Supply chain management method and corresponding product for coping with geopolitical risks

Abstract

The application relates to the field of supply chain intelligent management, and provides a supply chain management method for coping with geopolitical risks and a corresponding product. The method comprises the following steps: acquiring multi-source heterogeneous geopolitical risk related data; performing semantic analysis and vectorization processing on the geopolitical risk related data by using a cross-modal natural language processing model to generate a geopolitical risk feature vector; dynamically calculating a geopolitical risk level of a target region based on the geopolitical risk feature vector; triggering a supply chain reconstruction response mechanism of a corresponding level according to a numerical interval of the risk level by a multi-level response strategy generation engine; simulating an execution process of the supply chain reconstruction response mechanism in a virtual supply chain network in a digital twin environment, verifying material flow continuity and node fault tolerance, and optimizing a backup path topology structure according to a simulation result; and outputting the verified reconstruction strategy to a cross-border supply chain management system to realize automatic adjustment of a supplier layout, capacity allocation and a transportation route.

Description

Technical Field

[0001] This invention relates to the field of intelligent supply chain management, and in particular to a supply chain management method and related products for addressing geopolitical risks. Background Technology

[0002] As global economic integration deepens, supply chain networks have extended to multiple countries and regions, exhibiting high complexity and interdependence. Geopolitical risks (such as natural disasters, pandemics, changes in trade policies, and logistical disruptions) are external uncertainties that can easily lead to disruptions at key nodes in the supply chain, triggering chain reactions such as material shortages and production halts, posing a serious threat to the sustainable operation of enterprises. Against this backdrop, enhancing supply chain resilience (i.e., the ability of the supply chain to quickly recover to a stable state after being disrupted) has become an important goal of enterprise risk management, necessitating the development of intelligent methods capable of dynamically assessing geopolitical risks and automatically reconstructing the supply chain.

[0003] Currently, management methods for supply chain disruption risks mainly include risk warning models based on historical data analysis, emergency plan database matching mechanisms, and supplier redundancy layout design. However, these methods generally suffer from limitations such as a single dimension of risk perception, passive and lagging response mechanisms, reliance on human experience for strategy generation, and verification methods that are detached from real-world environments. Summary of the Invention

[0004] Based on this, it is necessary to address the technical problems of the existing technologies, such as the single dimension of risk perception, the passive and lagging response mechanism, the reliance on human experience for strategy generation, and the fact that the verification methods are detached from the real environment. Therefore, a supply chain management method and corresponding products for dealing with geopolitical risks are proposed.

[0005] Firstly, a supply chain management method for addressing geopolitical risks is provided, the method comprising: Acquire multi-source heterogeneous geopolitical risk-related data; The geopolitical risk-related data are semantically parsed and vectorized using a cross-modal natural language processing model to generate geopolitical risk feature vectors. Based on the aforementioned geopolitical risk feature vector, and combined with a preset six-dimensional geopolitical risk index system, the geopolitical risk level of the target area is dynamically calculated. The risk level is input into a multi-level response strategy generation engine built on a large language model, and the multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level. The execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network is simulated in a digital twin environment to verify the continuity of material flow and the fault tolerance of nodes, and the backup path topology is optimized based on the simulation results. The validated restructuring strategy is exported to the cross-border supply chain management system to automate the adjustment of supplier locations, capacity allocation, and transportation routes.

[0006] Secondly, a supply chain management device for addressing geopolitical risks is provided, the device comprising: The acquisition module is used to acquire multi-source heterogeneous geopolitical risk-related data; The generation module is used to perform semantic parsing and vectorization processing on the geopolitical risk-related data using a cross-modal natural language processing model to generate geopolitical risk feature vectors. The calculation module is used to dynamically calculate the geopolitical risk level of the target area based on the geopolitical risk feature vector and in combination with a preset six-dimensional geopolitical risk index system. The triggering module is used to input the risk level into the multi-level response strategy generation engine built based on the large language model, and the multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level. The optimization module is used to simulate the execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network in a digital twin environment, verify the continuity of material flow and the fault tolerance of nodes, and optimize the backup path topology based on the simulation results. The output module is used to output the validated restructuring strategy to the cross-border supply chain management system, enabling automated adjustments to supplier layout, capacity allocation, and transportation routes.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described supply chain management method for addressing geopolitical risks.

[0008] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described supply chain management method for addressing geopolitical risks.

[0009] As can be seen from the technical solution provided in this application, on the one hand, by acquiring multi-source heterogeneous geopolitical risk-related data and using a cross-modal natural language processing model to perform semantic parsing and vectorization processing on the data, it is possible to overcome the problems of inconsistent data formats and semantic ambiguity in traditional methods, and generate more representative geopolitical risk feature vectors, thereby providing a more comprehensive and reliable data foundation for subsequent risk assessment. Based on the above geopolitical risk feature vectors, combined with a preset six-dimensional geopolitical risk indicator system, the geopolitical risk level of the target area is dynamically calculated, so that the risk assessment not only covers multi-dimensional factors, but also responds to changes in the external environment in real time, avoiding misjudgments caused by single indicators or delayed updates. On the other hand, by inputting the risk level into a multi-level response strategy generation engine built on a large language model, and according to... The numerical range of risk levels triggers corresponding supply chain restructuring response mechanisms, automatically matching countermeasures under different risk levels, reducing delays caused by manual intervention, and improving response speed and targeting. Thirdly, by simulating the execution process of the restructuring response mechanism in a virtual supply chain network within a digital twin environment, the continuity of material flow and node fault tolerance are verified. Based on the simulation results, the backup path topology is optimized, enabling the pre-identification of strategy defects, avoiding secondary risks in actual implementation, and enhancing the overall resilience of the supply chain system. Furthermore, by outputting the verified restructuring strategy to the cross-border supply chain management system, automated adjustments to supplier layout, capacity allocation, and transportation routes are achieved, significantly reducing human error, improving supply chain restructuring efficiency, and ensuring rapid recovery of business continuity under geopolitical risks. In summary, the technical solution of this application, based on multi-source data fusion and intelligent simulation, realizes dynamic assessment of geopolitical risks and automated, verifiable restructuring of the supply chain, significantly improving the overall resilience of the supply chain system. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is an application scenario diagram of a supply chain management method for addressing geopolitical risks in one embodiment; Figure 2 A flowchart illustrating a supply chain management method for addressing geopolitical risks in one embodiment; Figure 3 This is a structural block diagram of a supply chain management device for addressing geopolitical risks in one embodiment. Figure 4 This is a structural block diagram of a computer device in one embodiment. Detailed Implementation

[0011] 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, not all, of the embodiments of the present invention. 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.

[0012] Currently, management methods for supply chain disruption risks mainly include risk warning models based on historical data analysis, emergency plan database matching mechanisms, and supplier redundancy layout design. However, these methods generally have the following limitations: 1) Single dimension of risk perception: Most systems only focus on economic indicators or logistics status (e.g., port congestion), lacking comprehensive consideration of deep-seated structural factors such as geopolitics, security incidents, and socio-cultural factors; 2) Passive and lagging response mechanisms: Existing solutions usually only initiate emergency procedures after an disruption occurs, lacking the ability to predict and proactively intervene; 3) Strategy generation relies on human experience: Although some systems have introduced rule engines for path recommendation, their logic is fixed and their generalization ability is poor, making it difficult to cope with complex and ever-changing real-world scenarios; 4) Verification methods are divorced from the real environment: The effectiveness of strategy execution mostly relies on post-event review or simple simulation, lacking the support of high-fidelity, iterative virtual simulation platforms.

[0013] To address the aforementioned problems in existing technologies, this application proposes a supply chain management method for mitigating geopolitical risks, which can be applied to... Figure 1 The example application scenarios mainly involve entities such as the geopolitical risk monitoring system 101, the supply chain digital twin platform 102, and the cross-border supply chain management system (SCM) 103. Figure 1In the example application scenario, suppose the core products of global consumer electronics brand AlphaTech rely on chips shipped from country A to country B via a key strait. When external interference occurs in the area between country A and its neighboring country C, AlphaTech's geopolitical risk monitoring system 101 automatically activates. The system first acquires real-time satellite images showing abnormal activity in the area from commercial satellite service providers. Simultaneously, it captures multilingual policy texts, including shipping warnings and announcements of specific activities issued by the official authorities of countries A and C, and combines this with international news and analysis reports to complete the acquisition of multi-source, heterogeneous geopolitical risk-related data. Next, the system's built-in cross-modal Natural Language Processing (NLP) analysis engine begins operation. It performs integrated analysis of visual information in satellite images, legal wording in policy texts, and semantics in news reports, identifying key risk signals such as "extremely high probability of temporary navigation ban," and fusing this heterogeneous information to generate a unified, quantifiable geopolitical risk feature vector. The feature vector was fed into the system's six-dimensional geopolitical risk assessment module in real time. This module dynamically calculates risks based on six preset indicators: geopolitical risk, security incidents, economy, logistics, technology, and socio-cultural factors. Due to a sharp increase in the scores for security risk intensity and logistical vulnerability, the system instantly upgraded the geopolitical risk level of the "Country A to Country B" route from Level 2 (low risk) to Level 7 (high risk). This risk level immediately triggered a multi-level response strategy generation engine built on a Large Language Model (LLM). According to preset rules (Level > 6 triggers Level 3 emergency response), the engine automatically generated a preliminary restructuring strategy, with core recommendations including: immediately activating the backup chip factory in Country D, initiating an air freight solution to replace some sea freight, and releasing strategic chip inventory. Before implementing the strategy, the draft strategy was imported into the company's supply chain digital twin platform 102 for simulation. The supply chain digital twin platform 102 constructed a network model in a virtual environment containing global factories, warehouses, and transportation routes, simulating the new supply plan. The simulation results showed that air freight alone was too costly, but combining it with strategic inventory could allow for a smooth transition. Based on this, the supply chain digital twin platform 102 optimized the solution, generating the optimal combination and specific schedule for "air freight + strategic inventory". Ultimately, this validated and optimized restructuring strategy was automatically output to the enterprise's cross-border supply chain management system 103. The cross-border supply chain management system 103 then automatically executed: placing production orders with the factory in country D, reserving air freight space with logistics providers, and adjusting inventory allocation parameters. The entire process, from risk identification to execution adjustments, is fully automated by default. Simultaneously, to adapt to complex strategic needs, the system provides a human-machine collaboration interface, allowing managers to input high-level commands to fine-tune strategy generation preferences. The system will generate multiple alternative solutions and provide suggestions, with the final decision resting with the manager.

[0014] Please see Figure 2 As shown, Figure 2 A flowchart illustrating a supply chain management method for addressing geopolitical risks provided in an embodiment of the present invention mainly includes steps S201 to S206, detailed below: Step S201: Obtain geopolitical risk-related data from multiple sources and heterogeneity.

[0015] To comprehensively capture early signals of changing geopolitical risks, the system can access various data sources, covering three main categories: space observation data, textual bulletin data, and time-series dynamic data. Specifically, geopolitical risk-related data includes real-time satellite imagery, multilingual versions of publicly available industry data, and dynamic information on international relations, detailed below: 1) Real-time satellite imagery comes from commercial remote sensing satellite service providers (e.g., Maxar Technologies, PlanetLabs), covering key sensitive areas such as specific sea areas, critical transportation channels, and regional borders. The update frequency can be as high as once a day or even more, and the image resolution is no less than 0.5 meters, which is sufficient to identify maritime traffic patterns, airport take-off and landing activities, and temporary facility deployments.

[0016] 2) Multilingual versions of publicly available industry data include official documents such as diplomatic statements, import and export control lists, tax adjustment notices, and visa policy changes issued by various countries. The data is collected from international organization databases, the World Trade Organization website, the websites of the Ministry of Commerce of various countries, and public reports from mainstream media. The system supports the capture and caching of multiple languages ​​such as Chinese, English, Russian, and Arabic.

[0017] 3) Information on international relations dynamics comes from authoritative news agencies, think tank research reports, and content published by high-influence accounts on social media. After being screened by credibility scores, it is included in the analysis pool.

[0018] It is worth noting that due to differences in release time and delays in dissemination paths across different data sources, directly using raw data may lead to misjudgments. For example, news of a country announcing trade restrictions may first appear in English-language media, and only be officially published on the local government website several hours later. Therefore, Figure 2The example method further includes: after acquiring geopolitical risk-related data, synchronizing and calibrating data from different sources using timestamps, identifying differences in the expression of the same geopolitical event across multiple media, and merging the information content from each data source based on confidence weights to eliminate cognitive biases caused by language bias or publication delays. Specifically, this can be achieved by: attaching a precise timestamp to each data record and establishing an event-evidence association graph; when multiple data sources in the event-evidence association graph point to the same potential event (e.g., "a port in a certain region is closed due to a natural disaster"), automatically aligning the time series and identifying expression differences through semantic similarity calculations; assigning different weights to these data with expression differences based on the authority of the data source (e.g., documents from relevant international organizations have a weight of 0.9, mainstream media 0.7, and social media platforms 0.4); and using Dempster-Shafer evidence theory for fusion decision-making to output a unified factual statement, avoiding misleading information from a single source. The Dempster-Shafer evidence theory-based fusion decision-making process, resulting in a unified factual statement, involves: assigning initial weights based on data source type (e.g., 0.9 for official documents, 0.7 for mainstream media, and 0.4 for social media platforms); dynamically adjusting weights based on data source authority using semantic consistency detection; increasing weights if multiple sources describe the same event with highly overlapping semantics; and then using Dempster-Shafer evidence theory to fuse the weighted multi-source information to generate a unified factual statement. This embodiment, by acquiring multi-source heterogeneous geopolitical risk-related data, including real-time satellite imagery, multilingual government policy texts, and dynamic information on international relations, not only broadens the scope of risk monitoring but also enhances the authenticity and timeliness of information through data fusion mechanisms, laying a solid foundation for subsequent accurate assessments.

[0019] Step S202: Use a cross-modal natural language processing model to perform semantic parsing and vectorization on geopolitical risk-related data to generate geopolitical risk feature vectors.

[0020] After data acquisition and fusion, the next step is to uniformly represent the heterogeneous data. Due to the diverse forms of data—including low-structure natural language text and unstructured remote sensing images—traditional single-modal processing methods cannot effectively extract deep semantics. Therefore, this embodiment introduces a cross-modal natural language processing model to achieve joint text-image encoding. This model is based on an extension of the Transformer architecture and includes two branches: a text encoder and an image encoder. 1) The text encoder uses a Transformer-based Bidirectional Encoder Representations from Transformers (BERT-large) structure, responsible for converting policy texts, news statements, etc., into 768-dimensional semantic vectors; 2) The image encoder uses Vision Transformer (ViT), dividing satellite images into patches and encoding each patch into a spatial feature sequence. Subsequently, the two types of features are fused through a cross-attention mechanism, enabling the model to understand whether "the event described in a certain text is correspondingly reflected in the image." For example, if a news report states "intensified activity in a certain airspace," while concurrent satellite imagery shows significant fluctuations in the density of mobile units within critical infrastructure in the region, the two corroborate each other, enhancing the confidence level of the judgment. The final output geopolitical risk feature vector is a high-dimensional dense vector (e.g., 1024-dimensional), encapsulating the overall risk situation of the current region, which can be used for subsequent classification, clustering, or regression tasks.

[0021] Step S202 of the above embodiment realizes cross-modal semantic alignment, enabling the originally isolated text and image information to work together, significantly improving the accuracy and robustness of risk identification.

[0022] Step S203: Based on the geopolitical risk feature vector and combined with the preset six-dimensional geopolitical risk index system, dynamically calculate the geopolitical risk level of the target area.

[0023] As mentioned earlier, existing supply chain risk assessment indicators are singular and static, failing to comprehensively reflect the multifaceted and dynamic impact of geopolitical risks on the supply chain. Therefore, to achieve a multi-dimensional, quantitative, and automated comprehensive evaluation of geopolitical risks, this application adopts a technical solution based on geopolitical risk feature vectors, combined with a pre-defined six-dimensional geopolitical risk indicator system, to dynamically calculate the geopolitical risk level of a target region. Specifically, the geopolitical risk feature vectors are input into a multilayer perceptron (MLP)-based scoring model. This model learns the nonlinear mapping relationship between the feature vectors and the six-dimensional indicator scores through training, outputting initial scores for each dimension. Real-time dynamic data (e.g., the latest sanctions list, satellite activity trajectories) is used to dynamically correct the initial scores. The corrected six-dimensional indicator scores are normalized and weighted according to preset weights to generate the final risk level. The six-dimensional indicators include geopolitical stability, security risk intensity, economic dependence, logistical vulnerability, technological substitutability, and socio-cultural resilience. The design of the six-dimensional geopolitical risk indicator system fully considers the main causes of modern supply chain disruptions, quantifying and modeling them from six independent but interrelated dimensions, as detailed below: (1) Geopolitical stability is a dimension reflecting the potential impact of the target country's policy continuity risk, governance stability, and trade restriction risk on supply chain access. It is derived from international organization databases, the WTO / TBT notification system, government gazettes of various countries, etc. The geopolitical stability dimension is dynamically updated through the following operations: continuously monitoring the frequency of policy continuity risk, governance stability index, and changes in the trade restriction list of the target country; whenever a new sanctions announcement is detected, the system automatically calls the legal database to compare the trade compliance clauses between the country and the target country's enterprises, generates an impact score on the enterprise's supply chain access qualification, and feeds it back into the risk feature vector. In specific implementation, the system connects the Ministry of Foreign Affairs' foreign-related legal service platform with the WTO / TBT notification system to track changes in the sanctions list in real time. Once a new restrictive clause is discovered, the compliance impact analysis process is immediately initiated. For example, if a country is included in the entity list, the system will search the list of companies associated with it, assess whether there is a direct supply relationship, and generate a "probability of restricted access" between 0 and 1. Geopolitical stability can be quantified by monitoring the frequency of policy continuity risks, governance stability indices, and changes in trade restriction lists, generating a "probability of political instability" ranging from 0 to 1. For example, historical data shows that when the probability is greater than 0.7, the risk of supply chain disruption increases significantly. (2) Security risk intensity is a dimension for assessing the immediate threat level of security incidents to transportation corridors in a specific area. It can be derived from high-resolution satellite imagery (e.g., Maxar Technologies) and the AIS ship positioning system. Quantification of security risk intensity includes: extracting spatial distribution data such as the density of maritime mobile units, the activity trajectory of airspace mobile units, and facility deployment heatmaps from high-resolution satellite imagery of specific sea areas or border regions; converting the above spatial distribution data into a spatiotemporal sequence tensor, inputting it into a convolutional neural network to extract local risk activity features; performing similarity matching between the extracted features and historical conflict outbreak data to predict the probability of navigation restrictions or logistics disruptions occurring within a certain time period (e.g., 72 hours), and using this as the core input parameter for the military dimension. In the above embodiment, the convolutional neural network is trained on a historical conflict sample set from the past ten years and can identify typical "pre-war signs" patterns. In addition, to eliminate interference factors, the above spatiotemporal sequence tensor also incorporates meteorological and marine environmental data to exclude non-military aggregation phenomena caused by typhoons or severe sea conditions when analyzing the movement of mobile units in a specific sea area, thereby improving the accuracy of conflict identification.

[0024] (3) Economic dependence is a dimension that measures the proportion of the target market's contribution to the global industrial chain in terms of GDP and the import dependence rate of key raw materials. It can be derived from the World Bank database, statistics from the International Trade Centre (ITC), etc. Modeling or quantifying economic dependence includes: analyzing the proportion of the target market's contribution to the global key industrial chain in terms of GDP, and the changing trend of the import dependence rate of raw materials of enterprises in the target country; when the import dependence rate of a certain raw material is detected to exceed the threshold and the risk level of the region rises, a reverse stress test is initiated to simulate the scale of the capacity gap under the scenario of complete supply disruption, and the risk weight allocation is adjusted accordingly. For example, if 80% of the rare earth of a semiconductor company in country X comes from country Y, and the current risk level of country Y rises to level 5 or above, the system automatically estimates the number of days of production stoppage and revenue loss, and feeds back the financial impact to the high-level decision-making interface to realize the closed-loop management of the transmission of geopolitical risks from the physical level to the level of financial security.

[0025] (4) The dimension of logistical vulnerability is defined as the risk of disruption to shipping routes (e.g., shipping lines, ports) due to political or military events. Its main data sources are shipping company reports, port operation logs, and real-time weather data, etc. This dimension is used to measure the resilience of shipping routes, such as the likelihood of Houthi attacks on the Red Sea route and the stability of navigation in the Suez Canal. Logistical vulnerability can be quantified by overlaying historical delay data with current political events to calculate the probability of logistical delays.

[0026] (5) The dimension of technological substitutability is defined as assessing whether there are alternative technological paths or localized production capabilities for key components (e.g., chips). This dimension is used to assess whether there are alternative technological paths for key components. For example, if a chip is subject to export controls, the system will search open-source hardware communities (e.g., RISC-V) for compatible design solutions or whether domestic manufacturers have mass production capabilities. Technological substitutability mainly comes from open-source hardware communities (e.g., RISC-V), patent databases, and supplier technical white papers, etc. The quantification method mainly involves parsing technical documents using LLM (Large Language Model) and matching the technological compatibility scores of alternative solutions.

[0027] (6) The dimension of sociocultural resilience is defined as reflecting the local society's ability to withstand external shocks. Examples include strike frequency, cultural resilience indicators, and public trust surveys. Data sources are generally social media sentiment analysis and international organization survey reports (e.g., the World Bank Governance Index). Although this indicator is difficult to quantify, it can generally be obtained by extracting negative sentiment indices from public comments through Natural Language Processing (NLP) sentiment analysis and normalizing them to a resilience score of 0 to 1.

[0028] The scores of each dimension of the above six indicators are normalized and then weighted and summed according to preset weights to obtain the final risk level. The weights can be flexibly configured according to industry characteristics; for example, energy companies focus more on logistics and political dimensions, while high-tech companies emphasize technology and economic dimensions. Through the above embodiment, a systematic, interpretable, and scalable risk assessment framework is constructed, changing the previous practice of scoring based on experience and making risk rating more scientific and forward-looking.

[0029] Step S204: Input the risk level into the multi-level response strategy generation engine built on the large language model. The multi-level response strategy generation engine will trigger the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level.

[0030] Once the risk level is determined, the system enters the strategy generation phase. If only a simple "if-then" rule base is established, the rule base cannot handle unseen or highly complex risk scenarios, making maintenance difficult. To address the problems of manual reliance, low efficiency, and difficulty in systematically matching complex risk levels in response strategy generation, the technical solution of this application is to input the risk level into a multi-level response strategy generation engine built on a large language model. The multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism based on the numerical range of the risk level. Because the large language model can understand complex contexts and generate logically sound strategy solutions, the above solution can achieve a leap from "passive defense" to "active evolution," enabling the system not only to respond to crises but also to creatively find alternative paths, while ensuring the matching and proportionality between response measures and the severity of the risk.

[0031] Specifically, the implementation method of the supply chain restructuring response mechanism triggered by the multi-level response strategy generation engine based on the numerical range of risk level is as follows: When the risk level is ≤3, the Level 1 response mechanism is activated. Alternative production capacity with the same process standards is selected from existing partner manufacturers in nearby low-risk areas, and short-term contract manufacturing agreement negotiations are automatically initiated via smart contracts. This type of situation represents a minor disturbance and does not require large-scale structural adjustments. The system prioritizes existing partner suppliers to reduce setup costs. The smart contracts are deployed on a blockchain platform (e.g., Hyperledger Fabric), automatically locking in production timelines once the other party confirms order acceptance. After activating the Level 1 response mechanism, the order saturation and yield rate fluctuations of alternative manufacturers are monitored in real time. If their capacity utilization rate approaches its limit or quality indicators fall below the warning line for three consecutive days, the Level 2 response plan is activated in advance to prevent secondary interruption risks. When risk level 3 < risk level ≤ 6, a level 2 response mechanism is triggered, utilizing intercontinental production capacity resource pools. Based on equipment commonality index and patent sharing license status, target factories capable of rapid production line switching are matched, and new global logistics routes are planned. At this point, the risk has already shown a tendency to spread, requiring broader resource scheduling. The system accesses global manufacturing cloud platforms (e.g., Flexport Partner Network) to find factories with similar equipment models (e.g., consistent SMT placement machine brands) and the necessary intellectual property licenses, shortening the conversion preparation time.

[0032] When the risk level exceeds 6, a Level 3 emergency response is activated, deploying distributed small-batch production units and integrating additive manufacturing (3D printing) alternatives, while releasing strategic reserve inventory to maintain deliveries to key customers. This is the most severe scenario, where the conventional supply chain is almost paralyzed. The system shifts to a decentralized production model, utilizing edge manufacturing nodes to complete emergency replenishment.

[0033] It should be noted that the numerical range division of the above risk levels is derived through the following continuous technical process: based on the historical supply chain disruption event database (e.g., global supply chain risk event database), the actual disruption frequency corresponding to different risk indicator thresholds is statistically analyzed to generate a probability distribution model; the statistical results are corrected by combining an expert experience database (e.g., industry risk expert scoring table) to determine key threshold nodes; and the probability distribution and expert correction values ​​are mapped in multiple dimensions through a risk matrix model to generate scientific and reasonable range division rules. Specifically, the process of correcting statistical results by incorporating an expert experience database to determine key threshold nodes can be achieved as follows: Multiple rounds of expert consultation are organized using the Delphi Method to supplement and evaluate blind spots in historical data; a credibility coefficient (range 0-1) is calculated based on expert experience, domain matching, and historical prediction accuracy; expert suggestions for adjusting threshold nodes (e.g., "In conflicts in specific sea areas, the density threshold of maritime mobile units should be adjusted from the statistical value of 5 vessels / 100 square kilometers to 4 vessels") are structured and stored as a correction rule base; historical statistical probabilities are used as the basic probability allocation, and expert suggestions are used as an independent source of evidence; when statistical results highly conflict with expert judgments (e.g., statistics show a 20% interruption probability when the threshold is 3, but experts believe it should be increased to 30%), an adaptive weighting algorithm is activated; sensitivity analysis is performed on critical value regions (e.g., the risk level 3-4 boundary), and threshold intervals requiring key monitoring are marked. The implementation of generating scientific and reasonable interval division rules by mapping probability distributions and expert correction values ​​in a multidimensional way through a risk matrix model can be as follows: Design a 6×n-order risk matrix (n is the number of threshold nodes), where rows represent risk dimensions and columns represent threshold levels; set coupling rules between dimensions, for example, triggering composite risk coefficient weighting when "safety risk intensity > threshold A" and "logistics vulnerability > threshold B"; assign higher weights to recent risk events to avoid excessive influence from historical data; normalize the index values ​​of each dimension and map them to a preset evaluation set {VL (extremely low), L (low), M (medium), H (high), VH (extremely high)} through a membership function; perform matrix multiplication, i.e., comprehensive risk level = weight vector × The membership matrix, where the weight vector is determined by the dimensional importance ranking output by the expert experience base; the level boundary point is automatically adjusted according to the calculation results. For example, when the comprehensive risk value of 3 consecutive periods is >0.7, the high risk threshold is lowered from level 6 to level 5, and a scientific and reasonable risk level interval division rule is output, such as ≤3 is low risk, 3~6 is medium risk, >6 is high risk, etc.

[0034] To further enhance the feasibility of the strategy, after activating the Level 3 emergency response, the following measures are also included: identifying non-standardized parts layer by layer based on the product BOM, prioritizing the use of LLM to parse open-source design drawings and generate localized 3D printing instruction sets; distributing printing tasks to edge manufacturing centers of global collaborative nodes, and uploading quality inspection reports to the blockchain platform after each center completes the manufacturing of the parts to ensure traceability and credibility.

[0035] Step S205: Simulate the execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network in the digital twin environment, verify the continuity of material flow and the fault tolerance of nodes, and optimize the backup path topology based on the simulation results.

[0036] Regarding supply chain restructuring response mechanisms, skipping simulation and directly implementing the generated strategies, or using simplified flowcharts or spreadsheets for rough evaluation, presents several drawbacks. The former involves strategies that haven't been fully validated failing in practice due to unforeseen bottlenecks (e.g., customs delays, insufficient capacity), exacerbating supply chain disruptions. The latter's simplified evaluation cannot simulate the dynamic interactions and chain reactions in complex networks, leading to unreliable validation results. To address the risk of secondary losses due to insufficient consideration when implementing restructuring strategies directly in the actual supply chain (i.e., excessively high "trial and error costs"), this application introduces a digital twin environment for pre-implementation validation. This environment constructs a virtual network that fully maps to the real supply chain, including all nodes (factories, warehouses, ports), connecting edges (transportation routes), and their attributes (cost, timeliness, risk coefficient). Pre-implementation using this digital twin environment significantly reduces the risk of strategy implementation failure, while a continuous learning mechanism makes the system increasingly intelligent with use.

[0037] Specifically, simulating the execution process of the supply chain restructuring response mechanism in a virtual supply chain network within a digital twin environment verifies the continuity of material flow and the fault tolerance of nodes. Optimizing the backup path topology based on the simulation results can be achieved by: constructing a weighted directed graph model containing original nodes and candidate backup nodes, with edge weights determined by a composite function of transportation cost, customs clearance efficiency, and political risk; introducing random disturbance factors during the simulation to simulate the simultaneous failure of multiple nodes in extreme scenarios; and using reinforcement learning algorithms to iteratively optimize the node selection strategy, enabling the system to adaptively form a robust set of alternative paths after multiple failed attempts. For example, a simulation could present a triple fault scenario: "X port closure + Y factory power outage + Z route hijacking," to test whether the system can successfully circumvent the fault. Each simulation is run thousands of times, and the task completion rate, average delay time, and cost overrun ratio are statistically analyzed. The reinforcement learning agent uses "minimizing total cost + maximizing delivery success rate" as its objective function, continuously exploring new path combinations. The reward function is designed as follows: positive rewards are given to tasks that successfully complete end-to-end material flow; negative penalties are imposed on tasks that are blocked due to compliance issues, delayed for more than a set number of days, or whose total cost exceeds the budget threshold; after each round of training, the optimal strategy is fed back to the large language model to update its judgment logic on the feasibility of future strategies, forming a learning loop of "practice-reflection-improvement". In addition, the weighted directed graph model automatically updates the node reputation value after each simulation: a performance score is calculated based on the performance of each node in the simulation (e.g., on-time delivery rate, anti-interference ability); the score is included in the weight calculation of the next round of path planning, so that high-reliability nodes get higher scheduling priority in subsequent strategies.

[0038] Step S206: Output the validated restructuring strategy to the cross-border supply chain management system to automate the adjustment of supplier layout, capacity allocation and transportation routes.

[0039] Policies validated through digital twins are considered "trusted solutions" and can be directly implemented in an enterprise's ERP, SRM, or TMS systems. However, before deployment, it must be ensured that they comply with legal and regulatory requirements. Therefore, Figure 2 The example method could also include: dynamically loading the latest regulations and industry entry standards of the target market before outputting the reconstruction strategy; extracting technical barriers, environmental requirements, and localization ratio restrictions through a natural language understanding module; and converting these clauses into constraints and injecting them into the strategy generation process to ensure that the recommended backup solution meets local legal compliance requirements. For example, if new EU regulations require that the carbon footprint of electronic products not exceed a certain limit, the system will exclude high-emission transportation modes during route planning.

[0040] Furthermore, the dynamic loading of the latest regulations and industry access standards of the target market in the above embodiments can also include: when a new regulation is detected in the target market, a cross-language comparative analysis process is initiated to compare the differences between the old and new versions; if sensitive changes such as mandatory local storage of data or increased carbon footprint thresholds are found, the digital twin simulation is immediately rerun to assess whether the existing backup nodes still have sustainable operational capabilities, and a secondary strategy generation is triggered if the requirements are not met. This dynamic compliance verification mechanism ensures the long-term legality of the company's overseas operations. Finally, the strategy is output in the form of an API interface, driving the supply chain management system to automatically complete: 1) supplier contract switching; 2) production order reallocation; 3) logistics service provider change; 4) tariff declaration template update. The entire process requires no manual intervention, and the response time is shortened from several weeks to several hours.

[0041] To enhance human-machine collaboration, the system also supports managers issuing strategic instructions in natural language. Specifically, it receives priority instructions input by users in natural language (e.g., "Prioritize delivery to the North American market" or "Avoid entering related companies in Country X," etc.); parses these instructions into preference vectors, integrates them with a risk scoring matrix, and adjusts the ranking logic in response strategies at each level; and generates multiple candidate solutions with explanations, allowing decision-makers to choose the path that best aligns with their strategic intent. For example, inputting "Even if it costs more, ensure Q4 shipments" will automatically increase the delivery priority weight, even including some paths that exceed cost limits as alternatives.

[0042] From the above appendix Figure 2The example of supply chain management methods for addressing geopolitical risks demonstrates that, on the one hand, by acquiring multi-source, heterogeneous geopolitical risk-related data and utilizing cross-modal natural language processing models for semantic parsing and vectorization, it is possible to overcome problems such as inconsistent data formats and semantic ambiguity in traditional methods. This generates more representative geopolitical risk feature vectors, providing a more comprehensive and reliable data foundation for subsequent risk assessment. Based on these geopolitical risk feature vectors, and combined with a pre-set six-dimensional geopolitical risk indicator system, the geopolitical risk level of the target area is dynamically calculated. This ensures that the risk assessment not only covers multi-dimensional factors but also responds in real time to changes in the external environment, avoiding misjudgments caused by single indicators or delayed updates. On the other hand, by inputting the risk level into a multi-level response strategy generation engine built on a large language model, and... Based on the numerical range of risk levels, a corresponding supply chain restructuring response mechanism is triggered, which can automatically match countermeasures under different risk levels, reduce delays caused by manual intervention, and improve response speed and targeting. Thirdly, by simulating the execution process of the restructuring response mechanism in a virtual supply chain network within a digital twin environment, the continuity of material flow and the fault tolerance of nodes are verified. Based on the simulation results, the backup path topology is optimized, which can pre-identify strategy defects, avoid secondary risks in actual implementation, and enhance the overall resilience of the supply chain system. Furthermore, by outputting the verified restructuring strategy to the cross-border supply chain management system, the supplier layout, capacity allocation, and transportation routes are automatically adjusted, thereby significantly reducing human error, improving supply chain restructuring efficiency, and ensuring rapid recovery of business continuity under geopolitical risks. In summary, the technical solution of this application, based on multi-source data fusion and intelligent simulation, realizes dynamic assessment of geopolitical risks and automated, verifiable restructuring of the supply chain, significantly improving the overall resilience of the supply chain system.

[0043] Please see Figure 3 As shown, in one embodiment, a supply chain management device for addressing geopolitical risks is provided. This device may include an acquisition module 301, a generation module 302, a calculation module 303, a triggering module 304, an optimization module 305, and an output module 306, as detailed below: The acquisition module 301 is used to acquire multi-source heterogeneous geopolitical risk-related data; The generation module 302 is used to perform semantic parsing and vectorization processing on geopolitical risk-related data using a cross-modal natural language processing model to generate geopolitical risk feature vectors. The calculation module 303 is used to dynamically calculate the geopolitical risk level of the target area based on the geopolitical risk feature vector and in combination with the preset six-dimensional geopolitical risk index system. Trigger module 304 is used to input the risk level into the multi-level response strategy generation engine built on a large language model. The multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level. The optimization module 305 is used to simulate the execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network in the digital twin environment, verify the continuity of material flow and the fault tolerance of nodes, and optimize the backup path topology based on the simulation results. Output module 306 is used to output the validated restructuring strategy to the cross-border supply chain management system to achieve automated adjustment of supplier layout, capacity allocation and transportation routes.

[0044] From the above appendix Figure 3 As illustrated by the example of a supply chain management device for addressing geopolitical risks, on the one hand, by acquiring multi-source, heterogeneous geopolitical risk-related data and utilizing a cross-modal natural language processing model to perform semantic parsing and vectorization, it is possible to overcome problems such as inconsistent data formats and semantic ambiguity in traditional methods. This generates more representative geopolitical risk feature vectors, providing a more comprehensive and reliable data foundation for subsequent risk assessment. Based on these geopolitical risk feature vectors, and combined with a pre-set six-dimensional geopolitical risk indicator system, the geopolitical risk level of the target area is dynamically calculated. This ensures that the risk assessment not only covers multi-dimensional factors but also responds to changes in the external environment in real time, avoiding misjudgments caused by single indicators or delayed updates. On the other hand, by inputting the risk level into a multi-level response strategy generation engine built on a large language model, and... Based on the numerical range of risk levels, a corresponding supply chain restructuring response mechanism is triggered, which can automatically match countermeasures under different risk levels, reduce delays caused by manual intervention, and improve response speed and targeting. Thirdly, by simulating the execution process of the restructuring response mechanism in a virtual supply chain network within a digital twin environment, the continuity of material flow and the fault tolerance of nodes are verified. Based on the simulation results, the backup path topology is optimized, which can pre-identify strategy defects, avoid secondary risks in actual implementation, and enhance the overall resilience of the supply chain system. Furthermore, by outputting the verified restructuring strategy to the cross-border supply chain management system, the supplier layout, capacity allocation, and transportation routes are automatically adjusted, thereby significantly reducing human error, improving supply chain restructuring efficiency, and ensuring rapid recovery of business continuity under geopolitical risks. In summary, the technical solution of this application, based on multi-source data fusion and intelligent simulation, realizes dynamic assessment of geopolitical risks and automated, verifiable restructuring of the supply chain, significantly improving the overall resilience of the supply chain system.

[0045] Optionally, the above Figure 3 The example apparatus may also include: a fusion module, which, after acquiring multi-source heterogeneous geopolitical risk-related data, synchronously calibrates data from different sources using timestamps, identifies differences in the expression of the same geopolitical event in multiple media, and fuses the information content from each data source based on confidence weighting to eliminate cognitive biases caused by language bias or publication delays.

[0046] Optionally, the above Figure 3 The example trigger module 304 may include an activation unit, a calling unit, and a release unit, wherein: The activation unit is used to activate the first-level response mechanism when the risk level is ≤3, select backup production capacity points with the same process standards from existing cooperative manufacturers in the adjacent low-risk area, and automatically initiate short-term contract manufacturing agreement negotiation requests through smart contracts. The calling unit is used to trigger the secondary response mechanism when risk level 3 < risk level ≤ 6, call the intercontinental capacity resource pool, match the target factory that can quickly switch production lines based on the equipment generalization index and patent sharing license status, and plan a new global logistics route; The release unit is used to initiate a Level 3 emergency response when the risk level is greater than 6, enabling distributed small-batch production units and integrating additive manufacturing alternatives, while releasing strategic reserve inventory to maintain deliveries to core customers.

[0047] Optionally, the above Figure 3 The example device may also include a monitoring module and a contingency plan activation module, wherein: The monitoring module is used to monitor the order saturation and yield fluctuations of backup vendors in real time after the first-level response mechanism is activated. The contingency plan activation module is used to activate the Level II response plan in advance if it is found that the capacity utilization rate is close to the upper limit or the quality indicators are below the warning line for three consecutive days, so as to prevent the risk of secondary interruption.

[0048] Optionally, the above Figure 3 The example device may also include an identification module and an upload module, wherein: The identification module is used to identify non-standard parts layer by layer based on the product BOM after the Level 3 emergency response is activated, and to prioritize the use of LLM to parse open source design drawings and generate localized 3D printing instruction sets. The upload module is used to distribute printing tasks to edge manufacturing centers of global collaborative nodes. After each center completes the manufacturing of the parts, it uploads the quality inspection report to the blockchain platform to ensure traceability and credibility.

[0049] Optionally, the above Figure 3 The example optimization module 305 may include building blocks, simulation blocks, and iterative blocks, wherein: The building unit is used to construct a weighted directed graph model containing original nodes and candidate backup nodes. The edge weights are determined by a composite function of transportation cost, customs clearance efficiency, and political risk. The simulation unit is used to introduce random disturbance factors during the simulation process to simulate the situation where multiple nodes fail simultaneously in extreme scenarios. The iterative unit is used to iteratively optimize the node selection strategy using reinforcement learning algorithms, so that the system can adaptively form a robust set of alternative paths after multiple failed attempts.

[0050] Optionally, the reward function design of the above reinforcement learning algorithm includes: assigning positive rewards to tasks that successfully complete end-to-end material flow; imposing negative penalties on tasks that are intercepted due to compliance issues, are delayed for more than a set number of days, or have total costs exceeding the budget threshold; and after each round of training, feeding the optimal path strategy back to the large language model to update its judgment logic on the feasibility of future strategies.

[0051] In one embodiment, a computer device is provided, the computer device may be... Figure 1 The internal structure diagram of the example intelligent recovery solution generation system 102 can be shown as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When executed by the processor, the computer program implements the functions or steps of a supply chain management method for addressing geopolitical risks.

[0052] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, performs the following steps: Acquire multi-source heterogeneous geopolitical risk-related data; The geopolitical risk-related data are semantically parsed and vectorized using a cross-modal natural language processing model to generate geopolitical risk feature vectors. Based on geopolitical risk feature vectors and combined with a pre-set six-dimensional geopolitical risk index system, the geopolitical risk level of the target area is dynamically calculated. The risk level is input into a multi-level response strategy generation engine built on a large language model. The multi-level response strategy generation engine then triggers the corresponding level of supply chain restructuring response mechanism based on the numerical range of the risk level. The execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network is simulated in a digital twin environment to verify the continuity of material flow and the fault tolerance of nodes, and the backup path topology is optimized based on the simulation results. The validated restructuring strategy is exported to the cross-border supply chain management system to automate the adjustment of supplier locations, capacity allocation, and transportation routes.

[0053] The aforementioned computer program, based on multi-source data fusion and intelligent simulation, enables dynamic assessment of geopolitical risks and automated, verifiable reconfiguration of the supply chain, significantly improving the overall resilience of the supply chain system.

[0054] In one embodiment, a computer-readable storage medium is provided that stores a computer program, which, when executed by a processor, performs the following steps: Acquire multi-source heterogeneous geopolitical risk-related data; The geopolitical risk-related data are semantically parsed and vectorized using a cross-modal natural language processing model to generate geopolitical risk feature vectors. Based on geopolitical risk feature vectors and combined with a pre-set six-dimensional geopolitical risk index system, the geopolitical risk level of the target area is dynamically calculated. The risk level is input into a multi-level response strategy generation engine built on a large language model. The multi-level response strategy generation engine then triggers the corresponding level of supply chain restructuring response mechanism based on the numerical range of the risk level. The execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network is simulated in a digital twin environment to verify the continuity of material flow and the fault tolerance of nodes, and the backup path topology is optimized based on the simulation results. The validated restructuring strategy is exported to the cross-border supply chain management system to automate the adjustment of supplier locations, capacity allocation, and transportation routes.

[0055] The steps implemented by the computer program when executed by the processor are based on multi-source data fusion and intelligent simulation, which realizes dynamic assessment of geopolitical risks and automated and verifiable reconfiguration of the supply chain, significantly improving the overall resilience of the supply chain system.

[0056] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0057] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0058] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0059] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A supply chain management method for addressing geopolitical risks, characterized in that, Includes the following steps: Acquire multi-source heterogeneous geopolitical risk-related data; The geopolitical risk-related data are semantically parsed and vectorized using a cross-modal natural language processing model to generate geopolitical risk feature vectors. Based on the aforementioned geopolitical risk feature vector, and combined with a preset six-dimensional geopolitical risk index system, the geopolitical risk level of the target area is dynamically calculated. The risk level is input into a multi-level response strategy generation engine built on a large language model, and the multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level. The execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network is simulated in a digital twin environment to verify the continuity of material flow and the fault tolerance of nodes, and the backup path topology is optimized based on the simulation results. The validated restructuring strategy is exported to the cross-border supply chain management system to automate the adjustment of supplier locations, capacity allocation, and transportation routes.

2. The supply chain management method for addressing geopolitical risks as described in claim 1, characterized in that, The method further includes: After acquiring multi-source heterogeneous geopolitical risk-related data, the data from different sources are synchronized and calibrated using timestamps to identify differences in the expression of the same geopolitical event across multiple media. The information content from each data source is then fused based on confidence weighting to eliminate cognitive biases caused by language bias or publication delays.

3. The supply chain management method for addressing geopolitical risks as described in claim 1, characterized in that, The supply chain restructuring response mechanism triggered by the multi-level response strategy generation engine based on the numerical range of risk level includes: When the risk level is ≤3, the first-level response mechanism is activated, and backup production capacity points with the same process standards are selected from existing cooperative manufacturers in adjacent low-risk areas. Short-term contract manufacturing agreement negotiation requests are automatically initiated through smart contracts. When risk level 3 < risk level ≤ 6, the level 2 response mechanism is triggered, the intercontinental production capacity resource pool is called up, the target factory that can quickly switch production lines is matched based on the equipment generalization index and patent sharing license status, and a new global logistics route is planned. When the risk level is greater than 6, a Level 3 emergency response is initiated, distributed small-batch production units are activated and additive manufacturing alternatives are integrated, while strategic reserve inventory is released to maintain delivery to core customers.

4. The supply chain management method for addressing geopolitical risks as described in claim 3, characterized in that, The method further includes: After activating the first-level response mechanism, the order saturation and yield rate fluctuations of backup vendors are monitored in real time. If it is found that the capacity utilization rate is close to the upper limit or the quality indicators are below the warning line for three consecutive days, the Level II response plan will be activated in advance to prevent secondary interruption risks.

5. The supply chain management method for addressing geopolitical risks as described in claim 3, characterized in that, The method further includes: After the Level 3 emergency response is initiated, non-standard parts are identified layer by layer based on the product bill of materials (BOM). The Large Language Model (LLM) is used first to parse open-source design drawings and generate a localized 3D printing instruction set. Printing tasks are distributed to edge manufacturing centers of global collaborative nodes. After each center completes the manufacturing of the parts, it uploads the quality inspection report to the blockchain platform to ensure traceability and credibility.

6. The supply chain management method for addressing geopolitical risks as described in claim 1, characterized in that, The process of simulating the execution of the supply chain restructuring response mechanism in a virtual supply chain network within a digital twin environment, verifying the continuity of material flow and the fault tolerance of nodes, and optimizing the backup path topology based on the simulation results includes: Construct a weighted directed graph model containing original nodes and candidate backup nodes, where the edge weights are determined by a composite function of transportation cost, customs clearance efficiency, and political risk. A random perturbation factor is introduced during the simulation to simulate the situation where multiple nodes fail simultaneously under extreme conditions. The node selection strategy is iteratively optimized using a reinforcement learning algorithm, enabling the system to adaptively form a robust set of alternative paths after multiple failed attempts.

7. The supply chain management method for addressing geopolitical risks as described in claim 6, characterized in that, The reward function design of the reinforcement learning algorithm includes: Positive rewards are given for successfully completing end-to-end material flow tasks; Negative penalties will be imposed on tasks that are blocked due to compliance issues, delayed beyond the set number of days, or whose total cost exceeds the budget threshold; After each training round, the optimal path strategy is fed back to the large language model to update its judgment logic on the feasibility of future strategies.

8. A supply chain management device for addressing geopolitical risks, characterized in that, The device includes: The acquisition module is used to acquire multi-source heterogeneous geopolitical risk-related data; The generation module is used to perform semantic parsing and vectorization processing on the geopolitical risk-related data using a cross-modal natural language processing model to generate geopolitical risk feature vectors. The calculation module is used to dynamically calculate the geopolitical risk level of the target area based on the geopolitical risk feature vector and in combination with a preset six-dimensional geopolitical risk index system. The triggering module is used to input the risk level into the multi-level response strategy generation engine built based on the large language model, and the multi-level response strategy generation engine triggers the corresponding level of supply chain restructuring response mechanism according to the numerical range of the risk level. The optimization module is used to simulate the execution process of the supply chain reconfiguration response mechanism in the virtual supply chain network in a digital twin environment, verify the continuity of material flow and the fault tolerance of nodes, and optimize the backup path topology based on the simulation results. The output module is used to output the validated restructuring strategy to the cross-border supply chain management system, enabling automated adjustments to supplier layout, capacity allocation, and transportation routes.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the supply chain management method for addressing geopolitical risks as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the supply chain management method for addressing geopolitical risks as described in any one of claims 1 to 7.

Images