Knowledge graph-based cycle carrier full life cycle topology management method and device
By constructing a knowledge graph-based full lifecycle topology management method for cyclic vehicles, the problems of fragmented management dimensions, static topology modeling, and insufficient adaptability in cyclic vehicle management are solved. This enables refined control and intelligent scheduling of the entire vehicle process, improving the operational efficiency and reliability of warehousing and logistics.
Patent Information
- Application Number
- CN202512003495.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-29
- Publication Date
- 2026-02-06
AI Technical Summary
The existing management of cyclic vehicles suffers from fragmented management dimensions, static topology modeling, weak data and application linkage, and insufficient adaptability of general knowledge graphs. This results in a lack of full-process traceability and value assessment for vehicles, failing to meet personalized management needs.
A knowledge graph-based method for full lifecycle topology management of cyclic vehicles is constructed. This method involves collecting multi-source data to build a dedicated knowledge graph, constructing dynamic topology subgraphs according to lifecycle stages, and performing closed-loop optimization based on application feedback to achieve dynamic topology relationships and intelligent scheduling.
It enables refined management and control of the entire vehicle process, improves scheduling efficiency, reduces maintenance costs, increases vehicle utilization efficiency, adapts to different types of vehicles and scenarios, dynamically updates to avoid failure and congestion risks, and improves the continuity and operational reliability of warehousing and logistics operations.
Smart Images

Figure CN121480665A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of warehouse logistics circulating carrier management, in particular to a circulating carrier whole life cycle topology management method and device based on a knowledge graph. BACKGROUND
[0002] With the intelligent upgrading of the warehouse logistics industry, AGV, intelligent pallet and other circulating carriers have become the core equipment for improving operation efficiency, and their management level directly affects the overall operation efficiency of the logistics system. However, the current circulating carrier management technology still has many outstanding pain points.
[0003] Firstly, the management dimension is fragmented. The existing technology focuses on a single link of carrier operation scheduling, and fails to cover the whole life cycle of procurement, maintenance, and scrap / reuse, resulting in the lack of carrier whole process traceability and value assessment, and the inability to form a complete management closed loop. Secondly, the topology modeling is static. Although the existing dynamic space-time modeling technology considers the time and space dimensions, it fails to construct differentiated topology relationships for different life cycle stages of the carrier, making it difficult to adapt to the dynamic management and control needs of multiple scenarios, and the topology relationship cannot be adjusted in real time with the changes of the carrier state and the environment. Thirdly, the data-application linkage is weak. The multi-source data fusion technology focuses on data alignment and representation, and fails to establish a full-link linkage mechanism of data-knowledge-topology-application, and lacks the ability of continuous iteration and optimization based on business feedback, making it difficult to fully release the value of data. Fourthly, the adaptability is insufficient. The general logistics equipment knowledge graph is not designed for the exclusive attributes and associated relationships of circulating carriers, resulting in limited management accuracy and scheduling efficiency, and failing to meet the individualized management needs of circulating carriers.
[0004] To solve the above technical problems, the present application relies on the structured advantages of the knowledge graph "entity-relation-attribute" to construct a whole life cycle stage-based dynamic topology management system, realizes the fine and intelligent management and control of circulating carriers, and provides technical support for the efficient operation of the warehouse logistics industry. SUMMARY
[0005] The present application mainly solves the technical problems of fragmented management dimension, static topology modeling, weak data-application linkage and insufficient adaptability of general knowledge graph in the existing circulating carrier management.
[0006] In an embodiment of the present application, a circulating carrier whole life cycle topology management method based on a knowledge graph is provided, which comprises:
[0007] Collecting multi-source data of the whole life cycle of the circulating carrier, and constructing a circulating carrier exclusive knowledge graph after preprocessing and quality verification;
[0008] Based on the exclusive knowledge graph, a dynamic topology subgraph is constructed according to the whole life cycle stage of the circulating carrier, and the topological relationship is described through dynamic edge weight and topological distance calculation;
[0009] Based on the stage-based dynamic topology subgraph, whole life cycle intelligent application data is output, and the exclusive knowledge graph and topological modeling parameters are iteratively updated through a closed-loop optimization mechanism combined with application feedback.
[0010] In the above knowledge graph-based circulating carrier whole life cycle topology management method, as a preferred scheme, the whole life cycle stage includes the procurement and storage stage, the operation and use stage, the maintenance and repair stage, and the scrap / reuse stage.
[0011] The procurement and storage stage dynamic topology subgraph is associated with the circulating carrier, the manufacturer, the procurement process, the storage node and the acceptance personnel.
[0012] The operation and use stage dynamic topology subgraph is associated with the circulating carrier, the goods, the storage node, the operation personnel and other circulating carriers.
[0013] The maintenance and repair stage dynamic topology subgraph is associated with the circulating carrier, the fault type, the equipment accessory, the maintenance personnel and the maintenance node.
[0014] The scrap / reuse stage dynamic topology subgraph is associated with the circulating carrier, the scrap evaluation index, the reuse process, the recycling node and the manufacturer.
[0015] In the above knowledge graph-based circulating carrier whole life cycle topology management method, as a preferred scheme, the dynamic edge weight calculation formula of the operation and use stage dynamic topology subgraph is:
[0016] ;
[0017] Wherein, is the association strength of entity i and entity j, is the interaction frequency of entity i and entity j, is the total interaction frequency of entity i, and α is the congestion influence coefficient, is the real-time load rate of node j;
[0018] The topological distance calculation uses the formula:
[0019] ;
[0020] Wherein, is the total cost of the path, is the kth physical distance, and α is the congestion coefficient, is the real-time load rate of node k, is the adaptive speed of the circulating carrier in the kth path, and n is the number of path nodes.
[0021] In the above-mentioned knowledge graph-based full lifecycle topology management method for cyclic vehicles, as a preferred embodiment, the intelligent application data includes intelligent scheduling data, maintenance early warning data, and full lifecycle value assessment data;
[0022] The intelligent scheduling data is calculated using a multi-objective optimization function, which is:
[0023] ;
[0024] Where ω1~ω4 are dynamic weighting coefficients, and T is the estimated task completion time. H represents the unit task cost over the entire lifecycle, H represents the health of the cyclic vehicle, and L represents the target node load rate.
[0025] The estimated completion time of the task , where d is the total path distance. The speed is the speed adapted to the cyclic vehicle, and t is the loading and unloading time;
[0026] The health of the cyclic vehicle ,in The duration of use. Total design time The number of failures. Where M is the fault threshold and M is the degree of wear. This is the wear threshold.
[0027] In the aforementioned knowledge graph-based full lifecycle topology management method for cyclic vehicles, as a preferred embodiment, the closed-loop optimization mechanism achieves parameter iteration through a weight update formula, which is:
[0028] ;
[0029] in, For the updated weights, For the original weights, The step size is 1, and the total weight is constrained to 1.
[0030] In the above-mentioned knowledge graph-based full lifecycle topology management method for cyclic vehicles, as a preferred embodiment, the ontology structure of the cyclic vehicle-specific knowledge graph includes core entities, exclusive attributes, and full lifecycle relationships.
[0031] The core entities include reusable vehicles, storage nodes, goods, personnel, and equipment spare parts;
[0032] The exclusive attributes of the circulating carrier include the number of cycles, the degree of wear, the remaining maintenance time, the number of failures, and the scrap threshold, and the exclusive attributes of the storage node include the real-time load rate, the congestion state, and the maintenance capacity threshold;
[0033] The full life cycle association relationship includes a procurement relationship, a parking relationship, a maintenance relationship, an assembly relationship, a replacement relationship, and a scrap / reuse relationship.
[0034] In the above-mentioned circulating carrier full life cycle topology management method based on a knowledge graph, as a preferred solution, the construction of the exclusive knowledge graph includes knowledge extraction and entity disambiguation, and the knowledge extraction uses a BERT model combined with a circulating carrier exclusive dictionary;
[0035] Entity disambiguation is achieved through multi-dimensional similarity matching, and the matching formula is:
[0036] ;
[0037] wherein, is the weight of each dimension, is the similarity of each dimension, and Match≥0.85 is determined as the same entity;
[0038] The exclusive knowledge graph is stored in a Neo4j graph database, and a unique index is established for the circulating carrier ID, the number of cycles, and the fault type field;
[0039] The incremental update trigger condition of the exclusive knowledge graph is or Topo_Demand=True, wherein ΔD is the proportion of new data, is a preset threshold, and Topo_Demand is a topology modeling data demand identifier.
[0040] In the above-mentioned circulating carrier full life cycle topology management method based on a knowledge graph, as a preferred solution, the multi-source data includes sensing data, business system data, and document data, and the collection frequency is set as: real-time sensing data is collected at 10Hz, business system data is incrementally synchronized every 5 minutes, and document data is batch-converted every morning;
[0041] Data preprocessing includes sliding window filtering and denoising, and the filtering formula is:
[0042] ;
[0043] wherein, is the filtered load value at the ith point, N is the window length, is the original load value;
[0044] Data quality verification is determined by a comprehensive quality score Q, wherein 、 、 are weight coefficients, I is integrity, A is accuracy, T is timeliness, and Q is greater than or equal to 0.95.
[0045] In the knowledge graph-based cyclic carrier full life cycle topology management method described above, as a preferred solution, the intelligent application data is connected to the PDOS / WMS / AGV scheduling system through a RESTful API, and the scheduling instruction is issued through an MQTT protocol, with a delay of no more than 100 ms.
[0046] The cyclic carrier includes an AGV, an intelligent pallet, and a turnover box, and the warehouse node includes a storage location, a maintenance node, and an in-out warehouse node.
[0047] The scrap evaluation index includes the cycle number, the wear degree, and the failure frequency.
[0048] Embodiments of the present application also provide a knowledge graph-based cyclic carrier full life cycle topology management device, which comprises:
[0049] A data acquisition and preprocessing module is configured to acquire multi-source data of the cyclic carrier full life cycle and perform preprocessing and quality checking.
[0050] A dedicated knowledge graph construction module is configured to construct a cyclic carrier dedicated knowledge graph based on the checked data.
[0051] A phased topology modeling module is configured to construct a dynamic topology subgraph according to the full life cycle stage based on the dedicated knowledge graph, and to depict the topology relationship through dynamic edge weight and topology distance calculation.
[0052] An intelligent application output module is configured to output full life cycle intelligent application data based on the phased dynamic topology subgraph.
[0053] A closed-loop optimization module is configured to iteratively update the dedicated knowledge graph and topology modeling parameters through a closed-loop optimization mechanism in combination with application feedback.
[0054] According to the knowledge graph-based cyclic carrier whole life cycle topology management method of the above embodiment, through whole life cycle stage topology modeling and cyclic carrier exclusive knowledge graph, the carrier whole process fine management and control are realized, the dynamic topology relationship and intelligent scheduling algorithm effectively improve the scheduling efficiency, shorten the task completion time, and solve the problems of extensive management and scheduling lag in the prior art; the whole cycle cost accounting and maintenance early warning mechanism are integrated, invalid energy consumption is reduced through optimized scheduling, excessive maintenance is avoided through accurate early warning, carrier maintenance cost and unit task cost are reduced, and carrier utilization efficiency is improved; the cyclic carrier exclusive ontology design and modular architecture can quickly adapt to different types of carriers such as AGV, intelligent tray, and turnover box, and various warehouse logistics scenes, reduce the expansion cost of new carriers and scenes, and solve the problems of insufficient adaptability and poor expansibility of the prior art; the dynamic topology update and early warning mechanism can avoid failure and congestion risks in advance, reduce the carrier failure rate, shorten the operation interruption time, and improve the continuity and operation reliability of warehouse logistics business. BRIEF DESCRIPTION OF DRAWINGS
[0055] Figure 1 A flowchart of the knowledge graph-based cyclic carrier whole life cycle topology management method provided by the embodiment of the present application is shown in the figure.
[0056] Figure 2 A principle block diagram of the knowledge graph-based cyclic carrier whole life cycle topology management method provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION
[0057] The present application will be further described in detail by specific embodiments in conjunction with the accompanying drawings. In different embodiments, similar elements are associated with similar element reference numbers.
[0058] In the following embodiments, many details are described in order to make the present application better understood. However, those skilled in the art can easily recognize that some features can be omitted in different cases, or can be replaced by other elements, materials, or methods.
[0059] In some cases, some operations related to the present application are not shown or described in the specification, in order to avoid the core part of the present application being overwhelmed by too much description, and it is not necessary to describe these related operations in detail for those skilled in the art according to the description in the specification and general technical knowledge in the art.
[0060] In addition, features described in the specification, operations or characteristics can be combined in any appropriate manner to form various embodiments. Meanwhile, the steps or actions in the method description can also be sequentially exchanged or adjusted in a manner that can be apparent to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment, and do not mean that the sequence is necessary, unless otherwise stated that a certain sequence must be followed.
[0061] The serial numbers of components in the specification, such as "first", "second", etc., are only used to distinguish the described objects, and do not have any sequence or technical meaning. The "connection" and "coupling" in the present application include direct and indirect connection (coupling) unless otherwise specified.
[0062] Please refer to Figure 1 In order to solve the technical problems of management dimension fragmentation, topology modeling static, weak data and application linkage and insufficient adaptability of general knowledge graph in existing circulating carrier management, an embodiment of the present application provides a circulating carrier whole life cycle topology management method based on knowledge graph, which is suitable for whole process management and control of AGV, intelligent pallet, turnover box and other circulating carriers. By collecting multi-source data of the whole life cycle of the circulating carrier, the circulating carrier exclusive knowledge graph is constructed after preprocessing and quality checking. Based on the knowledge graph, dynamic topology subgraph is constructed according to the whole life cycle stage, and the topology relationship is accurately described through dynamic edge weight and topology distance calculation. Based on the dynamic topology subgraph in stages, application data such as intelligent scheduling, maintenance warning and other application data are output. Combined with application feedback, the knowledge graph and topology modeling parameters are iteratively updated through a closed-loop optimization mechanism to realize whole-link continuous optimization and effectively avoid the homogenization problem of existing technologies. The circulating carrier whole life cycle topology management method based on knowledge graph comprises the following steps:
[0063] Step S101, collecting multi-source data of the whole life cycle of the circulating carrier, and constructing a circulating carrier exclusive knowledge graph after preprocessing and quality checking.
[0064] When collecting and preprocessing multi-source data, multi-source data of the whole life cycle of the circulating carrier is collected, including sensing data, business system data and document data, and standardized data foundation is formed after preprocessing and quality checking, providing high-quality data support for knowledge graph construction.
[0065] When collecting data, the carrier end deploys a weighing sensor, an IMU inertial measurement unit, a GPS Beidou positioning module, and an RFID tag to collect sensor data such as load, speed, position, fault code, cycle number, and wear degree, with a collection frequency of 10 Hz; through an API interface, the PDOS / WMS / TMS system is connected to synchronize business data such as procurement time, maintenance cycle, cargo ID, location ID, and task priority in increments of 5 minutes; at dawn every day, batch OCR recognition and NLP conversion are performed on document data such as maintenance manuals and repair records to extract fault type and maintenance cycle association rules.
[0066] When pre-processing data, a sliding window filtering algorithm is used to denoise sensor data, and the filtered load value is calculated by the formula , where is the filtered load value at the ith point, and N is the window length. is the original load value, and when is greater than the set proportion, the original abnormal data is replaced by .
[0067] When quality checking, the data is determined to be admitted to the knowledge graph based on the comprehensive quality score Q, , where is the weight coefficient, I is the integrity, i.e., the ratio of the actual number of collected fields to the number of fields that should be collected; A is the accuracy, i.e., 1 minus the ratio of the number of abnormal data to the total number of data; T is the timeliness, i.e., 1 minus the ratio of data delay to the maximum allowed delay; when Q≥0.95, the data is admitted to the knowledge graph, and when Q<0.95, data backflow is triggered.
[0068] Step S102, based on the specialized knowledge graph, a dynamic topology subgraph is constructed according to the full life cycle stage of the circulating carrier, and the topology relationship is described through dynamic edge weight and topology distance calculation.
[0069] When constructing the circulating carrier specialized knowledge graph, the verified data is used to construct the circulating carrier specialized knowledge graph, and through knowledge extraction, entity disambiguation, storage update, etc., accurate modeling of entity relationship attributes is realized, providing structured knowledge support for topology modeling.
[0070] In the process of ontology structure design, the Protégé tool is used to build a dedicated ontology, and the core entities include circulating carriers, storage nodes, goods, operating personnel and equipment accessories. The circulating carriers include AGV, intelligent pallets, totes and other subclasses, and the dedicated attributes include the number of cycles, wear and tear, maintenance remaining time, failure frequency and scrap threshold; the storage nodes include storage locations, maintenance nodes, warehouse entry and exit nodes and other subclasses, and the dedicated attributes include real-time load rate, congestion state and maintenance capacity threshold; the full life cycle association relationship includes procurement relationship, parking relationship, maintenance relationship, assembly relationship, replacement relationship and scrap and reuse relationship.
[0071] In the process of knowledge extraction and fusion, the BERT model is used to extract knowledge in combination with the circulating carrier dictionary to improve the accuracy of entity and relationship extraction; entity disambiguation is achieved through multi-dimensional similarity matching, and the matching formula is wherein, is the weight of each dimension, is the similarity of each dimension, and Match≥0.85 is determined as the same entity.
[0072] In the process of storage and incremental update, the Neo4j graph database is used to store the knowledge graph, and unique indexes are established for the circulating carrier ID, cycle number and failure type fields to improve query efficiency; the incremental update trigger condition is the proportion of new data, ΔD≥preset threshold or Topo_Demand=True, and the update is completed within 10 seconds after triggering, and the data is pushed to the topological modeling module.
[0073] Step S103, based on the phased dynamic topology subgraph output full life cycle intelligent application data, and iteratively update the dedicated knowledge graph and topological modeling parameters through a closed-loop optimization mechanism combined with application feedback.
[0074] In the process of full life cycle phased topological dynamic modeling, based on the dedicated knowledge graph, the dynamic topological subgraph is constructed according to the full life cycle stage of the circulating carrier, the topological relationship is described through dynamic edge weight and topological distance calculation, and the accurate representation and dynamic linkage of the association relationship of each stage are realized.
[0075] In the process of phased topological subgraph construction, the procurement and warehousing stage topological subgraph is associated with circulating carriers, manufacturers, procurement processes, warehousing nodes and acceptance personnel, supporting full life cycle cost tracing, and the subgraph association strength is automatically updated after the carrier warehousing acceptance is completed, and the use stage topological subgraph is simultaneously triggered to initialize; the operating use stage topological subgraph is associated with circulating carriers, goods, storage nodes, operating personnel and other circulating carriers, and a dynamic edge weight is introduced, and the association strength between entities i and j is calculated through the formula wherein, is the interaction frequency, Let α be the total number of interactions for entity i, Lⱼ be the congestion impact coefficient, and Lⱼ be the real-time load rate of node j. The more severe the congestion, the lower the association strength. During the maintenance phase, the topology subgraph associates cyclical vehicles, fault types, equipment parts, maintenance personnel, and maintenance nodes to support intelligent matching of maintenance plans. After a vehicle fault is repaired, the fault maintenance relationship strength is automatically updated and synchronously fed back to the topology subgraph during the usage phase. During the scrapping and reuse phase, the topology subgraph associates cyclical vehicles, scrapping assessment indicators, reuse processes, recycling nodes, and manufacturers to support intelligent scrapping and reuse decisions. When a vehicle reaches the scrapping threshold, the subgraph association is automatically triggered, and the vehicle node in the topology subgraph during the usage phase is simultaneously deregistered.
[0076] When calculating topological distances, the real-time congestion status of nodes and vehicle-specific attributes are incorporated into Dijkstra's algorithm, using a formula... Calculate the total cost of the path, where Let be the physical distance of the k-th segment, and α be the congestion coefficient. Let k be the real-time load rate of node k. The adaptive speed of the cyclic vehicle on the k-th segment of the path is dynamically adjusted based on the degree of wear; n is the number of path nodes, and the total path distance is output. Used for subsequent scheduling time calculations.
[0077] When performing dynamic updates to the topology, the update is triggered when the vehicle status changes, the node congestion status changes, a fault occurs and is repaired, or a new task is completed. After a certain stage subgraph is updated, the associated stage subgraph is automatically synchronized to ensure the consistency of the topology throughout the entire lifecycle.
[0078] Following step S103, there is also a step of intelligent application output, which outputs core application data such as intelligent scheduling, maintenance early warning, and full life cycle value assessment based on a phased dynamic topology subgraph, providing intelligent support for warehousing and logistics operations.
[0079] When outputting intelligent scheduling data, the optimal vehicle path combination is calculated using a multi-objective optimization function, which is as follows: ,in The dynamic weighting coefficients are T, where T is the estimated task completion time. t represents the loading and unloading time. The unit task cost represents the total lifecycle cost, incorporating costs such as procurement, maintenance, energy consumption, and residual value upon disposal; H represents the health status of the cyclic vehicle, i.e. ; Duration of use Total design time Number of failures Where M is the fault threshold and M is the degree of wear. The wear threshold is L, and L is a target node load rate; the scheduling instruction is issued through an MQTT protocol, and a delay is less than or equal to 100 ms.
[0080] In the maintenance early warning data output, based on the maintenance stage topology subgraph and the vehicle health degree H, in combination with the maintenance period, the used time length, the wear degree and other parameters, the maintenance demand is predicted, and the maintenance early warning is triggered.
[0081] In the whole life cycle value evaluation data output, based on the whole life cycle data of the topology subgraph, the vehicle value is quantified, and the value evaluation result provides data support for the vehicle procurement, maintenance and scrap recycling decision.
[0082] In the above-mentioned knowledge graph-based cyclic vehicle whole life cycle topology management method, as a preferred scheme, the whole life cycle stage includes a procurement storage stage, an operation use stage, a maintenance stage and a scrap / recycling stage; the procurement storage stage dynamically associates the cyclic vehicle, the manufacturer, the procurement process, the storage node and the acceptance personnel; the operation use stage dynamically associates the cyclic vehicle, the goods, the storage node, the operation personnel and other cyclic vehicles; the maintenance stage dynamically associates the cyclic vehicle, the fault type, the equipment accessory, the maintenance personnel and the maintenance node; and the scrap / recycling stage dynamically associates the cyclic vehicle, the scrap evaluation index, the recycling process, the recycling node and the manufacturer.
[0083] In the above-mentioned knowledge graph-based cyclic vehicle whole life cycle topology management method, as a preferred scheme, the dynamic edge weight calculation formula of the operation use stage dynamic topology subgraph is:
[0084] ;
[0085] Wherein, is the association strength of entity i and entity j, is the interaction frequency of entity i and entity j, is the total interaction frequency of entity i, and a is a congestion influence coefficient, is the real-time load rate of node j;
[0086] The topology distance calculation adopts the formula:
[0087] ;
[0088] Wherein, is the total cost of the path, is the kth physical distance, a is a congestion coefficient, is the real-time load rate of node k, is the adaptive speed of the cyclic vehicle in the kth path, and n is the number of path nodes.
[0089] In the knowledge graph-based cyclic carrier whole life cycle topology management method, the intelligent application data includes intelligent scheduling data, maintenance warning data and whole life cycle value evaluation data.
[0090] The intelligent scheduling data is calculated by a multi-objective optimization function, and the multi-objective optimization function is:
[0091] ;
[0092] Wherein, ω1~ω4 are dynamic weight coefficients, T is a task estimated completion time, is a whole life cycle unit task cost, H is a cyclic carrier health degree, and L is a target node load rate.
[0093] The task estimated completion time , wherein d is a total path distance, is a cyclic carrier adaptive speed, and t is a loading and unloading time.
[0094] The cyclic carrier health degree , wherein is a used time length, is a total design time length, is a failure frequency, is a failure threshold value, M is a wear degree, is a wear threshold value.
[0095] In the knowledge graph-based cyclic carrier whole life cycle topology management method, as a preferred solution, the closed loop optimization mechanism realizes parameter iteration through a weight updating formula, and the weight updating formula is:
[0096] ;
[0097] Wherein, is an updated weight, is an original weight, is a step length, and the weight sum constraint is 1.
[0098] In the knowledge graph-based cyclic carrier whole life cycle topology management method, as a preferred solution, the ontology structure of the cyclic carrier exclusive knowledge graph includes core entities, exclusive attributes and whole life cycle correlation relationships.
[0099] The core entities include cyclic carriers, warehouse nodes, goods, operating personnel and equipment accessories.
[0100] The exclusive attributes of the cyclic carrier include cycle number, wear degree, maintenance remaining time length, failure frequency and scrap threshold value, and the exclusive attributes of the warehouse node include real-time load rate, congestion state and maintenance capacity threshold value.
[0101] The full life cycle correlation relationship includes a procurement relationship, a parking relationship, a maintenance relationship, an assembly relationship, a replacement relationship, and a scrap / reuse relationship.
[0102] In the above knowledge graph-based cyclic vehicle full life cycle topology management method, as a preferred scheme, the construction of the exclusive knowledge graph includes knowledge extraction and entity disambiguation, and the knowledge extraction adopts a BERT model combined with a cyclic vehicle exclusive dictionary;
[0103] Entity disambiguation is achieved through multi-dimensional similarity matching, and the matching formula is:
[0104] ;
[0105] wherein, is the weight of each dimension, is the similarity of each dimension, and Match≥0.85 is determined as the same entity;
[0106] The exclusive knowledge graph is stored in a Neo4j graph database, and a unique index is established for the cyclic vehicle ID, the cycle number, and the fault type field;
[0107] The incremental update trigger condition of the exclusive knowledge graph is or Topo_Demand=True, wherein ΔD is the proportion of new data, is a preset threshold, and Topo_Demand is a topology modeling data demand identifier.
[0108] In the above knowledge graph-based cyclic vehicle full life cycle topology management method, as a preferred scheme, the multi-source data includes sensing data, business system data, and document data, and the collection frequency is set as: real-time sensing data is collected at 10Hz, business system data is incrementally synchronized every 5 minutes, and document data is batch converted every morning;
[0109] Data preprocessing includes sliding window filtering and denoising, and the filtering formula is:
[0110] ;
[0111] wherein, is the filtered load value at the ith point, N is the window length, is the original load value;
[0112] Data quality verification is determined by a comprehensive quality score Q, wherein , , are weight coefficients, I is integrity, A is accuracy, T is timeliness, and Q≥0.95 is the data access exclusive knowledge graph.
[0113] In the above-mentioned knowledge graph-based cyclic carrier full-life cycle topology management method, as a preferred solution, the intelligent application data is connected to the PDOS / WMS / AGV scheduling system through RESTful API, and the scheduling instruction is issued through MQTT protocol, and the issuance delay is not greater than 100 ms.
[0114] The cyclic carrier includes AGV, intelligent pallet and turnover box, and the warehouse node includes storage location, maintenance node and warehouse access node.
[0115] The scrap evaluation index includes cycle number, wear degree and failure frequency.
[0116] In some embodiments, in addition to the conventional sensing devices, a pressure sensor is additionally arranged on the key load-bearing component of the AGV to assist in verifying the load data, the RFID tag of the intelligent pallet is integrated with temperature and humidity sensing function to adapt to the special storage requirements of goods, and exclusive collection fields such as carrier running time and charging frequency are added to enrich the data dimension; the sensing devices are deployed according to the density difference of the storage location, the deployment interval of the laser radar and the visual camera is reduced to 5 meters in the high-density operation area, and the interval is kept at 10 meters in the low-density area, and a spare parts inventory sensor is additionally arranged in the maintenance area to realize real-time collection of spare parts data. In the data preprocessing stage, the window length of the sliding window filtering algorithm can be dynamically adjusted according to the motion state of the carrier, which is set to 10 when moving at a constant speed, and automatically reduced to 5 when accelerating or turning to improve the response speed; in the quality check, the weight coefficients of integrity, accuracy and timeliness can be dynamically adapted according to the data type, the sensing data improves the accuracy weight to 0.5, the business data improves the integrity weight to 0.5, and the document data improves the timeliness weight to 0.3, and a secondary check mechanism is established after data backflow to ensure that the supplementary data meets the access standard, forming a full-link data quality management and control system of “collection-preprocessing-check-backflow-recheck”, which provides high-credibility data support for knowledge graph construction.
[0117] In some embodiments, the adaptation relationship attributes of the carrier and the goods in the ontology structure are added, the associated rules of the goods carrying restrictions, the fixing methods and the like of different types of carriers are clarified, the skill adaptation attributes of the warehouse node supplementary workers are added, and a three-dimensional associated system of "carrier-goods-personnel" is established; in the knowledge extraction stage, the BERT model is fine-tuned based on the circulating carrier industry corpus, the recognition accuracy of the exclusive terminology is optimized, the semantic similarity calculation of the text of the new fault description and maintenance operation and the like is added, and the implicit associated relationship is extracted; in the entity disambiguation process, in addition to the existing dimensions, the carrier running track similarity, the goods adaptation preference and the like are added, the matching dimensions are expanded to 7, the carrier running track similarity weight is set to 0.1, the goods adaptation preference weight is set to 0.05, and the accuracy of the same entity determination is further improved; in the storage level, the Neo4j cluster deployment mode is adopted, the data is partitioned according to the carrier type and the life cycle stage, the core business data is stored in the master node, the historical archive data is stored in the slave node, a data backup and recovery mechanism is established at the same time, and the safety and the availability of the knowledge graph data are ensured; in the incremental update, in addition to the existing trigger conditions, the data quality abnormal trigger update mechanism is added, when the data quality score of a certain type of data is less than 0.95 for 3 times in succession, the knowledge graph entity and the relationship corresponding to the data are triggered to update, and the knowledge graph and the actual data state are ensured to be real-time aligned.
[0118] In some embodiments, the procurement and warehousing stage topology subgraph adds related entities and relationships such as supplier qualification, procurement contract terms, etc., and subdivides the procurement cost into equipment cost, transportation cost, installation cost, etc. sub-items, supports the fine tracing of the full life cycle cost, and the subgraph correlation strength is dynamically adjusted according to the supplier performance quality and equipment acceptance qualification rate; in the dynamic edge weight calculation of the operation use stage topology subgraph, a time factor is added, and an intra-day operation peak coefficient is introduced, the congestion influence coefficient a is increased to 0.7 in the operation peak period (such as 10-12 am, 3-5 pm), and reduced to 0.3 in the non-peak period, at the same time, the edge weight is integrated with the adaptation degree parameter of the carrier and the goods, the higher the adaptation degree, the greater the correlation strength weight; the maintenance stage topology subgraph adds entities such as maintenance equipment and maintenance consumables, and correlates the quantitative relationship between maintenance cost, maintenance duration and maintenance effect, the dynamic update logic adds the feedback of the running state monitoring data of the carrier after maintenance, if the fault recurs within 72 hours after maintenance, the relationship strength corresponding to the maintenance scheme is automatically reduced; the scrap / reuse stage topology subgraph adds related factors such as environmental protection treatment standards and reuse processing costs, the scrap evaluation index is expanded to five items of cycle number, wear degree, fault frequency, energy consumption growth rate and maintenance cost proportion, and a multi-index comprehensive evaluation model is established; in the topology distance calculation, in addition to the existing parameters, the energy consumption coefficient of the carrier and the fragility coefficient of the goods are integrated, the higher the energy consumption coefficient and the stronger the fragility of the goods, the greater the path cost weight, at the same time, the path search algorithm is optimized, the improved Dijkstra algorithm combined with the advantages of A* algorithm is adopted, and the efficiency and accuracy of path planning are improved; when updating the topology relationship, an update priority mechanism is established, the update priority triggered by emergency events such as fault repair and carrier state change is the highest, the update of regular events such as node congestion state change and task addition is updated in order, and the real-time of key topology relationship is ensured.
[0119] In some embodiments, in the intelligent scheduling optimization, the dynamic weight coefficient of the multi-objective optimization function is dynamically adjusted according to the job scene, the in-warehouse job scene improves the target node load rate weight ω4 to 0.25, the out-of-warehouse job scene improves the task completion time weight ω1 to 0.5, and the cross-region transfer scene improves the whole life cycle unit task cost weight ω2 to 0.3. At the same time, the scheduling algorithm supports multi-task parallel processing, realizes multi-cargo collaborative scheduling through task priority sorting and resource allocation optimization, establishes a three-level early warning mechanism in the maintenance early warning data output, the first level early warning (cargo health degree H < 0.6) triggers the emergency maintenance process, synchronously locks the cargo to prohibit new task allocation, the second level early warning (0.6≤H<0.8) pushes the maintenance reminder and plans the maintenance window period, the third level early warning (0.8≤H<0.9) records the potential risk and continuously monitors, the early warning threshold is dynamically adjusted according to the cargo type and service life, and the early warning sensitivity of old cargo is appropriately improved; in the whole life cycle value evaluation data output, the quantitative model integrates external factors such as market supply and demand changes and equipment technology iteration, the job value calculation considers the timeliness value and safety value of goods transportation, the whole cycle cost adds depreciation cost and idle cost, and the recycling value combines the fluctuation of renewable resource price and the processing cost, forming a dynamically updated value evaluation system to provide more comprehensive data support for cargo procurement replacement, scrapping and recycling decision-making.
[0120] In some embodiments, the hardware deployment adopts the strategy of "core area full coverage + edge area blind filling", high-performance laser radar and visual camera are deployed in core areas such as warehouse entrances and exits and intersections, low-cost RFID readers are used to fill in the gaps and corners of shelves in edge areas to ensure that there is no sensing blind area; on the interface adaptation level, in addition to RESTful API, WebSocket interface is added to support real-time data stream transmission, compatible with OPCUA, ModbusTCP and other industrial standard protocols, seamless connection with different manufacturers' logistics systems is realized through a protocol conversion gateway, and interface access permission control and data encryption transmission mechanism are established to ensure the security of data interaction; the pilot scene promotion adopts a hierarchical strategy of "single scene verification - multi-scene fusion - full range landing", first verifies the accuracy of data acquisition and knowledge graph construction in a single storage location management scene, then expands to in-warehouse, out-of-warehouse, maintenance and other multi-scene fusion verification to verify the effectiveness of topology modeling and intelligent application, and finally optimizes system performance combined with pilot feedback to form a standardized deployment scheme to promote to the whole warehouse area. Modular installation and debugging are adopted in the deployment process to reduce the impact on existing warehouse operations.
[0121] In some embodiments, the weight update formula introduces an adaptive step mechanism, the step Δω is dynamically adjusted according to the deviation value of the actual effect and the target effect, when the deviation value is greater than 20%, the step is increased to 0.08, when the deviation value is less than 5%, the step is reduced to 0.02, to ensure the accuracy and stability of weight adjustment; In the model optimization process, in addition to weight adjustment, new topology modeling rule optimization and knowledge graph ontology structure optimization are added, unreasonable association rules in topology subgraph construction are identified based on application feedback data, entities, attributes and relationships of knowledge graph are updated combined with new technology and new business demand of the industry, for example, new constraint condition rules of intelligent scheduling algorithm, adaptation relationship of carriers and new goods; A multi-source feedback data fusion mechanism is established, feedback data not only includes system operation indicators, but also includes operation feedback of operation personnel and decision feedback of management personnel, the weights of different feedback sources are determined by AHP, and a comprehensive feedback evaluation result is formed, an optimization report is generated every quarter to clarify the optimization direction, implementation steps and expected effect, and an optimization effect verification mechanism is established, the system performance indicators before and after optimization are compared through A / B test to ensure the effectiveness of closed-loop optimization, and the adaptability and intelligent level of the system are continuously improved.
[0122] Please refer to Figure 2 Embodiments of the present application also provide a knowledge graph-based cyclic carrier full life cycle topology management device, the device comprises:
[0123] A data acquisition and preprocessing module is configured to acquire multi-source data of the cyclic carrier full life cycle, perform preprocessing and quality checking;
[0124] A dedicated knowledge graph construction module is configured to construct a dedicated knowledge graph of the cyclic carrier based on the checked data;
[0125] A phased topology modeling module is configured to construct dynamic topology subgraphs according to the dedicated knowledge graph and the full life cycle stages, and describe the topology relationship through dynamic edge weight and topology distance calculation;
[0126] An intelligent application output module is configured to output full life cycle intelligent application data based on the phased dynamic topology subgraphs;
[0127] A closed-loop optimization module is configured to iteratively update the dedicated knowledge graph and topology modeling parameters through a closed-loop optimization mechanism combined with application feedback.
[0128] Those skilled in the art can understand that all or part of the functions of various methods in the above embodiments can be realized by hardware or by a computer program. When all or part of the functions in the above embodiments are realized by a computer program, the program can be stored in a computer readable storage medium, which can include a read-only memory, a random access memory, a magnetic disk, an optical disk, a hard disk, and the like. The above functions are realized by executing the program by a computer. For example, the program is stored in a memory of a device, and when the program in the memory is executed by a processor, the above functions are realized. In addition, when all or part of the functions in the above embodiments are realized by a computer program, the program can also be stored in a storage medium such as a server, another computer, a disk, an optical disk, a flash disk, or a mobile hard disk, and is downloaded or copied into a memory of a local device or is updated in version to a system of the local device. When the program in the memory is executed by a processor, all or part of the functions in the above embodiments are realized.
[0129] The above application is described by using specific examples, which is only used to help understand the application and does not limit the application. According to the idea of the application, those skilled in the art can make several simple deductions, modifications, or substitutions.
Claims
1. A knowledge graph-based method for full lifecycle topology management of cyclic vehicles, characterized in that, The knowledge graph-based method for full lifecycle topology management of cyclic vehicles includes: Collect multi-source data on the entire lifecycle of reusable vehicles, and construct a dedicated knowledge graph for reusable vehicles after preprocessing and quality verification. Based on the exclusive knowledge graph, a dynamic topological subgraph is constructed according to the entire life cycle stage of the cyclic vehicle, and the topological relationship is characterized by dynamic edge weights and topological distance calculation. Based on the phased dynamic topology subgraph, the system outputs full-lifecycle intelligent application data, and combines application feedback to iteratively update the exclusive knowledge graph and topology modeling parameters through a closed-loop optimization mechanism.
2. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 1, characterized in that, The entire life cycle includes the procurement and warehousing stage, the operation and use stage, the maintenance and upkeep stage, and the scrapping / reuse stage; The dynamic topology subgraph of the procurement and warehousing stage associates the cyclical vehicles, manufacturers, procurement process, warehousing nodes, and acceptance personnel. The dynamic topology subgraph during the operation phase associates cyclic vehicles, goods, storage nodes, operators, and other cyclic vehicles. The maintenance phase involves a dynamic topology subgraph that associates cyclic vehicles, fault types, equipment parts, maintenance personnel, and maintenance nodes. The dynamic topology subgraph of the scrap / reuse phase associates cyclic vehicles, scrap assessment indicators, reuse processes, recycling nodes, and manufacturers.
3. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 2, characterized in that, The formula for calculating the dynamic edge weights of the dynamic topology subgraph during the operation phase is as follows: ; in, Let be the association strength between entity i and entity j. Let i be the frequency of interaction between entity i and entity j. Let α be the total number of interactions for entity i, and α be the congestion impact coefficient. Let be the real-time load rate of node j; The topological distance is calculated using the following formula: ; in, The total cost of the path. Let be the physical distance of the k-th segment, and α be the congestion coefficient. Let k be the real-time load rate of node k. Let n be the adaptive speed of the cyclic vehicle on the k-th path segment, and n be the number of path nodes.
4. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 3, characterized in that, The intelligent application data includes intelligent scheduling data, maintenance early warning data, and full lifecycle value assessment data; The intelligent scheduling data is calculated using a multi-objective optimization function, which is: ; Where ω1~ω4 are dynamic weighting coefficients, and T is the estimated task completion time. H represents the unit task cost over the entire lifecycle, H represents the health of the cyclic vehicle, and L represents the target node load rate. The estimated completion time of the task , where d is the total path distance. The speed is the speed adapted to the cyclic vehicle, and t is the loading and unloading time; The health of the cyclic vehicle ,in The duration of use. Total design time The number of failures. Where M is the fault threshold and M is the degree of wear. This is the wear threshold.
5. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 4, characterized in that, The closed-loop optimization mechanism achieves parameter iteration through a weight update formula, which is: ; in, For the updated weights, For the original weights, The step size is 1, and the total weight is constrained to 1.
6. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 1, characterized in that, The ontology structure of the knowledge graph specific to the cyclic vehicle includes core entities, specific attributes, and full lifecycle relationships; The core entities include reusable vehicles, storage nodes, goods, personnel, and equipment spare parts; The unique attributes of cyclic vehicles include the number of cycles, wear level, remaining maintenance time, number of failures, and scrap threshold; the unique attributes of storage nodes include real-time load rate, congestion status, and maintenance capacity threshold. The entire lifecycle relationships include procurement relationships, docking relationships, maintenance relationships, assembly relationships, replacement relationships, and scrapping / reuse relationships.
7. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 6, characterized in that, The construction of the exclusive knowledge graph includes knowledge extraction and entity disambiguation. Knowledge extraction adopts the BERT model combined with a cyclic vehicle exclusive dictionary. Entity disambiguation is achieved through multi-dimensional similarity matching, with the following matching formula: ; in, Weights for each dimension, For each dimension of similarity, if Match ≥ 0.85, they are determined to be the same entity; The dedicated knowledge graph is stored in the Neo4j graph database, and a unique index is established for the fields of cyclic vehicle ID, number of cycles, and fault type. The incremental update trigger condition for the proprietary knowledge graph is: Alternatively, Topo_Demand=True, where ΔD represents the percentage of newly added data. The preset threshold is used, and Topo_Demand is the identifier for topology modeling data requirements.
8. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 1, characterized in that, The multi-source data includes sensor data, business system data, and document data. The collection frequency is set as follows: real-time sensor data is collected at 10Hz, business system data is synchronized incrementally at 5-minute intervals, and document data is batch converted at midnight every day. Data preprocessing includes sliding window filtering for noise reduction; the filtering formula is as follows: ; in, Let N be the filtered load value at point i, and N be the window length. This is the original load value; Data quality is determined by a comprehensive quality score Q. ,in , , Where I is the weighting coefficient, A is the completeness, T is the timeliness, and Q ≥ 0.95 is required for data to be admitted into the dedicated knowledge graph.
9. The knowledge graph-based full lifecycle topology management method for cyclic vehicles as described in claim 1, characterized in that, The intelligent application data is connected to the PDOS / WMS / AGV scheduling system via RESTful API, and the scheduling instructions are sent via MQTT protocol with a sending delay of no more than 100ms. The circulating carriers include AGVs, smart pallets, and turnover boxes; the warehousing nodes include storage locations, maintenance nodes, and inbound / outbound nodes. The scrap assessment indicators include the number of cycles, the degree of wear and tear, and the frequency of failure.
10. A knowledge graph-based device for full lifecycle topology management of cyclic vehicles, characterized in that, The device includes: The data acquisition and preprocessing module is used to acquire multi-source data throughout the entire life cycle of the cyclic vehicle and perform preprocessing and quality verification. A dedicated knowledge graph construction module is used to construct a dedicated knowledge graph for a circular vehicle based on the verified data. The phased topology modeling module is used to construct a dynamic topology subgraph based on the exclusive knowledge graph according to the full life cycle stage, and to characterize the topology relationship through dynamic edge weights and topology distance calculation. The intelligent application output module is used to output full lifecycle intelligent application data based on the phased dynamic topology subgraph. A closed-loop optimization module is used to iteratively update the proprietary knowledge graph and topology modeling parameters by combining application feedback through a closed-loop optimization mechanism.
