Supply chain disruption risk early warning method fusing knowledge graph

Abstract

The present application relates to the technical field of intelligent risk management, and particularly relates to a supply chain interruption risk early warning method fusing a knowledge graph, a supply chain knowledge graph is constructed by mounting node buffer parameters and relationship edge transmission constraint parameters, real-time risk events are collected to match and locate risk source nodes and configure corresponding risk intensity, a single-path exclusive risk penetration threshold is calculated by traversing hop by hop along downstream supply relationships, the path interruption state is determined by comparing the risk intensity and the risk penetration threshold, after link side exclusive attenuation and node side shared quota attenuation, the remaining risk intensity is obtained and risk transmission is continued along the relationship edge, the node interruption risk value is obtained by collecting effective transmission paths, and grading early warning is realized in combination with multi-level early warning thresholds. The present application effectively reduces false early warning, improves risk transmission calculation accuracy and early warning level accuracy, and provides technical support for supply chain risk management disposal.

Description

Technical Field

[0001] This invention relates to the field of intelligent risk management technology, and in particular to a supply chain disruption risk early warning method that integrates knowledge graphs. Background Technology

[0002] In the field of manufacturing supply chain management, multi-level supporting and cross-regional collaborative supply networks have become the mainstream. With the development of knowledge graph technology, risk transmission reasoning based on supply chain topology graphs has gradually become the technical direction of intelligent risk control in the supply chain.

[0003] In existing technologies, the dependence between supply chain entities is used as the basis for determining risk transmission, and the risk transmission path and scope of impact are determined by the correlation of node structures. However, existing technical solutions do not embed factors with business buffering capabilities, such as enterprise safety stock, into the risk transmission reasoning process, ignoring the role of buffering capabilities in offsetting upstream risks. That is, for some entity nodes, they can completely offset the actual risks through their own business buffering capabilities. However, because the transmission reasoning ignores this influencing factor, false warnings of supply chain disruption risks occur, which not only increases the ineffective workload of risk control personnel but also easily leads to warning information overload. On the other hand, existing technologies assume that each transmission path can fully occupy the node's buffer resources, ignoring the shared nature of buffer resources. When multiple path risks overlap, the buffering and offsetting capabilities of nodes are significantly overestimated, resulting in inaccurate warning levels and failing to guarantee supply chain risk management and graded emergency response.

[0004] Therefore, it is necessary to propose a supply chain disruption risk early warning method that integrates knowledge graphs to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide a supply chain disruption risk early warning method that integrates knowledge graphs, in order to solve the problems of false early warnings and inaccurate early warning levels caused by ignoring the buffering capacity of entity nodes themselves and the shared attributes of buffer resources.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A supply chain disruption risk early warning method integrating knowledge graphs includes the following steps: S1: On the basic topology layer formed by the entity nodes and topological dependencies at all levels of the supply chain, configure node buffer parameters for entity nodes and transmission constraint parameters for relationship edges to obtain the supply chain knowledge graph; S2: Collect real-time risk event data, match and locate the risk event to the corresponding entity node in the knowledge graph, mark the entity node as the risk source node, and configure the corresponding risk intensity for the risk source node according to the event level of the risk event; S3: Starting from the risk source node, perform traversal along the relationship edge in the downstream direction. At each downstream entity node, use its corresponding node buffer parameter and the transmission constraint parameter of the corresponding upstream relationship edge to calculate the risk penetration threshold of the downstream entity node. S4: Compare the risk intensity with the risk penetration threshold to determine whether to cut off the current transmission path and terminate the downstream reasoning along the transmission path; if not cut off, calculate the risk attenuation coefficient to obtain the remaining risk intensity after penetration, and use the remaining risk intensity to continue traversing along the downstream relationship edge. S5: After completing the full transmission path traversal, gather all valid transmission paths that have not been truncated, and obtain the interruption risk value of the downstream entity node based on the remaining risk intensity of the downstream entity node in the valid transmission path. S6: Compare the interruption risk value with the preset warning threshold, and issue an interruption risk warning for downstream entity nodes based on the comparison result.

[0007] Preferably, the node buffer parameters include safety stock days and multi-source procurement quota ratio, and the transmission constraint parameters include the substitutability level of the corresponding supply channel and the buffer margin of order lead time.

[0008] Preferably, the step of matching and locating the risk event to the corresponding entity node in the knowledge graph is specifically executed progressively through the following matching conditions: When a risk event has information identifying the operating entity, the entity node is located by matching the entity's identity. When a risk event is a regionally impactful event, the entity node is located by matching the geographic scope. When a risk event is a category-impact event, the entity node is located by matching the category and industry affiliation. After executing at least one of the above matching conditions, the matching results are filtered by the supply chain role verification conditions to obtain the matched and located entity nodes.

[0009] Preferably, the role verification condition is to match the corresponding supply chain role type according to the risk event type, and filter out entity nodes whose supply chain role attributes do not match the risk event impact type.

[0010] Preferably, the step of calculating the risk penetration threshold of the downstream entity node includes: After standardizing the safety stock days and multi-source procurement quota ratio of the downstream entity nodes reached during the current traversal, the nodes are integrated according to the preset weights within the downstream entity nodes to obtain the node-side buffer score. For the upstream relationship edges of the downstream entity nodes through which the risk transmission path passes, after standardizing the substitutability level of the supply channel and the buffer margin of the order lead time of the upstream relationship edges, the link-side buffer score is obtained by merging according to the preset link weight. Based on the node-side buffer score and the link-side buffer score, the risk penetration threshold of the upstream relation edge corresponding to the current downstream entity node is calculated.

[0011] Preferably, the step of comparing the risk intensity with the risk penetration threshold to determine whether to cut off the current transmission path includes: Obtain the incident risk intensity from the upstream entity node to the downstream entity node through the corresponding relation edge in the current propagation path; When the incident risk intensity is less than the risk penetration threshold, it is determined that the downstream entity node can offset the risk impact of the current transmission path, the current transmission path is cut off, and the downstream inference along the transmission path is terminated. When the incident risk intensity is not less than the risk penetration threshold, it is determined that the risk of the current propagation path can penetrate the downstream entity node, the propagation path is retained, and the traversal continues downstream with the remaining risk intensity after penetration.

[0012] Preferably, the step of calculating the risk attenuation coefficient to obtain the remaining risk intensity after penetration includes: For the current propagation path, the incident risk intensity is attenuated on the link side using the propagation constraint parameters of the upstream relation edge, and the incident risk intensity after link attenuation is obtained. Based on the incident risk intensity after link-side attenuation, the incident risk intensity of all uninterrupted transmission paths of the current downstream entity node is aggregated, and the total incident risk intensity of the downstream entity node is calculated. The total buffer capacity of the nodes is determined based on the node buffer parameters of the downstream entity nodes. Based on the proportion of the incident risk intensity after link attenuation of each transmission path to the total incident risk intensity, the total buffer capacity of the nodes is allocated to each transmission path, generating a path buffer quota corresponding to each transmission path. For the current transmission path, the path buffer quota is used as the upper limit to attenuate the incident risk intensity after the link is attenuated at the node side. The part exceeding the path buffer quota is regarded as the remaining risk intensity after penetrating the downstream entity node.

[0013] The technical effects and advantages of the present invention in the above technical solution are as follows: 1. This invention, by configuring node buffer parameters for entity nodes and transmission constraint parameters for relation edges on the basis of the supply chain topology layer, differs from existing knowledge graphs that only depict topological dependencies. It utilizes the risk resistance of nodes and the transmission resistance of links to embed them into the graph structure, thereby reducing false warnings that only exist in terms of structural dependencies but whose actual risks can be completely absorbed at the source of knowledge graph modeling.

[0014] 2. By calculating the risk penetration threshold, and binding this threshold to each upstream supply relationship edge, the differences in risk resistance capabilities of different supply chains are reflected. The single upstream transmission path plus the corresponding supply relationship edge is used as the smallest independent judgment unit. Threshold comparison and truncation status judgment are performed independently for multiple upstream paths of the same downstream node. Invalid paths that have been completely offset are truncated in real time during the traversal process, thereby reducing the computational load of downstream multi-hop inference and improving the efficiency of early warning response.

[0015] 3. In this invention, link-side attenuation is executed independently based on the unique constraints of each link, and node-side attenuation is dynamically allocated to the global buffer quota according to the risk proportion of each path. This differs from existing methods that assume each path can fully occupy the node buffer resources. This solves the technical problem of overestimating the buffer offsetting capacity when multiple path risks overlap, improves the accuracy of the remaining risk intensity calculation, and avoids inaccurate early warning levels from the perspective of transmission calculation. Attached Figure Description

[0016] Figure 1 This is a flowchart of the supply chain disruption risk early warning method integrating knowledge graphs according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Currently, existing risk warning solutions that integrate knowledge graphs rely solely on topological dependencies as the basis for risk transmission. They fail to incorporate business buffer mechanisms, such as safety stock, into the transmission reasoning process. This leads to the frequent generation of false warnings that indicate structural dependencies exist, but the actual risks can be completely mitigated. Furthermore, when calculating the attenuation effect of multiple upstream risk paths on the same node independently, existing methods assume that each path can fully utilize the node's buffer resources, ignoring the shared nature of buffer resources. When multiple path risks overlap, the buffer offsetting capacity is significantly overestimated, resulting in inaccurate warning levels and an inability to guarantee risk management and tiered emergency response in the supply chain.

[0019] like Figure 1As shown, the purpose of this invention is to provide a supply chain disruption risk early warning method that integrates knowledge graphs. This embodiment takes the complete vehicle supply chain system of a passenger car OEM as an example. This supply chain has the OEM as the downstream node, and upstream it includes two Tier 1 suppliers (seat assemblies and chassis assemblies), five Tier 2 component suppliers (seat frames, stamped sheet metal parts, etc.), three Tier 3 raw material suppliers (hot-rolled steel, modified engineering plastics, etc.), and four logistics nodes (coastal ports, regional distribution warehouses, etc.). This embodiment addresses the risk event of a Tier 3 hot-rolled steel supplier experiencing a sudden safety accident and production stoppage. Based on the integrated supply chain knowledge graph, risk transmission reasoning is performed to achieve interruption risk prediction and tiered early warning output. The implementation process of this invention is explained in detail below with specific steps: In one embodiment of the present invention, existing supply chain knowledge graphs only depict the topological dependencies between entities, without incorporating risk buffering mechanisms such as safety stock, or transmission constraints such as supply channel substitution tolerance and delivery time margin into the knowledge graph. This results in a lack of quantitative judgment criteria for risk transmission reasoning. If warning results are output solely based on structural associations, a large number of false warnings are likely to be generated. Therefore, in step S1, node buffering parameters are configured for entity nodes and transmission constraint parameters are configured for relationship edges on the basic topology layer formed by the entity nodes and topological dependencies at all levels of the supply chain to obtain the supply chain knowledge graph. This is specifically implemented through the following method: For the basic topology layer of the supply chain, the vehicle OEM is the downstream entity node. Based on the supply dependence relationship at each level, two Tier 1 suppliers (seat assembly and chassis assembly), five Tier 2 component suppliers (such as seat frame factory and stamping sheet metal factory), three Tier 3 raw material suppliers (such as hot rolled steel factory and modified engineering plastics factory), and multiple logistics nodes are included as entity nodes. The entity nodes are connected by business dependence relationships such as raw material supply, component distribution, and assembly supply, forming the basic topology layer of the entire level. Based on the basic topology layer constructed by the above method, node buffer parameters are configured for each entity node. These parameters include safety stock days and multi-source procurement quota ratio. Taking the Tier 2 stamping sheet metal plant node as an example, its safety stock days for hot-rolled steel raw materials are configured to be 7 days, meaning that the node's standing inventory can support 7 days of continuous production consumption. The multi-source procurement quota ratio is 60% for the current main hot-rolled steel plant and 40% for the backup steel supplier. All entity nodes are configured with corresponding node buffer parameters according to their own business realities, serving as a buffer capacity carrier for the node to resist upstream supply interruptions. While configuring node buffer parameters, transmission constraint parameters are simultaneously configured for each upstream supply relationship edge. These transmission constraint parameters include the substitutability level of the supply channel and the order lead time buffer margin. For example, taking the hot-rolled steel supply relationship edge from the hot-rolled steel mill to the stamping sheet metal mill, the substitutability level of this channel is configured as level 2, meaning there are 2 switchable alternative suppliers for the corresponding steel product. The order lead time buffer margin is configured as 3 days, meaning the regular procurement and delivery cycle for this channel is 10 days, and a 3-day delivery tolerance window is reserved on the node side. All supply relationship edges are configured with corresponding transmission constraint parameters, serving as the risk transmission resistance carrier for each individual supply chain. After configuring the parameters of all nodes and relation edges, the entity nodes, relation edges and their respective attached business parameters are structured and stored to obtain a supply chain knowledge graph that integrates business buffer attributes.

[0020] In one embodiment of the present invention, existing methods for matching and locating risk events with supply chain knowledge graphs often rely on matching single enterprise name strings. This method can only be used for risk events that explicitly point to a specific business entity, which can easily lead to matching omissions. Furthermore, the matching process does not consider using a combination of risk event type and node supply chain role for verification, thus often including irrelevant entity nodes in the risk propagation scope. Therefore, in step S2, real-time risk event data is collected, the risk event is matched and located to the corresponding entity node in the knowledge graph, and this entity node is marked as a risk source node. Based on the event level of the risk event, a corresponding risk intensity is configured for the risk source node. This is specifically implemented through the following method: For real-time risk event data collection, we connect with industry safety supervision announcements, supply chain monitoring platforms, trunk logistics tracking systems and enterprise internal procurement anomaly reporting channels to collect multi-source real-time risk event data. Each event data entry includes attribute fields such as event subject identifier, location of occurrence, affected material category, event type, and expected duration. Based on the matching and location of risk events and entity nodes, the specific matching conditions include: When a risk event has information identifying the operating entity, the entity node is located by matching the entity's identity. When a risk event is a regionally impactful event, the entity node is located by matching the geographic scope. When a risk event is a category-impact event, the entity node is located by matching the category and industry affiliation. After executing at least one of the above matching conditions, the matching results are filtered by the supply chain role verification conditions to obtain the matched and located entity nodes.

[0021] Regarding the matching and positioning process, when a risk event contains clear business entity identification information, the entity node is located by matching the entity identity. In this embodiment, the production stoppage event contains a clear full name of the enterprise and a unified social credit code. These are matched one-to-one with the identity attribute fields of the entity node in the knowledge graph to match the Tier 3 hot-rolled steel plant entity node. When a risk event is a regionally impactful event, entity nodes are located by matching geographical scope. For example, when a notice of power rationing for the entire industrial area is collected, the geographical coverage of the risk event is extracted. At the same time, the geographical location attributes of all entity nodes in the knowledge graph are traversed to filter out entity nodes whose registered address or production plant is located within the geographical coverage area. When a risk event is a category-impact event, entity nodes are located by matching the category and industry affiliation. For example, when an event of a significant price increase across the entire industry of hot-rolled steel is collected, the material category "hot-rolled steel" associated with the event and its corresponding national economic industry classification code are extracted, and entity nodes in the knowledge graph that match the main product category and industry attributes are selected. After obtaining preliminary matching results by executing at least one of the above matching conditions, the matching results are filtered by supply chain role verification conditions. The role verification conditions are to match the corresponding supply chain role type according to the risk event type and filter out entity nodes whose supply chain role attributes do not match the risk event impact type. In this embodiment, the production safety accident belongs to the production interruption event, and its direct impact objects are limited to production and manufacturing nodes. Therefore, entity nodes belonging to logistics and warehousing or trade and distribution in the preliminary matching results are filtered out. For example, although the regional distribution warehouse node in the same area is within the geographical matching range, its supply chain role is a logistics node and is not directly affected by the production interruption event, so it is removed from the matching results, and the target entity node is finally obtained.

[0022] After matching and locating according to the above method, the obtained entity nodes are marked as risk source nodes, and the corresponding risk intensity is configured for the entity nodes according to the event level of the risk event. Specifically, supply chain risk events are pre-divided into four levels: general, major, serious, and extremely serious, which are mapped to risk intensity in the range of 0 to 100 points. The corresponding quantitative mapping rules are: general events correspond to risk intensity of 25, major events correspond to risk intensity of 45, serious events correspond to risk intensity of 70, and extremely serious events correspond to risk intensity of 95. Taking the hot-rolled steel plant full-line shutdown event in the above example, it is comprehensively evaluated as a serious event. Therefore, the initial incident risk intensity configured for this risk source node is 70.

[0023] By using progressive matching conditions, missed matches are effectively avoided. Through supply chain role verification, irrelevant entity nodes are filtered out from the matching source, which reduces the amount of invalid calculations in risk transmission reasoning and lowers the probability of false warnings.

[0024] In one embodiment of the present invention, existing risk transmission schemes typically set a uniform fixed risk threshold for entity nodes. However, the risk resistance capabilities of different upstream supply chains corresponding to the same entity node vary. Determining a fixed risk threshold can easily lead to overestimation of risk, resulting in false warnings, while underestimation can easily lead to missed warnings. Therefore, in step S3, starting from the risk source node, a traversal is performed along the relationship edge towards the downstream direction. At each downstream entity node, the risk penetration threshold of that downstream entity node is calculated using its corresponding node buffer parameters and the transmission constraint parameters of the corresponding upstream relationship edge. This is specifically implemented through the following method: Based on the risk source node identified by the above method, it is used as the starting point for traversal. Traversal is performed downstream along the relationship edge of the risk source node. Regarding the downstream direction, it should be understood that in this embodiment, when constructing the basic topology layer of the supply chain knowledge graph, the supply and demand dependency relationship is set as a unidirectional directed relationship edge. The unified rule of this relationship edge is that the material supplier (upstream entity) points to the material purchaser (downstream entity). Each relationship edge stores two attribute fields: upstream entity identifier and downstream entity identifier, thereby solidifying the upstream and downstream flow (i.e., direction) of the supply chain. During the traversal process starting from the risk source node, only all directed relationship edges with the risk source node as the upstream entity are retrieved. The purchaser entity pointed to by the arrow of the relationship edge is the downstream node. The traversal process in this embodiment only performs risk transmission traversal by jumping down along this unidirectional supply relationship, without tracing back to upstream raw materials or upstream suppliers, etc. During the traversal, the risk penetration threshold of each downstream entity node is calculated based on its node buffer parameters and the propagation constraint parameters of the corresponding upstream relation edges. The method is as follows: S31: After standardizing the safety stock days and multi-source procurement quota ratio of the downstream entity nodes reached during the current traversal, the nodes are merged according to the preset weights within the downstream entity nodes to obtain the node-side buffer score. S32: For the upstream relationship edges of the downstream entity nodes through which the risk transmission path passes, after standardizing the supply channel substitutability level and order lead time buffer margin of the upstream relationship edges, the link-side buffer score is obtained by merging according to the preset link weight. S33: Based on the node-side buffer score and the link-side buffer score, calculate the risk penetration threshold of the upstream relation edge corresponding to the current downstream entity node.

[0025] This embodiment follows the risk event of a sudden safety accident and production stoppage of a Tier 3 hot-rolled steel supplier in the above embodiment. Based on the marked risk source node of the Tier 3 hot-rolled steel plant as the traversal starting point, the traversal is performed along the downstream supply relationship edge, and the risk transmission direction is from the upstream entity to the downstream entity. The first hop of the traversal reaches the downstream Tier 2 stamping sheet metal plant entity node, and for this node and its corresponding upstream supply relationship edge of hot-rolled steel, the risk penetration threshold is calculated. As shown in steps S31 to S33 above, the calculation of the node-side buffer score is specifically implemented through the following method: For the current stamping and sheet metal plant entity nodes reached during the traversal, the configured safety stock days and multi-source procurement quota ratio are extracted, and standardized processing is performed separately to eliminate discrepancies. Then, they are merged according to the preset weights within the nodes to obtain the node-side buffer score. The standardization process uses a benchmark value normalization method to uniformly map the quantitative parameters of different businesses to a unified standard. For the safety stock days, a pre-defined industry benchmark safety stock period of 10 days is used as the normalization benchmark. The standardized value of the actual safety stock days is then the ratio of the actual days to the benchmark days, calculated using the following formula: ; in, The actual number of days of safety stock for physical nodes. Using the baseline safety stock days as an example, substitute the safety stock days for hot-rolled steel in the stamping sheet metal plant of this embodiment. Heavens, calculation ; For the standardization of multi-source procurement quota ratios, the maximum value of the reserve supplier procurement quota ratio (100%) is used as the benchmark. The standardized value is then the decimal form of the actual reserve quota ratio, calculated using the following formula: ; in, The proportion of procurement quotas allocated to backup suppliers for physical nodes; Substituting the proportion of spare steel supplier quota for the stamping sheet metal plant in this embodiment Calculations yielded .

[0026] After standardization is completed, based on the preset safety stock weights... Multi-source procurement quota weight The node-side buffer score is uniformly mapped to the range of 0-100 points, consistent with the risk intensity. The calculation formula is as follows: ; After substituting the values, the calculated result is 58, which means that the node-side buffer score of the stamping sheet metal plant node is 58 points. The higher the score, the stronger the node's ability to withstand upstream supply interruptions.

[0027] For the upstream relationship between hot-rolled steel mills and stamping sheet metal mills through which risk transmission occurs, the substitutability level of their supply channels and the buffer margin of order lead time are extracted. After standardization, these are merged according to preset link weights to obtain the link-side buffer score. This is specifically achieved through the following methods: The preset channel can be replaced up to 5 levels (corresponding to 5 switchable alternative suppliers). The standardized value is the ratio of the current level to the highest level, calculated using the following formula: ; in, As an alternative grade to the current supply channels, As the highest substitutable level, this substitutable level of the supply channel is used in this embodiment. Level, calculated ; If the preset baseline lead time buffer is 5 days, then the standardized value is the ratio of the actual buffer to the baseline buffer, with an upper limit of 1. The calculation formula is as follows: ; in, This provides a buffer for the lead time of orders from current channels. As a baseline buffer, based on the lead time buffer of orders in this channel. Heavens, calculation ; Integration is performed according to preset intra-link weights, and preset supply channels can replace grade weights. Order lead time buffer weight The link-side buffer score is also mapped to a score range of 0 to 100, and its calculation formula is as follows: ; The calculated buffer score for the hot-rolled steel supply relationship is 46 points. The higher the score, the stronger the risk transmission resistance of the supply chain itself.

[0028] Based on the node-side buffer score and the link-side buffer score, the weights of the node dimension are preset. Link dimension weight The summation yields the risk penetration threshold for the current downstream entity node corresponding to the upstream relation edge. The threshold score range is consistent with the risk intensity. The formula for calculating the risk penetration threshold is: ; After substituting the values, the calculated risk penetration threshold of the upstream supply side of the hot-rolled steel plant corresponding to the stamping sheet metal plant node is 53.2. This threshold represents the maximum risk intensity that the node can completely offset based on its own buffering capacity and corresponding link constraints.

[0029] It is important to understand that the same entity node will correspond to different risk penetration thresholds for different upstream supply relationships. For example, if another backup steel supply side of the stamping sheet metal plant has a channel substitutability level of 1 and a lead time buffer margin of 5 days, then the calculated link-side buffer score and the final risk penetration threshold will differ from those of the main supply side.

[0030] In one embodiment of the present invention, considering the differences in risk resistance capabilities among different upstream supply chain paths at the same node, existing methods cannot adapt to the differentiated transmission along different paths in a multi-source procurement scenario. Therefore, in step S4, the risk intensity is compared with the risk penetration threshold to determine whether to cut off the current transmission path and terminate downstream inference along the transmission path; this is specifically implemented through the following method: In this embodiment of the invention, the smallest independent judgment unit for risk transmission cutoff determination is: a single upstream transmission path (i.e., the entire risk path from the upstream node to the downstream node along the relationship edge) and the corresponding supply relationship edge (i.e., the unidirectional directed relationship edge in the knowledge graph). The two are combined into an independent judgment unit, and multiple upstream transmission paths of the same downstream node are independent of each other. According to the traversal process of the above embodiment, the current transmission path is from Tier3 hot-rolled steel plant to Tier2 stamping sheet metal plant. First, the incident risk intensity transmitted from the upstream hot-rolled steel plant to the stamping sheet metal plant through the hot-rolled steel supply relationship edge is obtained. In this embodiment, the initial risk intensity of the risk source node is 70, and it is the first hop of the transmission. Therefore, the incident risk intensity of this transmission path is 70. The intensity of the incident risk is compared one by one with the risk penetration threshold of the current upstream relation edge corresponding to the downstream entity node. The determination rule is as follows: When the incident risk intensity is less than the risk penetration threshold, it is determined that the downstream entity node can completely offset the risk impact from this path based on its own node buffering capacity and the link transmission constraints corresponding to the current transmission path. The risk will no longer continue to spread downstream. Therefore, the current transmission path is cut off, and the reasoning traversal along the transmission path to the more downstream nodes is terminated. When the incident risk intensity is not less than the risk penetration threshold, it is determined that the risk intensity of the current transmission path exceeds the maximum buffer consumption capacity of the node and link, and the risk can penetrate the downstream entity node. Therefore, the transmission path is retained and the subsequent risk attenuation calculation is performed to continue traversing downstream with the remaining risk intensity after penetration.

[0031] Substituting the values ​​from this embodiment for judgment, the incident risk intensity of the current main supply path is 70. The risk penetration threshold corresponding to this path calculated in the above embodiment is 53.2, which satisfies the condition that the incident risk intensity is greater than the risk penetration threshold. Therefore, it is determined that the risk can penetrate the stamping sheet metal plant node. The transmission path from the hot-rolled steel plant to the stamping sheet metal plant is retained without truncation. The risk attenuation calculation is performed based on this path, and after obtaining the remaining risk intensity, the transmission continues to the downstream Tier 1 seat assembly plant node.

[0032] It is important to understand that for the same entity node, if there are multiple upstream risk transmission paths, each transmission path independently performs the above threshold comparison and truncation judgment, and the results do not interfere with each other.

[0033] In this embodiment of the invention, if the transmission path is not interrupted, the risk attenuation coefficient is calculated based on the transmission path to obtain the remaining risk intensity after penetration, and the remaining risk intensity is used to continue traversing the downstream entity nodes. Because existing supply chain risk attenuation calculations mostly adopt a single-path independent attenuation mode, it is assumed that each upstream path can fully utilize all the node's buffer resources. However, node buffer capabilities such as safety stock and multi-source procurement quotas belong to a globally shared resource pool. When multiple upstream paths simultaneously transmit risks, buffer resources will be jointly consumed and mutually squeezed, leading to an overestimation of the risk mitigation effect and causing inaccurate warning levels. Therefore, in step S4, if not truncated, the risk attenuation coefficient is calculated to obtain the remaining risk intensity after penetration, and the remaining risk intensity is used to continue traversing along the downstream relationship edges. This is specifically implemented through the following method: The method for calculating the residual risk intensity after penetration includes the following steps: S41: For the current propagation path, use the propagation constraint parameters of the upstream relation edge to perform link-side attenuation on the incident risk intensity to obtain the incident risk intensity after link attenuation; S42: Based on the incident risk intensity after link-side attenuation, the incident risk intensity of all uninterrupted transmission paths of the current downstream entity node is aggregated, and the total incident risk intensity of the downstream entity node is calculated. S43: Determine the total buffer capacity of the nodes based on the node buffer parameters of the downstream entity nodes; S44: Based on the proportion of the incident risk intensity after link attenuation of each transmission path to the total incident risk intensity, allocate the total buffer capacity of the nodes to each transmission path, and generate the path buffer quota corresponding to each transmission path. S45: For the current transmission path, the node-side attenuation of the incident risk intensity after link attenuation is performed with the path buffer quota as the upper limit. The part exceeding the path buffer quota is regarded as the remaining risk intensity after penetrating the downstream entity node.

[0034] As shown in steps S41 to S45 above, the risk attenuation calculation in this embodiment adopts a two-tiered progressive method of link-side dedicated attenuation and node-side shared quota attenuation. Link-side attenuation is executed independently based on the unique transmission constraint parameters of each supply relationship edge, without resource contention. Node-side attenuation is performed by dynamically allocating quotas according to the risk proportion of each path based on the globally shared buffer resource pool of the nodes. Specifically, based on the determination results of the above embodiment, the current uncut path is from Tier 3 hot-rolled steel plant to Tier 2 stamping sheet metal plant. The incident risk intensity of this path is 70. The link-side buffer score corresponding to this upstream relationship edge is extracted as 46. Combined with the preset link dimension weight β=0.4, the maximum offsettable risk amount for this supply route is calculated using the following formula: ; in, This is a dedicated amount of risk that can be offset for a single link. This is the link-side buffer score for the relation edge. The weights are calculated by substituting the values ​​into the weights for the link dimension. ; Subtracting the link-counterable amount from the incident risk intensity, we obtain the incident risk intensity after link attenuation: ; in, The intensity of path incident risk. The incident risk intensity after link attenuation is obtained by substituting the values ​​into the numerical calculation. The link-side attenuation is executed independently for each supply relationship edge, with corresponding channel alternatives, delivery time margins, and other link-specific buffer capabilities, and the calculation results of each path do not affect each other.

[0035] Based on the attenuated incident risk intensity calculated above, the incident risk intensity of all uninterrupted transmission paths of the current downstream entity node is aggregated to calculate the total incident risk intensity of the downstream entity node. In this embodiment, the other backup steel supply path of the stamping sheet metal plant has been interrupted because the incident risk is lower than the corresponding penetration threshold, and only the main supply path retains an effective transmission state. Therefore, the total incident risk intensity of the downstream entity node is equal to the risk value of the main supply path after link attenuation, and the calculation formula is as follows: ; in, The total incident risk intensity of downstream entity nodes. This represents the total number of upstream paths that were not interrupted. For the first The incident risk intensity after link attenuation along the path is obtained by substituting the numerical values ​​and solving. If a node also has another unbroken upstream path, the risk after link attenuation is... The total incident risk intensity of the node is .

[0036] By extracting the node-side buffer score 58 corresponding to the stamping sheet metal plant, and combining it with the preset node dimension weights in the above embodiments, The total buffer capacity shared globally by the calculation node is the maximum total risk that the node can offset based on safety stock and multi-source procurement quotas. The corresponding calculation formula is: ; in, This represents the total buffer capacity of the node. This is the node-side buffer score for that node. The node weights are calculated by substituting the values. ; It is important to understand that the total buffer capacity of this node is a risk mitigation resource pool shared globally by the node, which is shared by all uninterrupted upstream transmission paths, and is not a full buffer enjoyed independently by each path.

[0037] Regarding the first The formula for calculating the path buffer quota allocated to an untruncated path is as follows: ; in, For the first The path buffer quota corresponding to each path, in this single-path scenario, has a risk ratio of 100% for the primary supply path. Therefore, the path buffer quota is: ; For quota calculation in a dual-path scenario, the total incident risk obtained from the above embodiments is 81.6, therefore the quota for the primary supply path is... Another backup path has a quota of approximately 12.8, thereby enabling the dynamic allocation of buffer resources according to the risk load.

[0038] For the current propagation path, using the path buffer quota calculated above as the upper limit, the incident risk intensity after link attenuation is attenuated at the node side. The portion exceeding the path buffer quota is considered as the remaining risk intensity after penetrating the downstream entity node. The calculation formula is as follows: ; in, The remaining risk intensity after penetrating the node is calculated as follows: That is, after the risk penetrates the stamping sheet metal plant node, the remaining risk intensity is This value will be used as the incident risk intensity for the next hop, and the process will continue to traverse downstream entity nodes such as Tier 1 seat assembly plants along the downstream supply relationship edge. It's important to understand that when the incident risk intensity is exactly equal to the risk penetration threshold, after link attenuation, the risk is completely equal to the total buffer capacity of the node, and the remaining risk intensity after attenuation on the node side is... At this point, the propagation path is cut off, and downstream traversal reasoning is no longer performed.

[0039] In one embodiment of the present invention, in step S5, after completing the full path traversal, all valid transmission paths that have not been truncated are collected, and the interruption risk value of the downstream entity node is obtained based on the remaining risk intensity of the downstream entity node in the valid transmission path; this is specifically implemented through the following method: For each downstream entity node, all transmission paths ending at that node and not interrupted are compiled to form a set of valid transmission paths for that node. Based on the remaining risk intensity of downstream entity nodes in the valid transmission paths, the corresponding interruption risk value is obtained. For example, for the OEM node of the terminal downstream entity node, a total of 2 valid transmission paths are obtained: Path 1: The risk is transmitted from the hot-rolled steel plant to the stamping sheet metal plant, from the stamping sheet metal plant to the seat assembly supplier, and then from the seat assembly supplier to the vehicle OEM. The residual risk intensity of this path to the vehicle OEM is 12.3. Path 2: The risk is transmitted from the hot-rolled steel plant to the chassis structural component plant, then from the chassis structural component plant to the chassis assembly supplier, and finally from the chassis assembly supplier to the vehicle OEM. The residual risk intensity of this path to the vehicle OEM is 8.7. Based on the aforementioned transmission path, the comprehensive interruption risk value of downstream entity nodes is calculated. The procurement weight of the materials corresponding to each path is introduced as an aggregation coefficient to weighted sum the remaining risk intensities of all effective paths, yielding the comprehensive interruption risk value of the downstream entity node. Specifically, the procurement weight is set according to the proportion of the corresponding material's annual procurement amount at that downstream entity node. In this embodiment, the procurement weight for the seat assembly is 0.4, and the procurement weight for the chassis assembly is 0.6, with a sum of weights of 1. Therefore, the interruption risk value is calculated using the following formula: ; in, This represents the overall interruption risk value for downstream entity nodes. This represents the total number of valid propagation paths for this node. For the first The procurement weights corresponding to each valid path For the first The remaining risk intensity of each effective path propagating to this node is obtained by substituting it into the numerical calculation. That is, the overall interruption risk value for vehicle OEMs is 10.14.

[0040] For nodes with only a single effective transmission path, such as Tier 2 stamping and sheet metal plants that only receive risk input from the main steel supply path, the interruption risk value of the node is directly equal to the remaining risk intensity of 16.8 after the path is penetrated, without the need for additional weighted aggregation calculation.

[0041] In one embodiment of the present invention, a multi-level threshold corresponds to a graded response to adapt to the graded handling of supply chain risk management, avoiding information overload and response mismatch. The structured early warning information can be directly connected to business systems such as ERP to achieve automated linkage between early warning and emergency handling. Therefore, in step S6, the interruption risk value is compared with the preset early warning threshold, and an interruption risk early warning is given to the downstream entity node based on the comparison result; this is specifically implemented through the following method: The system presets multiple warning thresholds, matching different risk response levels and business handling permissions at each level. By comparing the comprehensive interruption risk value of downstream entities such as OEMs and Tier 2 stamping and sheet metal plants with the warning thresholds at each level, the corresponding warning level is determined. The warning level, comprehensive interruption risk value, original risk event attributes, effective transmission path traceability details, and buffer penetration status of each node are structurally integrated to generate standardized warning information. According to the warning level, the warning information is pushed to the corresponding level of business processing port, and the corresponding level of emergency response process is triggered simultaneously to achieve the warning.

[0042] To calibrate the early warning threshold, the interruption risk value of the corresponding node was calculated by backtracking through supply interruption events that have occurred, and a mapping relationship between the risk score and the actual impact level was established. The risk score corresponding to different impact levels was used as the initial boundary of the threshold, and the early warning threshold was determined after multiple experimental verifications.

[0043] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A supply chain disruption risk early warning method integrating knowledge graphs, characterized in that, Includes the following steps: S1: On the basic topology layer formed by the entity nodes and topological dependencies at all levels of the supply chain, configure node buffer parameters for entity nodes and transmission constraint parameters for relationship edges to obtain the supply chain knowledge graph; S2: Collect real-time risk event data, match and locate the risk event to the corresponding entity node in the knowledge graph, mark the entity node as the risk source node, and configure the corresponding risk intensity for the risk source node according to the event level of the risk event; S3: Starting from the risk source node, perform traversal along the relationship edge in the downstream direction. At each downstream entity node, use its corresponding node buffer parameter and the transmission constraint parameter of the corresponding upstream relationship edge to calculate the risk penetration threshold of the downstream entity node. S4: Compare the risk intensity with the risk penetration threshold to determine whether to cut off the current transmission path and terminate the downstream reasoning along the transmission path; If not truncated, the remaining risk intensity after penetration is obtained by calculating the risk attenuation coefficient, and the remaining risk intensity is used to continue traversing along the downstream relationship edge; S5: After completing the full transmission path traversal, gather all valid transmission paths that have not been truncated, and obtain the interruption risk value of the downstream entity node based on the remaining risk intensity of the downstream entity node in the valid transmission path. S6: Compare the interruption risk value with the preset warning threshold, and issue an interruption risk warning for downstream entity nodes based on the comparison result.

2. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 1, characterized in that: The node buffer parameters include safety stock days and multi-source procurement quota ratio, while the transmission constraint parameters include the substitutability level of the corresponding supply channel and the buffer margin for order lead time.

3. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 1, characterized in that, The process of matching and locating risk events to corresponding entity nodes in the knowledge graph is executed progressively through the following matching conditions: When a risk event has information identifying the operating entity, the entity node is located by matching the entity's identity. When a risk event is a regionally impactful event, the entity node is located by matching the geographic scope. When a risk event is a category-impact event, the entity node is located by matching the category and industry affiliation. After executing at least one of the above matching conditions, the matching results are filtered by the supply chain role verification conditions to obtain the matched and located entity nodes.

4. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 3, characterized in that: The role verification condition is to match the corresponding supply chain role type according to the risk event type, and filter out entity nodes whose supply chain role attributes do not match the risk event impact type.

5. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 1, characterized in that, The step of calculating the risk penetration threshold of the downstream entity node includes: After standardizing the safety stock days and multi-source procurement quota ratio of the downstream entity nodes reached during the current traversal, the nodes are integrated according to the preset weights within the downstream entity nodes to obtain the node-side buffer score. For the upstream relationship edges of the downstream entity nodes through which the risk transmission path passes, after standardizing the substitutability level of the supply channel and the buffer margin of the order lead time of the upstream relationship edges, the link-side buffer score is obtained by merging according to the preset link weight. Based on the node-side buffer score and the link-side buffer score, the risk penetration threshold of the upstream relation edge corresponding to the current downstream entity node is calculated.

6. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 1, characterized in that, The step of comparing the risk intensity with the risk penetration threshold to determine whether to cut off the current transmission path includes: Obtain the incident risk intensity from the upstream entity node to the downstream entity node through the corresponding relation edge in the current propagation path; When the incident risk intensity is less than the risk penetration threshold, it is determined that the downstream entity node can offset the risk impact of the current transmission path, the current transmission path is cut off, and the downstream inference along the transmission path is terminated. When the incident risk intensity is not less than the risk penetration threshold, it is determined that the risk of the current propagation path can penetrate the downstream entity node, the propagation path is retained, and the traversal continues downstream with the remaining risk intensity after penetration.

7. The supply chain disruption risk early warning method integrating knowledge graphs according to claim 1, characterized in that, The step of calculating the risk attenuation coefficient to obtain the remaining risk intensity after penetration includes: For the current propagation path, the incident risk intensity is attenuated on the link side using the propagation constraint parameters of the upstream relation edge, and the incident risk intensity after link attenuation is obtained. Based on the incident risk intensity after link-side attenuation, the incident risk intensity of all uninterrupted transmission paths of the current downstream entity node is aggregated, and the total incident risk intensity of the downstream entity node is calculated. The total buffer capacity of the nodes is determined based on the node buffer parameters of the downstream entity nodes. Based on the proportion of the incident risk intensity after link attenuation of each transmission path to the total incident risk intensity, the total buffer capacity of the nodes is allocated to each transmission path, generating a path buffer quota corresponding to each transmission path. For the current transmission path, the path buffer quota is used as the upper limit to attenuate the incident risk intensity after the link is attenuated at the node side. The part exceeding the path buffer quota is regarded as the remaining risk intensity after penetrating the downstream entity node.

Images