Intelligent supply chain risk early warning and resilience optimization management system

Abstract

The application provides a smart supply chain risk early warning and resilience optimization management system, and belongs to the technical field of supply chain risk management.The system comprises a data acquisition module, a supply chain graph network construction module, a multi-source data fusion module, a risk identification and early warning module, a resilience evaluation and optimization module and a decision support module.The system builds a supply chain relationship model based on a graph network, integrates multi-source heterogeneous data such as macroeconomics, geopolitics and meteorological environment, realizes the prospective identification of potential risks of a supply chain and the accurate tracking of a risk propagation path, evaluates the resilience of the supply chain based on a scenario simulation method, and generates targeted optimization strategies.The actual application shows that the system reduces the supply chain interruption loss by 65%, increases the inventory turnover rate by 30%, and provides a systematic risk management tool for enterprises.

Description

Technical Field

[0001] This invention belongs to the field of supply chain risk management technology, specifically relating to an intelligent supply chain risk early warning and resilience optimization management system. Background Technology

[0002] Supply chain risk management is a crucial component of modern enterprise operations management. With increasing globalization and the growing complexity of supply chain networks, supply chain disruptions are becoming more frequent, causing significant economic losses for businesses.

[0003] In the prior art, CN119740760A discloses an artificial intelligence-based method and system for assessing supply chain resilience. This technical solution uses the Gramian Field to convert supply chain time-series data into RGB image data, performs dimensionality reduction based on the Deer Herd Optimization Algorithm, and finally conducts resilience assessment using the EfficientHRNet model. This solution primarily focuses on the ex-post assessment of supply chain resilience, analyzing it by converting time-series data into images.

[0004] However, the aforementioned prior art has the following technical defects:

[0005] (1) Existing technologies focus on post-event resilience assessment and lack the ability to provide forward-looking early warning of supply chain risks. In a complex and ever-changing supply chain environment, relying solely on post-event assessment makes it difficult to identify potential risks early and intervene proactively. As a result, companies often only take countermeasures when risks have already occurred or are about to occur, missing the best opportunity to deal with them.

[0006] (2) Existing technologies fail to effectively capture the complex interdependencies among entities in the supply chain network. The supply chain is a network system composed of multiple entities such as suppliers, manufacturers, logistics providers, and distributors, with intricate relationships among them. While existing methods for converting time-series data into images can capture time-series features, they cannot fully express the topological characteristics of the supply chain network and the interdependencies between entities, resulting in insufficient ability to identify risk propagation paths.

[0007] (3) Existing technologies lack the ability to comprehensively utilize multi-source heterogeneous data. The generation and evolution of supply chain risks are influenced by various factors such as macroeconomics, geopolitics, weather conditions, and the financial status of suppliers. These factors correspond to data from different sources and in different formats. Existing technologies mainly focus on operational data within the supply chain and fail to fully integrate external environmental data, resulting in insufficient comprehensiveness and accuracy in risk identification.

[0008] (4) Existing technologies do not provide targeted resilience optimization solutions. Even if the resilience level of the supply chain can be assessed, the practical value of the assessment results will be greatly reduced if specific and feasible optimization strategies are lacking. Enterprises need not only to know which weak links exist in the supply chain, but also to understand how to specifically improve the risk resistance of these weak links.

[0009] Therefore, there is an urgent need for an intelligent supply chain risk warning and resilience optimization management system that can achieve early risk warning, fully consider the characteristics of the supply chain network structure, integrate multi-source data, and provide specific optimization strategies. Summary of the Invention

[0010] To address the problems existing in the prior art, this invention provides an intelligent supply chain risk warning and resilience optimization management system, which aims to solve the technical problems of insufficient risk warning capabilities, lack of network relationship modeling, insufficient utilization of multi-source data, and lack of targeted optimization strategies in the prior art.

[0011] The technical solution of the present invention is as follows:

[0012] The intelligent supply chain risk warning and resilience optimization management system includes:

[0013] The data acquisition module is used to collect supply chain operation data, external environment data, and historical risk event data. The supply chain operation data includes supplier information data, production data, logistics data, and market data. The external environment data includes macroeconomic indicator data, geopolitical event data, weather forecast data, and supplier financial data. The historical risk event data includes supply chain disruption record data, recovery time data, and loss assessment data.

[0014] A supply chain graph network construction module, connected to the data acquisition module, is used to construct a supply chain graph network model based on the supply chain operation data. In the supply chain graph network model, supplier nodes, production nodes, logistics nodes, and market nodes are graph nodes, and supply relationships, production relationships, transportation relationships, and sales relationships are graph edges. Each graph node contains node attribute information, and each graph edge contains edge weight information. The node attribute information reflects the operational status characteristics of the corresponding entity, and the edge weight information reflects the strength of the relationship between entities.

[0015] A multi-source data fusion module, connected to the data acquisition module and the supply chain graph network construction module, is used to fuse the external environment data with the supply chain graph network model. It sets data weight coefficients for different data sources, determines these coefficients based on the reliability and timeliness levels of the data sources, and assigns a first data weight coefficient if both the reliability and timeliness levels of the data sources meet a first reliability threshold range. If neither the reliability nor timeliness level of the data sources meets a first timeliness threshold range, a second data weight coefficient is assigned. The first data weight coefficient is greater than the second data weight coefficient. The external environment data is integrated as an additional feature into the node attribute information of the corresponding graph nodes in the supply chain graph network model. The external environment feature values ​​in the node attribute information are multiplied by the corresponding data weight coefficients to update the node attribute information of all relevant graph nodes in the supply chain graph network model, thereby obtaining the fused enhanced graph network model.

[0016] The risk identification and early warning module, connected to the multi-source data fusion module, is used to identify potential risk points and risk propagation paths in the supply chain based on the enhanced graph network model. It calculates the vulnerability index of each node, which reflects the importance and anti-interference ability of the node in the supply chain network. If the vulnerability index of a node exceeds the first vulnerability threshold, the node is marked as a high-risk node. The risk impact is propagated from the high-risk node to adjacent nodes along the graph edges using a graph network propagation algorithm. During the propagation process, the intensity of the risk impact is attenuated according to the edge weight and the node's resistance ability. If the intensity of the risk impact received by an adjacent node exceeds the threshold that the adjacent node can withstand, the adjacent node is marked as an affected node. The module simulates the diffusion process of risk in the supply chain network, identifies the critical path of risk propagation, and if the number of affected nodes on a certain propagation path exceeds the first node number threshold or the impact range exceeds the first range threshold, the path is marked as a critical risk propagation path. Risk early warning information is generated and sent to the decision support module.

[0017] The resilience assessment and optimization module, connected to the risk identification and early warning module, is used to assess the resilience level of the supply chain under different risk scenarios based on scenario simulation methods. It sets multiple risk scenario parameter combinations, including risk type, risk intensity, and risk duration. For each risk scenario, it simulates the response process of the supply chain network and calculates supply chain resilience assessment indicators, including recovery time indicators, performance loss indicators, and adaptability indicators. If the recovery time indicator exceeds a first time threshold or the performance loss indicator exceeds a first loss threshold, the supply chain is deemed to be insufficiently resilient under that scenario. For scenarios with insufficient resilience, a resilience optimization strategy is generated, including supplier diversification strategies, inventory buffer strategies, alternative logistics route strategies, and emergency response strategies. These strategies are ranked according to their cost-effectiveness ratio, with priority given to optimization strategies whose cost-effectiveness ratio falls within the range of the first benefit threshold.

[0018] The decision support module, connected to the risk identification and early warning module and the resilience assessment and optimization module, is used to integrate the risk early warning information and the resilience optimization strategy, providing a visual decision support interface for supply chain managers. The visual decision support interface displays a supply chain network topology map, a risk heat map, a key risk propagation path map, and an optimization strategy comparison map, supporting managers to conduct interactive scenario analysis and strategy selection.

[0019] Preferably, the first reliability threshold range is a data source reliability score greater than or equal to 80 points, and the first timeliness threshold range is a data update cycle less than or equal to 24 hours.

[0020] Preferably, the value range of the first data weight coefficient is 0.8 to 1.0, and the value range of the second data weight coefficient is 0.3 to 0.7.

[0021] Preferably, the first correlation threshold range is where the absolute value of the correlation coefficient is greater than or equal to 0.6.

[0022] Preferably, the first influence weight ranges from 0.7 to 1.0, and the second influence weight ranges from 0.3 to 0.6.

[0023] Preferably, the first degree centrality threshold is a normalized degree centrality value greater than or equal to 0.5, the first betweenness centrality threshold is a normalized betweenness centrality value greater than or equal to 0.6, and the first clustering coefficient threshold is a clustering coefficient value less than or equal to 0.4.

[0024] Preferably, the first connection density threshold is a ratio of the number of connections between nodes within the community to the maximum possible number of connections greater than or equal to 0.5.

[0025] The beneficial effects of this invention are as follows:

[0026] (1) This invention constructs a supply chain relationship model based on graph networks, which can comprehensively capture the complex dependencies and topological features among entities in the supply chain network. Compared with the existing technology of converting time-series data into images, the graph network model can more naturally express the multi-level relationships between entities such as suppliers, manufacturers, and logistics providers in the supply chain, effectively solving the shortcomings of the existing technology in network relationship modeling.

[0027] (2) This invention innovatively integrates multi-source heterogeneous external environmental data, including macroeconomic indicators, geopolitical events, and weather forecasts, and achieves differentiated fusion of different data sources through a data weighting coefficient mechanism. This multi-source data fusion strategy significantly improves the comprehensiveness and accuracy of risk identification, overcoming the limitations of existing technologies that only focus on internal supply chain operational data. According to actual application cases, after integrating external environmental data, the system's early warning accuracy for supply chain disruption events increased by approximately 40%, and the average early warning time increased by 5 to 7 days.

[0028] (3) This invention, through a risk propagation path identification mechanism, can simulate the diffusion process of risks in the supply chain network and identify the potential scope and key propagation paths of risks in advance. This proactive risk warning capability enables enterprises to take preventive measures before risks actually occur, effectively reducing losses caused by supply chain disruptions. Implementation cases show that enterprises using this system have reduced their average losses by approximately 65% ​​when facing sudden supply chain risks.

[0029] (4) This invention uses scenario simulation to assess resilience, quantifying the performance of the supply chain under different risk scenarios and generating specific and feasible optimization strategies for identified weaknesses. Through cost-benefit analysis and strategy ranking, this system provides practical decision support for enterprises, not only pointing out the problems but, more importantly, providing solutions. Application data shows that after implementing the optimization strategies recommended by this system, the average recovery time of the enterprise's supply chain is shortened by about 50%, and the inventory turnover rate is increased by about 30%.

[0030] (5) The visual decision support interface provided by the present invention can present complex supply chain networks, risk situations and optimization solutions to managers in an intuitive and easy-to-understand way, support interactive scenario analysis and strategy selection, and significantly improve decision efficiency and decision quality. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the overall architecture of the intelligent supply chain risk early warning and resilience optimization management system of the present invention.

[0032] Figure 2This is a schematic diagram of the workflow of the supply chain graph network construction module of the present invention.

[0033] Figure 3 This is a schematic diagram of the data fusion process of the multi-source data fusion module of the present invention.

[0034] Figure 4 This is a schematic diagram of the risk propagation path identification process of the risk identification and early warning module of the present invention.

[0035] Figure 5 This is a schematic diagram of the scenario simulation and optimization strategy generation process of the toughness assessment and optimization module of the present invention.

[0036] Figure 6 This is a schematic diagram of data flow in the practical application of the system of the present invention. Detailed Implementation

[0037] Please refer to the attached document. Figures 1-6 To make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] like Figure 1 As shown, the intelligent supply chain risk early warning and resilience optimization management system provided by the present invention includes a data acquisition module 1, a supply chain graph network construction module 2, a multi-source data fusion module 3, a risk identification and early warning module 4, a resilience assessment and optimization module 5, a decision support module 6, and a dynamic monitoring module 7.

[0039] Data acquisition module 1 is responsible for collecting supply chain operation data, external environment data, and historical risk event data from multiple data sources. In one possible implementation, data acquisition module 1 interfaces with the enterprise's ERP system, WMS system, TMS system, and other business systems via API interfaces to obtain supply chain operation data such as supplier information, production plans, inventory status, order information, and logistics tracking in real time. The collection frequency is set according to the changing characteristics of the data type; for example, for inventory data, the collection frequency is set to once per hour; for order data, the collection frequency is set to real-time collection.

[0040] External environmental data is collected through multiple channels. Collected indicators include GDP growth rate, inflation rate, and exchange rate fluctuations, with a collection frequency set to once daily. Geopolitical event data is obtained through news APIs and event databases. The system is configured with a keyword monitoring list, automatically capturing event information when major geopolitical events related to the supply chain occur. Weather forecast data is obtained from meteorological service APIs, focusing on extreme weather warnings in areas where key nodes in the supply chain are located, with a collection frequency set to once every 6 hours. Supplier financial data is obtained through corporate credit systems and publicly available financial reports. The financial health of key suppliers is regularly assessed, with an assessment frequency set to once quarterly. For suppliers showing financial warning signals, the assessment frequency is increased to once a month.

[0041] Historical risk event data includes detailed records of supply chain disruptions encountered by enterprises in the past, including the time of the disruption, the type of disruption (such as supplier bankruptcy, natural disasters, logistics disruptions, etc.), the affected supply chain nodes, the duration of the disruption, the direct and indirect losses caused, the countermeasures taken, and the recovery process. This historical data provides important reference for subsequent risk identification and resilience assessment. In a specific example, historical data from a manufacturing company shows that during a typhoon disaster in 2023, its three suppliers located in the southeastern coastal region were severely affected, resulting in a 12-day disruption of the supply of key components and causing production line shutdown losses of approximately 5 million yuan. The complete record of this event is stored in the historical risk event database.

[0042] like Figure 2 As shown, the supply chain graph network construction module 2 constructs a graph network model based on the collected supply chain operation data. In this embodiment, the supply chain of an electronics manufacturing company is used as an example. The company's supply chain includes upstream chip suppliers, electronic component suppliers, and packaging material suppliers; midstream assembly plants; and downstream logistics and distribution centers and retailers.

[0043] First, the system identifies various entities in the supply chain and creates corresponding graph nodes for each entity. In this example, the system identifies 15 Tier 1 suppliers, 8 Tier 2 suppliers, 3 self-owned production facilities, 5 logistics and distribution centers, and 200 retail outlets. Different types of entities are represented by different types of nodes in the graph network: supplier nodes are labeled Type A, production facility nodes are labeled Type B, logistics center nodes are labeled Type C, and retailer nodes are labeled Type D.

[0044] Secondly, the system establishes graph edges based on the business relationships between entities. If supplier S1 supplies parts to production facility P1, a directed edge E (S1→P1) is established between node S1 and node P1. The direction of the edge indicates the direction of material or information flow. In this example, the system establishes approximately 350 graph edges, forming a multi-layered supply chain network topology.

[0045] Next, a node attribute vector is set for each graph node. Taking a chip supplier node as an example, its node attribute vector includes: (1) historical supply volume, with an average monthly supply of 10,000 wafers over the past 12 months; (2) supply stability, with an on-time delivery rate of 95% calculated based on historical data; (3) current inventory level, with a current inventory of 15,000 wafers; (4) production capacity, with a monthly production capacity of 12,000 wafers; (5) geographical location, located at XX degrees North latitude and XX degrees East longitude; (6) cooperation period, with a cooperation period of 5 years with this company; and (7) financial health score, with a score of 85 out of 100 based on the latest financial report. These attribute information constitute a multidimensional vector, which is stored in the attribute fields of the graph node.

[0046] Then, an edge weight is assigned to each graph edge. The calculation of the edge weight comprehensively considers three factors: trading volume, trading frequency, and trading stability. In this embodiment, the edge weight is calculated using the following formula:

[0047] ,

[0048] in: For the node To the node Edge weights; For nodes and nodes The normalized transaction volume between them, with a value range of [0, 1]; For nodes and nodes The normalized transaction frequency between them, with a value range of [0, 1]; For nodes and nodes The transaction stability coefficient is obtained by calculating the coefficient of variation of historical transaction volume. The lower the coefficient of variation, the higher the stability. The value range is [0, 1]. , , For the weighting coefficients, satisfying In this embodiment, it is set , , .

[0049] Taking a specific edge as an example, the normalized transaction volume of the edge between chip supplier S1 and production facility P1... (This indicates that the supplier accounts for 85% of the total chips required by the production facility), Normalized Trading Frequency (This indicates that the supplier's monthly delivery frequency reaches 90% of the frequency stipulated in the contract), Transaction stability coefficient. (This indicates that the supply volume fluctuates relatively little). The edge weight is calculated based on the above formula. An edge weight exceeding 0.8 indicates that this is an important supply chain path, and its disruption would have a significant impact on the supply chain.

[0050] Furthermore, in an optimized implementation, the supply chain graph network construction module 2 also employs a community detection algorithm to identify the community structure in the supply chain network. A community refers to a group of nodes in a graph network, where the connections between these nodes are relatively dense, while the connections to nodes outside the community are relatively sparse. In a supply chain scenario, communities typically correspond to relatively independent functional modules or business units. This embodiment uses the Louvain algorithm for community detection, clustering nodes whose connection density exceeds a first connection density threshold into communities. Preferably, the first connection density threshold is the ratio of the number of connections between nodes within a community to the maximum possible number of connections greater than or equal to 0.5. In the aforementioned supply chain network of the electronics manufacturing enterprise, the algorithm identifies five main communities: Community 1 contains chip suppliers and related second-tier suppliers; Community 2 contains a group of electronic component suppliers; Community 3 contains a group of packaging material suppliers; Community 4 contains production facilities and their closely related first-tier suppliers; and Community 5 contains logistics and distribution networks and retail outlets.

[0051] After identifying the community structure, the system pays special attention to the edges connecting different communities, as these cross-community edges represent critical interfaces in the supply chain. Disruptions to these cross-community edges could lead to functional fragmentation of the supply chain network. In this example, the edge connecting community 1 and community 4 (i.e., the connection from the chip supply chain to the production stage) is marked as a critical interface edge, and these edges will be closely monitored during risk assessment.

[0052] like Figure 3 As shown, the multi-source data fusion module 3 is responsible for integrating external environmental data with the supply chain graph network model to form a more comprehensive risk perception capability.

[0053] First, the system preprocesses the external environment data. Data cleaning operations include removing duplicate records, imputing missing values, and correcting outliers. For missing values, if the missing rate is less than 5%, interpolation is used to impute them using data from nearby time points; if the missing rate exceeds 5%, the data source is marked as a low-reliability data source. Data standardization transforms data of different dimensions to a uniform scale range, for example, normalizing GDP growth rate (usually between -5% and 10%) and exchange rate fluctuation range (possibly between -20% and 20%) to the [0, 1] interval. Time alignment ensures that data from different sources remain consistent in the time dimension; for data with different sampling frequencies, upsampling or downsampling is used to unify them to a daily-level time granularity.

[0054] Secondly, the system establishes the correlation between external environmental factors and supply chain nodes. This process is based on domain knowledge and historical data analysis. For example, political instability in a country may affect all supplier nodes located in that country; therefore, the system establishes a correlation mapping between this political event data and supplier nodes in the relevant geographical area. Similarly, a rise in raw material prices may affect all supplier nodes using that raw material; the system establishes a correlation between raw material price data and relevant supplier nodes. In a specific example, the system detects that a typhoon orange warning has been issued in a certain region. This region has two Tier 1 suppliers and one Tier 2 supplier for the company; the system automatically establishes a correlation between this weather warning data and these three supplier nodes.

[0055] Next, the system calculates the impact weight of each external environmental factor on the supply chain node. The impact weight is determined by analyzing the correlation between changes in external factors and changes in the operational status of supply chain nodes in historical data. Specifically, the system collects historical data from the past three years and calculates the correlation coefficient between the change in each external environmental factor and the change in the operational status of each relevant node. If the absolute value of the correlation coefficient meets the first correlation threshold range (i.e., the absolute value of the correlation coefficient is greater than or equal to 0.6), a strong influence relationship is considered to exist, and a first impact weight is assigned. If the absolute value of the correlation coefficient does not meet the first correlation threshold range, a second impact weight is assigned, where the first impact weight is greater than the second impact weight. Preferably, the first impact weight ranges from 0.7 to 1.0, and the second impact weight ranges from 0.3 to 0.6. In this embodiment, the impact weight ranges from [0, 1], and the closer the impact weight is to 1, the more significant the impact of the external factor on the node.

[0056] For example, for a chip supplier node, the system analysis found that: the correlation coefficient between the monthly manufacturing PMI index of the supplier's region and the supplier's actual output is 0.75, which meets the first correlation threshold range. Therefore, the system assigns a first influence weight of 0.80 to the PMI index; the correlation coefficient between the region's quarterly inflation rate and the supplier's supply price is 0.55, which does not meet the first correlation threshold range. Therefore, the system assigns a second influence weight of 0.50 to the inflation rate; the correlation coefficient between the number of extreme weather events (typhoons, rainstorms, etc.) in the region and the supplier's on-time delivery rate is -0.82 (negative correlation, the absolute value meets the first correlation threshold range). Therefore, the system assigns a first influence weight of 0.85 to extreme weather events.

[0057] Then, the system incorporates external environmental factors as additional features into the attribute vector of the corresponding node. Based on the original node attribute vector, a new dimension of environmental impact features is added. For the aforementioned chip supplier node, its original attribute vector was 7-dimensional; after incorporating external environmental data, it expands to 10 dimensions. The three newly added dimensions are: the current PMI index weighted value, the current inflation rate weighted value, and the recent extreme weather risk weighted value. Taking the PMI index as an example, if the current PMI index is 51.5 (indicating manufacturing expansion), after normalization, it becomes 0.65. Multiplying this by the impact weight of 0.80 yields a PMI index weighted value of 0.52, which serves as a new dimension of the node's attribute vector.

[0058] During the integration process, the differences in reliability and timeliness among different data sources are reflected through data weighting coefficients. The determination of these coefficients is based on an assessment of two dimensions: reliability level and timeliness level. The reliability level is comprehensively evaluated based on the data source's authority, historical accuracy, and data integrity, using a percentage-based scoring system. The timeliness level is assessed based on the data's update frequency and timeliness. Preferably, the first reliability threshold ranges from a data source reliability score greater than or equal to 80 points, and the first timeliness threshold ranges from a data update cycle less than or equal to 24 hours.

[0059] In this embodiment, the system assigns data weight coefficients according to the following rules: If the reliability level of the data source meets the first reliability threshold range and the timeliness level meets the first timeliness threshold range, a first data weight coefficient is assigned, with a value range of 0.9-1.0; if the reliability level meets the first reliability threshold range but the timeliness level does not, or if the reliability level does not meet the first reliability threshold range but the timeliness level meets the first timeliness threshold range, an intermediate level data weight coefficient is assigned, with a value range of 0.7-0.9; if neither meets the corresponding threshold range, a second data weight coefficient is assigned, with a value range of 0.3-0.5. Preferably, the first data weight coefficient ranges from 0.8 to 1.0, and the second data weight coefficient ranges from 0.3 to 0.7. When applying the data weight coefficients, the external environment feature values ​​in the node attribute vector need to be multiplied by the corresponding data weight coefficient to reflect the differentiated contribution of the data source.

[0060] The reliability level is high (score 95, within the first reliability threshold range), the timeliness level is medium (quarterly updates, not within the first timeliness threshold range), and the assigned data weight coefficient is 0.85. Weather forecast data obtained from a commercial meteorological service API has a medium reliability level (score 70, not within the first reliability threshold range) and a high timeliness level (updated every 6 hours, within the first timeliness threshold range), and the assigned data weight coefficient is 0.80. Geopolitical event information obtained from social media monitoring has a low reliability level (score 55, not within the first reliability threshold range) and a high timeliness level (real-time, within the first timeliness threshold range), and the assigned data weight coefficient is 0.45.

[0061] Finally, the system updates the attribute information of all relevant nodes in the graph network model, forming an enhanced graph network model that incorporates external environmental information. This enhanced graph network model not only includes the internal structure and operational information of the supply chain but also integrates the influence of external factors such as macroeconomics, geopolitics, and meteorological conditions, providing a more comprehensive data foundation for subsequent risk identification and resilience assessment.

[0062] like Figure 4 As shown, the risk identification and early warning module 4 identifies potential risk points and risk propagation paths in the supply chain based on an enhanced graph network model.

[0063] First, the system calculates the vulnerability index of each node. The vulnerability index comprehensively reflects the node's importance and resilience within the supply chain network. The specific calculation process is as follows:

[0064] (1) Calculate the degree centrality index of a node. Degree centrality refers to the sum of the in-degree and out-degree of a node, reflecting the number of direct connections between that node and other nodes. For a directed graph, the formula for calculating degree centrality is:

[0065] ,

[0066] in: For nodes Degree centrality; For nodes The in-degree of a node is the number of edges pointing to that node. For nodes The out-degree is the number of edges that point from that node.

[0067] (2) Calculate the betweenness centrality index of a node. Betweenness centrality reflects the importance of a node in the shortest path of the network, and the formula is as follows:

[0068] ,

[0069] in: For nodes Betweenness centrality; For nodes and nodes The total number of shortest paths between them; For nodes and nodes Between nodes The number of shortest paths.

[0070] (3) Calculate the clustering coefficient of the nodes. The clustering coefficient reflects the connection density between adjacent nodes of a node, and the calculation formula is:

[0071] ,

[0072] in: For nodes Clustering coefficients; For nodes The actual number of edges between adjacent nodes; For nodes The degree.

[0073] (4) Calculate the vulnerability index comprehensively. The formula for calculating the vulnerability index is:

[0074] ,

[0075] in: For nodes Vulnerability indicators; It has the highest degree centrality in the network; It has the highest betweenness centrality in the network; For nodes The operational stability score ranges from [0, 1], with a higher score indicating more stable operation. , , , As a weighting coefficient, it is set in this embodiment. , , , .

[0076] The meaning of vulnerability indicators is as follows: the higher the degree centrality and betweenness centrality of a node, the more important it is in the network, and its interruption will affect more other nodes, so the vulnerability is higher; the higher the clustering coefficient of a node, the denser the connection between its neighboring nodes, and when the node is interrupted, its function may be shared by the group of neighboring nodes, so the vulnerability is relatively low; the lower the operational stability of a node, the more likely it is to have problems, so the vulnerability is higher.

[0077] When determining whether a node has high vulnerability, the system adopts the following rule: if a node's degree centrality exceeds a first degree centrality threshold and its betweenness centrality exceeds a first betweenness centrality threshold, while its clustering coefficient is lower than a first clustering coefficient threshold or its operational stability data does not meet the range of a first stability threshold, then the node's vulnerability index is determined to exceed the first vulnerability threshold. Preferably, the first degree centrality threshold is a normalized degree centrality value greater than or equal to 0.5, the first betweenness centrality threshold is a normalized betweenness centrality value greater than or equal to 0.6, and the first clustering coefficient threshold is a clustering coefficient value less than or equal to 0.4.

[0078] In the aforementioned case of an electronics manufacturing company, the key indicators of a critical chip supplier node were: degree centrality. (This node has 3 upstream tier-2 suppliers and 5 downstream production facilities), the normalized value is 0.62, exceeding the first-degree centrality threshold of 0.5; betweenness centrality... After normalization, it is 0.75, exceeding the first betweenness centrality threshold of 0.6 (indicating that a large number of supply chain paths pass through this node); clustering coefficient The score is below the first clustering coefficient threshold of 0.4 (sparse connections between adjacent nodes); Operational stability score The vulnerability index is calculated based on the formula. Because the node's degree centrality exceeds the first degree centrality threshold, its betweenness centrality exceeds the first betweenness centrality threshold, and its clustering coefficient is lower than the first clustering coefficient threshold, the node's vulnerability index is determined to exceed the first vulnerability threshold. Therefore, this chip supplier node is marked as a high-risk node, requiring close monitoring and the development of contingency plans.

[0079] After identifying high-risk nodes, the risk identification and early warning module uses a graph network propagation algorithm to identify critical paths for risk propagation. Specifically, the system selects an initial risk node and propagates the risk impact along the graph edges to adjacent nodes. During propagation, the intensity of the risk impact decays according to the edge weights and node resistance capabilities. If the intensity of the risk impact received by an adjacent node exceeds its tolerance threshold, the adjacent node is marked as an affected node, and the complete risk propagation path is recorded. The system calculates the risk level of the propagation path based on the total number of nodes affected by the propagation path, the sum of the impact intensity, and the criticality of the affected nodes. If the risk level exceeds the first risk level threshold, the propagation path is marked as a critical risk propagation path, and risk warning information is generated based on high-risk nodes and critical risk propagation paths.

[0080] like Figure 5 As shown, the resilience assessment and optimization module 5 assesses the resilience level of the supply chain under different risk scenarios based on the scenario simulation method and generates targeted optimization strategies.

[0081] First, the system establishes a risk scenario library. This library contains various typical supply chain risk scenarios, each with clearly defined definitions and parameter settings. In this embodiment, the risk scenario library includes:

[0082] Scenario 1: Natural Disaster Scenario. Simulates the impact of natural disasters such as earthquakes, typhoons, and floods on a specific geographical area. The set of affected nodes is all supply chain nodes in the area. The duration of the impact is set to 3-30 days, and the intensity of the impact is set to 0.5-1.0 (complete disruption).

[0083] Scenario 2: Supplier Bankruptcy Scenario. Simulates the bankruptcy of one or more suppliers due to financial problems. The affected nodes are specific suppliers and their directly served downstream nodes. The duration of the impact is set to long-term (more than 90 days), and the impact strength is set to 1.0.

[0084] Scenario 3: Logistics Disruption Scenario. Simulates the disruption of major logistics channels (such as port blockade or highway interruption). The affected nodes are all nodes that depend on the logistics channel. The duration of the impact is set to 7-60 days, and the impact intensity is set to 0.6-1.0.

[0085] Scenario 4: Sudden Demand Fluctuation Scenario. This scenario simulates a sudden and significant fluctuation (surge or drop) in market demand. The affected nodes are production and logistics nodes. The duration of the impact is set to 14–90 days, and the impact strength parameter reflects the magnitude of the demand change, with a value range of 0.3–0.8.

[0086] Scenario 5: Geopolitical Impact Scenario. This scenario simulates the impact of changes in external environmental factors in a specific country or region on supply chain nodes. The affected nodes are those located in the specific country or region. The duration of the impact is set to 30–180 days, and the impact intensity is set to 0.5–0.9.

[0087] In a specific application example, the system sets up a specific instance of Scenario 2 for the aforementioned electronic product manufacturing company: simulating the sudden bankruptcy of a key chip supplier. The specific parameters of this scenario are set as follows: affected node = {chip supplier S1}, duration of impact = permanent (S1 cannot recover), and impact intensity = 1.0 (complete interruption).

[0088] Secondly, the system simulates the impact of risk scenarios in a supply chain graph network model. For this instance of scenario 2, the system performs the following simulation process:

[0089] (1) Set the state of node S1 to interrupt, and the node stops providing products to downstream.

[0090] (2) Identify the downstream nodes that directly depend on node S1, which in this example are three production facility nodes P1, P2 and P3.

[0091] (3) Calculate the performance degradation level of downstream nodes. For production nodes, the performance degradation level depends on the proportion of the supplier's supply to its total demand and whether there are alternative supply sources. The specific calculation formula is as follows:

[0092] ,

[0093] in: For nodes The degree of performance degradation, with a value ranging from [0, 1]; For the node (Interruption node) to node Supply volume; For nodes Total demand; For nodes The substitutability coefficient reflects the ability of a node to obtain alternative supplies from other supply sources. The value ranges from [0, 1]. The stronger the substitutability, the larger the coefficient.

[0094] In this example, production node P1 obtains 80% of its chip supply from supplier S1. Since S1 provides high-end, specialized chips, alternative suppliers are difficult to find in the short term, resulting in a high substitutability coefficient. Therefore, the performance degradation of P1 is... This means a 72% decrease in production capacity. Similarly, calculations show... , .

[0095] The system is configured to determine that a node is functionally impaired if its performance degradation exceeds 0.5. In this example, all three production nodes, P1, P2, and P3, are determined to be functionally impaired.

[0096] (4) Simulating the cascading propagation of the impact. Damaged production nodes cannot meet the needs of downstream logistics centers, thus affecting the operations of logistics centers and final retailers. The system continues to calculate the secondary and tertiary affected nodes until the impact no longer spreads. In this example, the system simulation shows that a total of 8 nodes are ultimately affected, including 1 supplier node, 3 production nodes, 2 logistics nodes, and 2 major retail areas.

[0097] Secondly, the system simulates the self-recovery process of the supply chain network. The recovery process considers the following factors:

[0098] (1) Availability of backup suppliers. The system checks whether there are backup suppliers that can provide similar products. If backup suppliers exist, the time required to activate them is assessed. In this example, although the company has a backup chip supplier B1, the chip specifications provided by B1 are not exactly the same as those of S1, requiring adaptation adjustments to the product design, with an estimated adjustment period of 45 days.

[0099] (2) Node Redundancy Resources. The system assesses whether the affected nodes have sufficient safety stock to buffer against supply disruptions. In this example, production node P1 has enough chip safety stock to sustain production for 10 days, P2 for 7 days, and P3 for 12 days. These stocks can partially mitigate the impact until an alternative supply source is found.

[0100] (3) Network reconfiguration capability. The system assesses whether the supply chain network can self-adjust by reallocating resources and adjusting production plans. In this example, since P1, P2, and P3 use the same type of chip, the limited inventory can be concentrated to prioritize ensuring the full-load operation of one production node, while other nodes reduce their capacity.

[0101] Based on the above factors, the system calculates the time required for the supply chain to recover to normal operating conditions (defined as recovering to 90% of the pre-disruption operating level). In this example, the calculated recovery time is 52 days (including a 10-day inventory buffer period, a 45-day product adjustment period, and a 7-day capacity ramp-up period).

[0102] The system then calculates the supply chain performance loss throughout the entire scenario simulation. The performance loss includes three aspects:

[0103] (1) Production Loss. Calculate the difference between actual and normal production during the simulation period. In this example, during the 52-day recovery period, production decreased by approximately 60% in the first 10 days, by approximately 80% in the middle 35 days (maintaining only one production node), and gradually recovered in the last 7 days, with an average production loss of approximately 70%. If the company's normal daily production is 1000 units, the total production loss is approximately... tower.

[0104] (2) Delivery Delays. Calculate the order delivery delays caused by insufficient production capacity. In this example, approximately 15,000 orders were delayed during the simulation, with an average delay time of 25 days.

[0105] (3) Additional costs. These include emergency procurement costs, overtime production costs, and customer compensation costs. In this case, the unit price of the chip from the backup supplier B1 was 15% higher than that of S1, the additional engineering costs resulting from product design adjustments were approximately RMB 500,000, and the estimated customer compensation and reputational damage were approximately RMB 2 million.

[0106] Taking into account both recovery time and performance loss, the system assesses the resilience level of the supply chain under this scenario. The resilience assessment uses a comprehensive resilience index, calculated as follows:

[0107] ,

[0108] in: This is the toughness index, with a value range of [0, 1]. A higher value indicates better toughness. This refers to the actual recovery time. The maximum acceptable recovery time is set to 90 days in this embodiment; This represents the actual performance loss. In this embodiment, the maximum acceptable performance loss is set to 100% of normal production. The adaptability score reflects the supply chain's ability to respond to and adjust during crises, with a value range of [0, 1]. , , As a weighting coefficient, it is set in this embodiment. , , .

[0109] In this example, sky, Equivalent to 36.4 days of normal production, adaptability score (The company has backup suppliers and can adjust its production strategy, but the adjustment speed is not fast enough.) The resilience index is calculated as follows:

[0110] ,

[0111] The system sets the first resilience threshold range to [0.6, 1.0]. Since the calculated resilience index of 0.527 is below this range, the system determines that the supply chain is not resilient enough in this scenario and an optimization strategy needs to be developed.

[0112] Finally, the system generates optimization strategies for scenarios with insufficient resilience. The specific process of generating optimization strategies is as follows:

[0113] (1) Analysis of weak links. System analysis revealed that the main reasons for the insufficient resilience of the supply chain in this scenario are: ① excessive dependence on a single supplier S1 (accounting for 80% of chip demand); ② mismatch in product specifications and long activation cycle of backup supplier B1; ③ insufficient safety stock level to cope with long-term interruptions.

[0114] (2) Generate candidate strategies for weak links:

[0115] Strategy A: Supplier diversification strategy. It is recommended to add 1-2 new chip suppliers to reduce the supply ratio of S1 from 80% to below 50%, while requiring new suppliers to provide chips with the same specifications as S1 to ensure that switching can be done at any time.

[0116] Strategy B: Increase safety stock. It is recommended to increase chip safety stock from the current 7-12 days to 30 days to cope with longer-term supply disruptions.

[0117] Strategy C: Flexible Product Design. It is recommended to redesign the product to be compatible with different chip specifications, reduce reliance on specific suppliers, and improve supply chain flexibility.

[0118] Strategy D: Deepen Supplier Cooperation Strategy. It is recommended to establish a long-term cooperative relationship with backup supplier B1, place small-batch orders regularly to maintain active cooperation, and require B1 to maintain a certain amount of emergency inventory.

[0119] (3) Evaluate the costs and benefits of each strategy:

[0120] Expected costs of Strategy A: Development costs for new suppliers are approximately RMB 200,000, and annual additional costs for managing multiple suppliers are approximately RMB 100,000. Expected benefits: If S1 is disrupted again, due to the reduced supply ratio and the availability of readily switchable alternative sources, the estimated recovery time can be shortened to 15 days, production loss reduced to 10 days, and performance loss reduced by approximately 70%. In this scenario, losses of approximately RMB 25 million can be avoided (based on the market value of production loss and customer compensation).

[0121] Expected costs of Strategy B: The additional inventory holding costs for increasing inventory levels. Assuming a chip price of 1000 yuan per unit, increasing the average inventory across the three production nodes from 10 days to 30 days requires an increase of approximately 6000 chips in inventory, with an annual holding cost of approximately [missing information]. RMB 10,000 (assuming an annualized cost of ownership of 15%). Expected benefits: It can extend the buffer period for dealing with supply disruptions from 10 days to 30 days, allowing more time to find alternative solutions. It is estimated that it can reduce performance loss by about 30%, and in this case, it can avoid losses of about RMB 8 million.

[0122] Expected costs of Strategy C: R&D costs for product redesign are approximately 1 million yuan, and production line adjustment costs are approximately 500,000 yuan, which may lead to a slight decrease in product performance. Expected benefits: After the product becomes compatible with multiple chips, it can significantly improve supply chain flexibility. If a similar interruption occurs again, the switchover time can be shortened to 7 days, and it is estimated that losses of approximately 30 million yuan can be avoided.

[0123] Expected costs of Strategy D: The annual order cost of maintaining cooperation with the backup supplier is approximately 500,000 yuan (small-batch orders typically have higher unit prices), and requiring the supplier to maintain inventory may require a certain inventory subsidy, approximately 200,000 yuan. Expected benefits: The activation cycle for the backup supplier can be shortened from 45 days to 10 days, and an estimated loss of approximately 15 million yuan can be avoided.

[0124] (4) Calculate and rank the cost-benefit ratios:

[0125] Cost-benefit ratio of Strategy A: Annual cost of 300,000 yuan (development cost amortized over 3 years), expected revenue of 25 million yuan, cost-benefit ratio = 25 million / 300,000 ≈ 83.3.

[0126] Cost-benefit ratio of strategy B: annual cost of 900,000 yuan, expected revenue of 8 million yuan, cost-benefit ratio = 800 / 90 ≈ 8.9.

[0127] Cost-benefit ratio of strategy C: annual cost of 500,000 yuan (one-time cost amortized over 3 years), expected revenue of 30 million yuan, cost-benefit ratio = 3000 / 50 = 60.0.

[0128] Cost-benefit ratio of strategy D: annual cost of 700,000 yuan, expected revenue of 15 million yuan, cost-benefit ratio = 15 million / 700,000 ≈ 21.4.

[0129] The system sets the first benefit threshold range as a cost-benefit ratio ≥ 5.0. All four strategies meet this range, and are ranked from highest to lowest cost-benefit ratio as follows: Strategy A (83.3) > Strategy C (60.0) > Strategy D (21.4) > Strategy B (8.9).

[0130] (5) Consider the difficulty of implementation and synergistic effects. Although strategy A has the highest cost-effectiveness ratio, its implementation requires a certain period (finding and certifying new suppliers typically takes 6-12 months). Strategy B can be implemented immediately but is more expensive. Strategy C has high returns but also high risks (product redesign may affect product competitiveness). Strategy D can be implemented relatively quickly and has moderate risks.

[0131] After comprehensive consideration, the recommended combination of optimization strategies is as follows: implement strategies B and D in the short term (0-6 months) to quickly improve the buffer capacity of the supply chain; implement strategies A and C in the medium to long term (6-18 months) to fundamentally improve the resilience of the supply chain.

[0132] The system will send the generated resilience optimization strategy to the decision support module 6 for managers to use as a reference for decision-making.

[0133] like Figure 6 As shown, Decision Support Module 6 integrates risk warning information and resilience optimization strategies, providing a visual decision support interface for supply chain managers.

[0134] The visual decision support interface is implemented in a web page format and includes the following main functional modules:

[0135] (1) Supply Chain Network Topology Diagram. This diagram presents the supply chain network using a force-directed layout algorithm. Different types of nodes are represented by different colors and shapes: supplier nodes are blue circles, production nodes are green squares, logistics nodes are orange triangles, and retail nodes are gray rhombuses. The size of a node is proportional to its importance in the supply chain (based on a combined score of degree centrality and betweenness centrality); more important nodes are displayed larger. The thickness of edges reflects the strength of business relationships, i.e., edge weights; the greater the edge weight, the thicker the edge line. Users can hover the mouse over a node or edge to view detailed attribute information, such as the node's historical operating data, current status, vulnerability indicators, etc., or the edge's transaction volume, transaction frequency, etc.

[0136] (2) Risk Heatmap. The risk heatmap is overlaid on the network topology map, and nodes are colored according to their vulnerability indices and current risk status. A gradient color scheme is used: green represents low risk (vulnerability index < 0.3), yellow represents medium risk (0.3 ≤ vulnerability index < 0.45), orange represents high risk (0.45 ≤ vulnerability index < 0.6), and red represents high risk (vulnerability index ≥ 0.6). In this way, managers can easily identify high-risk areas in the supply chain. In real-time monitoring mode, the node colors are dynamically updated based on the latest data, and when the risk status of a node changes, an animation effect (such as flashing) will attract the manager's attention.

[0137] (3) Key Risk Propagation Path Diagram. This diagram highlights the identified key risk propagation paths. Each path is represented by a flowing arrow of a different color; the arrow's flow speed reflects the speed of risk propagation, and the arrow's thickness reflects the intensity of the risk's impact. Users can click on a path to view its detailed information, including all nodes on the path, the impact intensity of each node, and the overall risk level of the path. The system also provides a path comparison function, which can display multiple propagation paths simultaneously, helping managers understand the differences in the impact of different risk sources.

[0138] (4) Scenario Analysis Tool. This tool provides interactive scenario simulation functionality. Managers can select a scenario from a pre-set risk scenario library or customize risk parameters (such as specifying a node interruption, setting the interruption duration and impact intensity). After clicking the "Run Simulation" button, the system calculates the supply chain response under that scenario in real time and dynamically displays the diffusion process of the impact on the network topology diagram. After the simulation is completed, the system displays the resilience assessment results, including indicators such as recovery time, performance loss, and resilience index, and displays the time series performance change curves in chart form. This interactive scenario analysis function enables managers to conduct hypothesis analysis, quickly understand the potential impact of different risk scenarios, and provide a basis for developing contingency plans.

[0139] (5) Optimization Strategy Comparison Chart. This chart uses bar charts, radar charts, etc., to compare different optimization strategies. The horizontal axis represents different optimization strategies, and the vertical axis represents various indicators of the strategies, including expected cost, expected benefit, cost-benefit ratio, implementation cycle, and implementation difficulty. Managers can use this chart to quickly compare the advantages and disadvantages of different strategies and choose the strategy most suitable for the company's current situation. The system also provides a strategy combination function, where managers can select multiple strategies, and the system calculates the overall cost, benefit, and synergy of the combined implementation, helping managers to formulate comprehensive optimization plans.

[0140] (6) Real-time Monitoring Dashboard. This dashboard displays key operational and risk indicators of the supply chain in card format, including: overall supply chain health score, number of current high-risk nodes, number of active risk warnings, list of recent risk events, and supply chain resilience index. Each card uses a combination of numbers and charts, with the changing trends of important indicators shown using mini line graphs. When an indicator exceeds the warning threshold, the card turns red and flashes. Managers can click on the card to access the corresponding detailed page for more information.

[0141] In a specific application scenario, a supply chain manager at an electronics manufacturing company logs into the decision support interface. The first thing they see on the real-time monitoring dashboard are two high-risk nodes and one active risk warning (regarding a typhoon that will impact suppliers along the southeast coast). The manager clicks on the warning, and the interface switches to a network topology diagram, automatically highlighting the threatened nodes and potential propagation paths. Using the scenario analysis tool, the manager selects the natural disaster scenario—typhoon—sets the affected area and intensity, and runs the simulation. A few seconds later, the system displays the simulation results: if the typhoon makes landfall as predicted, five nodes are expected to be affected, the supply chain recovery time is approximately 18 days, and the estimated production loss is about 15%. The system also displays optimization strategies for this scenario: recommending the activation of backup suppliers, increasing safety stock in advance, and coordinating alternative logistics routes. The manager reviews the optimization strategy comparison chart, selects the most cost-effective combination strategy, and sends the execution instructions to relevant departments through the system. The entire decision-making process is efficient, intuitive, and evidence-based.

[0142] The dynamic monitoring module 7 is responsible for real-time monitoring of the supply chain operation status and triggering the system to conduct real-time risk assessment when abnormal changes are detected.

[0143] The dynamic monitoring module 7 is connected to the data acquisition module 1, continuously receiving real-time updates of supply chain operation data and external environment data. The system sets a set of monitoring indicators and corresponding fluctuation thresholds, including:

[0144] (1) Supplier delivery rate. The normal fluctuation range is set at ±5%. If a supplier's delivery rate drops by more than 5% in a short period of time (such as within a week), an alert will be triggered.

[0145] (2) Inventory Level. Set a safety stock lower limit and an inventory upper limit for each node. If the inventory is lower than 110% of the safety stock lower limit, an alert is triggered; if the inventory exceeds 110% of the upper limit, an alert is also triggered (there may be a risk of decreased demand or overstocking).

[0146] (3) Logistics delay. The normal logistics timeliness is set to a baseline value, and the fluctuation threshold is set to ±20%. If the logistics delay exceeds 120% of the baseline value, an early warning will be triggered.

[0147] (4) External environment indicators. Thresholds are set for key external environment indicators, such as a PMI index below 50, a daily exchange rate fluctuation of more than 3%, and a weather warning reaching the orange or red level, all of which trigger an early warning.

[0148] When any indicator is detected to exceed the fluctuation threshold, the dynamic monitoring module 7 immediately notifies the risk identification and early warning module 4 to conduct a real-time risk assessment. The real-time risk assessment process is similar to that in Embodiment 4, but with higher priority and faster response. The system completes the risk assessment within seconds, updates the risk warning information, and displays the latest risk situation in real time on the decision support interface.

[0149] In a specific application instance, the system monitored that the delivery rate of a key chip supplier, S2, dropped from 98% to 88% in the past week, a decrease of 10%, exceeding the 5% fluctuation threshold. The dynamic monitoring module 7 immediately triggered an alert, and the risk identification and early warning module 4 initiated a real-time assessment. System analysis revealed a recent factory accident in the region where S2 is located (information obtained from news data sources), resulting in the shutdown of some production lines. The system immediately updated the operational stability score of the S2 node, recalculated its vulnerability indicators, and simulated the risk propagation path should S2 completely fail. The assessment results showed that if S2 could not recover within two weeks, it could affect four downstream production nodes. The system generated a level-two alert (orange), reminding the supply chain manager to closely monitor the recovery progress of S2 and prepare to activate backup supplier plans. Simultaneously, on the real-time monitoring dashboard of the decision support interface, the relevant risk indicator cards turned orange and updated their data, and the color of the S2 node on the network topology map changed from green to orange, alerting managers to the issue.

[0150] Through the real-time monitoring function of the dynamic monitoring module 7, the system can promptly capture changes in the supply chain environment, identify potential risks in advance, buy valuable response time for enterprises, and significantly improve the initiative and effectiveness of supply chain risk management.

[0151] The various technical features of this invention exhibit a synergistic mechanism. The data acquisition module collects multi-source heterogeneous data, providing a rich information foundation for subsequent modules; the supply chain graph network construction module transforms discrete supply chain entity information into a structured graph network model, enabling the quantitative analysis of complex supply chain relationships; the multi-source data fusion module achieves differentiated fusion of different data sources through a data weighting coefficient mechanism, ensuring that high-quality data contributes more to decision-making; the risk identification and early warning module identifies risk points and propagation paths based on an enhanced graph network model, achieving proactive early warning; the resilience assessment and optimization module quantifies supply chain resilience through scenario simulation and generates optimization strategies; and the decision support module integrates all analysis results and presents them to managers in a visual manner. These modules work collaboratively to form a complete closed loop from data acquisition, model building, risk identification to decision support, significantly improving the systematic nature and effectiveness of supply chain risk management.

[0152] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of the claims of the present invention.

Claims

1. An intelligent supply chain risk early warning and resilience optimization management system, characterized in that, include: The data acquisition module is used to collect supply chain operation data, external environment data, and historical risk event data. The supply chain operation data includes supplier information data, production data, logistics data, and market data. The external environment data includes macroeconomic indicator data, geopolitical event data, weather forecast data, and supplier financial data. The historical risk event data includes supply chain disruption record data, recovery time data, and loss assessment data. The supply chain graph network construction module is connected to the data acquisition module and is used to construct a supply chain graph network model based on the supply chain operation data. In the supply chain graph network model, supplier nodes, production nodes, logistics nodes and market nodes are graph nodes, and supply relationships, production relationships, transportation relationships and sales relationships are graph edges. Each graph node contains node attribute information, and each graph edge contains edge weight information. A multi-source data fusion module, connected to the data acquisition module and the supply chain graph network construction module, is used to fuse the external environment data with the supply chain graph network model. It determines data weight coefficients based on the reliability and timeliness levels of the data sources. If the reliability level and timeliness level of the data source meet a first reliability threshold range, a first data weight coefficient is assigned. If the reliability level or timeliness level of the data source does not meet the first reliability threshold range, a second data weight coefficient is assigned. The first data weight coefficient is greater than the second data weight coefficient. The external environment data is integrated as an additional feature into the node attribute information of the corresponding graph nodes in the supply chain graph network model. The external environment feature values ​​in the node attribute information are multiplied by the corresponding data weight coefficients to update the node attribute information of all relevant graph nodes in the supply chain graph network model, thereby obtaining the fused enhanced graph network model. The risk identification and early warning module, connected to the multi-source data fusion module, is used to identify potential risk points and risk propagation paths in the supply chain based on the enhanced graph network model, calculate the vulnerability index of each node, and mark the node as a high-risk node if the vulnerability index of a node exceeds the first vulnerability threshold. The risk impact is then propagated from the high-risk node to adjacent nodes along the graph edges using a graph network propagation algorithm. During the propagation process, the intensity of the risk impact is attenuated according to the edge weight and the node's resistance capacity. If the intensity of the risk impact received by an adjacent node exceeds the threshold that the adjacent node can withstand, the adjacent node is marked as an affected node. This simulates the diffusion process of risk in the supply chain network, identifies the critical path of risk propagation, and generates risk early warning information. The resilience assessment and optimization module, connected to the risk identification and early warning module, is used to assess the resilience level of the supply chain under different risk scenarios based on the scenario simulation method. It sets multiple risk scenario parameter combinations, simulates the response process of the supply chain network for each risk scenario, calculates the supply chain resilience assessment index, and generates a resilience optimization strategy if the supply chain resilience assessment index does not meet the preset threshold range. The optimization strategies are then ranked according to their cost-benefit ratio. The decision support module, connected to the risk identification and early warning module and the resilience assessment and optimization module, is used to integrate the risk early warning information and the resilience optimization strategy, and provides a visual decision support interface, which displays a supply chain network topology map, a risk heat map, a key risk propagation path map and an optimization strategy comparison map.

2. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When constructing the supply chain graph network model, the supply chain graph network construction module is also used to identify various entities in the supply chain, treating suppliers, production facilities, logistics centers, and retailers as different types of graph nodes. Graph edges are determined based on the business relationships between entities. If entity A provides products or services to entity B, a directed graph edge is established between the graph node corresponding to entity A and the graph node corresponding to entity B. A node attribute vector is set for each graph node, and an edge weight is set for each graph edge. The edge weight is calculated based on transaction volume, transaction frequency, and transaction stability.

3. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When the multi-source data fusion module integrates external environmental data with the supply chain graph network model, it also preprocesses the external environmental data, establishes the correlation between external environmental factors and supply chain nodes, calculates the influence weight of each external environmental factor on the supply chain node, and assigns a first influence weight if the correlation coefficient between the change of a certain external environmental factor and the change of the operational status of a certain node meets the first correlation threshold range. If the correlation coefficient does not meet the first correlation threshold range, a second influence weight is assigned. The first influence weight is greater than the second influence weight, and the external environmental factors are integrated as additional features into the attribute vector of the corresponding node.

4. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When calculating node vulnerability indicators, the risk identification and early warning module also calculates the node's degree centrality, betweenness centrality, and clustering coefficient. The degree centrality is the sum of the node's in-degree and out-degree, where the in-degree is the number of graph edges pointing to the node and the out-degree is the number of graph edges pointing from the node. The betweenness centrality is determined based on the number of shortest paths between node pairs passing through the node and the total number of shortest paths between node pairs. The clustering coefficient is determined based on the actual number of graph edges between the node's neighboring nodes and the node's degree. Operational stability data is a score reflecting the node's operational stability and ranges from 0 to 1. Combining the degree centrality, betweenness centrality, and clustering coefficient with the node's operational stability data, the node's vulnerability indicator is calculated. If the node's degree centrality exceeds a first degree centrality threshold and its betweenness centrality exceeds a first betweenness centrality threshold, while its clustering coefficient is below a first clustering coefficient threshold or its operational stability data does not meet the first stability threshold range, then the node's vulnerability indicator is determined to exceed the first vulnerability threshold.

5. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When identifying risk propagation paths, the risk identification and early warning module also selects an initial risk node. Based on a graph network propagation algorithm, the risk impact is propagated from the initial risk node along the graph edges to adjacent nodes. During the propagation process, the intensity of the risk impact is attenuated according to the edge weights and the node's resistance capacity. If the intensity of the risk impact received by an adjacent node exceeds the node's tolerance threshold, the adjacent node is marked as an affected node. The complete path of risk propagation is recorded. The risk level of the propagation path is calculated based on the total number of nodes affected by the propagation path, the sum of the impact intensity, and the criticality of the affected nodes. If the risk level exceeds the first risk level threshold, the propagation path is marked as a critical risk propagation path.

6. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When performing scenario simulations, the resilience assessment and optimization module is also used to set up a risk scenario library, which includes natural disaster scenarios, supplier bankruptcy scenarios, logistics disruption scenarios, demand surge scenarios, and geopolitical shock scenarios. Specific scenario parameters are set for each risk scenario, and the impact of the risk scenario is simulated in the supply chain graph network model. The performance degradation of affected nodes is calculated, the self-recovery process of the supply chain network is simulated, the time required for the supply chain to recover to normal operating status is calculated, the supply chain performance loss during the entire scenario simulation process is calculated, and the supply chain resilience level is assessed by combining the recovery time and performance loss.

7. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When generating resilience optimization strategies, the resilience assessment and optimization module also analyzes the type and cause of identified supply chain weaknesses. If the weakness is a lack of alternative options for a critical node, a supplier diversification strategy is generated. If the weakness is insufficient node inventory buffer, an inventory buffer strategy is generated. If the weakness is a single logistics path, an alternative logistics path strategy is generated. For each optimization strategy, the expected cost and expected benefit of implementing the strategy are calculated, the cost-benefit ratio of each optimization strategy is calculated, the optimization strategies are ranked, and the optimization strategies with high cost-benefit ratio and implementation difficulty within the feasible range are given priority recommendation.

8. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When providing visual decision support, the decision support module is also used to present the supply chain network model in the form of a network topology diagram, overlay a risk heat map on the network topology diagram based on the vulnerability indicators of the nodes, highlight the identified key risk propagation paths, provide interactive scenario analysis functions, allow managers to select specific risk scenarios or customize risk parameters, and display a comparison diagram of optimization strategies for different risk scenarios.

9. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, When constructing the supply chain graph network model, the supply chain graph network construction module is also used to identify the community structure in the supply chain. The community discovery algorithm is used to cluster nodes with a connection density exceeding the first connection density threshold into communities. Each community represents a relatively independent functional module in the supply chain. When conducting risk analysis, the edges connecting across communities are evaluated first.

10. The intelligent supply chain risk early warning and resilience optimization management system according to claim 1, characterized in that, The system also includes a dynamic monitoring module, which is connected to the data acquisition module, the risk identification and early warning module, and the decision support module. It is used to monitor the supply chain operation status in real time. If the supply chain operation data or external environment data is detected to change beyond a preset fluctuation threshold, the system is triggered to perform a real-time risk assessment, update the risk warning information, and display the latest risk situation in the decision support interface in real time.

Images