Digital warehouse management method and system based on intelligent sorting

Abstract

The invention provides a digital warehouse management method and system based on intelligent sorting, and belongs to the technical field of data management. The method comprises the steps of obtaining real-time data of an external supply chain and an operation log of an internal warehouse to position an information lagging area, and determining a resource allocation matrix based on the information lagging area, so that resource waste is avoided, and the utilization rate is improved. The inventory state is quantified according to the inventory turnover rate and the inventory holding days, so that inventory overstock or inventory stockout early warning is triggered, and an inventory risk probability distribution diagram is generated accordingly. The model of the target bearing is accurately identified through the convolutional neural network and the radio frequency identification algorithm, and the position coordinates of the target bearing are obtained to plan the sorting path, so that the sorting efficiency and safety of automatic equipment are improved. The system achieves the internal and external deep collaboration of bearing storage, improves the management precision, efficiency and intelligent level, reduces the inventory risk, enhances the supply chain response capability, meets the demands of the modern manufacturing industry, and supports the intelligent upgrading of the bearing industry.

Description

Technical Field

[0001] This disclosure relates to the field of data management technology, specifically to a digital warehouse management method and system based on intelligent sorting. Background Technology

[0002] Currently, bearings, as core components of mechanical equipment, have their warehousing management directly impacting the stability and efficiency of the manufacturing supply chain. With the transformation of manufacturing towards intelligent manufacturing, bearing warehousing management faces unprecedented challenges and opportunities. Existing warehousing management methods often separate internal warehousing operations from external supply chain links, leading to delayed supply and demand information transmission and unreasonable resource allocation. This separated management model prevents the warehousing system from adjusting internal operational strategies in a timely manner according to changes in external demand, resulting in inventory backlogs or stockout risks. Furthermore, the lack of end-to-end collaboration mechanisms significantly reduces the responsiveness of the entire supply chain, thus failing to meet the stringent requirements of modern manufacturing for precision, speed, and cost control.

[0003] The information disclosed in the background section is only for enhancing the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] In view of this, this disclosure provides a digital warehouse management method and system based on intelligent sorting, which can improve the accuracy, efficiency and intelligence level of bearing warehouse management.

[0005] In a first aspect, embodiments of this application provide a digital warehouse management method based on intelligent sorting. The method includes: acquiring real-time demand data from an external supply chain and operation log data from an internal bearing warehouse, wherein the demand data includes order demand data, inventory change data, and delivery plan data, and the operation log data includes inbound data, outbound data, picking data, and packaging data; determining an information lag region between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and the operation log data; determining a resource allocation matrix for the internal bearing warehouse based on the information lag region; and determining inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix, wherein the inventory level evaluation indicators include inventory turnover rate. The system involves: determining the inventory holding days; triggering inventory warnings based on the inventory level evaluation indicators, including inventory backlog warnings and inventory shortage warnings; determining the inventory risk probability distribution map for different bearing models based on the inventory warnings; determining the inventory risk level based on the inventory risk probability distribution map, including high-risk, medium-risk, and low-risk levels; obtaining the set of bearing models corresponding to the high-risk level from the bearing specification database, and determining the target bearing model based on the set of bearing models using convolutional neural network and radio frequency identification algorithms; obtaining the location coordinates of the target bearing model, and determining a path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model.

[0006] Secondly, embodiments of this application provide a digital warehouse management system based on intelligent sorting. This system includes: an acquisition module, a first determination module, a second determination module, a third determination module, an early warning module, a fourth determination module, a fifth determination module, a sixth determination module, and a seventh determination module. The acquisition module is used to acquire real-time demand data from an external supply chain and operation log data from an internal bearing warehouse. The demand data includes order demand data, inventory change data, and delivery plan data. The operation log data includes inbound data, outbound data, picking data, and packaging data. The first determination module is used to determine the information lag region between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and the operation log data. The second determination module is used to determine the resource allocation matrix of the internal bearing warehouse based on the information lag region. The third determination module is used to determine the inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix. The inventory level evaluation indicators include inventory turnover rate and inventory holding days. The early warning module is used to determine the inventory level based on the inventory level... The evaluation indicators trigger inventory warnings, including inventory backlog warnings and inventory shortage warnings; the fourth determination module is used to determine the inventory risk probability distribution map of different bearing models based on the inventory warnings; the fifth determination module is used to determine the inventory risk level based on the inventory risk probability distribution map, including high-risk, medium-risk, and low-risk levels; the sixth determination module is used to obtain the set of bearing models corresponding to the high-risk level in the bearing specification database, and determine the target bearing model based on the set of bearing models using convolutional neural network algorithms and radio frequency identification algorithms; the seventh determination module is used to obtain the location coordinates of the target bearing model, and determine the path planning map from the starting point of the internal bearing warehouse to the target bearing model based on the location coordinates of the target bearing model.

[0007] This application provides a digital warehouse management method and system based on intelligent sorting. By acquiring real-time demand data from the external supply chain, including order demand data, inventory change data, and delivery plan data, and operational log data from the internal bearing warehouse, including inbound data, outbound data, picking data, and packaging data, it breaks down the data silo between internal operations and the external supply chain in traditional warehouse management, laying a comprehensive data foundation for subsequent internal and external process collaboration. Through analysis of the aforementioned demand data and operational log data, it accurately locates the information lag area between external supply chain demand and internal bearing warehouse response, effectively solving the problem of delayed supply and demand information transmission in existing warehouse management models, and providing a precise optimization direction for subsequent resource optimization. Based on the information lag area, it determines the resource allocation matrix for the internal bearing warehouse, enabling targeted adjustments to warehouse resource allocation, avoiding the unreasonable resource allocation problems caused by traditional separate management, and improving resource utilization efficiency. Based on the resource allocation matrix, using inventory turnover rate and inventory holding days as clear inventory level evaluation indicators, it can accurately quantify the inventory status of different bearing models, providing a basis for inventory management. Based on the inventory level evaluation indicators, it triggers inventory backlog warnings and inventory... The stockout warning mechanism can promptly detect inventory anomalies, preventing the risk of inventory backlog or stockouts from escalating due to failure to detect anomalies in time. Based on the inventory warning, it determines the probability distribution of inventory risk for different bearing models and further classifies them into high-risk, medium-risk, and low-risk levels. This accurately distinguishes the inventory risk level of different bearing models, facilitating key management of high-risk bearing models. It filters the bearing model set corresponding to the high-risk inventory level from the bearing specification database and uses a combination of convolutional neural network and radio frequency identification algorithms to accurately identify the target bearing model. This effectively solves the problems of insufficient accuracy of single identification technology due to the large number of bearing models and similar specifications, and the complexity of data processing through multi-technology fusion, achieving high-precision identification of high-risk bearing models. By obtaining the location coordinates of the target bearing model and determining the path planning map from the internal bearing storage starting point to the target bearing model, it provides precise sorting path guidance for automated warehouse equipment, overcoming the shortcomings of traditional static path algorithms in coping with dynamic environmental changes, and improving the sorting efficiency and operational safety of automated equipment. This enables deep collaboration between internal and external processes in bearing warehousing, significantly improving the accuracy, efficiency, and intelligence of warehousing management, reducing inventory backlog and stockout risks, thereby enhancing the responsiveness of the entire supply chain. It can meet the stringent requirements of modern manufacturing for bearing warehousing management in terms of bearing model identification accuracy, internal and external interaction speed, and management cost control, providing effective support for the intelligent upgrading of the bearing industry. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments or conventional technologies of this disclosure, the accompanying drawings used in the description of the embodiments or conventional technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting provided in an exemplary embodiment of this application.

[0010] Figure 2 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0011] Figure 3 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0012] Figure 4 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0013] Figure 5 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0014] Figure 6 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0015] Figure 7 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0016] Figure 8 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application.

[0017] Figure 9 This is a flowchart illustrating a digital warehouse management method based on intelligent sorting, provided in another exemplary embodiment of this application. Detailed Implementation

[0018] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are set forth to give a full understanding of embodiments of this disclosure.

[0019] The terms “a,” “one,” and “the” are used to indicate the existence of one or more elements / components / etc.; the terms “including” and “having” are used to indicate an open-ended inclusion and that other elements / components / etc. may exist in addition to those listed. The terms “first” and “second” are used only as markers and are not a limitation on the number of objects.

[0020] Currently, bearings, as core components of mechanical equipment, have their warehousing management directly impacting the stability and efficiency of the manufacturing supply chain. With the transformation of manufacturing towards intelligent manufacturing, bearing warehousing management faces unprecedented challenges and opportunities. Existing warehousing management methods often separate internal warehousing operations from external supply chain links, leading to delayed supply and demand information transmission and unreasonable resource allocation. This separated management model prevents the warehousing system from adjusting internal operational strategies in a timely manner according to changes in external demand, resulting in inventory backlogs or stockout risks. Furthermore, the lack of end-to-end collaboration mechanisms significantly reduces the responsiveness of the entire supply chain, thus failing to meet the stringent requirements of modern manufacturing for precision, speed, and cost control.

[0021] For example, bearing products come in a wide variety of models with similar specifications, making it difficult to guarantee accuracy using only a single identification technology. Furthermore, the integration of multiple identification technologies introduces complex data processing issues. When the identification system cannot accurately determine the bearing model and location, subsequent path planning for the target bearing model loses its foundational support. Simultaneously, optimal path planning in dynamic environments requires real-time responses to changes in bearing inventory, equipment status, and task priorities, demanding that the system possess autonomous learning and decision-making capabilities. For instance, when congestion occurs in a shelf area, automated equipment in the warehouse system needs to replan its sorting routes in real time. However, traditional static path algorithms cannot handle such dynamic changes, leading to inefficiencies and even collisions in the automated equipment.

[0022] Therefore, how to build a digital warehouse management system for bearings that can achieve deep collaboration between internal and external processes, while possessing high-precision identification and intelligent path optimization capabilities, has become a technical problem that needs to be solved to promote the intelligent upgrading of the bearing industry.

[0023] This disclosure provides a digital warehouse management method based on intelligent sorting, such as... Figure 1 The illustrated method is a digital warehouse management approach based on intelligent sorting. This method may include the following steps:

[0024] Step S110: Obtain real-time demand data from the external supply chain and operation log data from the internal bearing warehouse. The demand data includes order demand data, inventory change data, and delivery plan data. The operation log data includes inbound data, outbound data, picking data, and packaging data.

[0025] Step S120: Determine the information lag area between demand from the external supply chain and the response from the internal bearing warehouse based on demand data and operation log data;

[0026] Step S130: Determine the resource allocation matrix for the internal bearing storage based on the information lag area;

[0027] Step S140: Determine the inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix. The inventory level evaluation indicators include inventory turnover rate and inventory holding days.

[0028] Step S150: Trigger inventory alerts based on inventory level evaluation indicators. Inventory alerts include inventory backlog alerts and inventory shortage alerts.

[0029] Step S160: Determine the inventory risk probability distribution map for different bearing models based on the inventory warning;

[0030] Step S170: Determine the inventory risk level based on the inventory risk probability distribution map. The inventory risk level includes high-risk, medium-risk, and low-risk levels.

[0031] Step S180: In the bearing specification database, obtain the set of bearing models corresponding to the high-risk level of the inventory, and determine the target bearing model based on the set of bearing models using convolutional neural network algorithm and radio frequency identification algorithm;

[0032] Step S190: Obtain the location coordinates of the target bearing model, and determine the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model.

[0033] According to the intelligent sorting-based digital warehouse management method provided in this disclosure, the method can acquire real-time demand data from the external supply chain and operational log data from the internal bearing warehouse. The demand data includes order demand data, inventory change data, and delivery plan data, while the operational log data includes inbound data, outbound data, picking data, and packaging data. Based on the demand data and operational log data, the method determines the information lag area between the demand from the external supply chain and the response from the internal bearing warehouse. Based on the information lag area, it determines the resource allocation matrix for the internal bearing warehouse. Based on the resource allocation matrix, it determines inventory level evaluation indicators for different bearing models within the internal bearing warehouse. These inventory level evaluation indicators include inventory turnover rate and inventory level... The inventory is held for several days; inventory warnings are triggered based on inventory level evaluation indicators, including inventory backlog warnings and inventory shortage warnings; inventory risk probability distribution maps for different bearing models are determined based on inventory warnings; inventory risk levels are determined based on the inventory risk probability distribution maps, including high-risk, medium-risk, and low-risk levels; in the bearing specification database, a set of bearing models corresponding to the high-risk level is obtained, and the target bearing model is determined based on the set of bearing models using convolutional neural network algorithms and radio frequency identification algorithms; the location coordinates of the target bearing model are obtained, and a path planning map from the starting point of the internal bearing storage to the target bearing model is determined based on the location coordinates of the target bearing model.

[0034] This method, by acquiring real-time demand data from the external supply chain (including order demand data, inventory change data, and delivery plan data) and operational log data from the internal bearing warehouse (including inbound data, outbound data, picking data, and packaging data), breaks down the data silo between internal operations and the external supply chain in traditional warehouse management, laying a comprehensive data foundation for subsequent internal and external process collaboration. Analysis of the demand data and operational log data accurately identifies the information lag area between external supply chain demand and internal bearing warehouse response, effectively solving the problem of delayed supply and demand information transmission under the existing warehouse management model, and providing a precise optimization direction for subsequent resource optimization. Determining the internal bearing warehouse resource allocation matrix based on the information lag area allows for targeted adjustments to warehouse resource allocation, avoiding the unreasonable resource allocation problems caused by traditional separate management and improving resource utilization efficiency. Based on the resource allocation matrix, using inventory turnover rate and inventory holding days as clear inventory level evaluation indicators, the inventory status of different bearing models can be accurately quantified, providing a basis for inventory management. Early warnings for inventory backlog can be triggered based on the inventory level evaluation indicators. The inventory early warning mechanism, which includes stockout alerts, can promptly detect inventory anomalies, preventing the risk of inventory backlog or stockouts from escalating due to failure to detect anomalies in a timely manner. Based on inventory early warnings, it determines the probability distribution of inventory risk for different bearing models and further classifies them into high-risk, medium-risk, and low-risk levels. This allows for precise differentiation of the inventory risk level of different bearing models, facilitating focused management of high-risk bearing models. The mechanism filters the bearing model set corresponding to the high-risk inventory level from the bearing specification database and uses a combination of convolutional neural network and radio frequency identification algorithms to accurately identify the target bearing model. This effectively solves the problems of insufficient accuracy of single identification technologies due to the large number of bearing models and similar specifications, as well as the complexity of data processing through multi-technology fusion, achieving high-precision identification of high-risk bearing models. By obtaining the location coordinates of the target bearing model and determining the path planning map from the internal bearing storage starting point to the target bearing model, it provides precise sorting path guidance for automated warehouse equipment. This overcomes the shortcomings of traditional static path algorithms in coping with dynamic environmental changes, improving the sorting efficiency and operational safety of automated equipment. This enables deep collaboration between internal and external processes in bearing warehousing, significantly improving the accuracy, efficiency, and intelligence of warehousing management, reducing inventory backlog and stockout risks, thereby enhancing the responsiveness of the entire supply chain. It can meet the stringent requirements of modern manufacturing for bearing warehousing management in terms of bearing model identification accuracy, internal and external interaction speed, and management cost control, providing effective support for the intelligent upgrading of the bearing industry.

[0035] The following is a detailed description of each step of the digital warehouse management method based on intelligent sorting provided in this disclosure:

[0036] In one embodiment of this disclosure, step S110 involves acquiring real-time demand data from the external supply chain and operational log data from the internal bearing warehousing. The demand data includes order demand data, inventory change data, and delivery plan data. The operational log data includes inbound data, outbound data, picking data, and packaging data. Specifically, the real-time demand data from the external supply chain reflects the dynamic demand for bearings from the market and upstream links. Real-time data interaction with upstream supplier systems can be achieved through an API interface, with a collection frequency set to once every 5 minutes to ensure data timeliness. The order demand data can include key information such as the bearing order quantity, order priority, bearing model and specifications (e.g., 6205-2RS, 6305-2RS), and delivery deadline submitted by upstream manufacturers or downstream customers. For example, when an automotive parts manufacturer submits an order for 300 sets of 6205-2RS bearings with a delivery deadline of 72 hours after production line expansion, the API interface is used to capture the order's model, quantity, and delivery time data in real time, generating standardized order records and monitoring order response efficiency.

[0037] It should be noted that 6205-2RS, 6305-2RS, etc. are common models of deep groove ball bearings. Their model naming follows the common code rules in the rolling bearing industry, with each character corresponding to the bearing type, size series, inner diameter, and sealing form.

[0038] Inventory change data can include real-time inventory levels of bearings from upstream suppliers or transit warehouses, reasons for inventory increases or decreases (such as replenishment or transfer), and inventory warning status (such as below the safety stock threshold). For example, if upstream supplier A's inventory of 6005-2RS bearings decreases from 500 sets to 120 sets, the system collects this inventory change data in real time via API, labels the reasons for the inventory decrease, and simultaneously records the timestamp of the inventory change, providing a basis for subsequent analysis of supply and demand matching.

[0039] Delivery plan data can include information such as delivery batches of bearings from upstream suppliers to internal warehouses, delivery vehicle information, estimated arrival time, transportation routes, and transit status. For example, Supplier B generates a delivery plan for 200 sets of 6305-2RS bearings, showing the delivery batches. The estimated arrival time for the third batch is 9:00 AM the next day; the transportation route is from the Industrial Park Road to the main road of the warehouse center; the current transit status is "already left the factory." The system uses an API interface to capture all dimensions of the delivery plan data in real time, ensuring that internal warehouses can plan their warehousing preparations in advance (such as reserving storage space and arranging quality inspection personnel).

[0040] The internal bearing warehousing operation log is a dynamic ledger recording the entire internal warehousing operation process. It is automatically collected and timestamped by the internal warehousing management system to ensure the traceability of operation records. The inbound data includes information on the entire process after the bearings enter the warehouse, including receiving, quality inspection, location allocation, and inbound confirmation. This includes the inbound timestamp, bearing model (e.g., 6205-2RS), quantity, quality inspection results, assigned storage location number (e.g., shelf number 2 on the 3rd floor of area A), and operator number. For example, for the 200 sets of 6305-2RS bearings delivered by supplier B, the system records the inbound timestamp at 9:15 the next day: 202X-XX-XX 09:15, model: 6305-2RS, quantity: 200 sets, quality inspection results: 198 sets qualified, 2 sets unqualified (surface scratches), qualified storage location: shelf number 2 on the 3rd floor of area A, operator: 001, forming a complete inbound operation log.

[0041] Outbound data refers to the process information regarding the retrieval, verification, and confirmation of bearings from warehouse locations based on order requirements. This includes the outbound timestamp, associated order number, bearing model, outbound quantity, outbound location, verification result, and the receiving personnel's (e.g., logistics personnel) number. For example, when receiving an outbound order for 100 sets of 6205-2RS bearings, the outbound timestamp after completion would be recorded as: 202X-XX-XX 14:30, order number: DD202X001, model: 6205-2RS, quantity: 100 sets, outbound location: Shelf 5, 2nd floor, Zone B.

[0042] Picking data records the operational information of picking bearings from designated locations based on outbound orders. This includes picking timestamps, picking task numbers, target bearing models, target locations, picking quantities, picking time, and picking equipment (such as AGV robots) numbers. For example, a picking task (number JH202X008) targets 50 sets of 6005-2RS bearings. The system records the following picking timestamps: 202X-XX-XX 11:00-11:28, model: 6005-2RS, location: shelf 3, floor 1, zone C, quantity: 50 sets, picking time: 28 minutes, and equipment: AGV003.

[0043] Packaging data records information such as packaging specifications, packaging time, packaging quality inspection, and related orders before the bearings leave the warehouse. This includes packaging timestamp, packaging task number, bearing model, packaging quantity, packaging specifications (e.g., rust-proof cardboard box + foam cushioning), packaging quality inspection results (e.g., sealing qualified, labels correctly affixed), and related outbound order number. For example, for the aforementioned outbound order of 100 sets of 6205-2RS bearings, the system records the following: Packaging timestamp: 202X-XX-XX 14:15-14:25, Task number: BZ202X001, Model: 6205-2RS, Quantity: 100 sets, Specifications: Rust-proof cardboard box (10 sets per box), Quality inspection results: Sealing qualified, labels correct, Related order: DD202X001.

[0044] In one embodiment of this disclosure, step S120, which determines the information lag region between demand from the external supply chain and the response from the internal bearing warehouse based on demand data and operation log data, further includes the following steps: Figure 2 As shown, the specific content is as follows:

[0045] Step S210: Based on the time series analysis algorithm, obtain the update time of the demand data and the recording time of the operation log respectively;

[0046] Step S220: Determine whether the time difference between the update time of the demand data and the recording time of the operation log exceeds the preset deviation threshold;

[0047] Step S230: If it is determined that the preset deviation threshold is exceeded, it is determined that there is an information lag between the demand data and the operation log data from the demand of the external supply chain to the response of the internal bearing warehouse, and the information lag node is identified.

[0048] Step S240: Determine the information lag area based on the information lag node.

[0049] Specifically, taking a bearing warehousing scenario on May 20, 2024 as an example, the upstream supplier system updated the order demand data for 6205-2RS bearings (the order quantity increased from 200 sets to 300 sets) through the API (Application Programming Interface) at 09:10 on May 20, 2024. The update time of this demand data was accurately captured and recorded by the time series analysis algorithm. In response to this change in order demand, the internal warehouse management system needs to initiate a picking operation. After the AGV equipment completes the picking of 300 sets of 6205-2RS bearings, the system automatically generates an operation log, recording the completion time of the picking operation as 11:05 on May 20, 2024. Meanwhile, the supplier updated the delivery plan data for the 6305-2RS bearing at 14:30 on 2024-05-20 (the original estimated arrival time at 10:00 on 2024-05-21 was postponed to 15:00 on 2024-05-21). The internal warehouse needs to adjust the reserved storage location for this model of bearing accordingly. After the storage location adjustment operation is completed, the operation log will be recorded at 15:10 on 2024-05-20.

[0050] It should be noted that the preset deviation thresholds can include preset deviation thresholds for information transmission and preset deviation thresholds for operation execution. The preset deviation threshold for information transmission addresses the cross-system delay from external demand data to the internal warehouse response, and is set at 1.5 hours. The preset deviation threshold for operation execution addresses the delay in internal warehouse operations, and is set at 15 minutes.

[0051] For example, in the 6205-2RS bearing scenario, the time difference between the demand data update time (2024-05-20-09:10) and the picking operation record time (2024-05-20-11:05) is 1 hour and 55 minutes, exceeding the preset deviation threshold of 1.5 hours for information transmission. In the 6305-2RS bearing scenario, the time difference between the demand data update time (2024-05-20-01:30) and the location adjustment operation record time (2024-05-20-15:10) is 40 minutes, exceeding the preset deviation threshold of 15 minutes for operation execution. If the preset deviation threshold is exceeded, then for the lag scenario of the 6205-2RS bearing, it can be found that after the supplier demand data was synchronized to the warehouse system API interface at 09:10, the order demand parsing node of the warehouse system had insufficient computing power due to processing 5 types of bearing demand data at the same time. It took 50 minutes from receiving the data to generating the picking instruction (far exceeding the normal 20 minutes). In addition, the AGV picking scheduling node took 45 minutes for picking operation due to the allocation conflict of automated equipment (far exceeding the normal 30 minutes). Therefore, both the order demand parsing node and the AGV picking scheduling node are information lag nodes.

[0052] For example, regarding the aforementioned information lag nodes, information lag regions are divided according to the business modules to which these nodes belong. Then, using clustering algorithms, these regions are classified by lag severity (e.g., Level 1 high-risk information lag region, Level 2 medium-risk information lag region). The average lag time and lag frequency percentage of each region are labeled to form quantifiable information lag regions. Specifically, the order demand parsing node belongs to the external supply chain-internal warehouse data interaction region (responsible for cross-system data processing). This region has an average lag time of 1 hour and 40 minutes within 72 hours, and its lag frequency accounts for 30% of the total demand response frequency. The AGV picking scheduling node belongs to the internal warehouse picking operation region. Its average lag time within 72 hours is 55 minutes, and its lag frequency accounts for 30%. Through clustering algorithms, external supply chain-internal warehouse data interaction regions with an average lag time exceeding 1 hour are marked as Level 1 high-risk information lag regions, and internal warehouse location management regions with an average lag time of 30-60 minutes are marked as Level 2 medium-risk information lag regions.

[0053] In the above method, time series analysis algorithms are used to standardize and align the update time of demand data with the recording time of operation logs, solving the problem of misjudgment caused by chaotic time formats in traditional management. By tracing back the chain, information lag nodes are accurately located, achieving a breakthrough from overall lag perception to specific node location, avoiding blind investigation. Furthermore, clustering algorithms can be used to divide primary and secondary information lag areas and label the lag duration and proportion, transforming the abstract internal and external information transmission lag into quantifiable and locatable specific areas. This ensures that resource adjustments can focus on high-risk lag areas, avoiding the resource waste caused by blind optimization in traditional separate management, and laying the foundation for internal and external collaboration in bearing warehousing.

[0054] In one embodiment of this disclosure, step S130, determining the resource allocation matrix of the internal bearing storage based on the information lag region, further includes the following steps: Figure 3 As shown, the specific content is as follows:

[0055] Step S310: Obtain multi-source heterogeneous data for the information lag area. The multi-source heterogeneous data includes lagging node operation data, resource consumption data, and external environment data.

[0056] Step S320: Standardize the multi-source heterogeneous data to generate standardized data for information lag areas;

[0057] Step S330: Based on the random forest algorithm, determine the resource allocation matrix according to the standardized data of the information lag region;

[0058] Step S340: Determine whether the deviation rate between the resource allocation matrix and the historical benchmark matrix exceeds a preset deviation rate threshold;

[0059] Step S350: If it is determined that the deviation rate exceeds the preset threshold, then the adjustment parameters of the resource allocation matrix are obtained based on the linear regression algorithm;

[0060] Step S360: Optimize the resource allocation matrix based on the genetic algorithm and adjusting the parameters.

[0061] Specifically, lagging node operation data refers to the real-time operational status data of information lagging nodes (such as the order demand analysis module and AGV picking scheduling node), including the response time of node task processing, task queue length, and data throughput; resource consumption data refers to the hardware and software resource data consumed by the internal warehousing system to support the operation of the lagging area, including CPU utilization, memory usage, storage capacity consumption, and network bandwidth usage; external environment data refers to the data of external factors affecting the operational efficiency of the lagging area, including warehouse environment temperature, humidity, shelf area congestion, and equipment failure frequency. Taking the first-level high-risk information lagging area (external supply chain-internal warehousing data interaction area and internal warehousing picking operation area) identified on May 20, 2024 as an example, multi-source heterogeneous data is collected through sensor networks and system background monitoring modules. Among them, the lagging node operation data are: the task response time of the order demand analysis module is 50 minutes (normally 20 minutes), and the task queue length is 12 (normally 5); the task throughput of the AGV picking scheduling node is 8 orders / hour (normally 15 orders / hour). Resource consumption data: CPU utilization in this area is 85% (normal is 60%), storage capacity consumption is 75% (normal is 50%), and network bandwidth usage is 60Mbps (normal is 30Mbps). External environment data: warehouse picking area temperature is 28℃ (normal is 22-25℃), shelf A area congestion frequency is 3 times / hour (normal is 0-1 times / hour), AGV equipment failure frequency is 1 time / 4 hours (normal is 1 time / 8 hours). After standardizing the above data, the order demand parsing module response time is 50 minutes, with a minimum of 20 minutes and a maximum of 60 minutes. The standardized value is (50-20) / (60-20) = 0.75. CPU utilization of 85% is directly converted to 0.85. AGV equipment failure frequency is 1 time / 4 hours, with a mean of 0.25 times / 4 hours and a standard deviation of 0.1. The standardized value is (1-0.25) / 0.1 = 7.5. The final result is a standardized data matrix for information lag areas, with dimensions of 3 types of areas × 8 indicators (3 types of areas: data interaction area, picking operation area, and storage location management area; 8 indicators: response time, queue length, CPU utilization, storage consumption, bandwidth usage, temperature, congestion frequency, and failure frequency).

[0062] For example, using standardized data from information-lagging regions as input features (8 indicators) and historically optimal resource allocation ratios as output labels (such as the allocation ratios of CPU, storage, and bandwidth), 500 decision trees are constructed for multi-source data fusion. Each tree randomly selects 6 features for splitting, and the fusion result is output through a voting mechanism. The output resource allocation matrix A is:

[0063]

[0064] In this resource allocation matrix A, the rows represent the data interaction area and the picking operation area, respectively. The columns of resource allocation matrix A represent the CPU allocation ratio, storage allocation ratio, and bandwidth allocation ratio, respectively. The historical baseline matrix is ​​B= For the data interaction area: CPU allocation ratio, storage allocation ratio, and bandwidth allocation ratio are 0.25, 0.40, and 0.20, respectively. For the picking operation area: CPU allocation ratio, storage allocation ratio, and bandwidth allocation ratio are 0.45, 0.30, and 0.15, respectively. The historical baseline matrix shows CPU allocation ratios of 0.30, 0.35, and 0.25, respectively.

[0065] For example, the resource allocation matrix output above can be compared with the historical baseline matrix, and a preset deviation rate threshold of 15% can be set (i.e., if the deviation rate of a single matrix element exceeds 15% or the average deviation rate of the matrix exceeds 10%, it is determined that adjustment is needed). Data interaction area and CPU allocation ratio: The deviation rate between the current CPU allocation ratio of 0.25 and the historical CPU allocation ratio of 0.30 is calculated as (|0.25-0.30| / 0.30) × 100% ≈ 16.67% (exceeding the 15% threshold). Data interaction area and storage allocation ratio: The deviation rate between the current storage allocation ratio of 0.40 and the historical storage allocation ratio of 0.35 is 14.29% (not exceeding the threshold). Therefore, a linear regression model can be constructed for the CPU allocation ratio deviation in the data interaction area. The independent variable X of the linear regression model is: the standardized value of the order demand analysis module response time (0.75) and the standardized value of CPU utilization (0.85). The dependent variable Y of the linear regression model is: the CPU allocation ratio deviation rate (16.67%). The fitted linear equation is: Y = 0.2X1 + 0.15X2 - 0.05 (X1 = response time, X2 = CPU utilization). Then, the adjustment parameters are calculated using the linear regression equation: to reduce the deviation rate to below 10%, the CPU allocation ratio needs to be increased by 0.02 from 0.25 (adjustment parameter is +0.02), and the storage allocation ratio needs to be decreased by 0.01 from 0.40 (adjustment parameter is -0.01).

[0066] For example, a genetic algorithm based on a population size of 200, a crossover probability of 0.8, and a mutation probability of 0.1 can be used. The fitness function is ≤15% for the resource allocation matrix deviation rate and ≥20% for the improvement of indicators in the information lag area. After 150 iterations, the algorithm converges and finally outputs the optimized resource allocation matrix: the CPU allocation ratio in the data interaction area is adjusted to 0.27 and the storage allocation ratio is adjusted to 0.39; the CPU allocation ratio in the picking operation area is adjusted to 0.43 and the storage allocation ratio is adjusted to 0.30.

[0067] In the above method, the resource allocation matrix is ​​generated based on the random forest algorithm, ensuring the scientific nature and accuracy of resource allocation. Next, a linear regression algorithm is used to mine the linear relationship between resource allocation and lagging indicators, providing a quantitative basis for adjustments. Finally, a genetic algorithm is used to optimize the resource allocation matrix, ultimately reducing the average deviation rate between the resource allocation matrix and the historical benchmark matrix, improving operational efficiency in information-lagging areas, solving the problem of unreasonable internal bearing warehousing resource allocation, providing an optimized resource foundation for subsequent inventory level indicator calculations, and simultaneously improving operational efficiency in information-lagging areas, thus promoting enhanced collaboration between internal warehousing and the external supply chain.

[0068] In one embodiment of this disclosure, step S140 involves determining inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix. These inventory level evaluation indicators include inventory turnover rate and inventory holding days. The process also includes the following steps: Figure 4 As shown, the specific content is as follows:

[0069] Step S410: Determine the average allocated inventory for different bearing models based on the resource allocation matrix;

[0070] Step S420: Obtain the sales volume of the corresponding bearing model within the period from the external supply chain;

[0071] Step S430: Calculate the ratio between the sales volume of the corresponding bearing model and the average allocated inventory during the period, and use the ratio result as the inventory turnover rate of the corresponding bearing model.

[0072] Step S440: Calculate the ratio between the cycle days and the inventory turnover rate, and use the ratio result as the inventory holding days for the corresponding bearing model;

[0073] Step S450: Obtain the weights corresponding to inventory turnover rate and inventory holding days, respectively;

[0074] Step S460: Based on the weights corresponding to inventory turnover rate and inventory holding days, perform a weighted calculation on inventory turnover rate and inventory holding days, and use the weighted calculation result as the inventory level evaluation index for the corresponding bearing model.

[0075] Specifically, for three core bearing models (6205-2RS, 6305-2RS, and 6005-2RS) in the internal bearing inventory, based on the optimized resource allocation matrix, a linear programming algorithm is used, with the objective function being to minimize the total cost function Z = 0.5X1 + 0.3X2 + 0.7X3. Here, X1, X2, and X3 represent the inventory allocation quantities for the three models, respectively. The calculated average inventory allocation is 1200 sets for 6205-2RS bearings, 800 sets for 6305-2RS bearings, and 1500 sets for 6005-2RS bearings. Next, the data statistics period is set to May 2024 (30 days). Data collected through an external supply chain API interface shows that the sales volume of 6205-2RS bearings during this period was 400 sets (mainly supplied to an automotive parts manufacturer), 600 sets (supplied to an engineering machinery company), and 300 sets (supplied to a motor manufacturer). Based on the above data, the inventory turnover rate for each bearing model is calculated as follows: Inventory turnover rate for 6205-2RS bearing = Sales volume during the period (400 units) / Average allocated inventory (1200 units) ≈ 0.33. Inventory turnover rate for 6305-2RS bearing = Sales volume during the period (600 units) / Average allocated inventory (800 units) = 0.75. Inventory turnover rate for 6005-2RS bearing = Sales volume during the period (300 units) / Average allocated inventory (1500 units) = 0.20.

[0076] Specifically, inventory holding days is an indicator that measures the time inventory is held. The longer the holding period, the higher the risk of inventory backlog. The calculation formula is: Inventory Holding Days = Cycle Days / Inventory Turnover Rate. For example, using a cycle of 30 days, the inventory holding days for various bearing models are calculated as follows: 6205-2RS bearing: 30 days / 0.33 ≈ 90.9 days; 6305-2RS bearing: 30 days / 0.75 = 40 days; 6005-2RS bearing: 30 days / 0.20 = 150 days.

[0077] Specifically, by combining bearing industry warehousing management standards with internal historical operational data (the correlation coefficient between inventory turnover rate and stockout rate over the past 12 months is 0.72, and the correlation coefficient between inventory holding days and backlog costs is 0.58), the weight corresponding to inventory turnover rate can be determined to be 0.6, and the weight corresponding to inventory holding days to be 0.4. Since inventory turnover rate and inventory holding days have different dimensions, inventory holding days need to be normalized before weighted calculation to avoid dimensional interference leading to result bias. Therefore, the final comprehensive evaluation index ranges from [0,1], with values ​​closer to 1 indicating a better inventory level and values ​​closer to 0 indicating a worse inventory level. For example, setting the reasonable range for industry inventory holding days as 30-120 days, the normalization formula is: Standardized Value = (Actual Value - Minimum Value) / (Maximum Value - Minimum Value). The calculated standardized inventory holding days for the 6205-2RS bearing are: (90.9 - 30) / (120 - 30) ≈ 0.677. The standardized inventory holding days for the 6305-2RS bearing are: (40.0 - 30) / (120 - 30) ≈ 0.111. The standardized inventory holding days for the 6005-2RS bearing are: (150.0 - 30) / (120 - 30) ≈ 1.333. Finally, through weighted calculation, the inventory level evaluation index is obtained as: Inventory Turnover Rate × 0.6 + (1 - Standardized Inventory Holding Days) × 0.4. For example, the inventory level evaluation index for bearing 6205-2RS is 0.33 × 0.6 + (1 - 0.677) × 0.4 ≈ 0.327. The inventory level evaluation index for bearing 6305-2RS is 0.75 × 0.6 + (1 - 0.111) × 0.4 ≈ 0.806. The inventory level evaluation index for bearing 6005-2RS is 0.20 × 0.6 + (1 - 1.0) × 0.4 ≈ 0.12.

[0078] The above method utilizes external supply chain cycle sales data to ensure that inventory level evaluation is linked to market demand, thereby generating inventory level evaluation indicators and enabling horizontal comparability of inventory status for different bearing models. This effectively avoids misjudgments caused by single-indicator evaluation (such as focusing only on turnover rate while ignoring the risk of overstocking), improves the scientific nature and accuracy of internal bearing warehousing inventory management, and provides inventory-level decision support for internal and external supply chain collaboration.

[0079] In one embodiment of this disclosure, step S150 involves triggering an inventory warning based on an inventory level evaluation index. The inventory warning includes both an inventory backlog warning and an inventory shortage warning, and further includes the following steps: Figure 5 As shown, the specific content is as follows:

[0080] Step S510: Determine whether the inventory level evaluation index exceeds the preset inventory level evaluation index threshold;

[0081] Step S520: If the inventory level exceeds the preset threshold, an inventory backlog warning is triggered.

[0082] Step S530: If it is determined that the preset inventory level evaluation index threshold has not been exceeded, an inventory shortage warning will be triggered.

[0083] Specifically, for example, if the preset inventory level evaluation threshold is 0.5, and the 6305-2RS bearing's inventory level evaluation index is 0.806 > 0.5, it is determined to exceed the preset threshold, triggering an inventory backlog warning. The 6205-2RS bearing's inventory level evaluation index is 0.327 < 0.5, determined to not exceed the preset threshold, triggering an inventory shortage warning. Similarly, the 6005-2RS bearing's inventory level evaluation index is 0.12 < 0.5, also determined to not exceed the preset threshold, triggering an inventory shortage warning. Thus, for scenarios exceeding the preset threshold, an inventory backlog warning is triggered, coordinating with internal and external supply chains to quickly allocate and adjust resources for backlogged inventory, avoiding the occupation of inventory funds and wasted storage space. For scenarios not exceeding the preset threshold, an inventory shortage warning is triggered, automatically generating replenishment suggestions with the external supply chain to ensure timely replenishment of high-risk bearing models, reducing the risk of supply chain disruptions due to stockouts. It enables accurate identification and rapid response to risks of inventory backlog and stockout, effectively connects internal warehousing and inventory management with external supply chain replenishment / allocation processes, and improves the sensitivity and response efficiency of the entire bearing warehousing management system to changes in supply and demand.

[0084] In one embodiment of this disclosure, step S160, which involves determining the inventory risk probability distribution map for different bearing models based on inventory warnings, further includes the following steps: Figure 6 As shown, the specific content is as follows:

[0085] Step S610: Based on the Bayesian network model, determine the inventory backlog probability of different bearing models according to the inventory backlog warning;

[0086] Step S620: Based on the normal distribution function, determine the inventory shortage probability of different bearing models according to the inventory shortage warning;

[0087] Step S630: Based on the Monte Carlo simulation algorithm, determine the inventory risk probability distribution map according to the inventory backlog probability and inventory shortage probability.

[0088] Specifically, the Bayesian network model calculates the posterior inventory backlog probability using prior probabilities (historical inventory backlog data) and conditional probabilities (the probability of the impact of inventory warning-related indicators on backlog), thus transforming qualitative warnings into quantitative probabilities. The final output is the inventory backlog probability for each bearing model (range 0-1, with values ​​closer to 1 indicating higher backlog risk). The normal distribution function describes the probability distribution of demand using the periodic projected demand and the standard deviation of historical demand fluctuations, thereby calculating the probability of stockouts when actual inventory is lower than demand. Furthermore, based on the Monte Carlo simulation algorithm, a large number of random samples are used to simulate the probability distribution of inventory risk, generating an inventory risk probability distribution map. The horizontal axis represents the risk level, and the vertical axis represents the probability. For example, the calculated results show: the backlog probability of bearing 6305-2RS is 0.78, and the stockout probability is 0.12; the backlog probability of bearing 6205-2RS is 0.45, and the stockout probability is 0.28; and the backlog probability of bearing 6005-2RS is 0.62, and the stockout probability is 0.65. The inventory risk probability distribution chart shows a distribution of high backlog risk (backlog probability 75%-85%), medium backlog risk (backlog probability 40%-50%), medium stockout risk (stockout probability 25%-30%), and high stockout risk (stockout probability 60%-70%).

[0089] In the aforementioned method, a Bayesian network model is used to transform inventory backlog warnings from qualitative judgments into quantitative inventory backlog probabilities. The causal relationship between historical data and warning indicators is leveraged to improve the accuracy of probability calculations, addressing the problem of traditional inventory risk assessment relying solely on experience and lacking data support. Simultaneously, based on the normal distribution function and combined with external supply chain demand forecast data and historical fluctuation standard deviations, the probability of inventory shortages for each bearing model is accurately quantified, providing a unified probability standard for risk classification of different models and avoiding subjectivity in shortage risk assessment. Finally, based on the Monte Carlo simulation algorithm, the probabilities of inventory backlog and shortages are integrated to generate a visualized inventory risk probability distribution map, clearly presenting the probability distribution characteristics of different risk levels. This addresses the pain points of traditional methods, such as vague risk results and the inability to intuitively guide decision-making. It provides a precise probabilistic basis for subsequent determination of inventory risk levels, transforming inventory risk from an abstract concept into quantifiable and visualized concrete data. This supports internal warehousing inventory adjustment decisions (such as backlog inventory transfer and shortage inventory replenishment) and external supply chain collaborative responses (such as adjusting replenishment cycles and optimizing demand forecasts), further improving the technical support of the end-to-end collaborative framework in the inventory risk management process.

[0090] In one embodiment of this disclosure, in step S170, the inventory risk level is determined according to the inventory risk probability distribution map. The inventory risk level includes a high-risk inventory level, a medium-risk inventory level, and a low-risk inventory level. If the backlog probability of the current bearing model is greater than the backlog probability percentage threshold (e.g., the threshold is 60%) or the out-of-stock probability of the current bearing model is greater than the out-of-stock probability percentage threshold (e.g., the threshold is 60%), it is determined to be a high-risk inventory level. If the backlog probability of the current bearing model is within the range of the backlog probability percentage threshold (e.g., the threshold range is 30%~60%) or the out-of-stock probability of the current bearing model is within the range of the out-of-stock probability percentage threshold (e.g., the threshold range is 30%~60%), it is determined to be a medium-risk inventory level. If the backlog probability of the current bearing model is less than the backlog probability percentage threshold (e.g., the threshold is 30%) or the out-of-stock probability of the current bearing model is less than the out-of-stock probability percentage threshold (e.g., the threshold is 30%), it is determined to be a low-risk inventory level. For example, parameters are extracted from the inventory risk probability distribution chart for the 6305-2RS bearing: the backlog probability is 78%, and the stockout probability is 12%. Since the backlog probability of 78% is greater than or equal to the backlog probability threshold of 60% for high-risk inventory levels, the inventory risk level of the 6305-2RS bearing is determined to be high-risk.

[0091] In one embodiment of this disclosure, step S180 involves obtaining a set of bearing models corresponding to high-risk inventory levels from a bearing specification database, and determining the target bearing model based on the set of bearing models using a convolutional neural network algorithm and a radio frequency identification algorithm. The step also includes the following steps: Figure 7 As shown, the specific content is as follows:

[0092] Step S710: Obtain the bearing image data and bearing label data from the bearing model set respectively;

[0093] Step S720: Based on a convolutional neural network, extract features from the bearing image data to generate an image feature matrix;

[0094] Step S730: Determine the radio frequency characteristic parameters based on the bearing tag data using the radio frequency identification algorithm;

[0095] Step S740: Based on the support vector machine algorithm, determine the multimodal feature fusion vector according to the image feature matrix and radio frequency feature parameters;

[0096] Step S750: Determine the target bearing model based on the multimodal feature fusion vector.

[0097] Specifically, bearing image data and bearing tag data can be extracted from the bearing specification database (which stores the basic specifications, image templates, and label information of all bearing models) to identify the high-risk bearing models in the inventory. The bearing image data can be generated by using an industrial high-definition camera to capture images of the bearing from three key angles (end face, side, and rolling element area), collecting visual information such as surface texture (e.g., seal groove patterns), dimensional characteristics (e.g., markings of 25mm inner diameter and 52mm outer diameter), and appearance defects (e.g., absence of scratches / rust), generating RGB format image data. The bearing tag data can be obtained by reading the RFID tags attached to the bearing surface using a Radio Frequency Identification (RFID) reader (e.g., operating at 915MHz), retrieving structured data stored within the tag, such as the unique bearing identifier (e.g., B-6305-2RS-20240501-001), production batch (e.g., 202405), and specifications (e.g., 25mm inner diameter, 52mm outer diameter, and 15mm thickness).

[0098] Specifically, a technical solution can be implemented based on Convolutional Neural Networks (CNN) to automatically extract visual features. The solution uses the ResNet-50 model to obtain a recognition result with a confidence score of 0.94 in a 2048-dimensional feature space. Through multi-layer convolution and pooling operations, the pixel information of the bearing image data is transformed into a quantized feature matrix, highlighting the differentiated features of the bearing (such as the number of sealing grooves and the outer diameter).

[0099] Specifically, RFID tag data can be parsed using radio frequency identification (RFID) algorithms to convert unstructured tag signals into quantified radio frequency characteristic parameters, and then decoded to obtain the bearing's unique identification code and production batch information. For example, the RFID tag of the 6305-2RS bearing was read 10 times. The signal strength of the 10 reads was -45dBm, -46dBm, -44dBm, ..., -47dBm. After removing the minimum value of -47dBm and the maximum value of -44dBm, the average value was -45.2dBm. The unique identification code and production batch were successfully decoded in all 10 reads, with a decoding accuracy of 100%. The response times of the 10 reads were 8ms, 7ms, 9ms, ..., 10ms, with an average of 8.5ms. The final radio frequency characteristic parameters of the 6305-2RS bearing were determined to be: [signal strength -45.2dBm, decoding accuracy 100%, response time 8.5ms].

[0100] Specifically, the Support Vector Machine (SVM) algorithm can be used to achieve cross-modal fusion of visual and radio frequency (RF) features, overcoming the limitations of single-modal data. Through weight allocation and feature mapping, a fusion vector of uniform dimension is generated. For example, features of the 6305-2RS bearing can be fused. The 2048 elements of its image feature matrix are Z-score normalized to a mean of 0 and a standard deviation of 1. In the RF feature parameters, the signal strength of -45.2 dBm is normalized to 0.85, the decoding accuracy of 100% is normalized to 1.0, and the response time of 8.5 ms is normalized to 0.7. The image feature matrix (weight 0.6) and the normalized RF feature parameters (weight 0.4) are input into the SVM. After mapping using the RBF kernel function, a weighted multimodal feature fusion vector of dimension 1×2051 (2048+3) is generated. The first 2048 dimensions are the weighted image features, and the last 3 dimensions are the weighted radio frequency features. The overall confidence level of the fused vector is 0.96.

[0101] Specifically, the target bearing model can be accurately determined by comparing the multimodal fusion vector with database templates. The model with the highest similarity (≥95%) between the multimodal feature fusion vector and the standard model template vector in the bearing specification database is the target bearing model. For example, the multimodal feature fusion vector of bearing 6305-2RS has a cosine similarity of 98.7% (≥95%) with the standard template vector of 6305-2RS in the database, and a similarity of 72.3% (<95%) with the standard template vector of 6005-2RS. Therefore, the target bearing model is determined to be 6305-2RS. Similarly, the fusion vector of bearing 6005-2RS has a similarity of 97.5% (≥95%) with the standard template vector of 6005-2RS, and a similarity of 70.1% (<95%) with the standard template vector of 6305-2RS. Therefore, the target bearing model is determined to be 6005-2RS.

[0102] In the above method, a complete data source for dual-modal recognition is constructed by collecting bearing image data (multi-angle high-definition shooting) and bearing tag data (RFID reading), solving the problem of traditional single image recognition lacking electronic tag support. Convolutional neural networks are used to extract image feature matrices, achieving high-precision extraction of bearing appearance texture and size features. Radio frequency identification algorithms are used to analyze radio frequency feature parameters such as signal strength, decoding accuracy, and response time, ensuring the reliability of electronic tag data. Support vector machine algorithms are used to generate multi-modal feature fusion vectors, solving the complexity of cross-modal data fusion. Finally, cosine similarity comparison is used to determine the target bearing model, achieving accurate identification of high-risk bearings in inventory. This effectively solves the problem of insufficient accuracy of single identification technologies due to the large number of bearing models and similar specifications, providing a model basis for the precise operation of subsequent automated equipment, further strengthening the technical support of the end-to-end collaborative framework in the precise control of high-risk bearings, and improving the accuracy and efficiency of bearing warehouse management.

[0103] In one embodiment of this disclosure, step S190, which involves obtaining the location coordinates of the target bearing model and determining a path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model, further includes the following steps: Figure 8 As shown, the specific content is as follows:

[0104] Step S810: Determine standardized position parameters based on position coordinates;

[0105] Step S820: Determine the environmental condition data for the target bearing model based on standardized position parameters;

[0106] Step S830: Based on deep learning algorithms, train a path planning model from the starting point of the internal bearing storage to the target bearing model according to environmental state data;

[0107] Step S840: Determine the path planning map from the starting point of the internal bearing storage to the target bearing model based on the path planning model;

[0108] Step S850: Determine whether there are obstacles on the path planning map. If obstacles are found, update the environmental status data.

[0109] Specifically, the precise coordinates of the target bearing in three-dimensional space can be determined using laser rangefinders and machine vision technology. After obtaining point cloud data through laser radar scanning and filtering, the bearing's center point coordinates are calculated using a binocular stereo vision algorithm, achieving a positional accuracy of ±0.1 mm, thus realizing high-precision coordinate acquisition. For example, the three-dimensional coordinates of the bearing in the world coordinate system obtained by the laser rangefinder and binocular stereo vision algorithm are: 1275.3 mm, 893.1 mm, 452.2 mm, with a positional accuracy of ±0.1 mm. Then, the coordinate transformation module is called, inputting a rotation matrix of 15° around the Z-axis and translation vectors offset by -500 mm in the X direction and -300 mm in the Y direction, mapping them to the robotic arm's workspace coordinate system to obtain standardized position parameters: X: 775.3 mm, Y: 593.1 mm, Z: 452.2 mm. Environmental status data is collected based on the standardized position parameters, including static environmental features (the target bearing is located on shelf A, layer 3, shelf 2, coordinate range 750mm-800mm, 550mm-600mm, 400mm-500mm, aisle width 1200mm, coordinates of nearby load-bearing columns 1000mm, 600mm, 0mm-5000mm) and dynamic environmental features (AGV-03 real-time position 650mm, 593.1mm, 0mm, movement speed 0.5m / s, robotic arm ARM-02 is in idle state).

[0110] Next, a deep reinforcement learning algorithm was employed, using environmental state data as input (dimension 1×12) and the key node coordinate sequence of the path as output. After 150 iterations of training, the model converged, outputting the optimal path's key node sequence as follows: starting point coordinates 0mm, 0mm, 0mm → node 1 coordinates 500mm, 500mm, 0mm → node coordinates 2700mm, 593.1mm, 0mm → ending point coordinates 775.3mm, 593.1mm, 452.2mm. Simultaneously, a path planning map was generated based on the trained path planning model, marking the starting point (internal warehouse scheduling center), key nodes and their coordinates, and the ending point (standardized position of the 6305-2RS bearing). The total path length was calculated to be 1.2014m (749.2mm AGV ground path + 452.2mm robotic arm Z-axis movement), with an estimated execution time of 9.5 seconds (1.5 seconds AGV travel + 8 seconds robotic arm retrieval). The designated execution devices were AGV-01 and ARM-02. By jointly detecting the path planning map using a lidar sensor array (sampling frequency 10Hz) and an infrared sensor, it was found that AGV-03 had entered the path segment from node 1 to node 2 due to scheduling adjustments (current position 550mm, 550mm, 0mm). It was determined that there was a dynamic obstacle, so the environmental status data was updated, and the information of AGV-03 was updated to position 550mm, 550mm, 0mm, speed 0.5m / s, entering the path node 1 to node 2'. The path planning model was then retrained to avoid the risk of collision.

[0111] In the above method, by mapping the three-dimensional spatial coordinates of the target bearing to the workspace coordinate system of the robotic arm, standardized position parameters are generated, solving the problem of path planning deviation caused by the inconsistency of multiple coordinate systems in traditional warehousing, and ensuring that the coordinate accuracy meets the requirements. By collecting static environmental features (shelves, aisles, load-bearing columns) and dynamic environmental features (AGV, personnel, equipment status) around the target bearing, complete environmental status data is constructed, providing comprehensive environmental input for path planning and avoiding path infeasibility due to incomplete environmental perception. The path planning model is trained based on deep reinforcement learning algorithm, and through multiple iterations and multi-weight reward functions, an optimal path model that takes into account the shortest path, safety and collision-free operation, and equipment efficiency is generated. By converting the output of the path planning model into a visualized path planning map including key nodes, path length, and estimated execution time, a directly executable operational basis is provided for automated equipment. Finally, obstacles are detected in real time by LiDAR and infrared sensors, the environmental status data is updated, and the path planning model is iteratively optimized, solving the pain point that traditional static path algorithms cannot cope with dynamic interference. It achieves high precision, dynamism and intelligence in internal bearing storage path planning, further improves the technical support of the end-to-end collaborative framework in the operation path optimization of automated equipment, enhances the operation efficiency and safety of automated equipment, and avoids time waste caused by collision risks and path redundancy.

[0112] In one embodiment of this disclosure, after determining the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model in step S190, the following steps are also included: Figure 9 As shown, the specific content is as follows:

[0113] Step S910: Based on the path planning diagram, obtain the device execution log data, and determine the state transition matrix according to the device execution log data;

[0114] Step S920: Based on the sliding window algorithm, determine the efficiency index value within the time window according to the state transition matrix;

[0115] Step S930: Determine whether the efficiency index value shows a continuous downward trend;

[0116] Step S940: If a continuous downward trend is determined, update the environmental status data.

[0117] Specifically, real-time operating data of automated equipment (such as AGVs and robotic arms) executing path planning diagrams can be collected to construct a state transition matrix reflecting the changing patterns of equipment states. Among these, the equipment execution log data records key information throughout the entire process of the equipment execution path task, including equipment number, task start time, task completion time, task status (success / failure / interruption), resource consumption rate (such as AGV power consumption and robotic arm computing power usage), and data collection frequency synchronized with the equipment execution cycle (e.g., one log entry is recorded for each completed target bearing sorting task). For example, consider the path planning diagram for target bearing model 6305-2RS (start point → node 1 → node 2 → end point, executed by AGV-01 and robotic arm ARM-02). Based on the path planning map, obtain 100 device execution log data. For example, a log record of AGV-01 starts at 2024-06-01 09:02:30, finishes at 09:02:45, and changes from idle (S0) to execution (S1) to fault interruption (S3). The fault code is F01 - path conflict. Count the number of transitions between the four states of idle (S0), execution (S1), completion (S2), and fault (S3) and generate a 4×4 dimension state transition matrix (e.g., S0→S1 probability 0.90, S1→S3 probability 0.05).

[0118] For example, a sliding window algorithm (window size of 10 execution cycles, step size of 1 execution cycle) can be used. For each sliding window, the values ​​of three indicators are calculated, and the weighted average (average task completion time weighted at 0.4, resource utilization weighted at 0.3, and task success rate weighted at 0.3) is taken as the final efficiency indicator value. For example, the efficiency indicator value of window 1 (cycles 1-10) is 0.8645, window 2 (cycles 2-11) is 0.851, window 3 (cycles 3-12) is 0.8345, and window 4 (cycles 4-13) is 0.818.

[0119] For example, an efficiency indicator value can be defined as showing a continuous downward trend if the efficiency indicator value decreases sequentially for three or more consecutive sliding windows, with a total decrease of ≥3%. Here, sequential decrease means that the indicator value of the next window is less than the indicator value of the previous window, and the total decrease is calculated as (first window indicator value - last window indicator value) / first window indicator value × 100%. For example, if window 2 (efficiency indicator value 0.851) < window 1 (efficiency indicator value 0.8645), and window 3 (efficiency indicator value 0.8345) < efficiency indicator value window 2 (0.851), it satisfies two consecutive sequential decreases. The total decrease is calculated as (window 1 indicator value 0.8645 - window 4 indicator value 0.818) / 0.8645 × 100% ≈ 5.38% ≥ 3%. This demonstrates that the efficiency indicator value shows a continuous downward trend. At this point, the corresponding fields in the environmental status data can be corrected by checking the fault codes and location records in the device execution log to determine the reasons for the decreased positioning efficiency (such as an increase in AGV failures and interruptions due to new obstacles, or an increase in task completion time due to channel congestion). This ensures that the data is consistent with the actual environment.

[0120] In the above method, a state transition matrix is ​​constructed by collecting equipment execution log data, transforming the dynamic process of equipment execution path into a quantifiable state change pattern. This solves the lag problem caused by traditional path planning's emphasis on planning over execution monitoring, providing a precise data foundation for efficiency evaluation. Based on the sliding window algorithm (window size of 10 execution cycles, step size of 1 cycle), efficiency index values ​​(average task completion time, equipment resource utilization, and task success rate) within the time window are calculated, eliminating the interference of single data fluctuations and ensuring the continuity and objectivity of efficiency evaluation. By tracing back the logs to locate the root cause of efficiency decline and updating environmental state data (such as adding AGV docking information and channel congestion frequency), the latest environmental input is provided for the path planning model retraining, forming an effective closed loop and solving the problem of equipment execution efficiency decline caused by path planning map lag. This strengthens the dynamic linkage capability of path planning, equipment execution, and environment adaptation in the end-to-end collaborative framework, ensuring that the path planning of internal bearing storage is always synchronized with the actual equipment status and environmental changes, significantly improving the execution efficiency and path feasibility of automated equipment, reducing the risk of failure and interruption due to planning lag, and further improving the intelligence and adaptive capability of bearing digital storage management.

[0121] This disclosure also provides a digital warehouse management system based on intelligent sorting. The system may include an acquisition module, a first determination module, a second determination module, a third determination module, an early warning module, a fourth determination module, a fifth determination module, a sixth determination module, and a seventh determination module. The acquisition module is used to acquire real-time demand data from the external supply chain and operation log data from the internal bearing warehouse. The demand data includes order demand data, inventory change data, and delivery plan data. The operation log data includes inbound data, outbound data, picking data, and packaging data. The first determination module is used to determine the information lag area between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and operation log data. The second determination module is used to determine the resource allocation matrix of the internal bearing warehouse based on the information lag area. The third determination module is used to determine the inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix. The inventory level evaluation indicators include inventory turnover rate and inventory holding days. The early warning module is used to determine the inventory level evaluation indicators based on the inventory level evaluation... The system is structured as follows: 1) Price indicators trigger inventory alerts, including inventory backlog alerts and inventory shortage alerts; 2) A fourth determination module determines the inventory risk probability distribution map for different bearing models based on the inventory alerts; 3) A fifth determination module determines the inventory risk level based on the inventory risk probability distribution map, including high-risk, medium-risk, and low-risk levels; 4) A sixth determination module retrieves the set of bearing models corresponding to the high-risk level from the bearing specification database, and determines the target bearing model based on the set of bearing models using convolutional neural network and radio frequency identification algorithms; 5) A seventh determination module obtains the location coordinates of the target bearing model and determines the path planning map from the starting point of the internal bearing warehouse to the target bearing model based on the location coordinates.

[0122] It should be noted that the embodiments of the intelligent sorting-based digital warehouse management system provided in this application can be used to execute the processing flow of the embodiments of the intelligent sorting-based digital warehouse management method in the above embodiments. Its functions will not be repeated here, but can be referred to the detailed description of the above method embodiments.

[0123] As described above, the digital warehouse management system based on intelligent sorting provided in this disclosure breaks down the data silo between internal operations and the external supply chain in traditional warehouse management by acquiring real-time demand data from the external supply chain, including order demand data, inventory change data, and delivery plan data, and operation log data from the internal bearing warehouse, including inbound data, outbound data, picking data, and packaging data. This lays a comprehensive data foundation for subsequent internal and external process collaboration. Through analysis of the aforementioned demand data and operation log data, the system accurately identifies the information lag area between external supply chain demand and internal bearing warehouse response, effectively solving the problem of delayed supply and demand information transmission under the existing warehouse management model and providing a precise optimization direction for subsequent resource optimization. Based on the information lag area, the system determines the resource allocation matrix for the internal bearing warehouse, enabling targeted adjustments to warehouse resource allocation, avoiding unreasonable resource allocation problems caused by traditional separate management, and improving resource utilization efficiency. Based on the resource allocation matrix, using inventory turnover rate and inventory holding days as clear inventory level evaluation indicators, the system can accurately quantify the inventory status of different bearing models, providing a basis for inventory management. Based on the inventory level... The evaluation indicators trigger an inventory early warning mechanism that includes inventory backlog warnings and inventory shortage warnings. This mechanism can promptly capture inventory anomalies and prevent the risk of inventory backlog or shortages from escalating due to failure to detect anomalies in a timely manner. Based on inventory early warnings, a probability distribution map of inventory risk for different bearing models is determined, and further divided into high-risk, medium-risk, and low-risk levels. This allows for precise differentiation of the inventory risk level of different bearing models, facilitating key management of high-risk bearing models. The bearing model set corresponding to the high-risk inventory level is selected from the bearing specification database, and a combination of convolutional neural network algorithm and radio frequency identification algorithm is used to accurately identify the target bearing model. This effectively solves the problems of insufficient accuracy of single identification technology due to the large number of bearing models and similar specifications, as well as the complexity of data processing for multi-technology fusion, achieving high-precision identification of high-risk bearing models. By obtaining the location coordinates of the target bearing model and determining the path planning map from the internal bearing storage starting point to the target bearing model, precise sorting path guidance is provided for warehouse automation equipment. This overcomes the shortcomings of traditional static path algorithms in coping with dynamic environmental changes, improving the sorting efficiency and operational safety of automated equipment. This enables deep collaboration between internal and external processes in bearing warehousing, significantly improving the accuracy, efficiency, and intelligence of warehousing management, reducing inventory backlog and stockout risks, thereby enhancing the responsiveness of the entire supply chain. It can meet the stringent requirements of modern manufacturing for bearing warehousing management in terms of bearing model identification accuracy, internal and external interaction speed, and management cost control, providing effective support for the intelligent upgrading of the bearing industry.

[0124] This disclosure also provides an electronic device including one or more processors and memory resources, represented by a memory, for storing instructions executable by the processor, such as application programs. The application programs stored in the memory may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor is configured to execute instructions to perform the aforementioned digital warehouse management method based on intelligent sorting.

[0125] The electronic device may also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on operating devices stored in memory, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0126] In one embodiment, a computer device, which may be a server, is also provided. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device stores data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a digital warehouse management method based on intelligent sorting.

[0127] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a digital warehouse management method based on intelligent sorting. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0128] This disclosure also provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the aforementioned electronic device, the electronic device is able to execute a digital warehouse management method based on intelligent sorting, including: acquiring real-time demand data from an external supply chain and operation log data from an internal bearing warehouse, wherein the demand data includes order demand data, inventory change data, and delivery plan data, and the operation log data includes inbound data, outbound data, picking data, and packaging data; determining the information lag region between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and the operation log data; determining the resource allocation matrix of the internal bearing warehouse based on the information lag region; and determining the different bearing models in the internal bearing warehouse based on the resource allocation matrix. Inventory level evaluation indicators include inventory turnover rate and inventory holding days. Inventory alerts are triggered based on these indicators, including inventory backlog alerts and inventory shortage alerts. Inventory risk probability distribution maps for different bearing models are determined based on these alerts. Inventory risk levels are determined based on these probability distribution maps, including high-risk, medium-risk, and low-risk levels. From the bearing specification database, a set of bearing models corresponding to the high-risk level is obtained, and the target bearing model is determined based on the set of bearing models using convolutional neural network and radio frequency identification algorithms. The location coordinates of the target bearing model are obtained, and a path planning map from the starting point of the internal bearing storage to the target bearing model is determined based on these coordinates.

[0129] This disclosure can take the form of a computer program product implemented on one or more storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0130] It should be noted that although the steps of the digital warehouse management method based on intelligent sorting in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps, such as omitting certain steps, combining multiple steps into one step, and / or breaking down one step into multiple steps, should all be considered part of this disclosure.

[0131] It should be understood that this disclosure is not limited to the detailed structure and arrangement of the modules of the intelligent sorting-based digital warehouse management system proposed in this specification. This disclosure can have other implementations and can be implemented and performed in various ways. The foregoing variations and modifications fall within the scope of this disclosure. It should be understood that this disclosure, as disclosed and defined in this specification, extends to all alternative combinations of two or more individual features mentioned or apparent in the text and / or drawings. All these different combinations constitute multiple alternative aspects of this disclosure. The embodiments described in this specification illustrate the best known mode for implementing this disclosure and will enable those skilled in the art to utilize this disclosure.

Claims

1. A digital warehouse management method based on intelligent sorting, characterized in that, include: Real-time demand data from the external supply chain and operational log data from the internal bearing warehouse are acquired separately. The demand data includes order demand data, inventory change data, and delivery plan data, while the operational log data includes inbound data, outbound data, picking data, and packaging data. Based on the demand data and the operation log data, determine the information lag area between the demand from the external supply chain and the response from the internal bearing warehouse; The resource allocation matrix of the internal bearing storage is determined based on the information lag area; Based on the resource allocation matrix, inventory level evaluation indicators for different bearing models in the internal bearing warehouse are determined, including inventory turnover rate and inventory holding days. Inventory alerts are triggered based on the aforementioned inventory level evaluation indicators, including inventory backlog alerts and inventory shortage alerts. Based on the inventory warning, determine the inventory risk probability distribution map for the different bearing models; The inventory risk level is determined based on the inventory risk probability distribution map, and the inventory risk level includes high inventory risk level, medium inventory risk level and low inventory risk level. In the bearing specification database, a set of bearing models corresponding to the high-risk level of the inventory is obtained, and the target bearing model is determined based on the set of bearing models using a convolutional neural network algorithm and a radio frequency identification algorithm. Obtain the location coordinates of the target bearing model, and determine the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model.

2. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The step of determining the information lag area between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and the operation log data includes: Based on time series analysis algorithms, the update time of the demand data and the recording time of the operation log are obtained respectively; Determine whether the time difference between the update time of the demand data and the recording time of the operation log exceeds a preset deviation threshold; If the preset deviation threshold is exceeded, it is determined that there is an information lag between the demand data and the operation log data, from the demand of the external supply chain to the response of the internal bearing warehouse, and the information lag node is identified. The information lag region is determined based on the information lag node.

3. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, Determining the resource allocation matrix of the internal bearing storage based on the information lag region includes: Acquire multi-source heterogeneous data in the information lag area, including lagging node operation data, resource consumption data, and external environment data; The multi-source heterogeneous data is standardized to generate standardized data for information lag regions; Based on the random forest algorithm, the resource allocation matrix is ​​determined according to the standardized data of the information lag region. Determine whether the deviation rate between the resource allocation matrix and the historical benchmark matrix exceeds a preset deviation rate threshold; If the deviation rate exceeds the preset threshold, the adjustment parameters of the resource allocation matrix are obtained based on the linear regression algorithm. The resource allocation matrix is ​​optimized based on the genetic algorithm and the adjustment parameters.

4. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The step of determining the inventory level evaluation indicators for different bearing models in the internal bearing warehouse based on the resource allocation matrix includes: The average allocated inventory of the different bearing models is determined based on the resource allocation matrix. Obtain the sales volume of the corresponding bearing model within the period from the external supply chain; The sales volume of the corresponding bearing model within the period is calculated as a ratio to the average allocated inventory, and the ratio result is used as the inventory turnover rate of the corresponding bearing model. The ratio of the cycle days to the inventory turnover rate is calculated, and the ratio result is used as the inventory holding days for the corresponding bearing model. Obtain the weights corresponding to the inventory turnover rate and the inventory holding days, respectively. Based on the weights corresponding to the inventory turnover rate and the inventory holding days, a weighted calculation is performed on the inventory turnover rate and the inventory holding days, and the weighted calculation result is used as an evaluation index of the inventory level for the corresponding bearing model.

5. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The step of triggering an inventory warning based on the inventory level evaluation index includes: Determine whether the inventory level evaluation index exceeds the preset inventory level evaluation index threshold; If the inventory level exceeds the preset threshold, an inventory backlog warning will be triggered. If the inventory level assessment threshold is not exceeded, an inventory shortage warning will be triggered.

6. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The step of determining the inventory risk probability distribution map for different bearing models based on the inventory warning includes: Based on the Bayesian network model, the inventory backlog probability of different bearing models is determined according to the inventory backlog warning. Based on the normal distribution function, the probability of stock shortage for different bearing models is determined according to the stock shortage warning. Based on the Monte Carlo simulation algorithm, the inventory risk probability distribution map is determined according to the inventory backlog probability and the inventory shortage probability.

7. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The method of determining the target bearing model based on the set of bearing models, using convolutional neural network and radio frequency identification algorithms, includes: Obtain the bearing image data and bearing label data from the bearing model set, respectively; Based on a convolutional neural network, feature extraction is performed on the bearing image data to generate an image feature matrix; Based on the radio frequency identification algorithm, radio frequency characteristic parameters are determined according to the bearing tag data; Based on the support vector machine algorithm, a multimodal feature fusion vector is determined according to the image feature matrix and the radio frequency feature parameters; The target bearing model is determined based on the multimodal feature fusion vector.

8. The digital warehouse management method based on intelligent sorting according to claim 1, characterized in that, The step of determining the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model includes: Determine standardized position parameters based on the stated position coordinates; The environmental condition data for the target bearing model are determined based on the standardized position parameters. Based on deep learning algorithms, a path planning model is trained from the starting point of the internal bearing warehouse to the target bearing model according to the environmental state data; The path planning model is used to determine the path planning map from the starting point of the internal bearing storage to the target bearing model; Determine whether there are obstacles on the path planning map. If the obstacles are found to exist, update the environmental status data.

9. The digital warehouse management method based on intelligent sorting according to claim 8, characterized in that, After determining the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model, the method further includes: Based on the path planning diagram, obtain device execution log data, and determine the state transition matrix according to the device execution log data; Based on the sliding window algorithm, the efficiency index value within the time window is determined according to the state transition matrix. Determine whether the efficiency index value exhibits a continuous downward trend; If the continuous downward trend is determined to exist, the environmental status data is updated.

10. A digital warehouse management system based on intelligent sorting, characterized in that, include: The acquisition module is used to acquire real-time demand data from the external supply chain and operation log data from the internal bearing warehouse. The demand data includes order demand data, inventory change data, and delivery plan data. The operation log data includes inbound data, outbound data, picking data, and packaging data. The first determining module is used to determine the information lag area between the demand from the external supply chain and the response from the internal bearing warehouse based on the demand data and the operation log data. The second determining module is used to determine the resource allocation matrix of the internal bearing storage based on the information lag area; The third determining module is used to determine the inventory level evaluation index of different bearing models in the internal bearing warehouse according to the resource allocation matrix. The inventory level evaluation index includes inventory turnover rate and inventory holding days. The early warning module is used to trigger inventory early warnings based on the inventory level evaluation indicators, including inventory backlog warnings and inventory shortage warnings. The fourth determining module is used to determine the inventory risk probability distribution map of the different bearing models based on the inventory warning; The fifth determining module is used to determine the inventory risk level based on the inventory risk probability distribution map, wherein the inventory risk level includes high inventory risk level, medium inventory risk level and low inventory risk level; The sixth determining module is used to obtain a set of bearing models corresponding to the high-risk level of the inventory from the bearing specification database, and determine the target bearing model based on the set of bearing models using a convolutional neural network algorithm and a radio frequency identification algorithm. The seventh determining module is used to obtain the location coordinates of the target bearing model and determine the path planning map from the starting point of the internal bearing storage to the target bearing model based on the location coordinates of the target bearing model.

Images