chain relationship decay coefficient is 1.0 when the shortest path step size is 1, 0.6 when the shortest path step size is 2, 0.3 when the shortest path step size is 3, 0.1 when the shortest path step size is 4, and 0 when the shortest path step size is greater than or equal to 5 or when there are no two related product nodes.

5. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 1, characterized in that: The range of the city's industrial distribution advantage in S3 is [0,1].

6. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 1, characterized in that: S5 contains the total logistics space edge weight matrix from city i to city j. matrix elements The sum of the correlation strengths for all products: 。 7. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 1, characterized in that: Total logistics space edge weight matrix from city i to city j For an N×N square matrix, the specific expression is as follows: ； Wherein, the total logistics space edge weight matrix from city i to city j matrix elements The specific expression is as follows: ； Wherein, the diagonal matrix elements W(i,i)=0 and W(i,j)≠W(j,i), and P is the product set.

8. The method for constructing a manufacturing industry spatial map based on artificial intelligence graph reasoning according to claim 7, characterized in that: Step S5 specifically includes the following sub-steps: S51. Matrix initialization: Set an NxN zero matrix, where N is the number of cities, and all elements are initialized to 0. S52. Traversal and Accumulation: Traverse all ordered product pairs (p1, p2) in the product set P, calculate the directed association strength from city i to city j for each product pair, and accumulate them sequentially into the total logistics space edge weight matrix from city i to city j. matrix elements middle; S53. Normalization Process: Weighting the total logistics space edge matrix from city i to city j Min-Max normalization is performed, mapping all elements to the interval [0,1]. The calculation formula is as follows: ； in, These are the normalized matrix elements. It represents the maximum value of all elements in the matrix; S54. Remove low-weight edge connections: Set a minimum threshold θ, and apply the following to the elements of the normalized matrix: Perform the following operations: 。 9. The method for constructing a spatial map of the manufacturing industry based on artificial intelligence graph reasoning according to claim 8, characterized in that: Minimum threshold The range of values ​​is .

10. An industrial spatial map construction system for the manufacturing industrial spatial map construction method based on artificial intelligence graph reasoning as described in claim 1, characterized in that: It includes a shortest path step length calculation unit, an industrial chain relationship attenuation coefficient definition unit, an urban industrial distribution advantage calculation unit, a directed correlation strength calculation unit, and a directed graph construction unit; The shortest path step length calculation unit extracts the supply chain topology network of logistics-sensitive manufacturing industries, combines supply chain knowledge graphs and artificial intelligence relationship reasoning to identify the core product nodes of the target supply chain and their upstream and downstream relationships, constructs a product directed graph, and calculates the shortest path step length between any two product nodes. The definition unit for the supply chain relationship attenuation coefficient is based on the characteristics of logistics-sensitive manufacturing industries and defines and maps the supply chain relationship attenuation coefficient. Decrease as the shortest path step size increases; The urban industrial distribution advantage calculation unit extracts the production capacity or number of enterprises of each city at each product node, and calculates the relative advantage of city i on product p as the urban industrial distribution advantage. ; The directed correlation strength calculation unit is used for any two cities i and j, and product pairs in the industrial chain. Calculate the directed correlation strength ; The directed graph construction unit traverses all product pairs and accumulates them to obtain the total logistics space edge weight matrix from city i to city j. matrix elements middle, It is an asymmetric matrix, based on the total logistics space edge weight matrix from city i to city j. A complex spatial network of industrial cities is generated, resulting in a spatial map of logistics-sensitive manufacturing industries.

Images