A warehouse management method and system based on big data processing

Abstract

The application discloses a kind of based on big data processing's warehouse management method and system, comprising the following steps: step one: data acquisition, form warehouse state mirror;Step two: generate goods node, location node and order witness hyperedge, form order witness hypergraph;Step three: execute mapping pairing and area collection, form initial location division result;Step four: difference calculation split change, path span change and path return change, form tensor cut gain rule;Step five: closed matching migration relationship and migration relationship, form conservation exchange ring rule;Step six: stack arrangement forms exchange evidence stack rule, and constitute improved Kernighan-Lin algorithm;Step seven: execute node exchange search and loop check, form target conservation exchange ring;Step eight: execute residual gain determination, output warehouse management result.The application improves the path continuity and exchange stability of location division.

Description

Technical Field

[0001] This invention relates to the field of warehouse management technology, and in particular to a warehouse management method and system based on big data processing. Background Technology

[0002] With the expansion of warehousing operations and the increasing demands for order fulfillment timeliness, warehouse management technologies targeting product location allocation, operational path organization, and dynamic location adjustment have received widespread attention. Existing warehouse management systems primarily rely on inventory turnover rate, order co-occurrence frequency, and location distance for product zoning or storage allocation; however, these systems commonly suffer from the following problems in practical applications: The fragmented sources of warehouse-related data, coupled with the lack of a unified state mirror representation among product occupancy status, storage location occupancy status, and operational flow status, make it difficult for the storage location partitioning process to accurately reflect real-time occupancy relationships and operational path relationships within the warehouse space. Existing product association modeling methods are mostly based on pairwise association edges or ordinary graph structures, which are insufficient to express the order witnessing relationships formed by multiple product nodes along the operational path in the same task. This leads to the easy neglect of order-level path continuity, storage location area span, and path backtracking changes during storage location optimization. Traditional graph partitioning or heuristic exchange methods typically use single node exchanges or simple gain sorting as the optimization basis, lacking a closed-loop verification mechanism for storage location migration relationships, migration out relationships, and exchange dependencies. This results in storage location adjustment schemes prone to problems such as non-closed exchange chains, non-conservative occupancy relationships, and difficulty in determining residual gains after adjustment, affecting the stability and executability of warehouse management results.

[0003] Therefore, how to provide a warehouse management method and system based on big data processing is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose a warehouse management method and system based on big data processing. This invention fully utilizes warehouse status mirroring, order witness hypergraph, and improved Kernighan-Lin algorithm to map and divide product nodes and storage location nodes, perform conservation exchanges, and determine residual gains, thereby achieving dynamic optimization management of warehouse storage locations. It has the advantages of high executability of storage location adjustments, strong order path continuity, and high efficiency of warehouse operations.

[0005] A warehouse management method based on big data processing according to an embodiment of the present invention includes the following steps: Step 1: Collect warehouse-related data, perform state extraction and spatial mapping on the warehouse-related data to form a warehouse state mirror; Step 2: Generate product nodes and storage location nodes in the warehouse status mirror, and arrange product nodes associated with the same task into order witness hyperedges according to the operation path to form an order witness hypergraph; Step 3: Perform mapping and pairing on the product nodes and storage location nodes in the order witness supermap, and aggregate the mapping and pairing results according to the storage location area to form the initial storage location division results; Step 4: Perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping to form tensor cut gain rules; Step 5: Perform closed-loop matching on the migration-in and migration-out relationships of the storage location nodes to form a conserved exchange loop rule; Step 6: Perform stack-style arrangement on the exchange order and exchange dependencies of the goods nodes to form the exchange evidence stack rules, and the improved Kernighan-Lin algorithm is composed of the tensor cut gain rule, the conservation exchange ring rule, and the exchange evidence stack rules. Step 7: Use the improved Kernighan-Lin algorithm to perform node swap search and loop verification in the order witness hypergraph to form a target-conserving swap cycle; Step 8: Adjust the initial warehouse location allocation results according to the objective conservation exchange loop, perform residual gain determination on the adjusted order witness hypergraph, and output the warehouse management results.

[0006] Optionally, step one specifically includes: Collect warehouse-related data, establish a collection record set according to data source and collection time, and perform unified coding on the product identifier, storage location identifier and operation task identifier in the collection record set to form an identifier-unified record set; The identifier-normalized record set is sorted by collection time, and the sorted identifier-normalized record set is merged by task identifier to form a time-series associated record set. Extract the goods occupancy status, storage location occupancy status, and work flow status from the time-series associated record set, and establish a status arrangement relationship according to the collection time to form a status extraction record set; Map the location identifiers in the status extraction record set to warehouse space coordinates, and bind the goods occupancy status, location occupancy status, and operation flow status to the warehouse space coordinates to form a space mapping record set; The spatial mapping record set is mirrored according to the collection time and storage space coordinates to form a storage status mirror.

[0007] Optionally, step two specifically includes: Extract the product occupancy status and storage location occupancy status from the warehouse status image, generate product nodes and storage location nodes respectively, and establish the occupancy mapping relationship between product nodes and storage location nodes; Extract the task identifier and task path relationship from the warehouse status image, and arrange the goods nodes associated with the same task according to the execution order of the task path relationship to form a task goods node sequence; Connect the goods nodes in the task goods node sequence to form an order witness superedge, and associate the order witness superedge with the occupancy mapping relationship to form a superedge association record; Arrange the execution graph structure of product nodes, storage location nodes, occupancy mapping relationships, order witness hyperedges, and hyperedge association records to form the order witness hypergraph.

[0008] Optionally, step three specifically includes: Extract product nodes, storage location nodes, and order witness hyperedges from the order witness hypergraph, and establish product path location records according to the path position of the product nodes in the order witness hyperedges; Extract the storage area and spatial location of the storage location node, match the goods path location record with the execution location of the storage location node, and form a mapping pairing result between the goods node and the storage location node; Based on the storage location area, the mapping and pairing results are aggregated by region, and the product nodes, storage location nodes and order witness superedges within the same storage location area are associated and organized to form a regional storage location division record; Based on the regional storage location division records, an initial division relationship is established between the product nodes and storage location nodes in the order witness supermap, forming the initial storage location division result.

[0009] Optionally, step four specifically includes: Extract the mapping and pairing results between the goods node and the storage location node from the initial storage location division results, and extract the order witness hyperedge and the path position of the goods node in the order witness hyperedge from the order witness hypergraph to form the hyperedge state record before the exchange. Select the goods node to be exchanged in the super-edge status record before the exchange, perform the exchange simulation on the storage location node corresponding to the goods node to be exchanged, and fill the storage location area after the exchange simulation back into the order witness super-edge to form the super-edge status record after the exchange. Perform a region attribution comparison on the super-edge status records before and after the exchange, and count the changes in the number of warehouse location regions crossed by the super-edge for the same order to form a split change record; Compare the path positions of the super-edge status records before and after the exchange, and count the changes in the warehouse location span and the number of path reversals of the same order's witness super-edge in the operation path to form path span change records and path reversal change records; The split change records, path span change records, and path return change records are aggregated according to the nodes of the goods to be exchanged, and the aggregation results are configured as the tensor cut gain rules corresponding to the nodes of the goods to be exchanged.

[0010] Optionally, step five specifically includes: Extract product nodes, storage location nodes, and mapping pairing results from the order witness hypermap and initial storage location division results. Determine the storage location nodes occupied by the product nodes according to the mapping pairing results to form storage location occupancy records. Select the goods node to be exchanged in the storage location occupancy record, mark the storage location node occupied by the goods node to be exchanged before the exchange as the moving-out storage location node, and mark the storage location node to which the goods node to be exchanged enters after the exchange as the moving-in storage location node, thus forming the moving-out relationship and the moving-in relationship. According to the direction of goods node exchange, the migration-out relationship and migration-in relationship are matched end to end, so that the previous migration-in warehouse location node is consistent with the next migration-out warehouse location node, forming a warehouse location node exchange chain; Perform closed matching on the storage location node exchange chain. When the starting storage location node and the ending storage location node of the exchange chain are the same, a closed exchange chain is formed. Perform conservation checks on the quantity of goods nodes and storage locations occupied in the closed exchange chain, and configure the closed exchange chain that passes the conservation check as a conservation exchange ring rule.

[0011] Optionally, step six specifically includes: Extract the item nodes to be exchanged and the tensor cut gain value from the tensor cut gain rule, and sort the item nodes to be exchanged in descending order according to the tensor cut gain value to form the item node exchange order; Extract the closed exchange chain from the conservation exchange ring rule, and perform position matching on the goods nodes to be exchanged in the closed exchange chain according to the exchange order of the goods nodes to form an exchange ring matching record; Dependency labeling is performed on the nodes of goods to be exchanged in the exchange ring matching record, and the storage location acceptance relationship between the outgoing storage location node, the incoming storage location node and the adjacent outgoing storage location node of the node to be exchanged is organized into exchange dependency relationship; According to the order of goods node exchange, push the exchange ring matching record into the exchange evidence stack, and write the tensor cut gain value, closed exchange chain and exchange dependency relationship into the same stack position to form the stack position evidence record. The stack evidence records are sorted according to the push order, and the connection index between adjacent stack positions is established according to the exchange dependency relationship to form the exchange evidence stack rules. The tensor cut gain rule is determined as the node exchange evaluation rule, the conserved exchange loop rule is determined as the node exchange constraint rule, and the exchange evidence stack rule is determined as the node exchange backtracking rule, thus forming the improved Kernighan-Lin algorithm.

[0012] Optionally, step seven specifically includes: The improved Kernighan-Lin algorithm is called to process the order witness hypergraph and the initial storage location partitioning results, and the product nodes, storage location nodes and order witness hyperedges participating in the node exchange search are marked in the order witness hypergraph to form the node exchange search state; The tensor cut gain rule is applied to the node exchange search state. The tensor cut gain value corresponding to the goods node to be exchanged is read, and the goods nodes to be exchanged are sorted in descending order according to the tensor cut gain value to form candidate node exchange records. Apply the conservation exchange ring rule to the candidate node exchange record, read the outgoing and incoming warehouse locations corresponding to the goods to be exchanged, and connect the outgoing and incoming warehouse locations according to the warehouse location acceptance relationship to form a candidate closed exchange chain. Perform loop verification on candidate closed exchange chains, compare the consistency of the starting and ending storage locations, and compare the continuity of the storage location acceptance relationship between adjacent goods to be exchanged nodes to form a loop verification record; The rules for exchanging evidence stacks are applied to the loop verification record. The loop verification record is matched with the stack evidence record, and consistency checks are performed on the tensor cut gain value, candidate closed exchange chain, and exchange dependency to form a candidate conserved exchange loop. The candidate conserved commutative rings are sorted according to their tensor cut gain values, and the candidate conserved commutative ring with the highest ranking is selected to form the target conserved commutative ring.

[0013] Optionally, step eight specifically includes: Read the exchange order of goods nodes and the storage location acceptance relationship in the target conservation exchange ring, and adjust the mapping and pairing results in the initial storage location division result according to the exchange order of goods nodes; Write the adjusted mapping pairing results back to the order witness hypergraph and update the product nodes and storage location nodes associated with the order witness hyperedge to form the adjusted order witness hypergraph; Extract the tensor cut gain, closed commutation chain, and commutation dependency from the target-conserving commutation ring, generate commutation evidence records, and write the commutation evidence records into the commutation evidence stack. The residual calculations for the splitting changes, path span changes, and path return changes of the adjusted order witness hypergraph are performed to form the residual gain determination results; The location allocation status of the adjusted order witness supermap is determined based on the residual gain determination results, and the warehouse management results containing the location allocation status and exchange evidence records are output.

[0014] A warehouse management system based on big data processing according to an embodiment of the present invention includes: The image generation module is used to collect warehouse-related data and perform state extraction and spatial mapping to form a warehouse state image; The hypergraph construction module is used to generate product nodes and storage location nodes, and to arrange product nodes associated with the same job task into order witness hyperedges according to the job path, forming an order witness hypergraph; The initial partitioning module is used to perform mapping and pairing of product nodes and storage location nodes and to aggregate them according to storage location areas to form the initial storage location partitioning results; The gain construction module is used to perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping, forming tensor cut gain rules; The exchange ring construction module is used to perform closed matching on the migration-in and migration-out relationships of storage location nodes to form a conserved exchange ring rule; The algorithm construction module is used to perform stack-style arrangement of the exchange order and exchange dependency of goods nodes, form the exchange evidence stack rules, and constitute the improved Kernighan-Lin algorithm; The target search module is used to perform node exchange search and loop verification in the order witness supergraph to form a target conservation exchange loop; The results output module is used to adjust the initial storage location allocation results according to the target conservation exchange loop and perform residual gain determination, and output the storage management results.

[0015] The beneficial effects of this invention are: This invention extracts and spatially maps the execution status of warehouse-related data to form a warehouse status mirror. Within this mirror, it constructs an order witness hypergraph composed of product nodes, storage location nodes, and order witness hyperedges, enabling the representation of product occupancy status, storage location occupancy status, and operational flow status within a unified graph structure. Consequently, warehouse management no longer relies solely on single inventory turnover rates or pairwise product relationships. Instead, it incorporates product path relationships, storage location area affiliations, and order-level relationships within the same operational task into the storage location allocation process, improving the matching degree between the initial storage location allocation results and the actual operational paths.

[0016] This invention employs an improved Kernighan-Lin algorithm, constructed using tensor cut gain rules, conserved exchange loop rules, and exchange evidence stack rules. This algorithm performs node exchange search and loop verification within the order witness hypergraph, enabling warehouse location adjustments to simultaneously consider changes in order witness hyperedge splitting, path span, and path return. Furthermore, the closed-loop matching of warehouse location node migration in and out relationships ensures the conservation of exchange processes. This reduces issues such as incomplete exchange chains, non-conserved occupancy relationships, and difficulty in determining residual gains after warehouse location adjustments, thereby improving the stability, executability, and picking efficiency of warehouse management results. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a warehouse management method based on big data processing proposed in this invention; Figure 2 This is a schematic diagram of the order witness hypergraph construction for a warehouse management method based on big data processing proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-2 A warehouse management method based on big data processing includes the following steps: Step 1: Collect warehouse-related data, perform state extraction and spatial mapping on the warehouse-related data to form a warehouse state mirror; Step 2: Generate product nodes and storage location nodes in the warehouse status mirror, and arrange product nodes associated with the same task into order witness hyperedges according to the operation path to form an order witness hypergraph; Step 3: Perform mapping and pairing on the product nodes and storage location nodes in the order witness supermap, and aggregate the mapping and pairing results according to the storage location area to form the initial storage location division results; Step 4: Perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping to form tensor cut gain rules; Step 5: Perform closed-loop matching on the migration-in and migration-out relationships of the storage location nodes to form a conserved exchange loop rule; Step 6: Perform stack-style arrangement on the exchange order and exchange dependencies of the goods nodes to form the exchange evidence stack rules, and the improved Kernighan-Lin algorithm is composed of the tensor cut gain rule, the conservation exchange ring rule, and the exchange evidence stack rules. Step 7: Use the improved Kernighan-Lin algorithm to perform node swap search and loop verification in the order witness hypergraph to form a target-conserving swap cycle; Step 8: Adjust the initial warehouse location allocation results according to the objective conservation exchange loop, perform residual gain determination on the adjusted order witness hypergraph, and output the warehouse management results.

[0020] In this embodiment, step one specifically includes: Warehouse-related data is formed by the data stream generated during warehousing operations. The data source is used to distinguish the data collection entry point, and the collection time is used to determine the data entry order. Each record in the collection record set retains the product identifier, storage location identifier, and operation task identifier. The unified coding converts identifiers with different expressions but pointing to the same object into a consistent code, so that products, storage locations, and operation tasks form a unified identifier record set under the same data caliber. The identifier-unified record set is arranged from front to back according to the collection time, so that the records corresponding to the same product identifier and the same storage location identifier are kept in time continuity. Then, the records belonging to the same operation task are merged according to the operation task identifier, so that the product flow relationship and storage location occupancy relationship within the operation task are concentrated in the same task scope, forming a time-series associated record set. The product identifier and storage location identifier in the time-series association record set are used to determine the product occupancy status and storage location occupancy status, while the task identifier and collection time are used to determine the task flow status. The status arrangement relationship is established according to the collection time, so that the product occupancy status, storage location occupancy status and task flow status remain corresponding in the time direction, forming a status extraction record set. The location identifiers in the status extraction record set are mapped to warehouse space coordinates according to the warehouse space layout. The occupancy status of goods is bound to the warehouse space coordinates where the corresponding goods are located, the occupancy status of storage locations is bound to the warehouse space coordinates where the corresponding storage locations are located, and the workflow status is bound to the warehouse space coordinates through which the work tasks pass, forming a spatial mapping record set. The spatial mapping record set establishes a time dimension according to the collection time and a spatial dimension according to the warehouse space coordinates. The occupancy status of goods, storage location, and operation flow corresponding to the same collection time are filled into the corresponding spatial locations, so that the warehouse space forms a calculable mirrored state expression under continuous collection time, forming a warehouse state mirror.

[0021] In this embodiment, step two specifically involves: The occupancy status of goods in the warehouse status mirror is used to determine the location of goods at the time of collection, and the occupancy status of storage locations is used to determine the carrying status of storage locations at the time of collection. Goods nodes are established with goods identifiers as node indexes, and storage location nodes are established with storage location identifiers as node indexes. Goods nodes and storage location nodes that are occupied at the same time of collection are paired to form an occupancy mapping relationship. The task identifier is used to define the scope of the same task, and the task path relationship is used to determine the flow order of goods nodes in the storage space. The goods nodes associated with the same task are arranged according to the order of passage in the task path relationship, so that the goods nodes form a continuous path position expression within the task scope, forming a task goods node sequence. The product nodes in the task product node sequence form the same hyperedge range according to the arrangement order. The order witness hyperedge is used to represent the common flow relationship between product nodes within the same task. The order witness hyperedge is bound to the occupancy mapping relationship so that the order witness hyperedge corresponds to the storage location node where the product node is located, forming a hyperedge association record. The graph structure is arranged with product nodes and storage location nodes as node units, occupancy mapping relationships as mapping edges between nodes, order witness hyperedges as task association edges, and hyperedge association records as connection indexes between order witness hyperedges and storage location nodes, so that the warehouse state mirror is transformed into an order witness hypergraph containing node relationships and hyperedge relationships.

[0022] In this embodiment, step three specifically includes: In the order witness hypergraph, the product node is used to represent the object occupied by the product in the warehouse status mirror, the storage location node is used to represent the space carrying object that the product can be mapped, and the order witness hyperedge is used to represent the associated range formed by the product nodes along the operation path within the same operation task. The path order value is extracted according to the arrangement position of the product node in the order witness hyperedge, and the product node identifier is bound to the path order value to form a product path position record. The storage location area of ​​a storage location node is used to characterize the storage zone to which the storage location node belongs, and the spatial location is used to characterize the location of the storage location node in the storage space coordinates. When matching the location of goods path location records with storage location nodes, the proximity relationship between the path order value and the spatial location is used as the matching basis to assign goods nodes to the corresponding storage location nodes, forming a mapping and pairing result between goods nodes and storage location nodes. When the storage location area performs regional aggregation on the mapping and pairing results, the product nodes and storage location nodes mapped to the same storage location area are included in the same region, and the corresponding order witness hyperedge is associated with the product nodes within the same region, so that the node mapping relationship and hyperedge association relationship within the same storage location area are consistent, forming a regional storage location division record; The regional storage location division record is used to determine the initial correspondence range between product nodes and storage location nodes in the order witness hypergraph, and to establish the initial division relationship between product nodes and storage location nodes according to the mapping and pairing results in the regional storage location division record, so that the order witness hypergraph obtains the storage location division status before entering the node exchange search, forming the initial storage location division result.

[0023] In this embodiment, step four specifically includes: Extract the mapping and pairing results between product nodes and storage location nodes from the initial storage location division results. Determine the storage location node corresponding to each product node according to the mapping and pairing results. Then, determine the storage location area belonging to the product node by the storage location node. Extract the order witness hyperedge and the path position of the product node in the order witness hyperedge from the order witness hypergraph. Organize the storage location area belonging and path position according to the order witness hyperedge to form the hyperedge status record before the exchange. Select the goods node to be exchanged in the super-edge status record before the exchange, read the storage location node corresponding to the goods node to be exchanged, simulate the exchange of storage location nodes between the goods nodes to be exchanged, and redetermine the storage location area affiliation of the goods node to be exchanged based on the storage location node after the exchange simulation. Then, fill the redetermined storage location area affiliation back into the order witness super-edge where the goods node to be exchanged is located to form the super-edge status record after the exchange. Perform a region attribution comparison on the super-edge status records before and after the exchange, and count the number of warehouse locations crossed by the same order witness super-edge before and after the exchange. Divide the number of warehouse locations crossed after the exchange by the number of warehouse locations crossed before the exchange to form a split change record. Perform path position comparison on the super-edge status record before and after the exchange, read the warehouse area belonging to the adjacent goods node according to the path position in the order witness super-edge, count the difference in warehouse area span before and after the exchange to form a path span change record, and count the difference in the number of warehouse area back-turns before and after the exchange to form a path back-turn change record. The split change records, path span change records, and path return change records are aggregated according to the goods to be exchanged nodes, so that each goods to be exchanged corresponds to a set of exchange impact records consisting of split changes, path span changes, and path return changes, and the exchange impact records are configured with the tensor cut gain rules corresponding to the goods to be exchanged nodes.

[0024] In this embodiment, step five specifically includes: Extract product nodes and storage location nodes from the order witness supergraph, extract mapping pairing results from the initial storage location division results, expand the mapping pairing results by product nodes, determine the storage location node currently occupied by each product node, and organize the product nodes, storage location nodes and occupancy correspondence into storage location occupancy records; Select the goods node to be exchanged in the storage location occupancy record, determine the outgoing storage location node according to the occupancy correspondence of the goods node to be exchanged before the exchange, determine the incoming storage location node after the exchange according to the exchange direction of the goods node, and organize the goods node to be exchanged and the outgoing storage location node, and the goods node to be exchanged and the incoming storage location node into outgoing relationship and ingoing relationship respectively. According to the direction of goods node exchange, the migration-out relationship and migration-in relationship are matched end to end. The migration-in storage location node corresponding to one goods node to be exchanged is matched with the migration-out storage location node corresponding to another goods node to be exchanged in the adjacent exchange location. The migration-out relationship and migration-in relationship are connected according to the matching order to form a storage location node exchange chain. When performing closed matching on the storage location node exchange chain, the starting storage location node and the ending storage location node in the storage location node exchange chain are read, and the consistency comparison between the starting storage location node and the ending storage location node is performed. If the consistency comparison passes, the storage location node exchange chain is determined to be a closed exchange chain. When performing conservation checks on the number of goods nodes and storage locations occupied in a closed exchange chain, the number of goods nodes occupied before the exchange and after the exchange are counted separately. Similarly, the number of storage locations occupied before the exchange and after the exchange are counted separately. When both types of occupancy remain consistent, the closed exchange chain is configured as a conservation exchange ring rule.

[0025] In this embodiment, step six specifically includes: Extract the item nodes to be exchanged and their corresponding tensor cut gain values ​​from the tensor cut gain rules. The tensor cut gain values ​​are obtained from the split change records, path span change records, and path return change records corresponding to the item nodes to be exchanged. Arrange the item nodes to be exchanged in descending order of tensor cut gain values, so that the item nodes that have a greater improvement in the degree of order witness hyperedge splitting, path span, and path return are placed in the front position, thus forming the item node exchange order. The closed exchange chain is extracted from the conservation exchange ring rule. The closed exchange chain contains the outgoing and incoming storage locations corresponding to the goods nodes to be exchanged. The goods nodes to be exchanged in the closed exchange chain are located one by one according to the exchange order of the goods nodes. The arrangement position of the goods nodes to be exchanged in the exchange order of the goods nodes is matched with the link position in the closed exchange chain to form an exchange ring matching record. When performing dependency calibration on the nodes of goods to be exchanged in the exchange ring matching record, extract the outgoing storage location node and the incoming storage location node corresponding to each node of goods to be exchanged, and perform consistency sorting on the incoming storage location node of a node of goods to be exchanged and the outgoing storage location node of the adjacent node of goods to be exchanged to generate the storage location acceptance relationship between adjacent nodes of goods to be exchanged. Then sort outgoing storage location node, incoming storage location node and storage location acceptance relationship into exchange dependency relationship; According to the order of goods node exchange, push the exchange ring matching record into the exchange evidence stack, so that the stack position order of the exchange ring matching record in the exchange evidence stack is consistent with the order of goods node exchange. Write the tensor cut gain value, closed exchange chain and exchange dependency relationship corresponding to the same goods node to be exchanged into the same stack position, so that each stack position saves the evaluation information, constraint information and dependency information required for one node exchange, forming a stack position evidence record. The stack evidence records are sorted according to the push order to ensure that the stack positions in the exchange evidence stack correspond to the exchange order of the goods nodes to be exchanged. An index is established between adjacent stack positions according to the exchange dependency relationship so that the connection relationship between the warehouse entry node in the previous stack position and the warehouse exit node in the next stack position can be retrieved, thus forming the exchange evidence stack rules. The tensor cut gain rule is defined as the node exchange evaluation rule, which is used to evaluate the impact of the goods node to be exchanged on the exchange of the order witness superedge. The conserved exchange ring rule is defined as the node exchange constraint rule, which is used to restrict the goods node to be exchanged from forming a closed exchange chain. The exchange evidence stack rule is defined as the node exchange backtracking rule, which is used to record the exchange order and exchange dependency of goods nodes. The node exchange evaluation rule, node exchange constraint rule and node exchange backtracking rule together constitute the improved Kernighan-Lin algorithm.

[0026] This invention structurally improves the Kernighan-Lin algorithm through tensor cut gain rules, conserved exchange loop rules, and exchange evidence stack rules. It expands the single cut edge gain in traditional node exchange into a tensor cut gain characterized by splitting changes, path span changes, and path back changes, enabling node exchange evaluation to reflect the overall change of the order witness superedge in the operational path. It extends traditional two-node exchange into a closed exchange chain with outgoing and incoming warehouse nodes, ensuring conservation in the number of warehouse nodes and goods nodes occupied during the goods node exchange process. Finally, it expands the traditional exchange sequence into an exchange evidence stack with stack evidence records and acceptance indexes, allowing tensor cut gain values, closed exchange chains, and exchange dependencies to be reflected in the same exchange. The process is synchronously recorded and traced back, resulting in an improved Kernighan-Lin algorithm that prioritizes the nodes of goods to be exchanged in the order witness hypergraph that have significantly improved the degree of hyperedge splitting, path span, and path backtracking. This avoids the problem of goods being scattered within the same task due to warehouse location adjustments based solely on product popularity. At the same time, by restricting the inbound and outbound relationships of the nodes to be exchanged through closed exchange chains, it reduces the risk of exchange chain breaks, occupancy conflicts, and increased path backtracking during warehouse location adjustments. Furthermore, by storing node exchange evaluation information, constraint information, and dependency information through an exchange evidence stack, it provides a traceable basis for residual gain determination, thereby improving the path continuity, exchange executability, adjustment stability, and warehouse management efficiency of the warehouse location division results.

[0027] In this embodiment, step seven specifically includes: The improved Kernighan-Lin algorithm is called to process the order witness hypergraph and the initial storage location partitioning results. The product nodes and storage location nodes in the order witness hypergraph are organized according to the initial storage location partitioning results. The product nodes participating in the exchange search in the order witness hyperedge, the storage location nodes corresponding to the product nodes, and the order witness hyperedges where the product nodes are located are marked together, so that the node exchange search range keeps corresponding to the initial storage location partitioning results, forming a node exchange search state. The tensor cut gain rule is applied to the node exchange search state. The split change, path span change, and path return change corresponding to each item node to be exchanged are read. The split change, path span change, and path return change are converted into tensor cut gain values ​​according to the tensor cut gain rule. Then, the item nodes to be exchanged are arranged from largest to smallest according to the tensor cut gain values, so that the exchange priority of the item nodes to be exchanged forms an order record, forming a candidate node exchange record. The conservation exchange ring rule is applied to the candidate node exchange record. The storage location node occupied by the goods to be exchanged in the initial storage location division result is read as the outgoing storage location node. The storage location node that the goods to be exchanged intends to enter according to the candidate node exchange record is read as the incoming storage location node. The outgoing storage location node and the incoming storage location node are connected according to the relationship between the incoming storage location node and the corresponding outgoing storage location node of the next goods to be exchanged, forming a candidate closed exchange chain. Perform loop verification on the candidate closed exchange chain, read the starting and ending storage location nodes in the candidate closed exchange chain and perform consistency comparison, read the incoming and outgoing storage location nodes between adjacent goods to be exchanged and perform continuity comparison, organize the consistency results of the starting and ending storage location nodes and the continuity results of the storage location acceptance relationship according to the candidate closed exchange chain to form a loop verification record; The rules of the exchange evidence stack are applied to the loop verification record. The stack evidence record is retrieved according to the stack position order in the rules of the exchange evidence stack. The loop verification record is matched with the stack evidence record to ensure that the tensor cut gain value, candidate closed exchange chain and exchange dependency are consistent within the same item node to be exchanged. After the consistency verification is completed, a candidate conserved exchange loop is formed. Candidate conservation exchange rings are sorted according to their tensor cut gain values. The candidate conservation exchange ring with the largest tensor cut gain value is determined as the first-ranked candidate conservation exchange ring, and the first-ranked candidate conservation exchange ring is used as the target conservation exchange ring for adjusting the initial storage location division results.

[0028] In this embodiment, step eight specifically includes: Read the exchange order of goods nodes and the storage location acceptance relationship in the target conservation exchange loop. The exchange order of goods nodes is used to determine the order in which goods nodes participate in the adjustment, and the storage location acceptance relationship is used to determine the connection direction between the outgoing storage location node and the incoming storage location node. According to the exchange order of goods nodes, locate the mapping pairing results in the initial storage location division result item by item, and adjust the goods nodes from the outgoing storage location node to the corresponding incoming storage location node to complete the adjustment of the mapping pairing result; Write the adjusted mapping and pairing results back to the order witness hypergraph, so that the product nodes and storage location nodes in the order witness hypergraph are re-matched according to the adjusted mapping and pairing results. Update the association between product nodes and storage location nodes according to the task association range corresponding to the order witness hyperedge, so that the node connection status within the order witness hyperedge is consistent with the adjusted storage location node, thus forming the adjusted order witness hypergraph. Tensor cut gain, closed exchange chain, and exchange dependency are extracted from the objective conservation exchange ring. The tensor cut gain is used to record the gain result brought about by node exchange, the closed exchange chain is used to record the closed acceptance path between storage location nodes, and the exchange dependency is used to record the storage location acceptance order between adjacent goods nodes. The tensor cut gain, closed exchange chain, and exchange dependency are aggregated according to the objective conservation exchange ring to generate exchange evidence records, and the exchange evidence records are written into the exchange evidence stack. When performing residual calculations on the adjusted order witness hypergraph, the order witness hyperedges in the adjusted order witness hypergraph are read, and the residuals of splitting change, path span change, and path return change of the order witness hyperedges after the adjustment are calculated respectively. The residuals of splitting change, path span change, and path return change are then aggregated to form the residual gain determination result. The location partitioning status of the adjusted order witness hypergraph is determined based on the residual gain determination result. The location partitioning status is used to characterize the partitioning result of the goods node and the location node after the target conservation exchange loop adjustment. The location partitioning status and the exchange evidence record are linked and organized to output the warehouse management result.

[0029] A warehouse management system based on big data processing includes: The image generation module is used to collect warehouse-related data and perform state extraction and spatial mapping to form a warehouse state image; The hypergraph construction module is used to generate product nodes and storage location nodes, and to arrange product nodes associated with the same job task into order witness hyperedges according to the job path, forming an order witness hypergraph; The initial partitioning module is used to perform mapping and pairing of product nodes and storage location nodes and to aggregate them according to storage location areas to form the initial storage location partitioning results; The gain construction module is used to perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping, forming tensor cut gain rules; The exchange ring construction module is used to perform closed matching on the migration-in and migration-out relationships of storage location nodes to form a conserved exchange ring rule; The algorithm construction module is used to perform stack-style arrangement of the exchange order and exchange dependency of goods nodes, form the exchange evidence stack rules, and constitute the improved Kernighan-Lin algorithm; The target search module is used to perform node exchange search and loop verification in the order witness supergraph to form a target conservation exchange loop; The results output module is used to adjust the initial storage location allocation results according to the target conservation exchange loop and perform residual gain determination, and output the storage management results.

[0030] Example 1: To verify the feasibility of this invention in practice, it was applied to a warehouse location optimization management scenario in a large e-commerce warehousing center. The warehousing center is divided into an ambient temperature goods area, a fast-turnover goods area, and a full-case temporary storage area, with approximately 6,000 types of goods and an average daily outbound order volume of approximately 15,000 orders. The original warehouse location management method adopted a manual rule-based warehouse location scheme. This scheme mainly arranged goods according to the turnover rate, the distance from the warehouse location to the outbound gate, and manual experience, with high-frequency goods placed near the outbound gate and low-frequency goods placed in the far warehouse area. Although this scheme can meet the basic outbound requirements, in actual operation, the goods to be picked continuously within the same task are often scattered in different warehouse areas. Pickers need to frequently move across areas to complete one order, and some routes even involve backtracking. This results in a long average picking distance, many warehouse location adjustment conflicts, and difficulty in continuously judging the benefits after adjustment.

[0031] When applying this invention, the warehousing system collects warehousing-related data generated during the processes of receiving, shipping, inventory counting, picking, and storage location occupancy. It then performs state extraction and spatial mapping on this data to form a warehousing state mirror. The warehousing state mirror records the occupancy status of goods, storage locations, and operational flow, and maps storage location identifiers to warehousing spatial coordinates. The system generates goods nodes and storage location nodes in the warehousing state mirror. Goods nodes picked consecutively in the same task are arranged along the operational path as order witness hyperedges, forming an order witness hypergraph. These order witness hyperedges preserve the order of goods nodes along the operational path, enabling the system to recognize order-level path continuity, rather than relying solely on pairwise co-occurrence relationships between goods.

[0032] After forming the order witness hypergraph, the system performs mapping and pairing on product nodes and storage location nodes, and aggregates the mapping and pairing results by storage location region to form the initial storage location allocation result. The system selects product nodes to be exchanged, and performs differential calculations on the splitting changes, path span changes, and path return changes of the order witness hyperedge before and after node exchange to form tensor cut gain rules. For product nodes with exchange needs, the system performs closed matching based on the migration in and migration out relationships of storage location nodes to form a conserved exchange loop rule, avoiding non-conservation of storage location occupancy after product node migration. The system forms the product node exchange order according to the tensor cut gain value, pushes the closed exchange chain and exchange dependency relationship into the exchange evidence stack to form the exchange evidence stack rule, and the tensor cut gain rule, the conserved exchange loop rule, and the exchange evidence stack rule constitute the improved Kernighan-Lin algorithm. This algorithm performs node exchange search and loop verification in the order witness hypergraph, filters the target conserved exchange loop, and adjusts the initial storage location allocation result according to the target conserved exchange loop. After the adjustment is completed, the system performs residual gain determination on the adjusted order witness supermap and outputs the warehouse management results.

[0033] The comparative scheme uses a manual rule-based warehouse location system, which does not construct an order witness hypergraph, perform conservation exchange loop verification, or record the exchange evidence stack. The present invention uses warehouse status mirroring, an order witness hypergraph, and an improved Kernighan-Lin algorithm for warehouse location adjustment. The two schemes are compared under the same order size, warehouse layout, and personnel and equipment conditions, as shown in Table 1: Table 1 Comparison of Storage Location Optimization Operations

[0034] As shown in Table 1, the present invention reduces the average picking distance per order from 284.6m to 221.8m, a reduction of approximately 22.07%; reduces the average picking time per order from 9.4min to 7.1min, a reduction of approximately 24.47%; reduces the proportion of cross-warehouse orders by 12.9 percentage points and the proportion of path-reversal orders by 9.3 percentage points, indicating that the order witness superedge can more accurately express the path continuity relationship of goods nodes within the same task. The warehouse location adjustment failure rate decreases from 8.6% to 2.1%, indicating that the conservation exchange ring rule can reduce adjustment failures caused by non-closed warehouse location node migration relationships. The average daily completed order volume increases from 15,120 orders to 17,840 orders, an increase of approximately 17.99%, indicating that the tensor cut gain rule, the conservation exchange ring rule, and the exchange evidence stack rule work together to improve the stability of warehouse location adjustment results and warehouse operation efficiency.

[0035] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A warehouse management method based on big data processing, characterized in that, Includes the following steps: Step 1: Collect warehouse-related data, perform state extraction and spatial mapping on the warehouse-related data to form a warehouse state mirror; Step 2: Generate product nodes and storage location nodes in the warehouse status mirror, and arrange product nodes associated with the same task into order witness hyperedges according to the operation path to form an order witness hypergraph; Step 3: Perform mapping and pairing on the product nodes and storage location nodes in the order witness supermap, and aggregate the mapping and pairing results according to the storage location area to form the initial storage location division results; Step 4: Perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping to form tensor cut gain rules; Step 5: Perform closed-loop matching on the migration-in and migration-out relationships of the storage location nodes to form a conserved exchange loop rule; Step 6: Perform stack-style arrangement on the exchange order and exchange dependencies of the goods nodes to form the exchange evidence stack rules, and the improved Kernighan-Lin algorithm is composed of the tensor cut gain rule, the conservation exchange ring rule, and the exchange evidence stack rules. Step 7: Use the improved Kernighan-Lin algorithm to perform node swap search and loop verification in the order witness hypergraph to form a target-conserving swap cycle; Step 8: Adjust the initial warehouse location allocation results according to the objective conservation exchange loop, perform residual gain determination on the adjusted order witness hypergraph, and output the warehouse management results.

2. The warehouse management method based on big data processing according to claim 1, characterized in that, Step one specifically involves: Collect warehouse-related data, establish a collection record set according to data source and collection time, and perform unified coding on the product identifier, storage location identifier and operation task identifier in the collection record set to form an identifier-unified record set; The identifier-normalized record set is sorted by collection time, and the sorted identifier-normalized record set is merged by task identifier to form a time-series associated record set. Extract the goods occupancy status, storage location occupancy status, and work flow status from the time-series associated record set, and establish a status arrangement relationship according to the collection time to form a status extraction record set; Map the location identifiers in the status extraction record set to warehouse space coordinates, and bind the goods occupancy status, location occupancy status, and operation flow status to the warehouse space coordinates to form a space mapping record set; The spatial mapping record set is mirrored according to the collection time and storage space coordinates to form a storage status mirror.

3. The warehouse management method based on big data processing according to claim 1, characterized in that, Step two specifically involves: Extract the product occupancy status and storage location occupancy status from the warehouse status image, generate product nodes and storage location nodes respectively, and establish the occupancy mapping relationship between product nodes and storage location nodes; Extract the task identifier and task path relationship from the warehouse status image, and arrange the goods nodes associated with the same task according to the execution order of the task path relationship to form a task goods node sequence; Connect the goods nodes in the task goods node sequence to form an order witness superedge, and associate the order witness superedge with the occupancy mapping relationship to form a superedge association record; Arrange the execution graph structure of product nodes, storage location nodes, occupancy mapping relationships, order witness hyperedges, and hyperedge association records to form the order witness hypergraph.

4. The warehouse management method based on big data processing according to claim 1, characterized in that, Step three specifically involves: Extract product nodes, storage location nodes, and order witness hyperedges from the order witness hypergraph, and establish product path location records according to the path position of the product nodes in the order witness hyperedges; Extract the storage area and spatial location of the storage location node, match the goods path location record with the execution location of the storage location node, and form a mapping pairing result between the goods node and the storage location node; Based on the storage location area, the mapping and pairing results are aggregated by region, and the product nodes, storage location nodes and order witness superedges within the same storage location area are associated and organized to form a regional storage location division record; Based on the regional storage location division records, an initial division relationship is established between the product nodes and storage location nodes in the order witness supermap, forming the initial storage location division result.

5. The warehouse management method based on big data processing according to claim 1, characterized in that, Step four specifically involves: Extract the mapping and pairing results between the goods node and the storage location node from the initial storage location division results, and extract the order witness hyperedge and the path position of the goods node in the order witness hyperedge from the order witness hypergraph to form the hyperedge state record before the exchange. Select the goods node to be exchanged in the super-edge status record before the exchange, perform the exchange simulation on the storage location node corresponding to the goods node to be exchanged, and fill the storage location area after the exchange simulation back into the order witness super-edge to form the super-edge status record after the exchange. Perform a region attribution comparison on the super-edge status records before and after the exchange, and count the changes in the number of warehouse location regions crossed by the super-edge for the same order to form a split change record; Compare the path positions of the super-edge status records before and after the exchange, and count the changes in the warehouse location span and the number of path reversals of the same order's witness super-edge in the operation path to form path span change records and path reversal change records; The split change records, path span change records, and path return change records are aggregated according to the nodes of the goods to be exchanged, and the aggregation results are configured as the tensor cut gain rules corresponding to the nodes of the goods to be exchanged.

6. The warehouse management method based on big data processing according to claim 1, characterized in that, Step five specifically involves: Extract product nodes, storage location nodes, and mapping pairing results from the order witness hypermap and initial storage location division results. Determine the storage location nodes occupied by the product nodes according to the mapping pairing results to form storage location occupancy records. Select the goods node to be exchanged in the storage location occupancy record, mark the storage location node occupied by the goods node to be exchanged before the exchange as the moving-out storage location node, and mark the storage location node to which the goods node to be exchanged enters after the exchange as the moving-in storage location node, thus forming the moving-out relationship and the moving-in relationship. According to the direction of goods node exchange, the migration-out relationship and migration-in relationship are matched end to end, so that the previous migration-in warehouse location node is consistent with the next migration-out warehouse location node, forming a warehouse location node exchange chain; Perform closed matching on the storage location node exchange chain. When the starting storage location node and the ending storage location node of the exchange chain are the same, a closed exchange chain is formed. Perform conservation checks on the quantity of goods nodes and storage locations occupied in the closed exchange chain, and configure the closed exchange chain that passes the conservation check as a conservation exchange ring rule.

7. The warehouse management method based on big data processing according to claim 1, characterized in that, Step six specifically involves: Extract the item nodes to be exchanged and the tensor cut gain value from the tensor cut gain rule, and sort the item nodes to be exchanged in descending order according to the tensor cut gain value to form the item node exchange order; Extract the closed exchange chain from the conservation exchange ring rule, and perform position matching on the goods nodes to be exchanged in the closed exchange chain according to the exchange order of the goods nodes to form an exchange ring matching record; Dependency labeling is performed on the nodes of goods to be exchanged in the exchange ring matching record, and the storage location acceptance relationship between the outgoing storage location node, the incoming storage location node and the adjacent outgoing storage location node of the node to be exchanged is organized into exchange dependency relationship; According to the order of goods node exchange, push the exchange ring matching record into the exchange evidence stack, and write the tensor cut gain value, closed exchange chain and exchange dependency relationship into the same stack position to form the stack position evidence record. The stack evidence records are sorted according to the push order, and the connection index between adjacent stack positions is established according to the exchange dependency relationship to form the exchange evidence stack rules. The tensor cut gain rule is determined as the node exchange evaluation rule, the conserved exchange loop rule is determined as the node exchange constraint rule, and the exchange evidence stack rule is determined as the node exchange backtracking rule, thus forming the improved Kernighan-Lin algorithm.

8. The warehouse management method based on big data processing according to claim 1, characterized in that, Step seven specifically involves: The improved Kernighan-Lin algorithm is called to process the order witness hypergraph and the initial storage location partitioning results, and the product nodes, storage location nodes and order witness hyperedges participating in the node exchange search are marked in the order witness hypergraph to form the node exchange search state; The tensor cut gain rule is applied to the node exchange search state. The tensor cut gain value corresponding to the goods node to be exchanged is read, and the goods nodes to be exchanged are sorted in descending order according to the tensor cut gain value to form candidate node exchange records. Apply the conservation exchange ring rule to the candidate node exchange record, read the outgoing and incoming warehouse locations corresponding to the goods to be exchanged, and connect the outgoing and incoming warehouse locations according to the warehouse location acceptance relationship to form a candidate closed exchange chain. Perform loop verification on candidate closed exchange chains, compare the consistency of the starting and ending storage locations, and compare the continuity of the storage location acceptance relationship between adjacent goods to be exchanged nodes to form a loop verification record; The rules for exchanging evidence stacks are applied to the loop verification record. The loop verification record is matched with the stack evidence record, and consistency checks are performed on the tensor cut gain value, candidate closed exchange chain, and exchange dependency to form a candidate conserved exchange loop. The candidate conserved commutative rings are sorted according to their tensor cut gain values, and the candidate conserved commutative ring with the highest ranking is selected to form the target conserved commutative ring.

9. A warehouse management method based on big data processing according to claim 1, characterized in that, Step eight specifically involves: Read the exchange order of goods nodes and the storage location acceptance relationship in the target conservation exchange ring, and adjust the mapping and pairing results in the initial storage location division result according to the exchange order of goods nodes; Write the adjusted mapping pairing results back to the order witness hypergraph and update the product nodes and storage location nodes associated with the order witness hyperedge to form the adjusted order witness hypergraph; Extract the tensor cut gain, closed commutation chain, and commutation dependency from the target-conserving commutation ring, generate commutation evidence records, and write the commutation evidence records into the commutation evidence stack. The residual calculations for the splitting changes, path span changes, and path return changes of the adjusted order witness hypergraph are performed to form the residual gain determination results; The location allocation status of the adjusted order witness supermap is determined based on the residual gain determination results, and the warehouse management results containing the location allocation status and exchange evidence records are output.

10. A warehouse management system based on big data processing, executing the warehouse management method based on big data processing as described in any one of claims 1 to 9, characterized in that, include: The image generation module is used to collect warehouse-related data and perform state extraction and spatial mapping to form a warehouse state image; The hypergraph construction module is used to generate product nodes and storage location nodes, and to arrange product nodes associated with the same job task into order witness hyperedges according to the job path, forming an order witness hypergraph; The initial partitioning module is used to perform mapping and pairing of product nodes and storage location nodes and to aggregate them according to storage location areas to form the initial storage location partitioning results; The gain construction module is used to perform differential calculations on the splitting changes, path span changes, and path return changes of the order witness superedge before and after node swapping, forming tensor cut gain rules; The exchange ring construction module is used to perform closed matching on the migration-in and migration-out relationships of storage location nodes to form a conserved exchange ring rule; The algorithm construction module is used to perform stack-style arrangement of the exchange order and exchange dependency of goods nodes, form the exchange evidence stack rules, and constitute the improved Kernighan-Lin algorithm; The target search module is used to perform node exchange search and loop verification in the order witness supergraph to form a target conservation exchange loop; The results output module is used to adjust the initial storage location allocation results according to the target conservation exchange loop and perform residual gain determination, and output the storage management results.

Images