An intelligent workshop material dynamic pull control method, system and device

Abstract

This invention provides a method, system, and device for dynamic material pull control in an intelligent workshop, relating to the field of intelligent manufacturing technology. The method includes: First, by acquiring the basic inventory buffer, equipment topology diagram, production scheduling characteristics, real-time production data, and material inbound rate of the intelligent workshop, and combining a multimodal prediction model to extract time-series production characteristics, equipment topology characteristics, and production scheduling characteristics, the predicted material consumption rate is output, improving prediction accuracy. Second, by combining the prediction results with equipment operating parameters, real-time inventory level and dynamic replenishment threshold are calculated to dynamically adapt to production fluctuations and equipment failure risks, avoiding inventory backlog or material shortages and shutdowns caused by traditional static inventory levels. Finally, if the real-time inventory level is lower than the dynamic replenishment threshold, a minimum delay path objective function is constructed based on material transportation parameters to generate the optimal path, and the automated guided vehicle is controlled to perform replenishment, ensuring timely replenishment, improving inventory turnover, and enhancing the lean management level of workshop materials.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a method, system and device for dynamic material pull control in an intelligent workshop. Background Technology

[0002] As the manufacturing industry moves towards intelligent and lean production, traditional material management models are no longer sufficient to meet the demands of high-efficiency production. Existing workshop material management technologies suffer from the following main shortcomings: First, the use of static inventory level settings, based on fixed maximum and minimum parameters, cannot adapt to fluctuations in production plans or abnormal equipment conditions, easily leading to inventory buildup or material shortages causing production stoppages. Second, replenishment relies on manual experience-based judgment, resulting in delayed responses and a failure to integrate dynamic variables such as equipment status and process defects, leading to untimely or excessive replenishment. Third, it focuses only on a single dimension, failing to quantify the coupled impact of multiple factors such as equipment failures and process abnormalities, resulting in a disconnect between replenishment strategies and actual production scenarios. Therefore, there is an urgent need to address core pain points such as inventory buildup and material shortages through data-driven intelligent prediction and dynamic control. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and device for dynamic material pulling control in an intelligent workshop, so as to solve one or more technical problems existing in the prior art, or at least provide a beneficial option or create conditions.

[0004] The solution to the technical problem of this invention is as follows: On the one hand, this invention provides a method for dynamic material pull control in an intelligent workshop, comprising the following steps:

[0005] The system acquires the basic inventory buffer, equipment topology, production scheduling characteristics, real-time production data, and material inbound rate of the smart workshop. The real-time production data includes current material inventory, equipment operating parameters, and material transportation parameters. The production scheduling characteristics are used to characterize information about production tasks and their corresponding product material usage.

[0006] Using a multimodal prediction model, time-series production features are extracted from the real-time production data, and equipment topology features are extracted from the equipment topology map. The time-series production features, equipment topology features, and production scheduling features are then fused through an attention mechanism to output a predicted material consumption rate.

[0007] By combining the predicted material consumption rate, the current material inventory, the material inbound rate, the basic inventory buffer, and the equipment operating parameters, the real-time inventory level and dynamic replenishment threshold are calculated.

[0008] If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and based on the material transportation parameters, a minimum delay path objective function is constructed to generate the optimal path, and the automated guided vehicle is controlled to perform replenishment.

[0009] Furthermore, the equipment topology diagram is used to characterize the process connections and fault propagation weights between equipment; the equipment topology diagram is constructed based on the production process flow diagram, equipment physical layout, and historical fault data of the smart workshop; the equipment topology diagram is in graph structure. The form of representation, in which Indicates a device node. This indicates the process connection relationships between equipment. Indicates the fault propagation weight between devices. The value is the probability of the cascading effect of equipment failures;

[0010] Furthermore, the multimodal prediction model includes a long short-term memory network, a graph neural network, and an attention mechanism;

[0011] The process of using a multimodal prediction model to extract time-series production features from real-time production data and equipment topology features from the equipment topology map, and then fusing the time-series production features, equipment topology features, and scheduling features through an attention mechanism to output a predicted material consumption rate, includes the following steps:

[0012] The time-series production features are extracted from the real-time production data using the long short-term memory network.

[0013] The device topology features are extracted from the device topology graph using the graph neural network.

[0014] The attention mechanism is used to fuse the time-series production features, the equipment topology features, and the production scheduling features to output a predicted material consumption rate.

[0015] Furthermore, the calculation of the real-time inventory level and dynamic replenishment threshold by combining the predicted material consumption rate, the current material inventory level, the material inbound rate, the basic inventory buffer, and the equipment operating parameters includes the following steps:

[0016] The dynamic risk compensation coefficient is calculated based on the average repair time and average interval between failures in the equipment operating parameters.

[0017] Based on the dynamic risk compensation coefficient and the basic inventory buffer, the dynamic inventory buffer and the dynamic replenishment threshold are calculated.

[0018] The real-time inventory level is calculated based on the dynamic inventory buffer, the predicted material consumption rate, the current material inventory, and the material inbound rate.

[0019] The real-time inventory level satisfies the following formula: ;

[0020] in, This indicates the real-time inventory level; This indicates the current inventory level of the material; This indicates the material receiving rate. Indicates the time window for material receipt; This indicates the predicted material consumption rate. Indicates the time window for material consumption; Indicates dynamic inventory buffer. The product of the dynamic risk compensation coefficient and the basic inventory buffer amount is given, and the dynamic risk compensation coefficient is calculated based on the average repair time and average failure interval time in the equipment operating parameters.

[0021] The dynamic replenishment threshold satisfies the following formula:

[0022] ;

[0023] in, This represents the dynamic replenishment threshold. Indicates the lead time for replenishment. This represents the average historical material consumption rate. This represents the average repair time among the operating parameters of the equipment. This refers to the mean time between failures (MTBF) in the equipment's operating parameters. This represents the preset average interval between failures (ABER) threshold. This indicates the preset average repair time threshold.

[0024] Furthermore, the dynamic risk compensation coefficient satisfies the following formula:

[0025] ;

[0026] in, This represents the dynamic risk compensation coefficient. Indicates the remaining time for the current work order. Indicates the process cycle constant;

[0027] The basic inventory buffer amount satisfies the following formula:

[0028] ;

[0029] in, This represents the basic inventory buffer level. This represents the tolerance coefficient for material breakage. Indicates the variance of replenishment lead time; This represents the standard deviation of historical material consumption rates.

[0030] Furthermore, the objective function for the minimum delay path satisfies the following formula:

[0031] ;

[0032] in, This represents the objective function value of the minimum delay path; This indicates the number of replenishment tasks in the smart workshop; This indicates that the automated guided vehicle has completed the first... The time required for each replenishment task Indicates the first The deadline for each replenishment task. Indicates the first The delayed completion time of each replenishment task; This represents the distance weighting coefficient, used to adjust the importance of the total travel distance of the automated guided vehicle in the minimum delay path objective function value; Indicates the first The automated guided vehicle completed the first... The driving distance for each replenishment task. This indicates the number of the automated guided vehicles; , Indicates the first The replenishment task was assigned to the first... The aforementioned automated guided vehicle performs this operation. Indicates the first The replenishment task is not assigned to the first... The automated guided vehicle (AGV) is used for this purpose.

[0033] Furthermore, the objective function of the minimum delay path must satisfy three constraints, including a first constraint, a second constraint, and a third constraint.

[0034] The first constraint condition satisfies the following formula:

[0035] ;

[0036] in, Indicates the first The task capacity of the automated guided vehicle;

[0037] The second constraint condition satisfies the following formula:

[0038] ;

[0039] The second constraint condition means that any replenishment task in the smart workshop is assigned to only one of the automated guided vehicles for execution;

[0040] The third constraint condition satisfies the following formula:

[0041] ;

[0042] in, Indicates the first The automated guided vehicle begins execution of the first... The time required for each replenishment task; Indicates the first The automated guided vehicle completed the first... The time required for each replenishment task; Indicates the first The location of the first replenishment task and the first The distance between the locations of each replenishment task; Indicates the first The moving speed of the automated guided vehicle; Indicates the first The number of tasks to be performed by the automated guided vehicle.

[0043] On the other hand, the present invention provides an intelligent workshop material dynamic pull control system, including a data acquisition module, a material consumption prediction module and a replenishment control module;

[0044] The data acquisition module is used to acquire the basic inventory buffer, equipment topology diagram, production scheduling characteristics, real-time production data, and material inbound rate of the smart workshop; wherein, the real-time production data includes the current material inventory, equipment operating parameters, and material transportation parameters; the production scheduling characteristics are used to characterize the information of production tasks and their corresponding product material usage;

[0045] The material consumption prediction module is used to extract time-series production features from the real-time production data using a multimodal prediction model, and extract equipment topology features from the equipment topology map. It then fuses the time-series production features, the equipment topology features, and the production scheduling features through an attention mechanism to output the predicted material consumption rate.

[0046] The replenishment control module is used to calculate the real-time inventory level and dynamic replenishment threshold by combining the predicted material consumption rate, the current material inventory, the material inbound rate, the basic inventory buffer, and the equipment operating parameters. If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and a minimum delay path objective function is constructed based on the material transportation parameters to generate the optimal path and control the automated guided vehicle to perform replenishment.

[0047] On the other hand, the present invention provides an intelligent workshop material dynamic pull control device, including: an Internet of Things sensing device, an edge computing device, a central decision-making device, and an automated guided vehicle;

[0048] The IoT sensing device is used to collect real-time production data in the smart workshop; wherein, the real-time production data includes current material inventory, equipment operating parameters, and material transportation parameters;

[0049] The edge computing device is equipped with an INT8-quantized multimodal prediction model, which is invoked via a gRPC interface.

[0050] The edge computing device is used to acquire the basic inventory buffer, equipment topology map, production scheduling characteristics, and material inbound rate of the smart workshop. Using the multimodal prediction model, it extracts time-series production characteristics from the real-time production data and equipment topology characteristics from the equipment topology map. It then fuses the time-series production characteristics, equipment topology characteristics, and production scheduling characteristics through an attention mechanism to output the predicted material consumption rate. The production scheduling characteristics are used to characterize the information of production tasks and their corresponding product material usage.

[0051] The central decision-making device is used to calculate the real-time inventory level and dynamic replenishment threshold by combining the predicted material consumption rate, the current material inventory, the material inbound rate, the basic inventory buffer, and the equipment operating parameters. If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and a minimum delay path objective function is constructed based on the material transportation parameters to generate the optimal path and control the automated guided vehicle to perform replenishment.

[0052] Furthermore, the IoT sensing device includes a line-side warehouse sensing unit, an equipment status sensing unit, and a logistics sensing unit;

[0053] The line-side warehouse sensing unit is used to obtain the current material inventory; the line-side warehouse sensing unit includes a smart bin, and the smart bin includes a weighing sensor and an RFID reader.

[0054] The equipment status sensing unit is used to collect the equipment operating parameters; the equipment status sensing unit is deployed on the production equipment in the smart workshop, and the equipment status sensing unit includes a vibration sensor and a PLC interface;

[0055] The logistics sensing unit is used to collect the position, speed and vehicle capacity of the automated guided vehicle as material transportation parameters; the logistics sensing unit is deployed on the automated guided vehicle and includes a lidar, a visual SLAM processor and a UWB chip.

[0056] The beneficial effects of this invention are as follows: The intelligent workshop material dynamic pull control method provided by this invention includes the following steps: First, by acquiring the basic inventory buffer, equipment topology diagram, production scheduling characteristics, real-time production data, and material inbound rate of the intelligent workshop, and combining this with a multimodal prediction model to extract time-series production characteristics, equipment topology characteristics, and production scheduling characteristics, the predicted material consumption rate is output, improving prediction accuracy. Second, by combining the prediction results with equipment operating parameters, real-time inventory level and dynamic replenishment threshold are calculated to dynamically adapt to production fluctuations and equipment failure risks, avoiding inventory backlog or material shortages and shutdowns caused by traditional static inventory levels. Finally, if the real-time inventory level is lower than the dynamic replenishment threshold, and based on material transportation parameters, a minimum delay path objective function is constructed to generate the optimal path, controlling the automated guided vehicle to perform replenishment, ensuring timely replenishment, improving inventory turnover, and enhancing the lean management level of workshop materials. This application also provides corresponding devices, systems, and vehicles. The beneficial effects of the devices, systems, and vehicles are similar to the method and will not be elaborated here.

[0057] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the description, claims and drawings. Attached Figure Description

[0058] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0059] Figure 1 This is a flowchart of the intelligent workshop material dynamic pull control method provided in this application;

[0060] Figure 2 This is a structural diagram of the intelligent workshop material dynamic pull control system provided in this application;

[0061] Figure 3 This is a structural diagram of the intelligent workshop material dynamic pull control device provided in this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0063] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0064] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0066] As the manufacturing industry transforms towards intelligent and lean manufacturing, the core objective of workshop material management has evolved from "ensuring supply" to "precisely matching demand, reducing inventory costs, and improving production continuity." In discrete manufacturing (such as automotive parts and 3C electronics) and process manufacturing (such as chemicals and food processing), material management needs to respond in real time to dynamic factors such as production plan fluctuations, equipment malfunctions, and changes in logistics and transportation. However, traditional material management models, due to technological limitations, are unable to meet these needs. Therefore, there is an urgent need to address core pain points such as inventory backlog, material shortages leading to shutdowns, and resource waste through data-driven intelligent forecasting and dynamic control technologies.

[0067] Traditional methods use fixed minimum / maximum inventory levels (Min / Max) based on historical average demand, failing to account for temporary adjustments to production plans (such as emergency orders or changes in work order priorities) or abnormal fluctuations in equipment status (such as sudden drops / rises in material consumption rates due to equipment failure). For example, when a sudden equipment failure causes production to stop, the material consumption rate at the line-side warehouse decreases, but the static level still triggers replenishment as originally planned, ultimately leading to inventory backlog. Conversely, if an emergency work order requires accelerated production, and the static level is not dynamically adjusted, it can easily lead to material shortages and production stoppages.

[0068] Secondly, the timing and quantity of replenishment rely on manual judgment based on experience (such as workers manually placing orders after observing the remaining materials in the line-side warehouse), lacking real-time data support. Replenishment quantities are calculated solely based on historical average demand, failing to integrate dynamic variables such as equipment operating status and process defects (e.g., the need for additional materials for rework of defective products). For instance, when equipment malfunctions lead to extended repair times, manual experience makes it difficult to adjust replenishment strategies in a timely manner, potentially causing material shortages on the production line due to delayed replenishment; or, due to the failure to consider the need for rework of defective products, insufficient replenishment quantities may necessitate a second emergency replenishment, increasing logistics costs.

[0069] Moreover, traditional methods only focus on the single dimension of material consumption, failing to quantify the coupled impact of multiple factors such as equipment operating status, logistics and transportation, and process anomalies. For example, a failure in equipment A may lead to material accumulation in equipment B, but traditional methods do not analyze this impact through equipment topology relationships and still replenish equipment B according to the original plan, resulting in inventory redundancy; or when Automated Guided Vehicle (AGV) path congestion causes replenishment delays, the path is not dynamically adjusted, further exacerbating the risk of material shortages.

[0070] Therefore, the core problem with existing technologies is that static and experience-based management models cannot adapt to the needs of dynamic production environments, resulting in low inventory turnover rates at production lines, high rates of downtime due to material shortages, and large proportions of stagnant materials, which seriously restrict the production efficiency and cost control of the manufacturing industry.

[0071] To address the aforementioned issues, this application provides a method, system, and device for dynamic material pull control in an intelligent workshop. The method acquires equipment topology maps, production scheduling characteristics, real-time production data, and quantities awaiting warehousing. It then utilizes a multimodal prediction model integrating LSTM, GNN, and attention mechanisms to extract temporal, topological, and production plan features, outputting a predicted material consumption rate. Combined with equipment operating parameters, it calculates a dynamic risk compensation coefficient, thereby generating real-time inventory levels and dynamic replenishment thresholds. If the inventory level is insufficient, it calls the minimum delay path objective function to generate the optimal AGV path and controls replenishment execution. The system adopts an architecture of "IoT sensing layer - edge computing layer - AI decision layer - execution control layer," achieving data acquisition, preprocessing, prediction, and execution through intelligent material bins, AGV transportation systems, and edge computing gateways. This effectively reduces line-side inventory, minimizes downtime due to material shortages, reduces AGV task response latency, and improves inventory turnover, effectively alleviating the core pain points of traditional material management.

[0072] First, the intelligent workshop material dynamic pull control method provided in the embodiments of this application will be described in detail below with reference to the accompanying drawings. (Refer to...) Figure 1 The implementation process of the intelligent workshop material dynamic pull control method provided in this application embodiment includes, but is not limited to, the following steps.

[0073] Step S100: Obtain the basic inventory buffer, equipment topology diagram, production scheduling characteristics, real-time production data, and material inbound rate of the smart workshop.

[0074] Real-time production data includes current material inventory, equipment operating parameters, and material transportation parameters.

[0075] In step S100, comprehensive multi-dimensional information is collected to support subsequent decision-making. The basic inventory buffer is a safety stock benchmark set based on historical production experience and normal demand fluctuations, providing a fundamental guarantee for handling unexpected daily production needs. The equipment topology diagram reflects the process connections between various equipment in the workshop and the risk of fault propagation, helping to identify the potential impact of equipment anomalies on material demand. Production scheduling characteristics, derived from the production planning system, clarify the specific needs of the current production task, ensuring that material supply aligns with production tasks. Real-time production data reflects the actual status of the line-side warehouse, equipment health, and material transportation progress. The material inbound rate records information on materials that have been issued but not yet delivered, avoiding duplicate replenishment or inventory shortages due to omissions. The integration of this data provides a comprehensive and real-time basis for subsequent forecasting and dynamic adjustments.

[0076] Step S200: Using a multimodal prediction model, time-series production features are extracted from real-time production data, and equipment topology features are extracted from the equipment topology diagram. Then, the time-series production features, equipment topology features, and scheduling features are fused through an attention mechanism to output the predicted material consumption rate.

[0077] In step S200, a multimodal prediction model transforms scattered production data into predictive results that can guide decision-making. Specifically, the model extracts time-series features reflecting material consumption trends (such as recent changes in the rate of material consumption) from real-time production data, extracts topological features reflecting the impact of equipment anomalies from the equipment topology diagram (such as changes in material demand for related equipment due to equipment failure), and combines these features with scheduling features (such as high-priority demands of urgent work orders). Through an attention mechanism, these features are dynamically fused to comprehensively determine the material consumption rate over a future period. This process breaks through the limitations of traditional experience-based judgments, making material consumption predictions more closely aligned with actual production scenarios and providing accurate demand data for subsequent dynamic inventory adjustments.

[0078] Step S300: By combining the predicted material consumption rate, current material inventory, material inbound rate, basic inventory buffer, and equipment operating parameters, the real-time inventory level and dynamic replenishment threshold are calculated.

[0079] In step S300, precise adjustments to inventory levels are achieved by integrating forecast results with real-time data. Specifically, this includes: calculating material demand over a future period based on the predicted material consumption rate and a preset time window; determining the total available inventory at the line-side warehouse based on current inventory levels (existing materials in the line-side warehouse) and incoming materials (materials in transit); and adjusting the basic inventory buffer based on equipment operating parameters (such as equipment failure frequency and maintenance time) to generate a real-time inventory level (i.e., the currently available safety stock). The dynamic replenishment threshold is dynamically set based on equipment operating status and production needs (e.g., raising the threshold when the risk of equipment failure is high), ensuring that inventory meets production demands while avoiding inventory backlog due to excessive replenishment. This process shifts inventory management from static response to dynamic adaptation, effectively improving inventory utilization efficiency.

[0080] In step S400, if the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and based on the material transportation parameters, a minimum delay path objective function is constructed to generate the optimal path and control the automated guided vehicle to perform replenishment.

[0081] It should be noted that the material transport parameters include the position, speed, and capacity of the automated guided vehicle.

[0082] In step S400, when the real-time inventory level falls below the dynamic replenishment threshold, the system generates a replenishment instruction and constructs a minimum delay path objective function based on material transportation parameters. Taking into account factors such as task urgency and AGV travel distance, the system generates the optimal replenishment path. Controlling the AGV to perform replenishment along this path prioritizes the material needs of urgent work orders (avoiding production stoppages due to material shortages) while reducing ineffective AGV travel through path optimization (lowering logistics costs). This step ultimately transforms data prediction and dynamic calculation into actual material transportation actions, ensuring real-time matching between line-side warehouse inventory and production needs, thus guaranteeing production continuity and efficiency.

[0083] In some embodiments of this application, the equipment topology diagram is used to characterize the process connections and fault propagation weights between devices. The equipment topology diagram is constructed based on the production process flow diagram of the smart workshop, the physical layout of the equipment, and historical fault data. The equipment topology diagram is in graph structure. The form of representation, in which Indicates a device node. This indicates the process connection relationships between equipment. Indicates the fault propagation weight between devices. The value is the probability of the cascading effect of equipment failures.

[0084] Equipment topology diagrams are a key data foundation for dynamic material pull control in smart workshops. Their core function is to provide accurate data for material demand forecasting and replenishment strategy optimization by quantifying the process relationships and fault propagation relationships between equipment. Specifically, equipment topology diagrams are constructed based on the smart workshop's production process flow diagram (clearly defining the upstream and downstream process relationships between equipment), equipment physical layout (reflecting the spatial relationships between equipment), and historical fault data (statistically calculating the cascading impact probability of equipment faults).

[0085] Device topology diagram in graphical structure The form of representation, in which It indicates equipment nodes and visually identifies various production equipment in the workshop, such as injection molding machines and assembly machines; It indicates the process connection relationships between equipment and clarifies the collaboration logic between equipment; Indicates the fault propagation weight between devices. The value is the probability of the cascading effect of equipment failures. For example, the probability that a failure of equipment A will cause equipment B to stop is 80%. This is used to quantify the impact of equipment malfunctions on related equipment. Through this structured expression, the system can accurately analyze the potential impact of equipment malfunctions (such as failures or maintenance delays) on material consumption (e.g., equipment A downtime leads to material accumulation in equipment B, requiring a reduction in replenishment to equipment B). This allows for dynamic adjustment of inventory levels and replenishment strategies, avoiding the inventory redundancy or material shortage risks caused by neglecting equipment correlation in traditional methods, and effectively improving the predictability and accuracy of material management.

[0086] In some embodiments of this application, scheduling features are used to characterize information about production tasks and their corresponding product material usage.

[0087] In some embodiments of this application, production scheduling features are extracted using a manufacturing execution system (MES) or an enterprise resource planning system (ERP) in a smart workshop.

[0088] The role of production scheduling features is to extract key information about production tasks through MES or ERP systems, transforming abstract production plans into specific material demand indicators to support the accuracy of material consumption forecasting and dynamic replenishment decisions. Specifically, production scheduling features include core information about production tasks (such as work order priority, planned output, and the corresponding bill of materials (BOM) usage for each product), clarifying key requirements such as how many products need to be produced, what materials are needed, and which tasks require priority. By extracting these features, the system can map dynamic changes in production tasks (such as emergency order insertions and output adjustments) to material management in real time: for example, the high priority feature of emergency work orders will increase the weight of their material demand, ensuring priority replenishment; the product BOM usage feature clarifies the quantity of materials required for each product, avoiding supply deviations caused by mismatched BOMs. This process highly aligns material supply with production tasks, avoiding inventory backlogs or material shortages caused by the disconnect between production planning and material management in traditional methods, effectively improving the coordination and responsiveness of material management.

[0089] In some embodiments of this application, the multimodal prediction model includes a long short-term memory network, a graph neural network, and an attention mechanism. In step S200, the process of using the multimodal prediction model to extract time-series production features from real-time production data and equipment topology features from the equipment topology graph, and fusing the time-series production features, equipment topology features, and scheduling features through the attention mechanism to output the predicted material consumption rate includes, but is not limited to, the following steps.

[0090] Step S210: Using a long short-term memory network, extract time-series production features from real-time production data.

[0091] In step S210, the dynamic changes in material consumption are captured from the time-series information of real-time production data. Real-time production data (such as changes in material weight in line-side warehouses over time, and time-series fluctuations in equipment vibration values) are time-dependent. Long Short-Term Memory (LSTM) networks, as a type of neural network adept at processing time-series data, can identify long-term trends (such as gradually accelerating material consumption), short-term fluctuations (such as consumption surges caused by temporary equipment acceleration), or periodic patterns (such as consumption peaks at fixed times each day). Through the time-series production features extracted by LSTM, the system can accurately grasp the actual dynamics of recent material consumption, providing crucial time-dimensional evidence for subsequent predictions and avoiding prediction biases caused by traditional methods neglecting time dependence.

[0092] Step S220: Using a graph neural network, extract device topology features from the device topology graph.

[0093] In step S220, a graph neural network (GNN) is used to analyze the equipment topology, extracting the potential impact features of equipment anomalies (such as malfunctions or maintenance delays) on the material demand of related equipment. For example, a malfunction in equipment A may lead to material accumulation in its downstream equipment B (because equipment A cannot supply semi-finished products), at which point the material demand of equipment B will decrease; or a malfunction in equipment C may cause its upstream equipment D to accelerate production, at which point the material demand of equipment D will increase. These topological features provide key information on the equipment association dimension for predicting material consumption rates, enabling the system to anticipate the impact of equipment anomalies on material demand in advance.

[0094] Step S230: By integrating time-series production characteristics, equipment topology characteristics, and scheduling characteristics through an attention mechanism, the predicted material consumption rate is output.

[0095] In step S230, multi-dimensional features are dynamically integrated to generate prediction results that fit the actual production scenario. Time-series production features reflect the time trend of material consumption (e.g., recent accelerated consumption), equipment topology features reflect the potential impact of equipment anomalies (e.g., equipment A failure may reduce material demand for equipment B), and scheduling features reflect the priority of production tasks (e.g., urgent work orders require priority material supply). An attention mechanism automatically adjusts the importance of these features based on the current production scenario (e.g., whether there are urgent work orders or equipment failures): for example, when there are urgent work orders, the weight of scheduling features increases to ensure that the prediction results prioritize the needs of urgent tasks; when a piece of equipment fails, the weight of equipment topology features increases to correct for demand changes caused by equipment anomalies. Through this dynamic fusion, the predicted material consumption rate output by the system can more accurately reflect actual production needs, providing a reliable basis for subsequent inventory adjustments and replenishment decisions.

[0096] In some embodiments of this application, the process of calculating the real-time inventory level and dynamic replenishment threshold in step S300 by combining the predicted material consumption rate, current material inventory, material inbound rate, basic inventory buffer, and equipment operating parameters includes, but is not limited to, the following steps.

[0097] Step S310: Calculate the dynamic risk compensation coefficient based on the average repair time and average fault interval in the equipment operating parameters.

[0098] In step S310, the reliability risk of the current operating status of the equipment is quantified, providing key risk quantification indicators for subsequent inventory adjustments. The Mean Time To Repair (MTTR) in the equipment operating parameters reflects the average time taken to recover after an equipment failure, while the Mean Time Between Failures (MTBF) reflects the average interval between two equipment failures. By analyzing the ratio of these two parameters, the potential risk of production anomalies caused by equipment failures can be assessed: the longer the MTTR or the shorter the MTBF, the lower the equipment reliability and the higher the risk of production fluctuations caused by failures. The calculation of the dynamic risk compensation coefficient transforms this risk into a quantifiable numerical indicator, providing a decision-making basis from the equipment status dimension for dynamically adjusting inventory buffers and replenishment thresholds.

[0099] Step S320: Calculate the dynamic inventory buffer and dynamic replenishment threshold based on the dynamic risk compensation coefficient and the basic inventory buffer.

[0100] In step S320, the basic inventory buffer is a safety stock benchmark set based on historical normal demand fluctuations and replenishment uncertainties, while the dynamic risk compensation coefficient reflects the quantitative result of current equipment anomaly risks. Through the combined calculation of these two, the dynamic inventory buffer is adjusted according to the equipment risk level based on the basic buffer. The dynamic replenishment threshold, as the critical value for triggering replenishment, is also adjusted based on the dynamic inventory buffer to ensure a high degree of matching with the current equipment status and production needs, thus improving the targeting of the replenishment strategy. The higher the equipment risk, the larger the dynamic inventory buffer and the higher the dynamic replenishment threshold, to reserve more inventory to cope with potential production fluctuations; when equipment operation is stable, the dynamic inventory buffer decreases accordingly, and the dynamic replenishment threshold also decreases accordingly to avoid inventory redundancy.

[0101] Step S330: Calculate the real-time inventory level based on the dynamic inventory buffer, predicted material consumption rate, current material inventory, and material inbound rate.

[0102] In step S330, by integrating multi-dimensional information, the current actual available inventory level of the line-side warehouse is calculated, providing a direct basis for replenishment decisions. The calculation of real-time inventory levels needs to consider: the predicted material consumption rate (reflecting material demand within a future preset time window), the current material inventory (existing material reserves in the line-side warehouse), the material inbound rate (materials issued but not yet delivered), and the dynamic inventory buffer (extra inventory to cope with equipment malfunction risks). Through the comprehensive calculation of the above parameters, the real-time inventory level can accurately reflect the actual supply capacity of the line-side warehouse after considering future demand, existing reserves, and risk buffers. If the real-time inventory level is lower than the dynamic replenishment threshold, it indicates that the existing inventory cannot meet future demand, and replenishment must be triggered immediately; conversely, it indicates that the inventory is sufficient, avoiding resource waste caused by excessive replenishment. This step transforms the prediction results and real-time data into actionable inventory indicators, ensuring a precise match between inventory management and actual production needs.

[0103] In some embodiments of this application, the real-time inventory level satisfies the following formula (1): (1);

[0104] In formula (1), Indicates real-time inventory level; Indicates the current material inventory level; Indicates the rate at which materials are received into the warehouse. Indicates the time window for material receipt; This indicates the predicted material consumption rate. Indicates the time window for material consumption; Indicates dynamic inventory buffer. It is the product of the dynamic risk compensation coefficient and the basic inventory buffer. The dynamic risk compensation coefficient is calculated based on the average repair time and average interval between failures in the equipment operating parameters.

[0105] As can be seen from formula (1), the real-time inventory level calculation formula integrates multi-dimensional data to achieve accurate quantification of the actual available inventory in the line-side warehouse, providing a key basis for dynamic replenishment decisions. Specifically, the calculation of the real-time inventory level comprehensively considers the following core elements: the current material inventory (the existing material reserves in the line-side warehouse) reflects the inventory that can be directly used at present; the material inbound rate (materials that have been issued but not yet delivered) reflects the inventory that will be replenished; the product of the predicted material consumption rate and the preset time window (the expected material consumption in the future period) clarifies the future demand; and the dynamic inventory buffer (the product of the risk compensation coefficient calculated based on equipment operating parameters and the basic inventory buffer) reserves a safety margin to cope with demand fluctuations caused by equipment anomalies (such as failures or maintenance delays). Through the comprehensive calculation of the above elements, the real-time inventory level can accurately reflect the actual supply capacity of the line-side warehouse after considering existing reserves, materials in transit, future demand, and risk buffers, ensuring that the inventory level meets the needs of production continuity while avoiding resource waste caused by excessive replenishment, effectively improving the dynamic adaptability and accuracy of material management.

[0106] In some embodiments of this application, the dynamic replenishment threshold satisfies the following formula (2):

[0107] (2);

[0108] In formula (2), Indicates the dynamic replenishment threshold. Indicates the lead time for replenishment. This represents the average historical material consumption rate. This refers to the average repair time in the equipment's operating parameters. This refers to the mean time between failures (MTBF) in the equipment's operating parameters. This represents the preset average interval between failures (ABER) threshold. This indicates the preset average repair time threshold.

[0109] As can be seen from formula (2), the calculation method of dynamic replenishment threshold realizes the dynamic adjustment of replenishment trigger conditions by combining equipment reliability status and historical demand characteristics, ensuring that inventory management is highly adapted to actual production needs. Specifically, the setting of dynamic replenishment threshold is based on the comparison results of the average repair time (reflecting the time taken to recover after equipment failure) and average failure interval time (reflecting the frequency of equipment failure) in the equipment operating parameters with the preset threshold, and is divided into two scenarios: When the average repair time of the equipment does not exceed the preset average repair time threshold and the average failure interval time of the equipment is higher than the preset average failure interval time threshold, the equipment reliability is high and the equipment operating status is relatively stable. The dynamic replenishment threshold is calculated by multiplying the replenishment lead time and the average of historical material consumption rates and adding the dynamic inventory buffer. When the average repair time of the equipment exceeds the preset average repair time threshold and the average failure interval time of the equipment is not higher than the preset average failure interval time threshold, the equipment reliability is low. In this case, the dynamic replenishment threshold is further increased on the above basic value to reserve more inventory to cope with potential material shortage risks.

[0110] Therefore, by dynamically increasing the buffer based on the ratio of equipment reliability parameters, production continuity is ensured while avoiding inventory backlog caused by excessive conservatism. This dynamic adjustment mechanism allows replenishment thresholds to flexibly adapt to changes in equipment status, effectively improving the accuracy and resilience of material management.

[0111] In some embodiments of this application, the dynamic risk compensation coefficient satisfies the following formula (3):

[0112] (3);

[0113] In formula (3), This represents the dynamic risk compensation coefficient. Indicates the remaining time for the current work order. This represents the process cycle constant.

[0114] As can be seen from formula (3), the dynamic risk compensation coefficient achieves accurate quantification of inventory risk by comprehensively considering the equipment reliability status and the remaining time of the work order, providing a key basis for dynamically adjusting the inventory buffer. Specifically, the calculation of this coefficient combines the average equipment repair time, average interval between failures, the remaining time of the current work order (reflecting the urgency of the production task), and the process cycle constant (the standard time for the equipment to complete a single process): the longer the average equipment repair time or the shorter the average interval between failures (the less reliable the equipment), the shorter the remaining time of the current work order (the more urgent the production task), the larger the dynamic risk compensation coefficient, indicating that more inventory buffer is needed to cope with the material demand fluctuations that may be caused by equipment failure (such as production suspension during maintenance, followed by a concentrated surge in demand when production resumes); conversely, the higher the equipment reliability or the more sufficient the remaining time of the work order, the smaller the dynamic risk compensation coefficient, avoiding inventory backlog caused by excessive buffering. Through this quantification mechanism, the system can dynamically adjust the inventory strategy according to the actual situation of equipment status and production tasks, effectively improving the accuracy and risk resistance of material management.

[0115] In some embodiments of this application, the basic inventory buffer satisfies the following formula (4):

[0116] (4);

[0117] In formula (4), This indicates the basic inventory buffer. This represents the tolerance coefficient for material breakage. This indicates the variance of the replenishment lead time. This represents the standard deviation of historical material consumption rates.

[0118] As can be seen from formula (4), the calculation method of basic inventory buffer provides a basic guarantee for inventory management to cope with routine risks by comprehensively considering the uncertainty of demand fluctuations and replenishment time. It is a key design to ensure production continuity and on-time delivery. Specifically, the determination of basic inventory buffer combines the following core elements: the material shortage tolerance coefficient is determined by the enterprise's tolerance for material shortage risk. For example, a material shortage tolerance coefficient of 97% means that the inventory will not be lower than the demand in 97% of the time, which means that there may be only 3 times out of every 100 replenishments that the material shortage may occur due to insufficient inventory; replenishment lead time (the average time from issuing a replenishment request to the delivery of materials), the standard deviation of historical material consumption rate (reflecting the degree of fluctuation in material demand; the larger the standard deviation, the more unstable the demand), and the variance of replenishment lead time (reflecting the degree of fluctuation in replenishment time; the larger the variance, the more uncertain the replenishment time).

[0119] By integrating and calculating these factors, the basic inventory buffer can quantify the combined impact of demand fluctuations and replenishment time uncertainties on inventory, setting a reasonable safety stock benchmark for line-side warehouses: when demand fluctuates greatly or replenishment time is unstable, the basic buffer increases accordingly to ensure sufficient inventory to cope with unexpected demand; when demand is stable or replenishment time is reliable, the basic buffer is appropriately reduced to avoid inventory backlog. This calculation method provides a scientific basis for dynamically adjusting inventory strategies, effectively balancing production continuity and inventory costs, and effectively improving the level of precision in materials management.

[0120] In some embodiments of this application, the minimum delay path objective function satisfies the following formula (5):

[0121] (5);

[0122] In formula (5), This represents the objective function value of the minimum delay path. This indicates the number of replenishment tasks in the smart workshop. This indicates that the automated guided vehicle has completed the first... The time required for each replenishment task Indicates the first The deadline for each replenishment task. Indicates the first The delay in completing each replenishment task. This represents the distance weighting coefficient, used to adjust the importance of the total travel distance of the automated guided vehicle in the minimum delay path objective function value. Indicates the first The automated guided vehicle completed the first... The driving distance for each replenishment task. This indicates the number of automated guided vehicles. , Indicates the first The replenishment task was assigned to the first... An automated guided vehicle is used for this purpose. Indicates the first The replenishment task is not assigned to the first... An automated guided vehicle is used for this purpose.

[0123] As can be seen from formula (5), the minimum delay path objective function provides a scientific optimization basis for the path planning of automated guided vehicles (AGVs) by comprehensively considering task timeliness and logistics costs, effectively improving the execution efficiency and resource utilization of replenishment tasks. Specifically, the core logic of this objective function is to minimize the weighted sum of two parts: the first part is the sum of the delay times of all replenishment tasks (only calculating the part where the actual completion time exceeds the deadline), ensuring that tasks are completed as on time as possible and avoiding production line material shortages due to delays; the second part is the weighted sum of the total travel distance of all AGVs performing tasks (the weights are adjusted by the distance weight coefficient β), controlling the ineffective travel of AGVs and reducing logistics energy consumption and equipment wear. Through the constraint of 0-1 variables (indicating whether a task is performed by a certain AGV), the system can dynamically allocate tasks to different AGVs and balance the load of each AGV. The design of this objective function enables AGV path planning to balance task timeliness and logistics costs. The distance weight coefficient value is adjusted according to actual needs (such as reducing the distance weight coefficient in emergency scenarios to prioritize timeliness), ultimately achieving efficient and economical execution of replenishment tasks and effectively improving the reliability of workshop material supply and resource utilization efficiency.

[0124] In some embodiments of this application, the minimum delay path objective function must satisfy three constraints, including a first constraint, a second constraint, and a third constraint.

[0125] The three constraints of the minimum delay path objective function, by clearly defining task allocation rules and execution restrictions, jointly ensure the feasibility and efficiency of Automated Guided Vehicle (AGV) scheduling, ensuring that the optimization results of the objective function meet the needs of actual production scenarios. Specifically: The first constraint limits the number of tasks or load performed by each AGV, avoiding equipment failure or efficiency decline due to AGV overload or too many tasks, ensuring the safety and stability of AGV operation; the second constraint stipulates that each replenishment task can only be performed by one AGV, avoiding task duplication or omission, ensuring the uniqueness and accuracy of material supply; the third constraint, by specifying the time sequence and spatial restrictions of task execution (such as the current task can only begin after the preceding task is completed, and the travel time must meet the relationship between distance and speed), avoids conflicts between multiple AGVs in the same time or space area, ensuring the continuity of task execution and the rationality of path planning.

[0126] These three constraints limit the objective function from three dimensions: equipment load, task allocation, and spatiotemporal coordination. This ensures that the optimized AGV path minimizes delay and travel distance while conforming to the actual operating rules of the workshop, effectively improving the reliability and efficiency of replenishment task execution.

[0127] In some embodiments of this application, the first constraint condition satisfies the following formula (6):

[0128] (6);

[0129] In formula (6), Indicates the first The task capacity of an automated guided vehicle.

[0130] As can be seen from formula (6), the first constraint, by limiting the task load of each AGV, ensures the safety and efficiency of AGV operation and is a key rule to ensure the reliable execution of replenishment tasks. Specifically, this constraint stipulates that the number of tasks assigned to each AGV (represented by 0-1 variables indicating whether a task is executed) must not exceed its preset task capacity (such as a maximum of 5 tasks executed at a time or a maximum load limit). This restriction avoids problems such as AGV overload, excessive travel time, or increased equipment wear caused by excessive task allocation, ensuring that the AGV can execute tasks in its optimal design state.

[0131] Meanwhile, the first constraint balances the task load among multiple AGVs, preventing resource waste where some AGVs are delayed due to excessive tasks while others remain idle, thus improving overall scheduling efficiency. Ultimately, this constraint, by standardizing the upper limit of AGV task allocation, ensures the stability and reliability of replenishment task execution, providing a fundamental guarantee for the continuity of material supply in the workshop.

[0132] In some embodiments of this application, the second constraint condition satisfies the following formula (7):

[0133] (7);

[0134] As can be seen from formula (7), the second constraint means that any replenishment task in the smart workshop is assigned to only one automated guided vehicle (AGV) for execution, ensuring the uniqueness and accuracy of task allocation, which is a key rule to ensure the orderly supply of materials. Specifically, this constraint requires that any replenishment task must be executed by only one AGV (limited by 0-1 variables), avoiding duplicate transportation caused by multiple AGVs executing the same task at the same time (such as multiple deliveries of the same batch of materials, resulting in redundant inventory in the line-side warehouse) or task omission (such as material shortage caused by no AGV executing a certain task).

[0135] Meanwhile, the unique assignment clearly defines the correspondence between tasks and AGVs, facilitating real-time tracking of task execution status (e.g., which AGV is responsible for which task, and its current progress). When a task is delayed or encounters an anomaly, the responsible AGV can be quickly located and emergency measures can be taken (e.g., adjusting the route or dispatching other AGVs to assist), effectively improving the efficiency of replenishment tasks and the ability to respond to anomalies. This constraint, by standardizing the uniqueness of task assignment, ensures the accuracy of material supply and the efficient use of logistics resources, providing an important guarantee for the continuity of workshop production.

[0136] The third constraint condition satisfies the following formula (8):

[0137] (8);

[0138] In formula (8), Indicates the first The automated guided vehicle began its first... The time allotted for each replenishment task. Indicates the first The automated guided vehicle completed the first... The time allotted for each replenishment task. Indicates the first The location of the first replenishment task and the first The distance between the locations of each replenishment task. Indicates the first The moving speed of an automated guided vehicle. Indicates the first The number of tasks that an automated guided vehicle (AGV) needs to perform.

[0139] As can be seen from formula (8), the third constraint, by regulating the temporal and spatial relationships of AGV task execution, ensures the continuity of task execution and the feasibility of path planning, and is a key rule to avoid AGV spatiotemporal conflicts. Specifically, this constraint stipulates that the AGV begins executing the first... The time required for each replenishment task must be at least equal to the time required to complete the previous task (the first replenishment task). The time for the first task, plus the time from the first task The mission location to the first The travel time to each task location is calculated based on the distance between the two locations and the AGV's moving speed. This rule ensures that the AGV can only start a subsequent task after completing the preceding task in terms of time, avoiding task confusion caused by time overlap. At the same time, by limiting the travel time, it ensures that the AGV has enough time to move from one task location to the next, avoiding path conflicts caused by insufficient spatial distance (such as appearing at two task locations simultaneously). Through this constraint, the system can generate AGV paths that conform to actual operating rules, improving the reliability and efficiency of task execution and ensuring the continuity and accuracy of material supply.

[0140] Secondly, this application provides an intelligent workshop material dynamic pull control system, including a data acquisition module 410, a material consumption prediction module 420, and a replenishment control module 430.

[0141] The data acquisition module 410 is used to acquire the basic inventory buffer, equipment topology diagram, production scheduling characteristics, real-time production data, and material inbound rate of the smart workshop. The real-time production data includes the current material inventory, equipment operating parameters, and material transportation parameters. The production scheduling characteristics are used to characterize information about production tasks and their corresponding product material usage.

[0142] The material consumption prediction module 420 uses a multimodal prediction model to extract time-series production features from real-time production data and equipment topology features from equipment topology diagrams. It then integrates time-series production features, equipment topology features, and scheduling features through an attention mechanism to output the predicted material consumption rate.

[0143] The replenishment control module 430 combines the predicted material consumption rate, current material inventory, material inbound rate, basic inventory buffer, and equipment operating parameters to calculate the real-time inventory level and dynamic replenishment threshold. If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated. Based on the material transportation parameters, a minimum delay path objective function is constructed to generate the optimal path and control the automated guided vehicle to perform the replenishment.

[0144] Furthermore, this application provides an intelligent workshop material dynamic pull control device, including: an Internet of Things sensing device 510, an edge computing device 520, a central decision-making device 530, and an automated guided vehicle 540.

[0145] The IoT sensing device 510 is used to collect real-time production data in the smart workshop. This real-time production data includes current material inventory levels, equipment operating parameters, and material transportation parameters.

[0146] The edge computing device 520 is equipped with an INT8-quantized multimodal prediction model, which is invoked via the gRPC interface.

[0147] Deploying INT8-quantized multimodal prediction models on edge computing devices and calling them via the gRPC interface is a key technological support for achieving real-time and efficient dynamic pull control of materials in intelligent workshops. Specifically, the INT8 quantization model effectively reduces the model size (typically by more than 75%) by compressing the weights and activation values ​​of the original 32-bit floating-point model into 8-bit integers, thereby reducing the storage and computing resource consumption of edge computing devices and improving inference speed (typically by more than 2 times). This enables the model to quickly process real-time production data (such as line-side warehouse inventory, equipment vibration, etc.) with limited hardware resources, meeting the stringent real-time requirements of workshop material management.

[0148] The gRPC interface, as a high-performance remote procedure call framework, achieves low-latency, high-throughput communication via the HTTP / 2 protocol. This ensures efficient and reliable data transmission and model invocation between edge computing devices and other devices, avoiding the prediction lag issues caused by latency or bandwidth limitations in traditional interfaces. Working together, the gRPC interface enables the multimodal prediction model to extract time-series production characteristics, equipment topology characteristics, and scheduling characteristics in real time, outputting accurate material consumption rates. This provides timely and reliable decision-making support for dynamic inventory level calculations and replenishment strategy generation, ultimately ensuring the continuity of material supply in the workshop and achieving lean management goals.

[0149] The edge computing device 520 is used to acquire the basic inventory buffer, equipment topology map, production scheduling characteristics and material inbound rate of the smart workshop. Using a multimodal prediction model, it extracts time-series production characteristics from real-time production data and equipment topology characteristics from equipment topology map. It then fuses time-series production characteristics, equipment topology characteristics and production scheduling characteristics through an attention mechanism to output the predicted material consumption rate.

[0150] The central decision-making device 530 combines the predicted material consumption rate, current material inventory, material inbound rate, basic inventory buffer, and equipment operating parameters to calculate the real-time inventory level and dynamic replenishment threshold. If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and based on the material transportation parameters, a minimum delay path objective function is constructed to generate the optimal path, controlling the automated guided vehicle 540 to perform replenishment.

[0151] In some embodiments of this application, the Internet of Things sensing device 510 includes a line-side warehouse sensing unit 511, an equipment status sensing unit 512, and a logistics sensing unit 513.

[0152] The line-side warehouse sensing unit 511 is used to obtain the current material inventory. The line-side warehouse sensing unit 511 includes a smart bin, which comprises a weighing sensor and an RFID reader / writer.

[0153] The function of the line-side warehouse sensing unit 511 is to obtain the current material inventory level of the line-side warehouse in real time and accurately, providing basic data support for inventory level calculation and replenishment decisions. This unit is implemented through a smart bin, which integrates a weighing sensor and an RFID reader: the weighing sensor detects the material weight in real time (e.g., with a range of 0-500kg) and, combined with material density or unit weight conversion, obtains the current material quantity in the line-side warehouse; the RFID reader reads the material tags to obtain information such as the material work order number and material ID, ensuring the binding relationship between inventory data and production work orders. The two work together to ensure the real-time nature of inventory quantity and achieve traceability of material source and destination (e.g., matching work order requirements through tag information), avoiding inventory management deviations caused by material mismatch or data errors.

[0154] The equipment status sensing unit 512 is used to collect equipment operating parameters. Deployed in the production equipment of the smart workshop, the equipment status sensing unit includes a vibration sensor and a PLC interface.

[0155] The equipment status sensing unit 512 is used to collect operating parameters of production equipment, providing key data for equipment reliability analysis and dynamic risk compensation. This unit is deployed on production equipment (such as injection molding machines and assembly machines) and integrates vibration sensors and a PLC interface: the vibration sensors sample equipment vibration values ​​at high frequencies (e.g., 1kHz) and combine this with Fast Fourier Transform analysis (e.g., frequency band 50-500Hz) to identify equipment malfunctions (e.g., bearing wear, loose parts); the PLC interface communicates with the equipment control system (e.g., Modbus TCP protocol) to directly obtain reliability parameters such as the mean time between failures (MTBF) and mean time to repair (MTTR). Using this data, the system can quantify equipment failure risks (e.g., the longer the MTTR and the shorter the MTBF, the less reliable the equipment), and dynamically adjust inventory buffers and replenishment thresholds to avoid material shortages or inventory redundancy caused by equipment malfunctions.

[0156] The logistics sensing unit 513 is used to collect the position, speed, and vehicle capacity of the automated guided vehicle 540 as material transport parameters. The logistics sensing unit is deployed on the automated guided vehicle 540 and includes a LiDAR, a visual SLAM processor, and a UWB chip.

[0157] The logistics sensing unit 513 is used to collect real-time status data of the automated guided vehicles (AGVs), providing a logistics-dimensional decision-making basis for route optimization and replenishment execution. Deployed on the AGV, this unit integrates a LiDAR, a visual SLAM processor, and a UWB chip: the LiDAR scans the environment (accuracy ±2cm) by emitting laser pulses to detect obstacles (such as temporarily stacked materials) in real time; the visual SLAM processor analyzes images captured by cameras (such as workshop scene features) to construct an environmental map and locate the AGV's position; and the UWB chip achieves centimeter-level high-precision positioning through communication with the workshop base station.

[0158] The three work together to obtain the AGV's position, speed (e.g., 0.5-2m / s) and vehicle capacity (e.g., 1.2m³ / 800kg in standard mode and 2.5m³ / 1500kg in extended mode) in real time. This provides logistics status input for the minimum delay path objective function, ensuring the real-time performance and accuracy of AGV path planning and avoiding replenishment delays or path conflicts caused by lagging logistics information.

[0159] In some embodiments of this application, the technical capabilities of the intelligent workshop material dynamic pull control method, system, and device are not limited to material management in a single workshop, but can be extended to material collaborative transportation and inventory control at the entire supply chain level. Through hierarchical functional positioning, parameterized dynamic calculation, and a collaborative replenishment process across the entire chain, precise control is achieved from suppliers to production lines, effectively improving the efficiency and reliability of the supply chain. Taking the supply chain of battery cells and protection boards as an example, a detailed explanation is provided from three aspects: supply chain hierarchical structure, core parameter definition, and collaborative replenishment process.

[0160] First, the supply chain used in this application embodiment includes four core levels, each with clearly defined functions and inventory types, forming a material flow link between supplier warehouses, central warehouse storage warehouses, central warehouse available warehouses, workshop line-side warehouses, and production lines, ensuring the precise flow of materials from the source of production to the end of the production line.

[0161] Specifically, the supplier warehouse, as the starting point of the supply chain, adopts a Just-In-Time (JIT) production model and delivers materials directly to the central warehouse. Its core function is to produce to order and deliver materials on time. The warehouse's inventory consists of active battery cells and protection boards, which must be strictly matched with the production plan to avoid overproduction and inventory backlog. Its core parameters include production cycle time and delivery reliability: production cycle time is the time to complete an order, for example, the supplier needs 3 days to complete production; delivery reliability is the probability of on-time delivery, such as a 95% on-time delivery rate, used to assess the supplier's stability and adjust replenishment lead time.

[0162] The central warehouse serves as a transit node between suppliers and available warehouses. For example, the central warehouse receives battery cells directly from suppliers and stores them refrigerated. Its core function is to complete the material settling process and quality inspection. The inventory type at this node is refrigerated battery cells, which need to undergo a 24-hour settling process (e.g., the battery cells need to stabilize their performance in a low-temperature environment) to ensure that the materials transferred to the available warehouse meet production requirements. Its key parameters include settling buffer and supplier replenishment cycle: the settling buffer is the inventory buffer that needs to be reserved during the settling process, for example, materials that cannot be released during the settling period need to be stockpiled in advance; the supplier replenishment cycle is the time period from placing an order with the supplier to the delivery of materials to the warehouse, for example, the supplier needs 3 days to complete production and transportation, which is used to dynamically adjust the warehouse's replenishment strategy.

[0163] The central warehouse, serving as an upstream replenishment node to the line-side warehouse, primarily stores ambient-temperature usable battery cells that have undergone 24-hour settling. Its core function is to provide rapidly responding usable materials to the line-side warehouse. This node's inventory type is ambient-temperature usable battery cells, and it's crucial to ensure that the materials have passed a settling test (e.g., the cells need to stabilize their performance) to prevent production anomalies caused by insufficient settling. Key parameters include the available inventory threshold and AGV response time: the available inventory threshold is the minimum inventory level required to trigger replenishment to the storage warehouse; when the inventory drops to this value, a replenishment request must be initiated to the central warehouse. The AGV response time is the time constraint from receiving the workshop request to completing material delivery; for example, the AGV delivers every 2 hours to ensure timely replenishment of the line-side warehouse's inventory.

[0164] The workshop-side warehouse directly serves the production line (e.g., 20 parallel production lines). Its core function is to replenish materials to the production line in real time as needed, ensuring production continuity. Its inventory types are battery cells and protection boards, stored at room temperature. Because they are directly exposed to the production site, they need to respond quickly to the immediate needs of the production line. Its core parameters include: replenishment point, the minimum inventory threshold that triggers replenishment; when inventory falls to this value, a replenishment request is automatically initiated from the available warehouse in the central warehouse; safety stock, a buffer against demand fluctuations to prevent material shortages due to short-term demand surges; and demand fluctuation, a parameter reflecting production uncertainty (e.g., equipment failure, production cycle fluctuations), used to dynamically adjust the safety stock.

[0165] Secondly, each level achieves precise inventory control through quantitative parameters. Taking the workshop-side warehouse as an example, the definition and calculation of its core parameters directly determine the scientific nature of the replenishment strategy.

[0166] For example, the average demand is 40,000 units per hour, calculated from the demand of 2,000 units per hour per line across 20 production lines. This reflects the average level of regular production demand and is based on historical production data (such as the average consumption rate over the past 30 days). It forms the basis for calculating safety stock and replenishment points. The standard deviation of demand is 4,800 units per hour, calculated by combining production fluctuations and equipment failure rates. Production fluctuations, such as temporary increases in orders, and equipment failures, such as a production line shutdown causing other lines to accelerate, can cause demand to deviate from the average. The larger the standard deviation, the higher the safety stock needs to be reserved. The replenishment lead time is 2 hours, referring to the transportation time of AGVs from the available warehouse in the central warehouse to the line-side warehouse in the workshop. This determines the time window from triggering replenishment to material delivery. The longer the lead time, the higher the safety stock needs to be reserved, because more material may have been consumed before it arrives. The material shortage tolerance coefficient is determined by the company's tolerance for the risk of material shortage. For example, a material shortage tolerance coefficient of 97% means that there is a 97% probability that the inventory will not be lower than the demand, which means that there may be only 3 out of every 100 replenishments that result in material shortage due to insufficient inventory.

[0167] Then, coordinated replenishment at each level is achieved through the linkage between dynamic inventory levels and replenishment thresholds. The specific process is as follows: Replenishment is triggered at the workshop line-side warehouse. When the inventory in the line-side warehouse drops to the replenishment point (less than or equal to 160,000 units), the system automatically sends a replenishment request to the available warehouse in the central warehouse. This threshold is set based on the average demand of 40,000 units / hour and the replenishment lead time of 2 hours, i.e., the consumption in 2 hours under normal demand is 40,000 × 2 = An 80,000-unit safety stock ensures sufficient inventory to sustain the material until delivery when replenishment is triggered. The central warehouse responds to and coordinates with the storage warehouse. Upon receiving a request, the available warehouse releases the dormant battery cells (available at room temperature) to complete delivery. Simultaneously, the system checks if the available warehouse inventory has fallen below the available inventory threshold (less than or equal to 880,000 units). If so, a replenishment request is triggered to the central warehouse storage warehouse. This threshold is based on the average demand of 40,000 units / hour for line-side warehouses and the AGV delivery frequency of once every 2 hours, ensuring the available warehouse can cover multiple delivery needs. For example, 880,000 units can support 22 deliveries, avoiding frequent requests to the storage warehouse. The central warehouse storage warehouse places an order with the supplier. After receiving the replenishment request from the available warehouse, the storage warehouse... Check if your own inventory has dropped to the storage warehouse replenishment threshold, which is less than or equal to 2,448,000 units. If it does, the system will place an order with the supplier. This threshold is set based on a 24-hour resting period and a 3-day supplier replenishment period to ensure that the storage warehouse still has enough inventory to support the demand of the available warehouse during the 24-hour resting period. For example, 2,448,000 units can support 61.2 hours of resting demand. The supplier completes the entire closed-loop production and delivery process. After receiving the order, the supplier completes production according to the 3-day production cycle and transports the materials to the central warehouse storage warehouse. After the 24-hour resting process is initiated, the materials are transferred to the central warehouse available warehouse to wait for replenishment requests from the workshop line-side warehouse, forming a complete closed-loop process of supplier warehouse, central warehouse storage warehouse, central warehouse available warehouse, workshop line-side warehouse and production line.

[0168] The dynamic pull control technology provided in this application enables real-time synchronization of inventory status and replenishment needs at each level of the supply chain, effectively improving the overall efficiency and reliability of the supply chain. It reduces inventory backlog by dynamically calculating demand mean, standard deviation, and material shortage tolerance coefficient. Each level initiates replenishment only when a threshold is triggered, avoiding inventory redundancy caused by advance stockpiling in traditional models. For example, the central warehouse only requests replenishment from the storage warehouse when inventory falls below a threshold, rather than periodically replenishing in bulk. It reduces the risk of material shortages. The high-frequency delivery by AGVs every two hours and the multi-level replenishment triggering mechanism (from line-side warehouse to available warehouse to storage warehouse to supplier) ensure the continuity of material supply to the production line. For example, when the inventory in the workshop line-side warehouse falls below a threshold, it responds immediately, avoiding material shortages due to transportation delays. It optimizes supply chain efficiency. The linkage between the JIT mode and the supplier's production cycle reduces material transit time and idle waiting time. For example, suppliers only produce when replenishment is triggered in the storage warehouse, avoiding overproduction. Simultaneously, the precise scheduling of AGV deliveries every two hours improves the utilization rate of logistics resources.

[0169] In summary, the embodiments of this application, through hierarchical functional positioning, parameterized dynamic calculation, and a collaborative replenishment process across the entire supply chain, achieve precise control of materials from suppliers to production lines. This provides key technical support for supply chain collaboration and lean management in the manufacturing industry and effectively solves the core contradiction of coexisting inventory backlog and material shortage risks in traditional supply chains.

[0170] In summary, the intelligent workshop material dynamic pull control method, system, and device provided in this application have the following technical effects.

[0171] The intelligent workshop material dynamic pull control method, system, and device provided in this application effectively improve the accuracy and dynamic adaptability of material management through multimodal data fusion and AI intelligent prediction technology. On the one hand, the multimodal prediction model, which integrates LSTM, GNN, and attention mechanisms, accurately captures the short-term trend of material consumption, the impact of equipment failure, and the priority of production tasks, controlling the prediction error within 8.5%, thus providing a reliable basis for dynamic decision-making. Furthermore, by dynamically adjusting the inventory buffer and replenishment threshold in conjunction with equipment operating parameters (MTBF / MTTR), the buffer is increased when equipment reliability is low to avoid material shortages, and decreased when equipment is stable to reduce backlogs. This reduces the proportion of line-side warehouse inventory and stagnant materials, minimizes downtime due to material shortages, and effectively balances inventory costs and production continuity.

[0172] On the other hand, based on the material transportation parameters, a minimum delay path objective function and multiple constraints (AGV capacity, task uniqueness, and spatiotemporal conflicts) are constructed. The system dynamically allocates replenishment tasks and generates the optimal path, balancing timeliness and logistics costs, reducing AGV task response delays, improving inventory turnover, and effectively enhancing logistics efficiency. At the same time, IoT sensing devices achieve full-scene data collection, edge computing devices ensure real-time performance through the INT8 quantization model and gRPC interface, and the central decision-making device dynamically responds to abnormal scenarios, providing overall system availability and comprehensively enhancing the risk resistance and production assurance level of workshop material management.

[0173] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0174] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of ordinary skill of an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary skill. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0175] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0176] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0177] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or, if necessary, processing in a suitable manner, and then stored in computer memory.

[0178] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0179] In the foregoing description of this specification, the reference to terms such as "one embodiment / implementation," "another embodiment / implementation," or "certain embodiments / implementations," etc., indicates that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in an embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0180] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0181] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of the present invention.

Claims

1. A smart plant material dynamic pull control method, characterized in that, The method comprises the following steps: Obtaining the basic inventory buffer of the smart workshop, the device topology graph, the scheduling characteristics, the real-time production data and the material storage rate; wherein the real-time production data comprises the current material inventory, the device operation parameters and the material transportation parameters; the scheduling characteristics are used to represent the information of the production task and the corresponding product material consumption; Using a multi-modal prediction model to extract the time-series production characteristics from the real-time production data, and to extract the device topology characteristics from the device topology graph, and to fuse the time-series production characteristics, the device topology characteristics and the scheduling characteristics through an attention mechanism to output the predicted material consumption rate; Combining the predicted material consumption rate, the current material inventory, the material storage rate, the basic inventory buffer and the device operation parameters to calculate the real-time inventory level and the dynamic replenishment threshold; If the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and according to the material transportation parameters, a minimum delay path objective function is constructed to generate an optimal path to control the automated guided vehicle to execute replenishment.

2. The smart plant material dynamic pull control method of claim 1, wherein, The device topology graph is used to represent the process connection between devices and the fault propagation weight; the device topology graph is constructed based on the production process flow chart, the device physical layout and the historical fault data of the smart workshop; The device topology graph is represented in a graph structure , wherein represents a device node, represents a process connection relationship between devices, represents a failure propagation weight between devices, the value of which is a device failure cascade influence probability.

3. The smart plant material dynamic pull control method of claim 1, wherein, The multi-modal prediction model comprises a long short-term memory network, a graph neural network and an attention mechanism; The multi-modal prediction model comprises a long short-term memory network, a graph neural network and an attention mechanism; The multi-modal prediction model comprises a long short-term memory network, a graph neural network and an attention mechanism; The long short-term memory network is used to extract the time-series production characteristics from the real-time production data; The graph neural network is used to extract the device topology characteristics from the device topology graph; 4. The smart plant material dynamic pull control method of claim 1, wherein, The attention mechanism is used to fuse the time-series production characteristics, the device topology characteristics and the scheduling characteristics to output the predicted material consumption rate. The dynamic risk compensation coefficient is calculated according to the average repair time and the average fault interval time in the device operation parameters; The dynamic inventory buffer and the dynamic replenishment threshold are calculated according to the dynamic risk compensation coefficient and the basic inventory buffer; The real-time inventory level is calculated according to the dynamic inventory buffer, the predicted material consumption rate, the current material inventory and the material storage rate; The real-time inventory level satisfies the following formula: ； wherein, represents the real-time inventory water level; represents the current material inventory amount; represents the material storage rate, represents a material storage time window; represents the predicted material consumption rate, represents a material consumption time window; represents a dynamic inventory buffer amount, is a product of a dynamic risk compensation coefficient and the base inventory buffer amount, the dynamic risk compensation coefficient being calculated according to an average repair time and an average failure interval time in the equipment operation parameters; The dynamic replenishment threshold satisfies the following formula: ； wherein, represents the dynamic replenishment threshold, represents a replenishment lead time, represents a mean of historical material consumption rate, represents a mean repair time in the equipment operating parameters, represents a mean time between failures in the equipment operating parameters; represents a preset mean time between failures threshold, represents a preset mean repair time threshold.

5. The smart plant material dynamic pull control method of claim 4, wherein, The dynamic risk compensation coefficient satisfies the following formula: ； wherein, represents a dynamic risk compensation factor, represents the remaining time of the current work order, represents a process cycle constant; The basic inventory buffer satisfies the following formula: ； wherein, represents the base inventory buffer quantity, represents the stock-out tolerance coefficient; represents the replenishment lead-time variance; represents the standard deviation of the historical material consumption rate.

6. The smart plant material dynamic pull control method of claim 1, wherein, The minimum delay path objective function satisfies the following formula: ； wherein, represents a minimum delay path objective function value; represents a number of replenishment tasks of the smart warehouse; represents a time for the automated guided vehicle to complete the th replenishment task, represents a deadline of the th replenishment task, represents a time for the automated guided vehicle to complete the th replenishment task with delay; represents a distance weight coefficient for adjusting importance of a total travel distance of the automated guided vehicle in the minimum delay path objective function value; represents a travel distance of the th automated guided vehicle to complete the th replenishment task, represents a number of the automated guided vehicles; , represents that the th replenishment task is assigned to the th automated guided vehicle to execute, represents that the th replenishment task is not assigned to the th automated guided vehicle to execute.

7. The smart plant material dynamic pull control method of claim 6, wherein, The minimum delay path objective function needs to meet three constraint conditions, including a first constraint condition, a second constraint condition, and a third constraint condition; The first constraint condition satisfies the following formula: ； wherein, represents the task capacity of the th automated guided vehicle. The second constraint condition satisfies the following formula: ； The second constraint condition indicates that any one replenishment task of the intelligent workshop is only assigned to one automated guided vehicle for execution; The third constraint condition satisfies the following formula: ； wherein, represents the time when the first of the automated guided vehicles starts to perform the first restocking task; represents the time when the first of the automated guided vehicles performs the first restocking task to completion; represents the time when the first of the automated guided vehicles performs the first restocking task to completion; represents the distance between the location of the first restocking task and the location of the second restocking task; represents the distance between the location of the first restocking task and the location of the second restocking task; represents the moving speed of the first of the automated guided vehicles; represents the number of tasks to be performed by the first of the automated guided vehicles.​​​​​​ 8. A smart plant material dynamic pull control system, characterized by, It comprises a data acquisition module, a material consumption prediction module, and a replenishment control module. The data acquisition module is used to acquire the basic inventory buffer of the intelligent workshop, a device topology graph, production scheduling characteristics, real-time production data, and a material storage rate; wherein the real-time production data includes a current material inventory, device operating parameters, and material transportation parameters; the production scheduling characteristics are used to represent information of a production task and corresponding product material consumption; The material consumption prediction module is used to extract time-series production characteristics from the real-time production data and device topology characteristics from the device topology graph using a multi-modal prediction model, and fuse the time-series production characteristics, the device topology characteristics, and the production scheduling characteristics through an attention mechanism to output a predicted material consumption rate; The replenishment control module is used to calculate a real-time inventory level and a dynamic replenishment threshold by combining the predicted material consumption rate, the current material inventory, the material storage rate, the basic inventory buffer, and the device operating parameters; if the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and a minimum delay path objective function is constructed according to the material transportation parameters to generate an optimal path for controlling the automated guided vehicle to perform replenishment.

9. An intelligent plant material dynamic pull control apparatus, characterized by, It comprises: an Internet of Things sensing device, an edge computing device, a central decision-making device, and an automated guided vehicle; The Internet of Things sensing device is used to collect real-time production data of the intelligent workshop; wherein the real-time production data includes a current material inventory, device operating parameters, and material transportation parameters; The edge computing device is deployed with an INT8 quantized multi-modal prediction model, and the multi-modal prediction model is called through a gRPC interface; The edge computing device is used to acquire the basic inventory buffer of the intelligent workshop, a device topology graph, production scheduling characteristics, and a material storage rate, extract time-series production characteristics from the real-time production data and device topology characteristics from the device topology graph using the multi-modal prediction model, and fuse the time-series production characteristics, the device topology characteristics, and the production scheduling characteristics through an attention mechanism to output a predicted material consumption rate; the production scheduling characteristics are used to represent information of a production task and corresponding product material consumption; The central decision-making device is used to calculate a real-time inventory level and a dynamic replenishment threshold by combining the predicted material consumption rate, the current material inventory, the material storage rate, the basic inventory buffer, and the device operating parameters; if the real-time inventory level is lower than the dynamic replenishment threshold, a replenishment instruction is generated, and a minimum delay path objective function is constructed according to the material transportation parameters to generate an optimal path for controlling the automated guided vehicle to perform replenishment.

10. The smart plant material dynamic pull control apparatus of claim 9, wherein, The Internet of Things perception device comprises a line-edge warehouse perception unit, a device state perception unit and a logistics perception unit; The line-edge warehouse perception unit is used for acquiring a current material inventory; the line-edge warehouse perception unit comprises an intelligent material box, and the intelligent material box comprises a weighing sensor and an RFID reader-writer; The device state perception unit is used for collecting device operation parameters; the device state perception unit is arranged on a production device of the intelligent workshop, and the device state perception unit comprises a vibration sensor and a PLC interface; The logistics perception unit is used for collecting a position, a speed and a carrier capacity of the automatic guided vehicle as the material transportation parameters; the logistics perception unit is arranged on the automatic guided vehicle, and the logistics perception unit comprises a laser radar, a visual SLAM processor and a UWB chip.

Images