CN-122022383-A - Intelligent supply chain optimization management method and system based on AI

CN122022383ACN 122022383 ACN122022383 ACN 122022383ACN-122022383-A

Abstract

The invention relates to the technical field of intelligent optimization of supply chains, in particular to an intelligent supply chain optimization management method and system based on AI, comprising the following steps: establishing a full-element knowledge graph of the supply chain with time attribute and confidence attribute, accessing into a multi-source heterogeneous supply chain operation data stream in real time, analyzing, fusing and extracting a state update event, driving the knowledge graph to dynamically evolve and update an attribute value, and generating a real-time supply chain state mirror image. Based on mirror image simulation scheduling strategy circulation, stock level and performance aging are predicted, potential abnormality is identified, a cause path is traced back, an optimization scheme is generated and decomposed into task lists, the task lists are distributed to an executing entity, task completion states are tracked, and results are fed back to a data analysis fusion step. The method realizes intelligent control and closed-loop adjustment of the whole flow of the supply chain through dynamic knowledge graph with attributes and anomaly traceability optimization.

Inventors

CHEN YINHONG
Zheng Qiaoting
WANG HAINAN

Assignees

中东集团物联网科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260327

Claims (10)

1. An AI-based intelligent supply chain optimization management method is characterized by comprising the following steps: Establishing a supply chain full-element knowledge graph containing material, capacity, storage, logistics and demand information, wherein time attributes and confidence attributes are attached to nodes and edges in the supply chain full-element knowledge graph; Accessing multi-source heterogeneous operation data streams from a provider system, a production execution system, a warehouse management system and a sales terminal system in real time; Analyzing and fusing the multi-source heterogeneous operation data flow, and extracting a state update event associated with a node in the full-element knowledge graph of the supply chain; dynamic evolution is carried out by using a state update event-driven supply chain full-element knowledge graph, and attribute values of corresponding nodes and edges are updated to generate a real-time supply chain state mirror image; Based on real-time supply chain state mirror image, simulating the circulation process of materials and orders in a network under different scheduling strategies, and predicting the inventory level and performance aging of key nodes; identifying potential anomalies which violate preset constraint conditions in the prediction result, and backtracking the cause paths of the potential anomalies; Generating an optimization scheme containing parameter adjustment suggestions and execution instructions according to the traced-back cause path; decomposing the optimization scheme into executable task lists and distributing the executable task lists to corresponding supply chain execution entities; continuously tracking the completion state of the executable task list, and feeding back the tracking result as a new state update event to the data analysis and fusion step.
2. The AI-based intelligent supply chain optimization management method of claim 1, wherein the analyzing and fusing the multi-source heterogeneous operation data stream extracts a state update event associated with a node in a supply chain full-element knowledge graph, specifically comprising: Configuring a special data adapter for each type of data source, wherein the data adapter converts an original data stream into an intermediate data format with unified field definition; entity identification is carried out on the data converted into the intermediate data format, and key information in the data record is mapped to a specific node in the full-element knowledge graph of the supply chain; Checking the logic integrity and the numerical rationality of the mapped data record, and marking the abnormal value record exceeding the normal fluctuation range; Extracting action types, variable amounts and time stamps describing node state changes from the verified data records, and packaging the action types, the variable amounts and the time stamps into standardized event objects; and (3) aligning and merging a plurality of event objects which are generated by different data sources and relate to the same node according to the time stamps of the event objects, eliminating redundant information and generating a final state update event.
3. The AI-based intelligent supply chain optimization management method of claim 2, wherein the dynamic evolution is driven by using a state update event to drive a supply chain full-element knowledge graph, and the updating of attribute values of corresponding nodes and edges specifically comprises: when the state update event arrives, locating the associated target node in the supply chain full-element knowledge graph; calculating updated values of the relevant attributes of the target nodes according to action types and variable quantities carried in the state updating events; According to the change of the attribute value of the target node, the influence is propagated along the edge in the knowledge graph, the weight or attribute of the edge directly related to the target node is recalculated, and the weight or attribute of the edge represents the strength or cost of the relationship between the nodes; Writing the updated target node attribute value and the recalculated weight or attribute of the edge into the storage of the full-element knowledge graph of the supply chain; According to the change of the attribute value of the target node, the influence is propagated along the edge in the knowledge graph, and the weight or the attribute of the edge directly related to the target node is recalculated, which comprises the following steps: determining the attribute type of the change of the attribute value of the target node, and inquiring the edge attribute type which can be influenced by the attribute type from the meta-model of the knowledge graph; traversing all edges directly connected with the target node, and screening out edges with affected edge attribute types; For each affected edge, recalculating the weight or attribute value of the edge according to the defined attribute calculation formula and combining the numerical value updated by the attribute of the target node and the numerical value of the node attribute at the other end of the edge; comparing the weight or attribute value of the recalculated edge with the original historical value of the edge, and if the change exceeds the sensitivity threshold value, marking the edge as 'attribute updated'; And writing the side information marked as 'attribute updated' into the full-element knowledge graph of the supply chain together with the new attribute value.
4. The AI-based intelligent supply chain optimization management method of claim 3, wherein the real-time supply chain state mirror image simulates the material and order circulation process in the network under different scheduling strategies, predicts the inventory level and performance aging of the key nodes, and specifically comprises: Loading a set of candidate scheduling policies from a decision knowledge base, the scheduling policies defining inventory allocation rules, transportation path selection rules, and production scheduling rules; Taking a real-time supply chain state mirror image as an initial state, taking future demand prediction data as input, and respectively loading each candidate scheduling strategy; in the simulation environment, pushing time step according to the rule of the loaded scheduling strategy, virtually executing the purchasing, moving and processing of materials and the matching, splitting and delivering processes of orders; at each simulation time point, recording the change of the number of virtual inventory of key nodes in the supply chain network, including an area warehouse, a production workshop and a distribution center, and recording the current completion state and the predicted completion time of each batch of virtual orders; and when the simulation time reaches a preset future time point, terminating the simulation, and summarizing and outputting the final predicted inventory level of each key node and the predicted performance aging distribution of all orders under each scheduling strategy.
5. The AI-based intelligent supply chain optimization management method of claim 4, wherein identifying potential anomalies in the predicted result that violate a preset constraint condition and backtracking a causal path thereof, comprises: Obtaining a simulation prediction result, wherein the simulation prediction result comprises a prediction stock level and a prediction performance aging distribution; comparing the predicted stock level with a preset safety stock threshold range, identifying nodes and time points of which the stock is lower than a lowest threshold or higher than a highest threshold, and marking the nodes and time points as stock abnormality; Comparing the predicted performance aging distribution with a service level agreement promised by a customer, identifying an order batch predicted to be delivered in a delayed manner, and marking the order batch as aging abnormality; aiming at stock abnormality or aging abnormality of each mark, positioning a simulation time point and a supply chain network position of the first occurrence of the abnormality in a corresponding simulation execution record; Starting from the position and the time point of the first occurrence of the abnormality, the reverse tracing leads to a virtual operation sequence of the simulation prediction result, wherein the virtual operation sequence comprises virtual consumption of materials, virtual distribution of orders and virtual delay of transportation until tracing to a simulated initial decision point or an external input event, so as to form a complete cause path.
6. The AI-based intelligent supply chain optimization management method of claim 5, wherein generating an optimization scheme including parameter adjustment suggestions and execution instructions according to the trace-back cause path comprises: Analyzing the cause path, and identifying key decision links and key resource bottlenecks which lead to abnormal results in the path; for a key decision link, searching an alternative decision rule from a strategy rule base or generating a parameter adjustment suggestion for an original rule, wherein the parameter comprises a re-ordering point, economic production batch and transportation priority; Generating resource allocation suggestions aiming at key resource bottlenecks, wherein the resource allocation suggestions comprise allocating materials from redundant nodes to bottleneck nodes, enabling alternative suppliers or logistics service providers and adjusting production shifts; Integrating the identified key decision links, key resource bottlenecks, corresponding parameter adjustment suggestions and resource allocation suggestions into a structured optimization scheme draft; and carrying out feasibility verification on the draft of the optimization scheme, evaluating the cost and time required by executing the scheme, and filtering out draft entries which do not meet the practical constraint to form the final optimization scheme.
7. The AI-based intelligent supply chain optimization management method of claim 6, wherein the performing feasibility check on the optimization scheme draft, evaluating the cost and time required for executing the scheme, specifically comprises: extracting each parameter adjustment suggestion and resource allocation suggestion in the optimization scheme draft, and quantifying the parameter adjustment suggestions and resource allocation suggestions into an executable operation action set; inquiring a full-element knowledge graph of a supply chain, and acquiring current state, resource availability and historical execution cost data of an entity involved in executing each operation action; estimating a minimum time period required to perform each of the operational actions based on the current state and the resource availability; Estimating an expected cost of performing each of the operational actions based on the historical execution cost data and the current market environmental parameters; Summarizing the time period and the expected cost required by all the operation actions, judging whether the sum is within the preset overall budget and time window constraint, and if the sum is beyond the preset overall budget and time window constraint, adjusting or removing part of the operation actions.
8. The AI-based intelligent supply chain optimization management method of claim 7, wherein the decomposing the optimization scheme into executable task lists and distributing to corresponding supply chain execution entities comprises: Converting each suggestion in the optimization scheme into one or more specific tasks with definite start-stop time, execution content, quality standard and responsible object; Distributing unique tracking codes for each specific task, and establishing a dependency relationship network between tasks, wherein the dependency relationship network defines the execution sequence and triggering conditions of the tasks; Determining a supply chain execution entity for receiving a task according to a responsible object defined in a specific task, wherein the supply chain execution entity comprises a purchasing system of a supplier, a production management system of a factory and a transportation system of a logistics carrier; pushing specific task information with tracking codes to a service system of a corresponding supply chain execution entity through an application programming interface or a message queue; After the task distribution is completed, recording the planned starting time, the planned finishing time and the target execution entity of each specific task.
9. The AI-based intelligent supply chain optimization management method of claim 8, wherein continuously tracking the completion status of the executable task list and feeding back the tracking result as a new status update event to the data parsing and fusion step, specifically comprising: Periodically polling or receiving task state callback information from each supply chain execution entity service system, wherein the task state callback information comprises task start execution, task progress percentage, task completion or task failure, matching the received task state callback information with specific task information recorded during task distribution, and updating the real-time state of the task; When the task state is updated to be finished, checking whether a finishing result accords with a quality standard, and recording deviation of actual finishing time and planned finishing time; when the task state is updated to fail or progress is stopped, triggering an alarm notification, and transmitting relevant information to an optimization scheme generating step to start a new round of analysis; And (3) whether the task succeeds or fails, packaging the final execution result of the task, including the actual completion time, the actual consumption resources and the output result, into a new state update event, wherein the action type of the new state update event is 'task execution feedback', and inputting the new state update event into a data analysis and fusion step.
10. An AI-based intelligent supply chain optimization management system comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of an AI-based intelligent supply chain optimization management method of any of claims 1-9.

Description

Intelligent supply chain optimization management method and system based on AI Technical Field The invention relates to the technical field of intelligent optimization of supply chains, in particular to an intelligent supply chain optimization management method and system based on AI. Background Most of the existing supply chain optimization management relies on static data to build a supply chain element association model, only integrates structural operation data of a single system, does not set time attribute and confidence attribute for nodes and association edges of all elements such as materials, productivity, storage, logistics, demands and the like, only completes basic analysis and simple splicing processing of multi-source heterogeneous data such as a supplier system, a production execution system and the like, cannot extract state update events accurately associated with knowledge map nodes, only replaces element basic information in data update, cannot drive the supply chain element model to realize dynamic adjustment by means of events, and is difficult to generate real-time supply chain state images matched with actual operation. In the traditional supply chain management, when the prediction of stock level and performance aging is abnormal by deducting materials and order circulation processes only based on static data in a scheduling strategy simulation link, adjustment measures are formulated only for abnormal symptoms, and the trace back of an abnormal causative path is avoided. The establishment of the optimization scheme takes no abnormal source as a support, the matching degree of the scheme and the actual supply chain problem is insufficient, the execution progress cannot be tracked after the optimization instruction is decomposed, and the execution result cannot be reversely integrated into the data updating flow. The method aims to solve the problems that the full element model of the supply chain lacks time and confidence coefficient attributes and cannot dynamically evolve, and meanwhile, the supply chain abnormality cannot trace back the cause path, the optimization scheme lacks pertinence and the execution is free of closed loop feedback. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides an AI-based intelligent supply chain optimization management method and system. In order to achieve the purpose, the invention adopts the following technical scheme that the intelligent supply chain optimization management method based on the AI comprises the following steps: Establishing a supply chain full-element knowledge graph containing material, capacity, storage, logistics and demand information, wherein time attributes and confidence attributes are attached to nodes and edges in the supply chain full-element knowledge graph; Accessing multi-source heterogeneous operation data streams from a provider system, a production execution system, a warehouse management system and a sales terminal system in real time; Analyzing and fusing the multi-source heterogeneous operation data flow, and extracting a state update event associated with a node in the full-element knowledge graph of the supply chain; dynamic evolution is carried out by using a state update event-driven supply chain full-element knowledge graph, and attribute values of corresponding nodes and edges are updated to generate a real-time supply chain state mirror image; Based on real-time supply chain state mirror image, simulating the circulation process of materials and orders in a network under different scheduling strategies, and predicting the inventory level and performance aging of key nodes; identifying potential anomalies which violate preset constraint conditions in the prediction result, and backtracking the cause paths of the potential anomalies; Generating an optimization scheme containing parameter adjustment suggestions and execution instructions according to the traced-back cause path; decomposing the optimization scheme into executable task lists and distributing the executable task lists to corresponding supply chain execution entities; continuously tracking the completion state of the executable task list, and feeding back the tracking result as a new state update event to the data analysis and fusion step. As a further aspect of the present invention, the parsing and fusing the multi-source heterogeneous operation data stream, extracting a state update event associated with a node in a full-element knowledge graph of a supply chain, specifically includes: Configuring a special data adapter for each type of data source, wherein the data adapter converts an original data stream into an intermediate data format with unified field definition; entity identification is carried out on the data converted into the intermediate data format, and key information in the data record is mapped to a specific node in the full-element knowledge graph of the supply chain; Checking the lo