Digital processing system for production scheduling and warehousing and ex-warehousing based on mobile terminal

Abstract

The invention relates to the technical field of informatization of the manufacturing industry, and discloses a production scheduling and warehouse in-out digital processing system based on a mobile terminal, and the system comprises a data collection and processing module which obtains original production data and carries out the preprocessing of the original production data; the constraint modeling module is used for obtaining a layered dynamic constraint model by adopting an incremental equipment constraint updating method and a triggered material constraint updating method; the production scheduling optimization module carries out task complexity evaluation and intelligent calculation distribution; the collaborative adjustment module is used for carrying out production scheduling storage collaborative demand prediction and matching on the production scheduling scheme; the warehousing operation module is used for generating a warehouse-in and warehouse-out task queue and guiding an operator to complete warehouse-in and warehouse-out operation; the abnormity identification module is used for identifying abnormal conditions by adopting a multi-modal data fusion algorithm; the hybrid decision module is used for dynamically switching an online mode, a weak network mode and an offline mode and adopting a decision and synchronization strategy of a corresponding mode; according to the invention, deep integration of production scheduling and warehouse management is realized, and the production efficiency and the inventory management level are improved.

Description

Technical Field

[0001] This invention relates to the field of information technology in manufacturing, and more specifically, to a digital processing system for production scheduling and warehousing inbound and outbound operations based on mobile terminals. Background Technology

[0002] As the manufacturing industry undergoes a profound transformation towards digitalization and intelligence, traditional production scheduling and warehouse management methods are no longer sufficient to meet the demands of modern manufacturing enterprises for rapid market response and improved operational efficiency. Traditional production scheduling systems typically employ a centralized architecture, relying on fixed servers and network environments, lacking flexibility and adaptability in the face of unexpected events such as equipment failures and network outages. Simultaneously, traditional warehouse management systems operate relatively independently from production scheduling systems, lacking effective information sharing and collaboration mechanisms, leading to inaccurate material demand forecasting, severe inventory backlogs, and inefficient inbound and outbound processes.

[0003] Existing production scheduling systems suffer from the following main problems: lack of real-time capability, failing to promptly detect dynamic changes on the production floor, resulting in scheduling plans often lagging behind actual conditions; high computational complexity, leading to excessively long calculation times when there are many orders and complex constraints, impacting decision-making efficiency; and lack of effective integration with warehousing systems, hindering the coordinated optimization of production demand and material supply. Warehouse management systems, on the other hand, suffer from the following shortcomings: reliance on manual operation, prone to errors such as incorrect or missed shipments; lack of intelligent task scheduling and path optimization, resulting in low inbound and outbound efficiency; and anomaly detection primarily relies on manual inspection, limiting real-time performance and accuracy.

[0004] Therefore, there is an urgent need for a digital processing system that can deeply integrate production scheduling and warehouse management, and has real-time perception and intelligent decision-making capabilities, in order to improve the overall operational efficiency and competitiveness of manufacturing enterprises. Summary of the Invention

[0005] This invention provides a digital processing system for production scheduling and warehousing inbound and outbound operations based on mobile terminals, which solves the technical problems of inaccurate material demand forecasting, serious inventory backlog, and low inbound and outbound efficiency in related technologies.

[0006] This invention provides a digital processing system for production scheduling and warehouse inbound / outbound operations based on mobile terminals, comprising: The data acquisition and processing module is used to acquire raw production data using a multi-source parallel acquisition strategy, perform timestamp alignment, cleaning and completion processing on the raw production data, and output a valid dataset. The constraint modeling module, based on an effective dataset, uses an incremental equipment constraint update method and a triggered material constraint update method to obtain a hierarchical dynamic constraint model. The production scheduling optimization module, based on a hierarchical dynamic constraint model, performs task complexity assessment and intelligent calculation allocation on the order demand data in the effective dataset, and outputs a production scheduling plan. The collaborative adjustment module is used to predict and match the production scheduling and warehousing collaborative demand of the production scheduling plan, and output the collaboratively adjusted production scheduling plan. The warehouse operation module is used to generate inbound and outbound task queues based on the collaboratively adjusted production schedule and guide operators to complete inbound and outbound operations, and output inbound and outbound operation records. The anomaly detection module, based on inbound and outbound operation records, uses a multimodal data fusion algorithm to identify abnormal situations and output early warning information; The hybrid decision-making module dynamically switches between online, weak network, and offline modes based on early warning information, adopts corresponding decision-making and synchronization strategies, and outputs consistent data that has been synchronized.

[0007] In a preferred embodiment, the multi-source data parallel acquisition strategy includes: Establish data channels with production equipment monitoring systems, warehouse management systems, manufacturing execution systems, and order management systems through the network and sensor interfaces of mobile terminals; A multi-threaded parallel data acquisition mechanism is adopted to simultaneously acquire equipment status data, process progress data, material inventory data, and order demand data.

[0008] In a preferred embodiment, the timestamp alignment, cleaning, and completion process includes: A sliding time window alignment method is adopted, and the width of the time window is set to perform interpolation processing on data records whose time difference exceeds a preset range; Outlier detection methods are used to identify outliers in numerical and categorical data and then replace them. Missing values ​​are filled using the forward filling method.

[0009] In a preferred embodiment, the incremental device constraint update method includes: A state difference comparison method is used to compare the current state of the equipment with its historical state to identify equipment whose state has changed. For devices whose state has not changed, directly copy the historical constraint parameters; For equipment whose status has changed, recalculate the available processing capacity parameters.

[0010] In a preferred embodiment, the triggered material constraint update method includes: Set inventory change trigger thresholds and safety stock trigger conditions; When the material inventory change rate exceeds the first threshold or the inventory level falls below the second threshold of the safety stock line, the material's constraint parameters are updated. For materials that have not triggered an update, keep the constraint parameters unchanged.

[0011] In a preferred embodiment, the task complexity assessment includes: Extract the number of processes, process complexity, number of material requirements, and delivery urgency of pending production orders; The computational complexity index of the production scheduling task is calculated using a linear weighting method; Based on the comparison between the complexity index and the preset threshold, the task is determined to be either lightweight or heavyweight.

[0012] In a preferred embodiment, the intelligent computing allocation includes: Detect the network connection status and computing resources between the mobile terminal and the edge server and cloud server; A comprehensive evaluation method combining computing resources and network conditions is used to compare the processing capabilities and transmission costs of different computing nodes; Based on the evaluation results, the optimal computing node is selected to execute the production scheduling optimization task.

[0013] In a preferred embodiment, the production scheduling and warehousing collaborative demand forecasting and matching includes: The material requirements decomposition method is used to calculate the required quantity and time of various materials based on the order information and process route in the production scheduling plan; A demand-inventory matching algorithm is used to compare the material demand with the current available inventory to identify materials with insufficient inventory. A collaborative adjustment strategy is adopted to adjust the order priority and timing in the production scheduling plan based on the material shortage situation.

[0014] In a preferred embodiment, generating the inbound / outbound task queue includes: The inbound and outbound task decomposition method is adopted to convert the collaboratively adjusted production scheduling plan into specific material inbound and outbound operation tasks; A task priority ranking algorithm is adopted, which comprehensively considers the urgency of tasks, material attributes, and storage location information to rank tasks. The task queue is displayed to the operator through the mobile terminal interface, guiding them to complete the inbound and outbound operations and recording the operation results.

[0015] A computer-readable storage medium for storing computer-readable instructions, which, when read by a computer, enable the operation of a mobile terminal-based digital processing system for production scheduling and warehouse inbound / outbound operations.

[0016] The beneficial effects of this invention are as follows: Through multi-source data acquisition and real-time constraint perception via mobile terminals, the system can promptly capture dynamic changes on the production floor. Employing incremental and trigger-based update strategies, it reduces computational overhead, enabling production scheduling plans to respond quickly to environmental changes. Simultaneously, through intelligent computing allocation via cloud-edge-device collaboration, the system can dynamically select the optimal computing resources based on task complexity, ensuring both rapid response for lightweight tasks and accurate calculation for complex tasks, thereby improving overall production scheduling efficiency.

[0017] By utilizing the scanning and positioning functions of mobile terminals, the system can accurately guide operators to complete inbound and outbound operations, effectively preventing issues such as misdelivery and omissions. Through task priority sorting and path optimization algorithms, the system can rationally arrange the order of operations, reduce unnecessary handling distances, and improve inbound and outbound efficiency. Through edge computing-assisted multimodal data fusion, the system can automatically identify problems such as abnormal material status and non-standard warehouse placement, improving the quality control capabilities of warehouse management and reducing the workload of manual inspections. Attached Figure Description

[0018] Figure 1 This is a block diagram of the digital processing system for production scheduling and warehousing inbound and outbound operations based on mobile terminals, as described in this invention. Detailed Implementation

[0019] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0020] At least one embodiment of the present invention discloses a digital processing system for production scheduling and warehouse inbound / outbound operations based on mobile terminals, such as... Figure 1 As shown, it includes: The data acquisition and processing module is used to acquire raw production data using a multi-source parallel acquisition strategy, perform timestamp alignment, cleaning and completion processing on the raw production data, and output a valid dataset. Includes the following: Based on the 4G / 5G network interface, Wi-Fi interface, Bluetooth interface, and local sensors of the mobile terminal, a multi-threaded parallel data acquisition mechanism is adopted to establish data channels with the production equipment monitoring system, warehouse management system, manufacturing execution system, and order management system. The mobile terminal connects to the workshop edge server through the Wi-Fi interface to obtain real-time operating status data of 20 CNC machining centers, including parameters such as equipment power-on status, current machining task, spindle speed, tool wear, and fault alarm information. Through the 4G network, it connects to the cloud manufacturing execution system to obtain the process progress data of 50 in-process orders, including order number, product type, current process, and completed processes. The data collection process includes: a process list, the start time of the current process, and the estimated completion time; material inventory data from the warehouse management system, including material numbers, names, current inventory levels, safety stock levels, and storage location numbers for 1500 types of materials, obtained through an edge server interface; and 30 new order request data from the order management system, including order numbers, customer names, product specifications, delivery dates, quantities, and special requirements. This data collection process is continuously performed in the background of the mobile terminal, with a collection frequency of once every 5 seconds. Data from each data source is simultaneously sent to the mobile terminal's data receiving buffer, resulting in a raw multi-source dataset containing approximately 200 records from four data sources.

[0021] Based on the collection timestamp of each data record in the original multi-source dataset, a sliding time window alignment method is adopted, with the time window width set to 10 seconds. The mobile terminal traverses all records in the dataset, extracts the collection time of each record, and calculates the time difference between the current time and the collection time.

[0022] For data records with a time difference of less than 10 seconds, they are directly retained and proceed to the next step of processing. For data records with a time difference between 10 and 30 seconds, it is determined whether the data type is slowly changing (such as inventory data or basic equipment parameters). If so, it is retained. If it is rapidly changing (such as real-time equipment load), a linear interpolation method is used for time compensation, and the value of the data at the current moment is estimated based on historical trends. For data records with a time difference exceeding 30 seconds, they are determined to be outdated data and are directly discarded. After time alignment processing, the effective time range of all retained data records is unified to the time window of 10 seconds before the current moment, resulting in a time-aligned dataset containing approximately 180 valid records.

[0023] The mobile terminal performs data quality checks on the time-aligned dataset, identifying outliers, missing values, and duplicates. For numerical data, the 3σ principle is used to detect outliers, i.e., the mean and standard deviation of the data are calculated, and data deviating from the mean by more than three times the standard deviation are marked as outliers. For categorical data, it checks for data items that do not conform to predefined categories. For detected outliers, different processing strategies are adopted according to the data type: numerical outliers are replaced with the historical mean or median, and categorical outliers are replaced with the most frequently occurring category. For missing values, a forward imputation method is used, i.e., the previous valid value of the data item is used to fill in the missing value; if the previous value is also missing, the historical average value of the data item is used to fill in the missing value. For duplicates, the record with the most recent timestamp is retained, and other duplicate records are deleted. After data cleaning and completion, a high-quality preprocessed dataset is obtained, which contains approximately 170 valid and complete data records.

[0024] The preprocessed effective dataset is normalized using a min-max normalization method to map all numerical data to the [0,1] interval. Specific processing steps include: traversing each numerical field in the dataset and calculating its minimum and maximum values; normalizing each value in the field using the formula (current value - minimum value) / (maximum value - minimum value); for percentage-type data such as equipment load rate and process completion rate, since the original data is already in the 0-100 range, dividing it by 100 completes the normalization; for count-type data such as inventory quantity and order quantity, the global maximum value for this type of data is first calculated, and each value is normalized by dividing it by this maximum value; for categorical data, a one-hot encoding method is used to convert each category into a binary vector representation; after normalization, a standardized dataset is obtained, in which all numerical data are within the [0,1] interval, facilitating subsequent machine learning algorithm processing and analysis.

[0025] The constraint modeling module, based on an effective dataset, uses an incremental equipment constraint update method and a triggered material constraint update method to obtain a hierarchical dynamic constraint model; Specifically, it includes the following: Based on the device status data extracted from the preprocessed valid dataset, the mobile terminal reads the device constraint vector and device status snapshot saved in the previous moment from local storage. Using a status difference comparison method, the current status of each device is compared with its historical status to check whether key parameters such as the device's operating status identifier (running / idle / fault), current task number, and cumulative runtime have changed. For devices whose status has not changed, the available machining capacity parameters of the device are directly copied from the historical constraint vector without recalculation. For devices whose status has changed, they are added to the device status change queue. For example, in 20 CNC machining centers, 17 have no status change, and 3 have changed status (1 changed from running to idle, 1 changed from idle to running, and 1 triggered a tool wear alarm). The device status change queue includes these 3 devices. Based on the equipment in the equipment status change queue, the equipment health assessment method is used to recalculate the available processing capacity for each changed piece of equipment: for equipment that changes from running to idle, its available processing capacity is restored to 100% of the rated capacity; for equipment that changes from idle to running, the remaining available time percentage of the equipment in the production scheduling period is calculated based on the estimated processing time of the current task; for equipment that has a tool wear alarm, the processing efficiency attenuation coefficient of the equipment is evaluated based on the wear level (light / medium / severe) and equipment maintenance records, and the rated capacity is multiplied by the attenuation coefficient to obtain the current available processing capacity.

[0026] The specific steps for calculating the available processing capacity of the equipment are as follows: read the rated processing capacity of the equipment from the equipment file as the baseline value; determine the health coefficient based on the current status of the equipment (1.0 for healthy equipment, 0 for faulty equipment, and 0.6-0.9 for equipment with wear); analyze the time utilization of the equipment during the production scheduling period to calculate the time availability rate; multiply the rated processing capacity by the health coefficient and the time availability rate to obtain the actual available processing capacity.

[0027] Based on the recalculation of the available processing capacity of the changed equipment, the parameter values ​​at the corresponding positions in the equipment constraint vector are updated, while the parameters of the unchanged equipment remain unchanged, resulting in an incrementally updated equipment constraint vector. This vector contains the latest available processing capacity of 20 equipment. Compared with the full update method, the incremental update only calculates the constraint parameters of 3 equipment, reducing the amount of computation by 85% and lowering the computational overhead of the mobile terminal.

[0028] Based on the material inventory data extracted from the preprocessed valid dataset, the mobile terminal reads the material constraint matrix and material inventory snapshot saved in the previous moment from local storage. A trigger threshold method is used, setting the inventory change trigger threshold to 10% or an absolute quantity change exceeding 50, and the safety stock trigger condition to current inventory being below 120% of the safety stock line. The inventory records of 1500 materials are traversed, comparing the current inventory with historical inventory levels item by item, calculating the inventory change rate and absolute inventory change. Materials with an inventory change rate exceeding 10% or an absolute change exceeding 50, and materials with current inventory below 120% of the safety stock line, are added to the material update trigger list. It is assumed that among the 1500 materials, 1420 materials have very small inventory changes and do not trigger updates, while 80 materials have experienced significant inventory changes or are close to the safety stock line, triggering updates.

[0029] Based on the materials in the material update trigger list, the constraint parameters of these materials are recalculated using inventory availability analysis: For each material that triggers an update, the current inventory, safety stock, number of orders in transit, and estimated arrival time of the material are read from the valid dataset; The specific calculation method for material availability is as follows: the actual available quantity is obtained by subtracting the safety stock from the current inventory (if the result is negative, it is set to 0); based on the estimated arrival time, it is determined whether the orders in transit can arrive within the production scheduling period. If the arrival time is earlier than the end time of the production scheduling period, the quantity in transit is included in the available quantity, and finally the total available quantity is obtained.

[0030] The parameters of these 80 materials in the material constraint matrix are updated while the parameters of the remaining 1420 materials remain unchanged, resulting in an on-demand updated material constraint matrix. This matrix contains the latest available quantities and delivery times of 1500 materials. The triggered update method reduces the computational load compared to the full update, thus improving update efficiency.

[0031] Based on incrementally updated equipment constraint vectors and on-demand updated material constraint matrices, combined with process schedule data from the effective dataset, a hierarchical constraint modeling method is adopted. Critical and secondary constraints are identified. Critical constraints are defined as those that significantly impact the production plan, such as the failure or severe wear of critical equipment (bottleneck equipment), insufficient inventory of critical materials (core raw materials with no substitutes), and process delays for urgent orders. Secondary constraints are defined as those that have a smaller impact on the production plan, such as minor wear of ordinary equipment, inventory fluctuations of routine materials, and process schedules for general orders.

[0032] The mobile terminal iterates through the equipment constraint vector, identifying equipment with less than 50% available processing capacity as critical equipment constraints. It assumes that 2 out of 20 devices have less than 50% available capacity. It then iterates through the material constraint matrix, identifying materials with actual available quantity less than 80% of demand and no orders in transit as critical material constraints. It assumes that 15 out of 1500 materials are in short supply. Finally, it iterates through the process progress data, identifying urgent orders with expected completion times later than the order delivery date as critical process delay constraints. It assumes that 5 out of 50 in-process orders are at risk of delay. The identified 2 critical equipment constraints, 15 critical material constraints, and 5 critical process constraints are grouped into a critical constraint set, totaling 22 critical constraints.

[0033] The constraints of the remaining 18 ordinary equipment units, 1485 types of common materials, and 45 orders with normal progress are grouped into a secondary constraint set, totaling 1548 secondary constraints. A hierarchical update strategy is adopted: the critical constraint set is updated in real time and immediately notifies the production scheduling system to trigger plan adjustments, with an update frequency of immediately updating whenever a critical constraint changes. The secondary constraint set is updated periodically in batches, with an update frequency of once every 30 seconds, reducing the system overhead of frequent updates.

[0034] Based on the sets of key and secondary constraints, a constraint relationship modeling method is used to construct the dependencies and influence relationships between constraints. For example, a failure of a key piece of equipment will affect all orders currently being processed and planned to be processed on that equipment, and a shortage of a key material will affect all orders that require that material. By constructing an association matrix of constraint orders and recording the list of orders affected by each constraint, a hierarchical dynamic constraint model is obtained. This model includes the constraint hierarchy, constraint parameter values, constraint update strategies, and constraint influence relationships, and can comprehensively describe the constraint state of the production system at the current moment.

[0035] Furthermore, a constraint slackness assessment method can be used to calculate the slackness index for each constraint in the hierarchical dynamic constraint model. Constraint slackness represents the margin of the constraint from the critical state; for example, a higher slackness occurs when the equipment availability is 80%, while a lower slackness occurs when the equipment availability is 20%.

[0036] The specific steps for calculating slackness include: determining constraint parameter values ​​and obtaining the current actual values ​​of constraints as the basis for calculation; setting a critical threshold, determining the critical value based on the constraint type, with constraints below this value considered critical constraints; setting a normal threshold, determining the normal value based on the constraint type, with constraints above this value considered sufficient constraints; calculating the slackness index, obtaining the constraint margin by calculating the difference between the current value and the critical value, obtaining the constraint range by calculating the difference between the normal value and the critical value, and dividing the constraint margin by the constraint range to obtain the slackness index, with the slackness value ranging from 0 to 1. A slackness close to 0 indicates a tense constraint requiring close attention, while a slackness close to 1 indicates a relaxed constraint that can be given less attention. A slackness label is attached to each constraint to facilitate differentiated processing based on the constraint tension during scheduling optimization, prioritizing the satisfaction of tense constraints with low slackness, resulting in a constraint model with slackness labels. This model provides more refined constraint status information for the scheduling algorithm.

[0037] The production scheduling optimization module, based on a hierarchical dynamic constraint model, evaluates the task complexity and intelligently calculates and allocates tasks based on the order demand data in the effective dataset, and outputs a production scheduling plan. Specifically, it includes the following: Based on the current hierarchical dynamic constraint model and the set of pending production orders obtained from the order management system (containing 30 newly received orders), a task complexity evaluation method is adopted. The mobile terminal extracts key attributes for each order, including product type, number of processes, process complexity, number of material requirements, delivery urgency, order priority, and other dimensions.

[0038] For the number of processes, the total number of processes included in each order is counted. The more processes, the more complex the order. For process complexity, a complexity score is calculated based on factors such as processing time, equipment requirements, and accuracy requirements for each process. The overall process complexity of the order is obtained by weighted summation. For the number of material requirements, the total number of different material types required by the order is counted. The more types, the greater the difficulty in material coordination. For delivery urgency, the difference between the order delivery date and the current time is calculated. The smaller the difference, the more urgent the delivery.

[0039] Based on the above multi-dimensional characteristics, a linear weighted method is used to calculate the computational complexity index of the production scheduling task. The specific complexity calculation steps include: collecting basic data to obtain four key indicators: the number of orders to be scheduled, the average process complexity per order, the average number of material requirements per order, and the number of key constraints; performing data preprocessing and normalization: for the order quantity, the maximum value normalization method is used, dividing it by the maximum order processing capacity designed by the system; for process complexity, a scoring system normalization method is used, directly dividing the score of 0-10 by 10 to normalize it to the 0-1 range; for the number of material requirements, the maximum value normalization method is used, dividing it by the maximum value of material requirements in the system; for the number of constraints, the maximum value normalization method is used, dividing it by the maximum number of constraints designed by the system; setting weight coefficients: based on practical experience, weight coefficients are set for each factor, namely, order quantity weight 0.4, process complexity weight 0.2, material type weight 0.2, and constraint quantity weight 0.2; performing weighted calculation: each normalized indicator is multiplied by its corresponding weight coefficient, and all weighted results are added together to obtain the final complexity index.

[0040] After normalization, all parameters are unified to the 0-1 range, eliminating the influence of different units and making the linear weighted calculation more reasonable. For the 30 orders in this embodiment, the average number of processes per order is 8, the average process complexity score is 6.5 (out of 10), the average number of material requirements is 12, and the number of key constraints is 22. The mobile terminal calculates the complexity using the following formula: the weight of the order quantity is 0.4 multiplied by 30 equals 12, the weight of the process complexity is 0.2 multiplied by 6.5 equals 1.3, the weight of the material types is 0.2 multiplied by 12 equals 2.4, and the weight of the number of constraints is calculated. The weight term is 0.2 multiplied by 22, which equals 4.4. Adding the four weight values ​​together gives a complexity index of 20.1. Based on the computational complexity index and the available computing resources of the mobile terminal, a complexity threshold of 15 is set. When the complexity index is less than or equal to 15, it is determined to be a lightweight task and computed locally on the mobile terminal. When the complexity index is greater than 15, it is determined to be a heavyweight task and assigned to edge or cloud servers for computation. In this embodiment, the complexity index is 20.1, which exceeds the threshold of 15. Therefore, it is determined to be a heavyweight task, and the task complexity evaluation result is "heavyweight". The scheduling optimization task needs to be assigned to edge or cloud servers for computation.

[0041] Based on the task complexity assessment result of "heavyweight" and the current network connection status of the mobile terminal, an intelligent computing allocation strategy is adopted. The mobile terminal detects the current network connection type and network quality, and measures the network latency and bandwidth with the edge server and cloud server. In this embodiment, the mobile terminal connects to the edge server via workshop Wi-Fi with a network latency of 10 milliseconds and a bandwidth of 100 Mbps; and connects to the cloud server via a 4G network with a network latency of 80 milliseconds and a bandwidth of 20 Mbps.

[0042] A comprehensive evaluation method considering computing resources and network conditions was used to compare the computing power and network transmission costs of edge servers and cloud servers. Edge servers, deployed on the workshop floor, are configured with 8-core CPUs and 32GB of memory, offering moderate computing power but low network latency. Cloud servers, configured with 64-core CPUs and 256GB of memory, offer powerful computing power but higher network latency. For this heavyweight production scheduling task, involving the comprehensive optimization of 30 orders, 20 devices, 1500 types of materials, and 22 key constraints, the estimated computation time is approximately 30 seconds on the edge server and approximately 10 seconds on the cloud server. Considering data upload time, uploading order and constraint data to the edge server takes 0.5 seconds, and uploading to the cloud server takes 2 seconds. Combining computation and transmission times, the total time for the edge server is 0.5 + 30 = 30.5 seconds, and the total time for the cloud server is 2 + 10 = 12 seconds.

[0043] Since the cloud server takes less time overall and the network latency of 80 milliseconds is within an acceptable range, the decision was made to allocate the production scheduling optimization task to the cloud server for computation. The mobile terminal generates a computation task allocation plan, which includes the following details: "Task type: heavyweight production scheduling optimization, execution node: cloud server, data transmission method: 4G network encrypted transmission, estimated completion time: 12 seconds."

[0044] Based on the task allocation scheme, the mobile terminal packages and compresses data such as the set of orders to be scheduled, the hierarchical dynamic constraint model, equipment parameters, and material information, and uploads it to the cloud server via HTTPS protocol and 4G network encryption. After receiving the data packet, the cloud server decompresses and verifies the data. After confirming the data integrity, it begins the scheduling optimization calculation. The cloud server uses a complete multi-objective optimization algorithm for scheduling calculation: features are extracted from 30 orders to construct an order feature matrix, which contains attributes such as the process sequence, processing time, bill of materials, delivery date, and priority of each order; based on the order feature matrix and constraint model, heuristic rules are used to generate an initial scheduling plan. Heuristic rules include prioritizing orders with high priority, prioritizing orders with tight delivery dates, prioritizing orders with fewer processes, and prioritizing allocation to equipment with high available capacity. According to these rules, the 30 orders are allocated to 20 pieces of equipment one by one, and the processing start and end times of each order on each piece of equipment are determined to form an initial Gantt chart, resulting in the initial scheduling plan.

[0045] Based on the initial production scheduling plan, a multi-objective genetic algorithm is used for optimization. The steps to construct the multi-objective optimization function include: determining three main optimization objectives (minimizing order delay penalty, minimizing equipment idle time, and minimizing inventory holding cost), assigning weight coefficients to each objective according to the actual needs of the enterprise (order delay penalty weight 0.5, equipment idle time weight 0.3, and inventory holding cost weight 0.2), and combining the three sub-objective functions into a comprehensive objective function by weighted summation.

[0046] The order delay penalty function is calculated as follows: iterate through all orders, compare the difference between the actual completion time and the scheduled delivery date, and use the square of the delay days as the penalty value for overdue orders. The total delay penalty is obtained by summing the delay penalty values ​​for all orders. The equipment idle time function is calculated as follows: iterate through all equipment, calculate the idle time period for each piece of equipment within the production scheduling period, and sum the idle times of all equipment to obtain the total system idle time. The inventory holding cost function is calculated as follows: determine the outbound time and actual usage time of each material according to the production scheduling plan, calculate the waiting time from material outbound to usage, multiply the material quantity, waiting time, and unit cost to obtain the holding cost of each material, and sum the holding costs of all materials to obtain the total inventory holding cost. A genetic algorithm was used for optimization, with a population size of 100 and an iteration number of 200. The chromosome encoding method was an order-equipment allocation sequence, where each gene bit represented the equipment number allocated to an order. Selection, crossover, and mutation operations were performed. Tournament selection was used for selection, two-point crossover was used for crossover, and random mutation was used for mutation. After 200 iterations, the algorithm converged, and the objective function value decreased from the initial 1250 to 580, resulting in the optimized production scheduling scheme.

[0047] The cloud server serializes the optimized production scheduling plan into JSON format, which includes detailed information such as equipment allocation, start time, end time, and material requirement time for each order. It then encrypts the data and sends it to the mobile terminal via HTTPS. After receiving the production scheduling plan, the mobile terminal parses it and stores it locally to obtain the production scheduling plan generated in the cloud.

[0048] Based on the cloud-generated production scheduling plan, the mobile terminal uses a visual display method, showing the scheduling results in the form of a Gantt chart on the touch screen interface. The horizontal axis of the Gantt chart represents time, and the vertical axis represents equipment. Each order is represented by a rectangle of a different color, and the position and length of the rectangle correspond to the start time and processing time of the order on the equipment. The production manager can intuitively see the allocation and time arrangement of 30 orders on 20 machines through the mobile terminal screen, and clearly identify issues such as equipment load, order on-time performance, and whether there are equipment conflicts.

[0049] The mobile terminal also displays critical path analysis results, highlighting key orders and equipment that affect the overall completion time in red. It shows a resource load chart, displaying the load rate of each piece of equipment in bar chart format; equipment with a load rate close to 100% is marked in orange to indicate it may become a bottleneck. It also displays order delivery date achievement status, showing a list comparing the estimated completion time and delivery date for each order, using green for on-time, yellow for tight deadlines, and red for delays.

[0050] Production managers can evaluate production scheduling plans holistically by reviewing the visualization results. If they find that certain orders are allocated to unsuitable equipment or that some equipment is overloaded, they can manually fine-tune the process through the mobile terminal's interactive interface. The mobile terminal provides a drag-and-drop function, allowing production managers to directly drag order rectangles on the Gantt chart to move orders from one piece of equipment to another or adjust order start times. After each drag-and-drop operation, the mobile terminal locally recalculates the objective function value and displays the real-time impact of the adjustments on delivery time, equipment utilization, and inventory costs, helping production managers determine whether the adjustments are appropriate.

[0051] After review and fine-tuning by the production manager, the production scheduling plan was confirmed to meet actual production needs. The plan was then officially activated by clicking the "Confirm and Execute" button on the mobile terminal interface. The mobile terminal saved the confirmed production scheduling plan to both local and cloud databases and sent a scheduling command to the Manufacturing Execution System (MES) to obtain the final confirmed production scheduling plan. This plan serves as the basis for subsequent warehousing material demand forecasting and inbound / outbound operations.

[0052] Furthermore, sensitivity analysis of the production scheduling plan can be used to conduct a robustness assessment of the final confirmed production scheduling plan. The purpose of sensitivity analysis is to evaluate how sensitive the production scheduling plan is to changes in constraints. When certain constraints change slightly (such as equipment failure, material delays, etc.), whether the production scheduling plan remains feasible or can be quickly adjusted. A mobile terminal simulates scenarios of changes in key constraints, such as reducing the availability of key equipment by 20% or delaying the arrival time of key materials by one day. Based on the changed constraints, the feasibility and objective function value of the production scheduling plan are recalculated. If the changed objective function value increases by no more than 10%, and all orders can still be completed within the tolerance range of delays, the production scheduling plan is considered to have good robustness. If the changed objective function value increases or there are serious order delays, the production scheduling plan is considered sensitive to that constraint, requiring the development of contingency plans or advance adjustments to the production scheduling strategy. The sensitivity analysis results are displayed in report form on the mobile terminal, helping production managers identify scheduling risks and take preventative measures in advance, resulting in a production scheduling plan with robustness assessment. This plan provides more reliable planning guidance for production execution.

[0053] The collaborative adjustment module is used to predict and match the production scheduling and warehousing collaborative demand of the production scheduling plan, and output the collaboratively adjusted production scheduling plan. Specifically, it includes the following: Based on the final confirmed production schedule, the mobile terminal uses a material requirements decomposition (MUD) method to calculate material requirements for each order. The production schedule includes detailed processing plans for 30 orders, each containing information such as order number, product type, production quantity, start and end times for each process, and allocated equipment. The mobile terminal first queries the product database for the bill of materials (BOM) for each order's corresponding product. The BOM describes the types, specifications, and quantities of all raw materials, semi-finished products, and auxiliary materials required to produce the product.

[0054] For the first order in the production schedule, the product type is engine block, and the production quantity is 50 units. The bill of materials shows that each unit requires 1 cast iron blank, 0.05 sets of cutting tools, 0.2 liters of coolant, and 1 packaging box. Based on the production quantity of 50 units, the material requirements for this order are calculated as 50 cast iron blanks, 2.5 sets of cutting tools, 10 liters of coolant, and 50 packaging boxes. The start time of the first process (rough machining) for this order is 2 hours after the current time, so the material requirement time for this order is also 2 hours after the current time, because the materials must be ready before production begins.

[0055] Following the same method, all 30 orders in the production scheduling plan were iterated through, and the bill of materials was expanded for each order. The quantities and required times for each material were calculated, resulting in a detailed material requirements list for each order. The material requirements for each order were then summarized and merged according to the material number, with the required quantities for the same material number added together, and the earliest required time being used. After summarizing and calculating, the 30 orders required a total of 1200 cast iron blanks, 800 gearbox housing blanks, 600 steering knuckle blanks, 95 sets of cutting tool sets, 350 liters of coolant, and several other auxiliary materials. This resulted in a production material requirements list containing information on 85 different materials, each item including three attributes: material number, required quantity, and required time.

[0056] Based on the production material requirements list and the obtained material constraint matrix, the mobile terminal uses a demand-inventory matching algorithm to determine whether the inventory of each material meets the production demand. It iterates through the 85 materials in the requirements list, and for each material, reads the current available inventory and the estimated arrival time of in-transit materials from the constraint matrix. For the first material, cast iron billets, the required quantity is 1200 pieces, and the required time is 2 hours after the current moment. The constraint matrix shows a current available inventory of 900 pieces, an in-transit order quantity of 500 pieces, and an estimated arrival time of 4 hours after the current moment. Comparing the required quantity of 1200 pieces with the available inventory of 900 pieces, it is found that the inventory is insufficient, with a shortage of 300 pieces. Checking the relationship between the in-transit order arrival time of 4 hours and the required time of 2 hours, it is found that the in-transit material's arrival time is later than the required time, and the shortage cannot be replenished in time. Therefore, [the missing item is removed from the original text]. The materials are recorded in the material shortage list. The shortage item is 300 cast iron blanks, with a demand time of 2 hours after the current time. For the second material, gearbox housing blanks, the demand quantity is 800 pieces, and the current available inventory is 1000 pieces, so the inventory is sufficient and there is no shortage. For the third material, steering knuckle blanks, the demand quantity is 600 pieces, and the current available inventory is 400 pieces, resulting in a shortage of 200 pieces. However, there are 300 pieces in transit orders that are expected to arrive in 1 hour, which is 2 hours earlier than the demand time. Therefore, the in-transit orders can cover the shortage, and the inventory of this material meets the demand. And so on. After going through all 85 types of materials, 18 types of materials are identified as having inventory shortages and in-transit orders that cannot be replenished in time. This results in a material shortage list containing 18 shortage records, including shortage items such as 300 cast iron blanks with a demand time of 2 hours after the current time and 20 sets of cutting tool sets with a demand time of 1 hour after the current time.

[0057] Based on a material shortage list containing 18 types of shortage materials, the mobile terminal employs a material delivery time verification method to further assess material supply risks. For each shortage material, the mobile terminal queries the procurement system and supplier database to obtain the material's typical procurement cycle and supplier delivery reliability.

[0058] For cast iron blanks, the normal procurement cycle is 3 days, and the fastest expedited procurement cycle is 1 day. The current demand is in 2 hours, and even if expedited procurement is initiated, the goods cannot arrive before the required time. Therefore, this material is deemed high-risk and added to the material supply risk list. For the tool kits, the warehouse has 50 sets in spare stock in another storage area, which can be replenished within 1 hour through inter-warehouse transfer. Although there is a shortage, the risk is manageable and they are not added to the risk list.

[0059] After reviewing all 18 types of materials in short supply, 8 types of materials were identified as having inventory that could not be replenished before the demand time due to excessively long procurement cycles or unreliable suppliers. These 8 materials constituted the real supply risk, resulting in a material supply risk list containing 8 risky materials, including risk items such as 300 pieces of cast iron blanks with a demand time of 2 hours and 100 sets of precision bearings with a demand time of 3 hours.

[0060] Based on a material supply risk list containing eight types of risky materials, the mobile terminal employs a production scheduling-warehousing collaborative adjustment strategy to link production scheduling plans and procurement plans. The mobile terminal analyzes the relationship between risky materials and orders in the production scheduling plan, identifies which orders depend on these risky materials, and assesses the adjustment space and cost for these orders.

[0061] For the high-risk material, cast iron blanks, a production schedule check revealed five orders requiring this material, with start times of 2 hours, 2.5 hours, 3 hours, 4 hours, and 5 hours respectively. The start times of the first three orders need adjustment before the material arrives. The mobile terminal evaluated adjustment options for these three orders. Option one was to uniformly postpone the start times of all three orders until the material arrives (assuming expedited procurement is initiated and the material arrives one day later). However, this would cause order delays, affecting delivery commitments. Option two was to prioritize using the existing inventory of 900 units to satisfy the first two orders (each order requires 250 units, totaling 500 units), and postpone the third order until the material arrives. This way, only one order is delayed, with less cost. The mobile terminal adopted Option two, adjusting the production schedule to postpone the start time of the third order by one day, and simultaneously sending an emergency procurement request to the purchasing system for expedited procurement of 300 cast iron blanks.

[0062] For the other seven risky materials, the same analysis and adjustment logic was applied. For some materials, adjusting the process sequence could stagger material demand times; for others, finding alternative materials could avoid stockout risks; and for still others, adjusting order priorities could ensure priority fulfillment of orders from key customers. After comprehensive coordination and adjustments, the mobile terminal modified the start time or process sequence of 12 orders in the production schedule, generated 8 emergency procurement requests, and obtained a collaboratively adjusted production schedule and emergency procurement list.

[0063] The mobile terminal updates the collaboratively adjusted production scheduling plan to the local and cloud databases, replacing the original plan, and sends an adjustment notification to the production manager, explaining the reasons for the adjustment and the affected orders. Simultaneously, an emergency procurement list is sent to the procurement system, triggering the procurement process to ensure timely material replenishment. This collaboratively adjusted production scheduling plan considers both production optimization goals and material supply constraints, achieving deep collaboration between production scheduling and warehousing. It avoids production interruptions due to material shortages and serves as the basis for subsequent inbound and outbound operations.

[0064] Furthermore, material demand forecasting methods can be employed, based on collaboratively adjusted production schedules and historical production data, to predict material demand trends for the coming week. Mobile terminals analyze historical production plans and material consumption data to establish a correlation model between material consumption and order type and production quantity. Based on order forecasts and production plans for the coming week, the expected consumption and inventory change trends of various materials are calculated, identifying materials that may experience future shortages. Proactive procurement recommendations or inventory replenishment reminders are then issued, resulting in a material demand forecast report. This report helps purchasing and warehousing departments plan material reserves in advance, avoiding reactive responses to material shortages and improving the foresight and proactiveness of the supply chain.

[0065] The warehouse operation module is used to generate inbound and outbound task queues based on the collaboratively adjusted production schedule and guide operators to complete inbound and outbound operations, and output inbound and outbound operation records. Specifically, it includes the following: Based on the collaboratively adjusted production schedule, the mobile terminal employs an inbound / outbound task decomposition method to extract the material outbound and product inbound tasks that need to be executed within the current time period (the next 2 hours). It iterates through the 30 orders in the production schedule, generating corresponding material outbound tasks for orders whose start time is within 2 hours of the current moment. For example, if the first order starts 30 minutes after the current moment and requires 50 cast iron blanks, 2.5 sets of cutting tool kits, and 10 liters of coolant, 3 outbound tasks will be generated.

[0066] Simultaneously, query the Manufacturing Execution System to obtain information on work-in-process orders expected to be completed within the current time period. Assuming five orders will complete their final processing within the next two hours, and the finished products from these orders need to be warehoused, generate a product warehouse entry task for each completed order. Each warehouse entry task records attributes such as task type (outbound / inbound), material number, material name, quantity, warehouse location number, priority (set according to order urgency), and target time.

[0067] After traversal and generation, a total of 58 outbound tasks (15 out of 30 orders start within 2 hours, with an average of 3-4 types of materials per order) and 5 inbound tasks were generated, totaling 63 tasks. This resulted in an outbound and inbound task queue, which contains all outbound and inbound operations that the warehouse manager needs to perform within the next 2 hours.

[0068] Based on the inbound / outbound task queue and warehouse layout information, the mobile terminal employs a task priority sorting algorithm to optimize the task queue. The sorting algorithm comprehensively considers three factors: task urgency, material weight, and storage location distance. Task urgency is calculated based on the time difference between the task's target time and the current time; the smaller the time difference, the more urgent the task. Material weight is retrieved from basic material data; heavier materials are prioritized to optimize manpower and handling equipment allocation. Storage location distance refers to the distance between the task's corresponding storage location and the current or previous task's storage location; closer tasks are prioritized to reduce handling path length. The mobile terminal calculates the comprehensive priority score for each task using the following steps: collecting basic task data (retrieving remaining task time, material weight information, and storage location distance); determining the maximum value for each evaluation dimension (traversing all tasks to find the maximum value for each dimension as a benchmark); calculating standardized scores for each dimension (time urgency score, weight priority score, and distance convenience score); setting weight coefficients and calculating the comprehensive score (urgency weight 0.5, weight weight 0.3, and distance weight 0.2).

[0069] To ensure the comparability of data across different dimensions and the rationality of calculations, data preprocessing is required for each parameter before calculating priority scores: remaining time is converted to minutes, material weight to kilograms, and storage distance to meters. Maximum value normalization is used to map the data of all three dimensions to the 0-1 range, eliminating dimensional differences and making the weighted summation calculation more scientific and reasonable. Priority scores are calculated for each of the 63 tasks, and they are sorted from highest to lowest score. When scores are the same, outbound tasks are executed first (outbound tasks affect production, while inbound tasks can be appropriately delayed). The mobile terminal also optimizes the path of the sorted queue. For two tasks that are very close (storage locations in the same or adjacent rows), even if their priorities differ slightly, they are adjusted to be executed consecutively to avoid back-and-forth handling. After sorting and optimization, a sorted task queue is obtained, providing warehouse managers with efficient work sequence guidance.

[0070] Based on the sorted task queue, the mobile terminal employs a guided workflow method, guiding the warehouse manager through inbound and outbound operations task by task. The mobile terminal screen displays the first task in the task queue, with the task description: "Outbound task: Cast iron blanks, quantity 50 pieces, target storage location A-12-03, high priority." Simultaneously, the screen displays a warehouse floor plan, using arrows and highlighted colors to indicate the shortest path from the current location to storage location A-12-03, providing navigation guidance for the operator.

[0071] The warehouse manager walks along the marked path using a handheld mobile terminal. The terminal tracks the operator's location in real time via built-in GPS or indoor positioning modules (based on Wi-Fi and Bluetooth beacons). When the operator approaches the target storage location, the mobile terminal vibrates and sounds to indicate arrival. Upon reaching storage location A-12-03, the operator sees the cast iron blanks stored there. The mobile terminal screen prompts, "Please scan the material label to confirm material information." The operator uses the mobile terminal's camera to scan the QR code label on the material box. The mobile terminal parses the QR code to obtain the material number, compares it with the material number in the task, and displays a green checkmark and "Material confirmed correct" on the screen after confirmation. The mobile terminal then prompts, "Please enter the actual quantity to be issued." A numeric input box is displayed on the screen, with a default value of 50 pieces. The operator counts the actual quantity... The quantity of materials to be issued is confirmed to be 50 pieces. Clicking the confirmation button, the mobile terminal records this issuance operation, generating an operation record containing information such as task number, material number, storage location number, operation type (issue), actual quantity 50 pieces, operator's employee number, and operation time. This record is added to the inbound / outbound operation record set. The mobile terminal screen updates to display "Task 1 completed, 62 tasks remaining," automatically redirecting to the next task's guidance interface. Following the same process, the operator completes the 63 tasks in the queue sequentially. For inbound tasks, the operation process is similar, except that the operator needs to move the finished product to the designated storage location, scan the finished product label and storage location label to bind them, input the inbound quantity, and the mobile terminal records the inbound operation information. After approximately 90 minutes of continuous operation, the operator completes all 63 tasks, obtaining an inbound / outbound operation record containing 63 records.

[0072] Based on 63 operation records in the inbound and outbound operation logs, the mobile terminal uses a real-time inventory data update method to synchronize each operation record to the inventory database. For each outbound record, the mobile terminal extracts the material number and outbound quantity, connects to the warehouse management system database, queries the current inventory record for that material, subtracts the outbound quantity from the inventory balance, updates the inventory balance field, and simultaneously updates the last outbound time field of that material to the current operation time. For inbound records, the inventory balance is added to the inbound quantity, and the last inbound time is updated.

[0073] For the outbound record of cast iron blanks, material number MAT-001, quantity outbound 50 units. A database query shows the current inventory of this material is 900 units. After calculation and updating, the inventory is 850 units. An SQL update statement is executed to update the inventory balance field to 850 units. This process is repeated for 63 operation records, completing the inventory data update for all involved materials. Outbound operations involve 42 types of materials, and inbound operations involve 5 types of products, updating a total of 47 inventory records.

[0074] The mobile terminal also updates the storage location usage status information. For outbound operations, if all materials in a storage location are removed, the location is marked as "free," making it available for storing new materials. For inbound operations, the storage location where the product is stored is marked as "occupied," and the product number and quantity are recorded. Real-time updates to storage location status ensure effective warehouse space management and prevent location conflicts or wasted space.

[0075] All updates are synchronized in real time to the cloud-based warehouse management system database via Wi-Fi, ensuring global data consistency. Other mobile terminals and PC-based management systems can instantly view the latest inventory data and receive updated inventory information. This data reflects the latest inventory status after inbound and outbound operations, providing an accurate data foundation for subsequent production scheduling decisions and material requirement calculations.

[0076] Furthermore, inventory turnover analysis can be used to calculate the inventory turnover rate of various materials based on updated inventory data and historical inbound and outbound records. Inventory turnover rate reflects the speed of material flow and inventory health; a high turnover rate indicates that materials are consumed quickly and inventory is low, while a low turnover rate indicates that materials are stockpiled and capital is tied up.

[0077] The mobile terminal uses the following steps to calculate inventory turnover: First, determine the statistical period and target material. The statistical period is typically one month. The target material to be analyzed can be a single material or all materials. Second, collect outbound data within the statistical period. Extract all outbound records of the target material within the statistical period from historical inbound and outbound records, organize the outbound quantity data in chronological order, and record the quantity and time of each outbound transaction. Third, calculate the total outbound quantity within the statistical period. Accumulate all outbound quantities of the target material within the statistical period to obtain the total outbound quantity of the material throughout the entire statistical period. Fourth, calculate the average inventory level within the statistical period. Collect inventory level change data of the target material within the statistical period and calculate the average of the beginning and ending inventory levels, or use a weighted average method to calculate the average inventory level for the entire period. Fifth, calculate the inventory turnover rate. Divide the total outbound quantity within the statistical period by the average inventory level to obtain the inventory turnover rate. A higher turnover rate indicates faster material flow and higher inventory efficiency.

[0078] To ensure the accuracy of turnover rate calculation, relevant data preprocessing is required. For outbound quantities, a unified time granularity is needed, converting outbound records from different time intervals (e.g., daily, weekly records) into daily statistics, and summing them to obtain the total outbound quantity within the statistical period. For average inventory levels, the issue of discontinuous inventory data over time needs to be addressed, using linear interpolation to fill in missing inventory records and ensure an accurate average value. Abnormal inventory data, such as negative inventory or abnormally high inventory due to system failures, needs to be handled, corrected using inventory data from previous and subsequent time points. After data preprocessing, the turnover rate calculation is ensured to be based on a complete, accurate, and consistent data foundation. The mobile terminal calculates the turnover rate of all materials, identifies stagnant materials with excessively low turnover rates, issues warnings to inventory managers, and suggests reducing procurement or adjusting inventory strategies, generating an inventory turnover analysis report. This report helps companies optimize their inventory structure, reduce inventory costs, and improve capital turnover efficiency.

[0079] The anomaly detection module, based on inbound and outbound operation records, uses a multimodal data fusion algorithm to identify abnormal situations and output early warning information; Specifically, it includes the following: Using a mobile terminal's camera, warehouse managers can capture real-time images of materials and storage locations during inbound and outbound operations. During these operations, when an operator scans a material's QR code, the mobile terminal automatically triggers the camera to take a picture of the material's overall appearance, capturing visual features such as shape, color, packaging condition, and label clarity. Simultaneously, it takes photos of the storage location, recording its cleanliness, the neatness of the materials' arrangement, and the presence of any foreign objects. For the outbound operation of cast iron billets, the mobile terminal's camera captures a picture of the material stacked on a pallet. The image resolution is 1920×1080 pixels, and the image size is approximately 2MB. To reduce data transmission and subsequent processing burden, the mobile terminal performs image preprocessing and feature extraction locally. Image compression algorithms were used to reduce the image resolution to 640×480 pixels, resulting in a compressed image size of approximately 200KB. Edge detection algorithms were employed to extract the contour edges of the materials in the image, color histogram algorithms were used to extract the main color distribution of the materials, and texture analysis algorithms were used to extract the texture features of the material surface. These low-level visual features were combined into a feature vector with a dimension of 128 and a data size of approximately 1KB, reducing the data volume by 99.95% compared to the original image. After executing 63 inbound and outbound tasks, the mobile terminal collected 126 images (one photo of the material and one photo of the storage location for each task), and extracted 126 visual feature vectors, resulting in a compressed set of visual feature vectors containing 126 feature vectors with a total data volume of approximately 126KB.

[0080] Based on the compressed visual feature vector set, the mobile terminal transmits the feature data to the edge computing node deployed on-site in the warehouse via a Wi-Fi network. The edge computing node is an industrial-grade server configured with a 16-core CPU, 64GB of memory, and a GPU accelerator card, and is equipped with a deep learning inference engine and a pre-trained object detection model.

[0081] After receiving 126 visual feature vectors, the edge nodes use a convolutional neural network model for target detection and recognition. The model has been pre-trained on a large number of warehouse scene images and can identify common material types (metal parts, plastic parts, packaging boxes, etc.), material status (intact, damaged, soiled, etc.), placement method (neat, messy), safety hazards (blocking fire exits, stacking too high), and other objects and abnormal situations.

[0082] For the feature vector of the material photo of cast iron blank, the edge nodes input the features into the convolutional neural network model. The convolutional layer of the model extracts high-level semantic features, and the fully connected layer performs classification and recognition. The output recognition result is "metal casting, intact, neatly stacked, no abnormalities", with a confidence level of 0.92. For the warehouse environment photo, the model recognition result is "warehouse is clean, clearly marked, and has no safety hazards", with a confidence level of 0.88.

[0083] The edge node analyzed each of the 126 feature vectors individually, with the total identification process taking approximately 15 seconds, averaging 0.12 seconds per feature vector. Of the identified results, 120 tasks were normal with no anomalies, while 6 tasks showed anomalies, including 2 instances of damaged material packaging, 3 instances of disorganized material placement, and 1 instance of unclear labeling. The edge node packaged the identification results, including the object category, status assessment, anomaly flag, and confidence level for each task, and transmitted it back to the mobile terminal via Wi-Fi. This resulted in the identification result data, containing 126 records and approximately 20KB in size.

[0084] Based on the identification results data, the spatial location data collected by the mobile terminal in step five (GPS coordinates or indoor positioning coordinates of the operator when performing each task), and the updated inventory data, the mobile terminal uses a lightweight data fusion algorithm to perform semantic association and consistency verification of multimodal data.

[0085] The mobile terminal traversed 63 inbound and outbound tasks. For each task, it extracted data from three aspects: visual recognition results, operation location, and inventory records, and performed correlation analysis. For the outbound task of cast iron blanks, the visual recognition result showed "metal casting, intact condition," the operation location showed storage location A-12-03, and the inventory record showed that the storage location stored material number MAT-001 (cast iron blanks) with a quantity of 850 pieces.

[0086] Using a table lookup matching method, the mobile terminal queries the material basic data table and confirms that the material type corresponding to material number MAT-001 is "metal castings," which matches the material category "metal castings" identified by visual recognition. This confirms that the visual recognition and system data match correctly. Using a rule-based judgment method, it checks whether the material status "intact" identified by visual recognition is consistent with the outbound operation logic (outbound materials should be in an intact state), confirming the operation is reasonable. Using a quantity comparison method, it compares the quantity of materials in the visual recognition image (estimated to be approximately 50 pieces based on material outline counting) with the 50 pieces entered by the operator for outbound. This confirms that the quantities match and there are no anomalies.

[0087] For the six tasks marked as abnormal in the identification results, the mobile terminal performed a detailed consistency check. For the two tasks involving damaged packaging, the mobile terminal checked whether the materials involved in these two tasks were raw materials or semi-finished products with low packaging requirements. If so, the abnormality level was set to "low"; if they were finished products, the abnormality level was set to "high," requiring quality inspection. After verification, both damaged tasks were raw materials, so the abnormality level was set to low, and a prompt message was generated: "The material packaging is slightly damaged, which does not affect use. It is recommended to strengthen material protection."

[0088] For the three tasks with disorganized placement, the mobile terminal checks the space utilization rate of these three storage locations. If the utilization rate exceeds 90%, it is judged as normal dense storage; if the utilization rate is below 70%, it is judged as a genuine problem of disorganization, requiring reorganization. Verification showed that the average utilization rate of the three storage locations was 65%, confirming the disorganization issue and generating the message "Materials in the storage location are not neatly arranged, affecting picking efficiency; reorganization is recommended." For the one task with unclear labeling, the mobile terminal determined that the label was worn or soiled, affecting subsequent identification, and generated the message "Storage location label is unclear; replacement of the label is recommended."

[0089] After multimodal data fusion and consistency verification, the mobile terminal identified 6 abnormal situations, including 4 general prompt anomalies and 2 operation anomalies that require handling. The anomaly identification results include information such as the anomaly task number, anomaly type, anomaly level, anomaly description, and suggested handling measures.

[0090] Based on the six anomalies identified in the anomaly detection results, the mobile terminal employs an anomaly handling strategy to display real-time warning information to the warehouse administrator and record the anomaly handling process. An anomaly notification window pops up on the mobile terminal screen, displaying "Six anomalies detected, please handle accordingly," and listing the anomalies sorted from highest to lowest anomaly severity.

[0091] For exceptions classified as "high" (there are no high-level exceptions in this example), the mobile terminal will block subsequent operations, requiring the exception to be resolved before continuing. For exceptions classified as "low" and "medium," the mobile terminal will display exception information but will not block operations, allowing operators to handle the exception after completing the current task, or record the exception for centralized handling by designated personnel.

[0092] The warehouse manager reviews the anomaly list. For two raw materials with damaged packaging, they click the "Confirmed, does not affect use" button, and the mobile terminal records the anomaly as confirmed but does not process it. For three disorganized storage locations, they click the "Arrange Organization" button, and the mobile terminal generates three organization work orders for these locations, which are assigned to other warehouse personnel for organization during their spare time. For one storage location with unclear labeling, they click the "Replace Label" button, and the mobile terminal generates a label replacement work order, notifying maintenance personnel to replace the label.

[0093] Each exception handling operation is recorded by the mobile terminal, including the exception number, exception discovery time, exception handling method (confirmation / rectification / replacement, etc.), handling personnel, handling time, and handling status (handled / pending). This results in an exception handling record, which serves as a quality archive for warehouse management and is used for subsequent quality analysis and improvement. The mobile terminal synchronizes the exception handling record to the cloud-based warehouse management system. Managers can view all exception records through the management backend, perform statistical analysis, identify weaknesses in warehouse management, and formulate improvement measures.

[0094] Furthermore, anomaly trend analysis can be employed, based on accumulated anomaly handling records, to analyze patterns in the frequency, type, and location of anomalies. Mobile terminals periodically (e.g., weekly) compile anomaly records, categorizing and summarizing them by anomaly type to identify high-frequency anomaly types. For example, if "packaging damage" anomalies account for 40%, it indicates insufficient material protection measures, requiring improvement of material packaging or handling processes. If the anomaly frequency in certain storage areas is significantly higher than in other areas, it suggests potential environmental issues or non-standard operational processes in those areas, requiring focused rectification, resulting in an anomaly trend analysis report. This report helps management identify the root causes of problems from the data, implement targeted improvement measures, and continuously improve warehouse management quality.

[0095] The hybrid decision-making module dynamically switches between online, weak network, and offline modes based on early warning information, adopts corresponding decision-making and synchronization strategies, and outputs consistent data that has been synchronized.

[0096] Specifically, it includes the following: Based on the network connection status of mobile terminals, a continuous network quality monitoring method is adopted. The mobile terminal starts a network monitoring thread in the background to evaluate the network connection quality with edge servers and cloud servers in real time. The monitoring thread sends heartbeat packets to the edge servers and cloud servers every second, and measures the round-trip time as a network latency indicator. At the same time, the packet loss rate and transmission bandwidth in the most recent 10 seconds are calculated.

[0097] Under normal operating conditions, the network latency between the mobile terminal and the edge server is 10-15 milliseconds, the packet loss rate is less than 1%, and the bandwidth is stable at around 100Mbps. The mobile terminal determines that the current network status is excellent and operates in "online mode." In online mode, all data collection, production scheduling calculations, and inventory update operations of the mobile terminal are synchronized with the server in real time, enjoying full system functionality.

[0098] When warehouse managers move to the metal-shielded area or underground storage area of ​​the warehouse, the Wi-Fi signal weakens, network latency increases to 50-80 milliseconds, packet loss rate rises to 5%-10%, and bandwidth drops to around 20Mbps. The mobile terminal detects this network quality degradation and, based on preset network quality thresholds (latency > 50ms or packet loss rate > 5% is considered a weak network), determines the current network status as poor and automatically switches to "weak network mode." In weak network mode, the mobile terminal employs data compression and incremental transmission strategies, prioritizing the transmission of critical business data (such as inbound / outbound operation records and key constraint changes) while delaying the transmission of non-critical data (such as operation logs and statistical reports) to ensure uninterrupted core business operations.

[0099] When the warehouse manager enters the deepest enclosed storage area or near the elevator shaft, the Wi-Fi signal completely disappears, network latency exceeds 5 seconds or there is no response at all, and the packet loss rate reaches 100%. The mobile terminal detects the network connection interruption and, based on a preset offline threshold (no heartbeat response for 5 consecutive seconds is considered offline), determines the current network status as unavailable and automatically switches to "offline mode." In offline mode, the mobile terminal activates its local decision engine; all operations and calculations are completed locally, and data is temporarily stored in a local cache, waiting to be synchronized after the network recovers, thus confirming the current working mode as "offline mode."

[0100] In offline mode, key business data and an offline decision engine stored locally on the mobile terminal are used to support the continuous execution of routine business operations through a rule-based decision-making method. During initialization, the mobile terminal downloads and caches key data from the cloud, including commonly used material information (basic data such as the number, name, specifications, storage location, and safety stock of 1500 materials), typical production scheduling rules (order priority rules, equipment allocation rules, and process sequencing rules), and basic equipment parameters (rated capacity of 20 machines, types of products that can be processed, and maintenance cycles), totaling approximately 50MB of data.

[0101] The warehouse administrator continues to perform inbound and outbound operations in offline mode. When a material's QR code is scanned, the mobile terminal retrieves material information from the local database, displays the material name and specifications, and prompts the operator for confirmation. After the operator enters the outbound quantity, the mobile terminal updates the inventory balance of that material in the local database and saves the operation record (material number, quantity, time, operation type) to a local cache table. Due to the inability to communicate with the server, the inventory update only takes effect locally; the inventory data on the server side has not yet been synchronized.

[0102] For simple production scheduling adjustments, such as delaying the delivery date of an order by one day, the production manager inputs the adjustment command via a mobile terminal. The mobile terminal's offline decision engine reads the locally cached production scheduling plan, modifies it locally according to preset adjustment rules (delaying the delivery date corresponds to delaying the order's start time), and displays the adjusted Gantt chart on the mobile terminal screen. The adjusted production scheduling plan is saved to the local cache and uploaded to the server once the network is restored.

[0103] For complex production scheduling optimization needs, such as adding 5 urgent orders requiring a new global production scheduling optimization, the mobile terminal's offline decision engine assesses the task complexity and finds that the task requires invoking a complete multi-objective optimization algorithm. However, the mobile terminal's local offline decision engine only supports simple rule-based decisions and cannot complete complex optimization calculations. The mobile terminal notifies the production manager that "the current task is complex and requires a network connection to complete. It has been added to the pending task queue and will be processed automatically after the network is restored." The mobile terminal saves the requirement information of this complex production scheduling task (new order data, current constraints) to the pending task queue and marks the task status as "pending."

[0104] After approximately 30 minutes of offline work, the mobile terminal completed 8 inbound / outbound operations, made 2 simple production scheduling adjustments, and recorded 1 complex production scheduling task to be processed. It obtained offline decision results (including 8 inbound / outbound records and 2 production scheduling adjustment records) and a queue of tasks to be synchronized (including 1 complex production scheduling task). These data are temporarily stored locally on the mobile terminal and will be uploaded and synchronized after the network is restored.

[0105] When the warehouse manager leaves the signal-blocked area and returns to the signal coverage area, the mobile terminal's network monitoring thread detects that the Wi-Fi signal has been restored, the network latency has dropped to the normal level of 10 milliseconds, and the packet loss rate has recovered to below 1%. The mobile terminal determines that the network has been restored and automatically switches its working mode from "offline mode" back to "online mode," thus obtaining the current working mode as "online mode."

[0106] Based on one pending task in the task queue and 10 locally cached operation records (8 inbound / outbound records + 2 production scheduling adjustment records), the mobile terminal employs an intelligent data synchronization strategy, uploading data in batches according to task priority and time order. The mobile terminal first uploads the highest priority pending task (a complex production scheduling task), packages the task data, and uploads it to the cloud server. Upon receiving the task, the cloud server immediately initiates production scheduling optimization calculations and sends the optimization results to the mobile terminal 10 seconds later. The production manager then reviews and confirms the new production scheduling plan.

[0107] The mobile terminal then uploads the 10 operation records generated during the offline period, one by one in chronological order. For inbound and outbound records, the cloud server receives the records, updates the inventory database, and compares the inventory balance with the balance calculated locally on the mobile terminal. If they match, the update is performed directly; otherwise, a conflict detection is triggered.

[0108] During synchronization, the mobile terminal detected a conflict in the inventory data for cast iron blanks. The mobile terminal's locally calculated inventory balance was 800 pieces (850 pieces before offline, with 50 pieces issued during the offline period), while the cloud server's inventory balance was 820 pieces (30 pieces were added during the offline period from other terminals). The mobile terminal detected a version conflict, using timestamps and version numbers for conflict detection. The mobile terminal's last update time was 30 minutes before the current time, while the cloud server's last update time was 10 minutes before the current time, indicating that the cloud data update time was later.

[0109] The mobile terminal displays detailed information about the conflicting data to the warehouse administrator, prompting: "Inventory data conflict detected: Local inventory 800 units, server inventory 820 units. Server data updated 10 minutes ago, local data updated 30 minutes ago. Please select a handling method: 1. Keep server data, 2. Keep local data, 3. Manually merge." The warehouse administrator selects "Manually merge." The mobile terminal displays a detailed data change history. After analysis, the warehouse administrator confirms that the 820 units on the server are accurate (other personnel performed inbound operations during this period), and the 50 units outbound operation on the local terminal is also genuine. Therefore, the correct inventory should be 770 units (820 minus 50). The warehouse administrator enters the merged inventory value of 770 units, and the mobile terminal uploads the merged result to the server. The server updates the inventory to 770 units, resolving the conflict.

[0110] After approximately two minutes of data synchronization and conflict resolution, the mobile terminal completed the upload of all locally cached data and the download and update of cloud data. The local database and the cloud database reached a consistent state, resulting in synchronized and consistent data. The mobile terminal then cleared its local cache queue and the queue of tasks awaiting synchronization, resuming normal online operation. All operations were resynchronized in real time to ensure global data consistency and business continuity.

[0111] Furthermore, data synchronization optimization methods can be employed to optimize usage scenarios that frequently enter offline mode. Mobile terminals can be used to statistically analyze historical offline work records, identifying patterns in the time periods, locations, and durations of offline occurrences. If signal issues frequently cause offline activity in certain warehouse areas, feedback should be provided to the IT department suggesting increased Wi-Fi access point coverage. If frequent conflicts occur during offline data synchronization, it indicates that multiple terminals are frequently operating on the same material; in this case, it is recommended to optimize job scheduling to avoid multiple users operating the same material simultaneously. Mobile terminals can also dynamically adjust their local cache data update strategies based on offline frequency and duration. For terminals that frequently go offline, the range and update frequency of locally cached data can be increased, improving offline decision-making capabilities and generating data synchronization optimization suggestion reports to help system administrators continuously improve system performance and user experience.

[0112] A computer-readable storage medium for storing computer-readable instructions that, when read by a computer, enable the execution of the aforementioned mobile terminal-based digital processing system for production scheduling and warehousing inbound / outbound operations.

[0113] In one embodiment of the present invention, a specific example is provided: A 30-day field test was conducted at an auto parts manufacturing company. During the test, a mobile terminal-based digital processing system for production scheduling and warehouse inbound / outbound operations was deployed, covering three production workshops and two warehouse areas of the company, with a test area of ​​approximately 8,000 square meters, involving 25 CNC machining machines, 15 warehouse locations, and 8 on-site operators.

[0114] Table 1 shows an example of how production equipment status data can be obtained. Table 1: Example of obtaining production equipment status data;

[0115] Table 2 shows an example of how to obtain material inventory data: Table 2: Example of material inventory data acquisition;

[0116] Through field testing, the invention has achieved excellent application results. The system can collect and process multi-source information such as production equipment status data and material inventory data in real time, realizing a deep integration of production scheduling and warehouse management, improving production efficiency and inventory management level, and verifying the practical value and application prospects of the method of the invention.

[0117] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A mobile terminal-based digital processing system for production scheduling and warehouse inbound / outbound operations, characterized in that: include: The data acquisition and processing module is used to acquire raw production data using a multi-source parallel acquisition strategy, perform timestamp alignment, cleaning and completion processing on the raw production data, and output a valid dataset. The constraint modeling module, based on an effective dataset, uses an incremental equipment constraint update method and a triggered material constraint update method to obtain a hierarchical dynamic constraint model. The production scheduling optimization module, based on a hierarchical dynamic constraint model, performs task complexity assessment and intelligent calculation allocation on the order demand data in the effective dataset, and outputs a production scheduling plan. The collaborative adjustment module is used to predict and match the production scheduling and warehousing collaborative demand of the production scheduling plan, and output the collaboratively adjusted production scheduling plan. The warehouse operation module is used to generate inbound and outbound task queues based on the collaboratively adjusted production schedule and guide operators to complete inbound and outbound operations, and output inbound and outbound operation records. The anomaly detection module, based on inbound and outbound operation records, uses a multimodal data fusion algorithm to identify abnormal situations and output early warning information; The hybrid decision-making module dynamically switches between online, weak network, and offline modes based on early warning information, adopts corresponding decision-making and synchronization strategies, and outputs consistent data that has been synchronized.

2. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The multi-source data parallel acquisition strategy includes: Establish data channels with production equipment monitoring systems, warehouse management systems, manufacturing execution systems, and order management systems through the network and sensor interfaces of mobile terminals; A multi-threaded parallel data acquisition mechanism is adopted to simultaneously acquire equipment status data, process progress data, material inventory data, and order demand data.

3. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The timestamp alignment, cleaning, and completion processes include: A sliding time window alignment method is adopted, and the width of the time window is set to perform interpolation processing on data records whose time difference exceeds a preset range; Outlier detection methods are used to identify outliers in numerical and categorical data and then replace them. Missing values ​​are filled using the forward filling method.

4. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The incremental device constraint update method includes: A state difference comparison method is used to compare the current state of the equipment with its historical state to identify equipment whose state has changed. For devices whose state has not changed, directly copy the historical constraint parameters; For equipment whose status has changed, recalculate the available processing capacity parameters.

5. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The triggered material constraint update method includes: Set inventory change trigger thresholds and safety stock trigger conditions; When the material inventory change rate exceeds the first threshold or the inventory level falls below the second threshold of the safety stock line, the material's constraint parameters are updated. For materials that have not triggered an update, keep the constraint parameters unchanged.

6. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The task complexity assessment includes: Extract the number of processes, process complexity, number of material requirements, and delivery urgency of pending production orders; The computational complexity index of the production scheduling task is calculated using a linear weighting method; Based on the comparison between the complexity index and the preset threshold, the task is determined to be either lightweight or heavyweight.

7. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The intelligent computing allocation includes: Detect the network connection status and computing resources between the mobile terminal and the edge server and cloud server; A comprehensive evaluation method combining computing resources and network conditions is used to compare the processing capabilities and transmission costs of different computing nodes; Based on the evaluation results, the optimal computing node is selected to execute the production scheduling optimization task.

8. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The production scheduling and warehousing collaborative demand forecasting and matching includes: The material requirements decomposition method is used to calculate the required quantity and time of various materials based on the order information and process route in the production scheduling plan; A demand-inventory matching algorithm is used to compare the material demand with the current available inventory to identify materials with insufficient inventory. A collaborative adjustment strategy is adopted to adjust the order priority and timing in the production scheduling plan based on the material shortage situation.

9. The production scheduling and warehousing inbound / outbound digital processing system based on mobile terminals according to claim 1, characterized in that, The queue for generating inbound and outbound tasks includes: The inbound and outbound task decomposition method is adopted to convert the collaboratively adjusted production scheduling plan into specific material inbound and outbound operation tasks; A task priority ranking algorithm is adopted, which comprehensively considers the urgency of tasks, material attributes, and storage location information to rank tasks. The task queue is displayed to the operator through the mobile terminal interface, guiding them to complete the inbound and outbound operations and recording the operation results.

10. A computer-readable storage medium, characterized in that, It is used to store computer-readable instructions, which, when read by a computer, enable the operation of the mobile terminal-based digital processing system for production scheduling and warehousing inbound and outbound operations as described in any one of claims 1-9.

Images