Method and system for early warning of automobile supply chain disruption risk based on multi-source data fusion

Abstract

This invention belongs to the field of risk warning technology, specifically relating to a method and system for early warning of automotive supply chain disruption risks based on multi-source data fusion. The method includes the following steps: acquiring and fusing enterprise resource planning system data, real-time logistics data, and external event data from each node in the automotive supply chain network to construct a corresponding cell for each node in the supply chain network; calculating the risk-bearing coefficient of each cell based on its own buffer inventory level and substitutability; defining the state of each cell as a multi-dimensional state vector containing operating status, material fulfillment rate, and risk accumulation value; and calculating the correlation strength between cells based on historical transaction frequency and real-time logistics data. This invention provides managers with intuitive and actionable early warning information by generating risk heatmaps, high-risk node lists, and risk propagation subchains, thereby enhancing the practical value of the early warning results.

Description

Technical Field

[0001] This invention belongs to the field of risk warning technology, specifically relating to a method and system for early warning of automotive supply chain disruption risks based on multi-source data fusion. Background Technology

[0002] Supply chain risk management methods include statistical analysis based on historical data, qualitative or semi-quantitative methods based on expert scoring, and quantitative analysis methods based on simulation modeling. The former is highly subjective and difficult to detect the evolution of risks; the latter, although it can simulate the behavior of the system, is difficult to build models and has high computational costs, making it difficult to meet the real-time risk warning requirements of modern supply chains.

[0003] Cellular automata, by defining simple local evolution rules, can emerge global behavior from the bottom up, making them suitable for simulating diffusion and evolution processes in distributed systems. However, these models typically employ von Neumann or Moore's neighborhoods, failing to accurately reflect the connections formed in the supply chain based on business and information flows. They often treat all nodes in the network as homogeneous cells, ignoring differences among enterprises in inventory levels, supplier substitutability, and risk resilience. The synchronous update mechanisms used are inconsistent with the reality that risks propagate asynchronously along different paths at different speeds. Furthermore, state transition rules based on probability or thresholds lack the ability to perceive and adjust to the global risk situation of the network, leading to room for improvement in the reliability of early warning results. Summary of the Invention

[0004] This invention provides a method and system for early warning of automotive supply chain disruption risks based on multi-source data fusion, in order to solve the technical problem that existing models cannot truly reflect the heterogeneous topological relationships formed by business flow and information flow in the supply chain, and lack the ability to perceive and adjust the overall risk situation of the network.

[0005] In a first aspect, the present invention provides a method for early warning of automotive supply chain disruption risks based on multi-source data fusion, comprising the following steps: Acquire and integrate enterprise resource planning system data, real-time logistics data, and external event data from each node of the automotive supply chain network to construct a corresponding cell for each node in the supply chain network; calculate the risk bearing coefficient of each cell based on its own buffer inventory level and substitutability. The state of each cell is defined as a multi-dimensional state vector that includes operating status, material fulfillment rate and risk accumulation value; the correlation strength between cells is calculated based on historical transaction frequency and real-time logistics data, and the set of cells with a correlation strength greater than a preset threshold with the central cell is defined as the propagation neighborhood of the central cell; The state transition probability of a negative change in the state vector of the central cell is calculated using Bayesian network inference; the spatial entropy of the cumulative risk value of all cells in the network is calculated as the global risk dispersion, and the state transition probability threshold is determined by combining the risk bearing coefficient of the central cell; when the state transition probability is greater than the state transition probability threshold, the state vector of the central cell is updated. An event-driven asynchronous update mechanism is adopted to calculate the base time for risk propagation based on the logistics distance and information delay between cells, and to correct the base time based on the risk-bearing coefficient of each intermediate cell in the propagation path to obtain the propagation time. After the cells evolve a predetermined number of steps, a supply chain interruption risk heat map is generated based on the state of all cells in the network, and a list of nodes whose risk accumulation value exceeds the warning threshold and a risk propagation sub-chain composed of high-risk nodes are output.

[0006] Furthermore, based on the buffer inventory level and substitutability of each cell, the risk-bearing coefficient of the cell is calculated, including: The inventory factor is obtained by normalizing the buffer inventory level according to the safety stock standard; the substitutability of suppliers or materials is divided into three levels: fully substitutable, partially substitutable, and non-substitutable, represented by 1.0, 0.5, and 0.1 respectively, to obtain the substitutability factor; the risk tolerance coefficient C is calculated according to the following formula:

[0007] in, For inventory factor, As a substitute factor.

[0008] Furthermore, the state of each cell is defined as a multi-dimensional state vector containing operating state, material fulfillment rate, and risk accumulation value, including: The operating status is discretized into three levels: normal, warning, and interruption; the material fulfillment rate is defined as the ratio of the actual received material quantity to the order demand quantity, with a value range of [0,1]; the risk accumulation value is initialized by the frequency and impact of historical interruption events, with a value range of [0,1].

[0009] Furthermore, the correlation strength between cells is calculated based on historical transaction frequency and real-time logistics data, including: The transaction frequency index is obtained by calculating the average monthly number of transactions between two cells over a specified period and normalizing the maximum value. ; Obtain the average logistics transportation days between two cells, and perform normalization and reverse processing to obtain the logistics timeliness index T; Calculate the correlation strength S between cells according to the following formula: S=0.7×F+0.3×T.

[0010] Furthermore, the state transition probability threshold is determined by combining the risk-bearing coefficient of the central cell, including: Calculate the state transition probability threshold using the following formula. :

[0011] in, This represents the global risk dispersion, with a value range of [0,1]. The risk-bearing coefficient of the central cell has a value range of [0,1]. and The preset weights for positive constants, The basic transition probability threshold is determined by the formula. The stronger the central cell's own carrying capacity, the higher the state transition probability threshold, while the higher the risk diffusion of the entire network, the lower the state transition probability threshold.

[0012] Furthermore, the base time is corrected based on the risk-bearing coefficient of each intermediate cell along the propagation path to obtain the propagation time, including: For each intermediate cell in the propagation path The delay time of intermediate cell generation Risk bearing coefficient of intermediate cells Decide:

[0013] in, The preset standard delay time; Transmission time Equal to base propagation time The sum of the delay times generated by all intermediate cells along the path: ,in This represents the set of intermediate cells along the propagation path.

[0014] Furthermore, after a predetermined number of cell evolution steps, a supply chain disruption risk heatmap is generated based on the state of all cells in the network, including: Let the cumulative risk value of a cell be... , and Thresholds for classifying risk levels, among which ;when When the cell is displayed in the topology graph, it is shown in the first color; when When, it displays the second color; when When the risk distribution is displayed, it will be shown in a third color to visualize the risk distribution.

[0015] Furthermore, the risk propagation subchain formed by high-risk nodes includes: The cumulative risk value R is greater than the risk level classification threshold. Nodes are defined as high-risk nodes. For all high-risk nodes that have not yet been assigned to any risk propagation subchain, iterative processing is performed: select the node with the largest cumulative risk value as the starting point, and use breadth-first search or depth-first search algorithms to find all nodes connected to the starting point that are also high-risk. The connected components formed by these nodes are identified as a risk propagation subchain, until all high-risk nodes are assigned to at least one risk propagation subchain.

[0016] Furthermore, in S1, external event data is obtained through web crawlers or subscription services.

[0017] Secondly, the present invention provides an automotive supply chain disruption risk warning system based on multi-source data fusion, including a memory and a processor. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned automotive supply chain disruption risk warning method based on multi-source data fusion is implemented.

[0018] The beneficial effects are as follows: This invention integrates multi-dimensional data from enterprise internal resource planning, real-time logistics, and external events to construct a multi-dimensional state vector for each node in the supply chain, including operational status, material fulfillment rate, and risk accumulation value. It also combines node buffer inventory and substitutability attributes to represent risk-bearing capacity, thus achieving a representation of node states. A propagation neighborhood reflecting real business relationships is defined based on historical transaction and logistics information, making the simulation of risk propagation paths more realistic. The judgment conditions for node state transitions comprehensively consider the local mutation probability based on Bayesian network inference, the global risk dispersion represented by the network-wide risk space entropy, and the node's own risk-bearing coefficient, improving the rationality of the state evolution rules. Furthermore, by utilizing asynchronous propagation time calculation based on logistics distance and node characteristics, it reflects the asynchronous and differentiated propagation process of risk in the network. By generating risk heatmaps, high-risk node lists, and risk propagation subchains, it provides managers with intuitive and actionable early warning information, enhancing the practical value of the early warning results. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for early warning of automotive supply chain disruption risks based on multi-source data fusion. Detailed Implementation

[0020] 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.

[0021] An embodiment of the automotive supply chain disruption risk early warning method based on multi-source data fusion provided by this invention: like Figure 1 As shown, the automotive supply chain disruption risk early warning method based on multi-source data fusion includes the following steps: S1 acquires and integrates enterprise resource planning system data, real-time logistics data, and external event data from each node in the automotive supply chain network to construct a corresponding cell for each node in the supply chain network; and calculates the risk-bearing coefficient of each cell based on its own buffer inventory level and substitutability.

[0022] By integrating real-time APIs with the enterprise resource planning (ERP) systems of Tier 1 suppliers, OEMs, and distribution centers, data on safety stock levels, in-transit materials, and production plans are obtained. Simultaneously, GPS data from logistics service providers and RFID data from warehouse management systems are integrated to acquire real-time material location and transportation status. Furthermore, external event data that may impact the supply chain, such as severe weather warnings for specific regions, port congestion reports, or changes in relevant trade policies, are obtained through web crawling or subscription services. After cleaning and standardizing the heterogeneous data, a unified data view for each node is formed, and each enterprise node in the supply chain network is abstracted as a cell. Based on this data view, the buffer inventory level and supplier substitutability of each cell are extracted. For example, substitutability can be represented by the number of qualified alternative suppliers. A risk tolerance coefficient is calculated using a weighted summation method, for example, multiplying the standardized buffer inventory level by a weighting coefficient of 0.6 and adding the standardized substitutability score by a weighting coefficient of 0.4, to obtain the overall risk tolerance capacity of the cell.

[0023] To represent the inherent resilience of each node in the supply chain network, in an optional embodiment, the risk-bearing coefficient of each cell is calculated based on its own buffer inventory level and substitutability, including: The inventory factor is obtained by normalizing the buffer inventory level according to the safety stock standard; the substitutability of suppliers or materials is divided into three levels: fully substitutable, partially substitutable, and non-substitutable, represented by 1.0, 0.5, and 0.1 respectively, to obtain the substitutability factor; the risk tolerance coefficient C is calculated according to the following formula:

[0024] in, For inventory factor, As a substitute factor.

[0025] Specifically, for example, if the safety stock standard for a cell is 1000 units, and the current actual buffer stock of a cell is 1200 units, then the cell stock factor... The inventory factor is 1.2; if the actual inventory of a cell is only 600 units, the inventory factor is 0.6. This value reflects the cell's ability to buffer against short-term supply disruptions.

[0026] For example, if a core component has only one supplier and no alternatives, then the component's substitutability level is non-substitutable, and the substitutability factor is... The value is 0.1; however, if a standard screw has multiple alternative suppliers, then the screw is fully substitutable, and the substitutability factor is 1.0. Substituting these two factors into the formula, we can calculate the substitutability factor. Assuming a cell has an inventory factor of 1.2, and the materials supplied by the cell are partially substitutable with a substitutability factor of 0.5, then the cell's risk tolerance coefficient C is 0.92, indicating that the cell has a strong risk tolerance capacity.

[0027] S2 defines the state of each cell as a multi-dimensional state vector containing operating state, material fulfillment rate and risk accumulation value; calculates the correlation strength between cells based on historical transaction frequency and real-time logistics data, and defines the set of cells with a correlation strength greater than a preset threshold with the central cell as the propagation neighborhood of the central cell.

[0028] The state vector of a cell is designed as a three-dimensional vector, where the operating state is a discrete value, for example, 0 represents normal, 1 represents interruption, and -1 represents warning; the material fulfillment rate is a continuous value between 0 and 1, representing the ratio of the quantity of materials recently received to the quantity of orders; and the risk accumulation value is a non-negative real number used to represent the cumulative effect of risk over time. To calculate the correlation strength between cells, the number of transactions between any two nodes in the past year is counted as the historical transaction frequency, and the average cargo transportation time between them is extracted from real-time logistics data. The correlation strength value between the two is obtained by weighted summing the standardized transaction frequency and the reciprocal of the transportation time. For example, for the central cell A, the correlation strength between A and all other cells B, C, D, ... in the supply chain network is calculated one by one. The correlation strength threshold is set to 0.7. If the correlation strength between A and B is 0.8 and the correlation strength between A and C is 0.5, then B is included in the propagation neighborhood of A, while C is not included.

[0029] In an optional embodiment, the state of each cell is defined as a multi-dimensional state vector containing operating state, material fulfillment rate, and risk accumulation value, including: The operating status is discretized into three levels: normal, warning, and interruption; the material fulfillment rate is defined as the ratio of the actual received material quantity to the order demand quantity, with a value range of [0,1]; the risk accumulation value is initialized by the frequency and impact of historical interruption events, with a value range of [0,1].

[0030] Specifically, to represent the real-time status of a cell, a three-dimensional vector is used for definition. The first dimension is the operating state. For example, if a production line is running smoothly under all conditions, the operating state of the production line is assigned a value of 1, representing normal operation. If the production line experiences a decrease in efficiency due to a delay in non-critical materials but does not stop production, the state changes to warning and is assigned a value of 2. When the supply of core materials is interrupted, causing the production line to stop completely, the state becomes interrupted and is assigned a value of 3.

[0031] The second dimension is the material fulfillment rate, which represents the degree to which upstream supply is met. For example, if an assembly unit orders 500 sets of components but only receives 450 sets due to logistical issues, then the material fulfillment rate is 0.9.

[0032] The third dimension is the risk accumulation value, which represents the historical risk characteristics of the cell. For example, based on historical data, if a supplier has experienced several minor delivery delays and one serious supply disruption in the past year, the initial risk accumulation value might be set to 0.6 through weighted calculation. Therefore, a cell in a warning state with a material fulfillment rate of 0.9 and high historical risk can be represented by a state vector of 2, 0.9, and 0.6.

[0033] To represent the degree of connectivity between upstream and downstream nodes in a supply chain network, in one optional embodiment, the correlation strength between cells is calculated based on historical transaction frequency and real-time logistics data, including: The transaction frequency index is obtained by calculating the average monthly number of transactions between two cells over a specified period and normalizing the maximum value. ; Obtain the average logistics transportation days between two cells, and perform normalization and reverse processing to obtain the logistics timeliness index T; Calculate the correlation strength S between cells according to the following formula: S=0.7×F+0.3×T.

[0034] Assume that the average monthly number of transactions between node A and node B was 10 over the past year, while the maximum average monthly number of transactions between all node pairs in the entire supply chain network was 20. Then the transaction frequency index F between node A and B is 0.5, which reflects the closeness of their business relationship.

[0035] Assuming the average logistics transportation time from node A to node B is 2 days, the shortest transportation time in the entire network is 1 day, and the longest is 11 days. Normalization yields a value of 0.1. Reverse normalization yields 0.9. The shorter the transportation time, the higher the value of the logistics timeliness index T, representing a closer logistics connection. Calculating the association strength S using the formula, and substituting the above example data, S is 0.62.

[0036] S3. Calculate the state transition probability of a negative change in the state vector of the central cell using Bayesian network inference; calculate the spatial entropy of the cumulative risk value of all cells in the network as the global risk dispersion, and determine the state transition probability threshold by combining the risk carrying coefficient of the central cell. When the state transition probability is greater than the state transition probability threshold, update the state vector of the central cell.

[0037] A Bayesian network model is constructed, where the parent node includes the current operating state of each cell in the propagation neighborhood of the central cell, as well as relevant external event variables, and the child node represents the next operating state of the central cell. The conditional probability table of the Bayesian network model is trained using historical data. When the state of a neighboring cell deteriorates or an external risk event is triggered, the posterior probability (state transition probability) of the central cell's operating state changing from normal to interrupted is calculated through Bayesian inference. Simultaneously, the current cumulative risk value of all cells in the network is obtained, normalized to a sum of 1, and the spatial entropy of this value is calculated using the Shannon entropy formula to obtain the global risk dispersion. The global risk dispersion ranges from 0 to 1; a larger value indicates a wider distribution of risk in the supply chain network. The threshold is calculated as: the base threshold multiplied by the reciprocal of the global risk dispersion, then divided by the risk-bearing coefficient of the central cell. For example, if the calculated state transition probability is 0.6 and the threshold is 0.5, a state update is triggered, the central cell's operating state becomes interrupted, the material satisfaction rate decreases accordingly, and the risk accumulation value increases.

[0038] To determine whether a cell's state will deteriorate due to the risk of its neighbors, in an optional embodiment, a state transition probability threshold is determined by combining the risk-bearing coefficient of the central cell, including: Calculate the state transition probability threshold using the following formula. :

[0039] in, This represents the global risk dispersion, with a value range of [0,1]. The risk-bearing coefficient of the central cell has a value range of [0,1]. and The preset weights for positive constants, The basic transition probability threshold is determined by the formula. The stronger the central cell's own carrying capacity, the higher the state transition probability threshold, while the higher the risk diffusion of the entire network, the lower the state transition probability threshold.

[0040] Specifically, system parameters are set, such as the base transition probability threshold. The weight is 0.6. It is 0.2. The risk tolerance coefficient is 0.3. Assume the risk tolerance coefficient of the current central cell A is... The calculated value of 0.8 indicates a relatively strong ability to withstand risks. Simultaneously, the overall risk dispersion across the entire supply chain network was monitored. A value of 0.4 indicates that the supply chain network as a whole faces a certain degree of risk.

[0041] Substitute the numerical values ​​into the formula to calculate the state transition probability threshold of cell A. The calculated result is 0.64. If the risk carrying capacity coefficient of another cell B is only 0.3, while the global risk diffusion increases to 0.7, then the state transition probability threshold will become 0.45. Risk propagation easily affects the vulnerable and harsh environment of cell B, and the threshold for state deterioration of cell B is lower.

[0042] S4 employs an event-driven asynchronous update mechanism to calculate the base time for risk propagation based on the logistics distance and information delay between cells, and corrects the base time based on the risk-bearing coefficient of each intermediate cell in the propagation path to obtain the propagation time. After a predetermined number of cell evolution steps, a supply chain interruption risk heat map is generated based on the state of all cells in the network, and a list of nodes whose risk accumulation value exceeds the warning threshold and a risk propagation sub-chain composed of high-risk nodes are output.

[0043] Instead of updating all cells at fixed time steps, a time-ordered event queue is maintained. When a cell's state undergoes a negative change, it generates a risk propagation event for every neighboring cell in its propagation neighborhood. The trigger time of this event, i.e., the propagation time, is calculated based on a corrected propagation delay. The logistics time is calculated based on the physical transport route length and average transport speed between two cells, and then added to an average information delay representing order processing and information transmission to obtain the base time for risk propagation. For example, the base time is 48 hours. All intermediate cells along the propagation path from the risk source to the target cell are examined, and the reciprocal of the risk-bearing coefficient of each intermediate cell is accumulated to obtain a correction. The base time is then added to this correction to obtain the propagation time. Intermediate nodes with higher risk-bearing coefficients have smaller reciprocals and less added delay, and vice versa, simulating the phenomenon that nodes can slow down the speed of risk propagation.

[0044] A cellular automaton model is set to evolve forward 100 time steps to simulate the risk propagation process over a future period. After the evolution, each cell is colored on a two-dimensional topology map of the supply chain based on its cumulative risk value; the higher the risk value, the darker the color, for example, from green and yellow to red, forming an intuitive risk distribution heatmap. A warning threshold is set, for example, 50. All cells are traversed, and the names, locations, and hierarchical information of cells with a cumulative risk value greater than 50 are extracted to form a list of high-risk nodes. Starting from the node with the highest risk in the list, the process traces back to which neighboring cells influenced the node's deterioration during the evolution process, continuing upstream until the initial risk source or the node from which the risk was externally input is found. This outlines one or more risk propagation paths, i.e., risk propagation sub-chains.

[0045] To calculate the specific time required for an interruption event to propagate from upstream to downstream, in an optional embodiment, the base time is corrected based on the risk-bearing coefficient of each intermediate cell along the propagation path to obtain the propagation time, including: For each intermediate cell in the propagation path The delay time of intermediate cell generation Risk bearing coefficient of intermediate cells Decide:

[0046] in, The preset standard delay time; Transmission time Equal to base propagation time The sum of the delay times generated by all intermediate cells along the path: ,in This represents the set of intermediate cells along the propagation path.

[0047] Set basic parameters, such as the base propagation time of an interrupt event under ideal conditions. The 5-day timeframe can be understood as pure logistics time. This is the set standard delay time. Two days represents the maximum additional delay that a cell with absolutely no risk-bearing capacity would cause.

[0048] Consider a propagation path where risk travels from node A through intermediate nodes B and C to node D. Assume node B has a risk carrying capacity coefficient of [missing information]. A value of 0.8 indicates that the node has a relatively strong capability; the risk carrying capacity coefficient of node C is... A value of 0.3 indicates that the node's capability is relatively weak. According to the formula, the delay caused by node B is 0.4 days. The delay caused by node C is 1.4 days. The propagation time of this interruption event from A to D is... It takes 6.8 days.

[0049] To transform complex risk data into intuitive visual images, in one optional embodiment, after a predetermined number of cell evolution steps, a supply chain disruption risk heatmap is generated based on the state of all network cells, including: Let the cumulative risk value of a cell be... , and Thresholds for classifying risk levels, among which ;when When the cell is displayed in the topology graph, it is shown in the first color; when When, it displays the second color; when When the risk distribution is displayed, it will be shown in a third color to visualize the risk distribution.

[0050] For example, setting a low-risk threshold The value is 0.3, which is the medium-risk threshold. The risk level is set to 0.7. The first color is green to represent low risk, the second color is yellow to represent medium risk, and the third color is red to represent high risk.

[0051] After the cellular automaton model finishes running, each cell will have a cumulative risk value R. The process iterates through every cell node in the supply chain topology. If a supplier cell A has a cumulative risk value of 0.2, it will be marked green on the heatmap because 0.2 is less than or equal to 0.3. If another core manufacturer cell B has a cumulative risk value of 0.65, it will be marked yellow because it is greater than 0.3 and less than or equal to 0.7. If a downstream distributor cell C has a cumulative risk value as high as 0.88, it will be marked red. Coloring all cells according to this rule creates a complete supply chain risk heatmap, identifying areas of concentrated risk and key nodes.

[0052] To extract key risk propagation paths from the global risk heatmap, in one optional embodiment, the risk propagation sub-chain composed of high-risk nodes includes: The cumulative risk value R is greater than the risk level classification threshold. Nodes are defined as high-risk nodes. For all high-risk nodes that have not yet been assigned to any risk propagation subchain, iterative processing is performed: select the node with the largest cumulative risk value as the starting point, and use breadth-first search or depth-first search algorithms to find all nodes connected to the starting point that are also high-risk. The connected components formed by these nodes are identified as a risk propagation subchain, until all high-risk nodes are assigned to at least one risk propagation subchain.

[0053] The high-risk threshold set according to the previous embodiment =0.7, filter out all nodes with a cumulative risk value R greater than 0.7. Assume that the set of high-risk nodes includes A, B, C, D, and E, with node risk values ​​of 0.9, 0.85, 0.95, 0.8, and 0.78, respectively.

[0054] The iterative process begins by selecting node C, with the highest risk value, from the unassigned nodes as the starting point for the first risk propagation sub-chain. Starting from node C, a breadth-first search algorithm is used to find its neighbors in the supply chain network. Assuming C is connected to B, and B's node risk value of 0.85 is also greater than 0.7, B is added to the current risk propagation sub-chain. Continuing the search from B, it is found that B is connected to A, and A's node risk value of 0.9 also meets the condition, so A is added to the sub-chain. If none of A and B's other neighbors are high-risk nodes, the search for the first risk propagation sub-chain ends; this sub-chain contains C, B, and A. Among the remaining unassigned high-risk nodes D and E, D, with the higher risk value, is selected as the new starting point, and the search process is repeated to identify all interconnected risk propagation sub-chains composed of high-risk nodes.

[0055] An embodiment of the automotive supply chain disruption risk early warning system based on multi-source data fusion provided by this invention: The automotive supply chain disruption risk warning system based on multi-source data fusion includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned automotive supply chain disruption risk warning method based on multi-source data fusion is implemented.

[0056] The automotive supply chain disruption risk early warning system based on multi-source data fusion also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0057] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.

[0058] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for early warning of automotive supply chain disruption risks based on multi-source data fusion, characterized in that, Includes the following steps: Acquire and integrate enterprise resource planning system data, real-time logistics data, and external event data from each node of the automotive supply chain network to construct a corresponding cell for each node in the supply chain network; calculate the risk bearing coefficient of each cell based on its own buffer inventory level and substitutability. The state of each cell is defined as a multi-dimensional state vector that includes operating status, material fulfillment rate and risk accumulation value; the correlation strength between cells is calculated based on historical transaction frequency and real-time logistics data, and the set of cells with a correlation strength greater than a preset threshold with the central cell is defined as the propagation neighborhood of the central cell; Calculate the state transition probability when the state vector of the central cell undergoes a negative change using Bayesian network inference; The spatial entropy of the cumulative risk value of all cells in the network is calculated as the global risk dispersion. The state transition probability threshold is determined by combining the risk bearing coefficient of the central cell. When the state transition probability is greater than the state transition probability threshold, the state vector of the central cell is updated. An event-driven asynchronous update mechanism is adopted to calculate the base time for risk propagation based on the logistics distance and information delay between cells, and to correct the base time based on the risk-bearing coefficient of each intermediate cell in the propagation path to obtain the propagation time. After a predetermined number of steps in cell evolution, a heatmap of supply chain disruption risk is generated based on the state of all cells in the network. A list of nodes whose cumulative risk value exceeds the warning threshold and a risk propagation subchain composed of high-risk nodes are also output.

2. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, Based on the buffer stock level and substitutability of each cell, the risk-bearing coefficient of the cell is calculated, including: The inventory factor is obtained by normalizing the buffer inventory level according to the safety stock standard; the substitutability of suppliers or materials is divided into three levels: fully substitutable, partially substitutable, and non-substitutable, represented by 1.0, 0.5, and 0.1 respectively, to obtain the substitutability factor; the risk tolerance coefficient C is calculated according to the following formula: in, For inventory factor, As a substitute factor.

3. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, The state of each cell is defined as a multi-dimensional state vector containing operating state, material fulfillment rate, and risk accumulation value, including: The operating status is discretized into three levels: normal, warning, and interruption; the material fulfillment rate is defined as the ratio of the actual received material quantity to the order demand quantity, with a value range of [0,1]; the risk accumulation value is initialized by the frequency and impact of historical interruption events, with a value range of [0,1].

4. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, The strength of the association between cells is calculated based on historical transaction frequency and real-time logistics data, including: The transaction frequency index is obtained by calculating the average monthly number of transactions between two cells over a specified period and normalizing the maximum value. ; Obtain the average logistics transportation days between two cells, and perform normalization and reverse processing to obtain the logistics timeliness index T; Calculate the correlation strength S between cells according to the following formula: S=0.7×F+0.3×T.

5. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, The state transition probability threshold is determined by combining the risk-bearing coefficient of the central cell, including: Calculate the state transition probability threshold using the following formula. : in, This represents the global risk dispersion, with a value range of [0,1]. The risk-bearing coefficient of the central cell has a value range of [0,1]. and The preset weights for positive constants, The basic transition probability threshold is determined by the formula. The stronger the central cell's own carrying capacity, the higher the state transition probability threshold, while the higher the risk diffusion of the entire network, the lower the state transition probability threshold.

6. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, The base time is corrected based on the risk-bearing coefficient of each intermediate cell along the propagation path to obtain the propagation time, including: For each intermediate cell in the propagation path The delay time of intermediate cell generation Risk bearing coefficient of intermediate cells Decide: in, The preset standard delay time; Transmission time Equal to base propagation time The sum of the delay times generated by all intermediate cells along the path: ,in This represents the set of intermediate cells along the propagation path.

7. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, After a predetermined number of cell evolution steps, a supply chain disruption risk heatmap is generated based on the state of all cells in the network, including: Let the cumulative risk value of a cell be... , and Thresholds for classifying risk levels, among which ;when When the cell is displayed in the topology graph, it is shown in the first color; when When, it displays the second color; when When the risk distribution is displayed, it will be shown in a third color to visualize the risk distribution.

8. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 7, characterized in that, The risk propagation subchain composed of high-risk nodes includes: The cumulative risk value R is greater than the risk level classification threshold. Nodes are defined as high-risk nodes. For all high-risk nodes that have not yet been assigned to any risk propagation subchain, iterative processing is performed: select the node with the largest cumulative risk value as the starting point, and use breadth-first search or depth-first search algorithms to find all nodes connected to the starting point that are also high-risk. The connected components formed by these nodes are identified as a risk propagation subchain, until all high-risk nodes are assigned to at least one risk propagation subchain.

9. The method for early warning of automotive supply chain disruption risks based on multi-source data fusion according to claim 1, characterized in that, In S1, external event data is obtained through web crawlers or subscription services.

10. A risk warning system for automotive supply chain disruptions based on multi-source data fusion, characterized in that: It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for early warning of automotive supply chain disruption risk based on multi-source data fusion as described in any one of claims 1-9 is implemented.

Images