Packaging carton production scheduling system based on cloud network fusion

Abstract

The present application relates to the field of industrial internet and intelligent manufacturing technology, specifically to a packaging carton production scheduling system based on cloud network fusion, comprising an order processing module, an edge collection module, an edge computing gateway and a data hub; the data hub constructs an ideal scheduling optimization model based on order data and generates an ideal scheduling result, generates a theoretical damaged scheduling result in combination with network disturbance parameters and equipment attenuation parameters, forms a real residual vector based on actual running state data and the ideal scheduling result, and carries out similarity coupling judgment with the theoretical residual vector, outputs a frozen scheduling instruction, an adaptive rearrangement instruction or a scheduling maintenance instruction, so as to distinguish data abnormalities from real abnormalities of the production line, reduce scheduling shock, and improve the consistency of scheduling decision and real production line state.

Description

Technical Field

[0001] This invention relates to the field of industrial internet and intelligent manufacturing technology, specifically a packaging carton production scheduling system based on cloud-network convergence. Background Technology

[0002] In the production process of packaging cartons, the production line usually includes multiple stages such as corrugation forming, printing, slotting, die cutting, and gluing. Different orders vary in delivery time, specifications, process routes, and equipment usage. The workshop also continuously receives rush orders, urgent orders, and regular orders from the business system. Therefore, production scheduling needs to be dynamically adjusted in combination with order flow, equipment operating status, and process connection relationships. The effectiveness of scheduling execution is directly related to the timeliness of production data collection and the accurate judgment of the actual operating status of the equipment. In traditional methods, scheduling systems primarily rely on business orders and workshop-reported progress for planning and adjustment. When network transmission experiences delays, jitter, packet loss, or out-of-order delivery, the completion feedback and equipment status received by the system are prone to lag or distortion. Simultaneously, equipment wear and tear, raw material abnormalities, and reduced operating speed can also lead to a decrease in actual production capacity. Under the combined effect of these two factors, existing methods struggle to distinguish between data anomalies and actual production line anomalies, easily resulting in accidental rescheduling, frequent oscillations, or improper scheduling. Consequently, the stability and accuracy of production scheduling remain relatively low. Summary of the Invention

[0003] The purpose of this invention is to provide a cloud-network convergence-based packaging carton production scheduling system to solve the following technical problems: It avoids frequent scheduling oscillations caused by confusing network disturbances with physical entity anomalies, and enables differentiated handling of network disturbances and physical bottlenecks, improving the consistency between scheduling decisions and actual production line status.

[0004] The objective of this invention can be achieved through the following technical solutions: The cloud-network convergence-based packaging carton production scheduling system includes: an order processing module, which communicates with the business system to provide order data; An edge acquisition module, communicating with production equipment and network nodes, is used to collect actual operating status data; an edge computing gateway; a data hub; both the order processing module and the edge acquisition module are communicating with the data hub via the edge computing gateway; the data hub is used to construct an ideal scheduling optimization model based on the order data and generate an ideal scheduling result, correct the ideal scheduling optimization model based on preset network disturbance parameters and preset equipment attenuation parameters to generate a theoretically damaged scheduling result, generate a real residual vector based on the comparison between the actual operating status data and the ideal scheduling result, generate a theoretical residual vector based on the comparison between the theoretically damaged scheduling result and the ideal scheduling result, perform a coupling decision based on the similarity between the real residual vector and the theoretical residual vector, and output a freeze scheduling instruction, an adaptive rescheduling instruction, or a schedule hold instruction to the edge computing gateway or business system according to the coupling decision result; The network disturbance parameters include at least one of network latency, jitter, packet loss, and out-of-order delivery; the equipment attenuation parameters include at least one of raw material abnormality, equipment wear, and equipment operating speed attenuation.

[0005] In one possible implementation, the order data includes at least a dynamic order pool, bill of materials, delivery date data, product specification data, and process type data; the actual operating status data includes at least equipment cycle time data, work-in-process inventory data, and network service quality data.

[0006] In one possible implementation, the data hub stores a mixed-integer linear programming solution model; The data hub is used to construct the ideal scheduling optimization model based on the dynamic order pool, the bill of materials, the delivery date data, the product specification data, and the process type data, and to solve it using the mixed integer linear programming solution model to obtain the planned start time, planned end time, and equipment allocation relationship of each order process on each equipment, as well as the baseline time series data with time as the horizontal axis and work-in-process quantity or process throughput rate as the vertical axis; the planned start time, planned end time, equipment allocation relationship, and the baseline time series data constitute the ideal scheduling result.

[0007] In one possible implementation, the data hub is further configured to map the network jitter parameters, packet loss parameters, and out-of-order parameters in the network disturbance parameters to an edge state reporting delay model, map the preset raw material anomaly parameters and equipment wear parameters to equipment operating speed attenuation parameters, and introduce the edge state reporting delay model and the equipment operating speed attenuation parameters into the ideal scheduling optimization model to correct the model constraint parameters in order to generate the theoretically damaged scheduling result.

[0008] In one possible implementation, the data hub is also used to compare the actual work-in-process throughput data, characterized by the equipment cycle time data and the work-in-process buffer amount data, with the baseline time-series data in the ideal scheduling result at corresponding times according to a preset sampling period or a unified time window, to obtain the actual throughput deviation data. The theoretical work-in-process throughput data corresponding to the theoretical damage scheduling result is compared with the baseline time series data at the corresponding time to obtain the theoretical throughput deviation data; The actual throughput deviation data constitutes the actual residual vector, and the theoretical throughput deviation data constitutes the theoretical residual vector. The actual residual vector and the theoretical residual vector are either one-dimensional time-series vectors formed based on unified time-series sampling points or multi-dimensional vectors formed based on device dimension and time dimension expansion.

[0009] In one possible implementation, the data hub stores a similarity decision algorithm, which includes a dynamic time warping algorithm or a multidimensional spatial cosine similarity algorithm. The data center is also used to calculate the normalized matching degree between the actual residual vector and the theoretical residual vector using the similarity decision algorithm. The normalized matching degree ranges from 0 to 1, with a larger value indicating a higher degree of matching. The data center is also used to generate a coupling decision result based on the normalized matching degree. The coupling decision result includes network perturbation matching degree and physical attenuation matching degree.

[0010] In one possible implementation, the data hub is further configured to generate the freeze scheduling instruction when the network disturbance matching degree is greater than or equal to a preset high threshold and the physical attenuation matching degree is less than the high threshold; generate the adaptive rearrangement instruction when the physical attenuation matching degree is greater than or equal to the high threshold and the network disturbance matching degree is less than the high threshold; generate a composite anomaly to be reviewed flag when both the network disturbance matching degree and the physical attenuation matching degree are greater than or equal to the high threshold; generate the schedule hold instruction when both the network disturbance matching degree and the physical attenuation matching degree are less than or equal to a preset low threshold; and generate a review flag when either the network disturbance matching degree or the physical attenuation matching degree is greater than the low threshold and less than the high threshold; wherein the high threshold is greater than the low threshold.

[0011] In one possible implementation, the data hub is also used to, after generating adaptive rescheduling instructions, reallocate the process execution order, equipment load, and process flow path of subsequent orders among optional equipment based on the dynamic order pool, equipment cycle time data, work-in-process buffer data, and coupling decision results, and output updated scheduling results.

[0012] In one possible implementation, the production equipment includes a corrugating machine, a printing machine, a slotting machine, a die-cutting machine, and a gluing machine; the actual operating status data includes equipment cycle time data and equipment process status data, wherein the equipment cycle time data includes corrugating machine speed data, printing cycle time data, slotting cycle time data, and gluing cycle time data, and the equipment process status data includes die-cutting pressure data; the process type data includes at least one of the following: corrugation type data, printing process parameters, slotting process parameters, die-cutting process parameters, and gluing process parameters.

[0013] The beneficial effects of this invention are: 1. This invention constructs an ideal scheduling result and introduces network disturbance and equipment attenuation parameters to generate a theoretically damaged scheduling result. The system performs similarity coupling judgment between the actual residual vector formed by actual operation and the theoretical residual vector. This mechanism effectively distinguishes between data anomalies caused by communication transmission and physical anomalies caused by equipment attenuation, avoids erroneous rescheduling due to network fluctuations, and significantly improves the accuracy and stability of scheduling decisions. 2. This invention comprehensively considers multi-dimensional order data and status data such as equipment cycle time and work-in-process buffer quantity, and uses a mixed integer linear programming model to solve for an ideal scheduling result containing baseline time series data; this establishes a unified comparison baseline for complex carton production flow, solves the problem of lack of absolute reference system due to frequent rush orders and complex production line links, and lays a solid data foundation for anomaly identification; 3. This invention maps network latency and jitter to a reporting latency model and equipment wear to a speed attenuation parameter, thereby correcting the ideal model to generate a theoretical damage schedule; combined with the actual and theoretical throughput deviation data extracted based on a unified time window, the system transforms the complex workshop status into a comparable residual vector, realizing a structured and accurate decomposition of network-side disturbances and physical-side attenuation. 4. This invention uses dynamic time warping or multidimensional spatial cosine similarity algorithm to calculate normalized matching degree, and outputs frozen scheduling, adaptive rescheduling or scheduling hold instructions according to preset high and low thresholds. This mechanism upgrades the scheduling action to confidence-driven, and configures differentiated response intensity for different matching results. It effectively prevents frequent oscillations in production line scheduling under slight data fluctuations, and ensures that no real production bottlenecks are missed. 5. After determining the existence of real physical attenuation and generating adaptive rearrangement instructions, the system will reallocate the execution order and flow path of subsequent orders based on the latest dynamic order pool, equipment cycle time, and work-in-process cache. This mechanism can flexibly avoid bottleneck workstations, effectively transform the anomaly identification results into executable production line adjustment actions, reduce the overall delay spread during the bottleneck period, and achieve targeted relief of physical bottlenecks. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of a cloud-network converged packaging carton production scheduling system provided in an embodiment of this application. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Please see Figure 1 The cloud-network converged packaging carton production scheduling system includes: an order processing module, which communicates with the business system to provide order data; and an edge acquisition module, which communicates with production equipment and network nodes to collect actual operating status data. Edge computing gateway; data hub; the order processing module and the edge acquisition module are all connected to the data hub via the edge computing gateway; the data hub is used to build an ideal scheduling optimization model based on order data and generate ideal scheduling results, correct the ideal scheduling optimization model based on preset network disturbance parameters and preset equipment attenuation parameters to generate theoretically damaged scheduling results, generate actual residual vectors based on the comparison between actual operating status data and ideal scheduling results, generate theoretical residual vectors based on the comparison between theoretically damaged scheduling results and ideal scheduling results, make coupling decisions based on the similarity between actual residual vectors and theoretical residual vectors, and output freeze scheduling instructions, adaptive rescheduling instructions or schedule hold instructions to the edge computing gateway or business system according to the coupling decision results; Among them, network disturbance parameters include at least one of network latency, jitter, packet loss, and out-of-order packets, and equipment attenuation parameters include at least one of raw material abnormalities, equipment wear, and equipment operating speed attenuation.

[0017] This embodiment provides a cloud-network convergence-based scheduling mechanism for packaging carton production. Specifically, the mechanism is deployed on a typical carton manufacturing production line, which includes corrugated forming, printing, post-forming, and gluing and shipping. On the business side, the enterprise resource planning system and manufacturing execution system continuously issue orders. On the workshop side, the status of equipment is collected through 5G mobile communication technology and industrial Ethernet. The edge computing gateway is responsible for collecting data from the nearest location and aligning it in time. The data hub is responsible for building scheduling references, identifying sources of anomalies, and outputting scheduling actions. Specifically, the system does not directly regard the current output being lower than the plan as a signal that a rescheduling is necessary. Instead, it first constructs an ideal reference. This ideal reference corresponds to an industrially interpretable, disturbance-free ideal state: there is no transmission delay in the network, no die wear in the equipment, stable paper quality, and order change losses are at a minimum level. The ideal scheduling results are used to construct a unified data comparison benchmark for the production line. The system then injects two of the most common yet most easily confused factors into the reference: one is network-side transmission disturbances, which manifest as late, lost, or out-of-order edge state reporting. Another type is the attenuation on the equipment side and the material side, which manifests as a decrease in machine speed, an increase in pressure compensation, an increase in the probability of paper breakage, or a slowdown in the process cycle time. The theoretical damage scheduling results obtained in this way are not simply statistical predictions, but engineering simulations of how the production line throughput curve should be deformed if the problem only comes from the network or if the problem only comes from the equipment. The data hub compares the actual collected equipment cycle time, work-in-process buffer amount, and network quality information with the aforementioned ideal reference to form a real residual vector; The actual residual vector represents how the actual production line deviates from the ideal production line; for example, if it is only a reporting delay, the workshop entity may have already completed production, but the completion signal on the system interface will arrive late, and the residual will show an overall delay on the time axis. If the die-cutting die wears down gradually, the residual will be more manifested as a continuous increase in the cycle time of the later process and the accumulation of buffer in the later stage. The system then couples the actual residual with various theoretical residuals to determine which cause the current deviation is closer to, and then outputs instructions to freeze the schedule, adaptively rearrange or keep the original schedule respectively. In addition, if the edge acquisition module experiences both device clock speed loss and network service quality loss within a certain period of time, the data center will not make a strong decision directly, but will mark the period as a low confidence period and only allow the original schedule to be maintained or the scheduling status confirmed in the previous cycle to be maintained, in order to prevent accidental rescheduling when the data confidence is insufficient. If the edge computing gateway detects that the timestamp base of the order stream and the device stream are inconsistent, it will perform clock correction first and then enter the residual calculation process. If the theoretically damaged scheduling results fail to cover the current abnormal form, the system can first output a conservative schedule hold instruction and send the abnormal sample to the expert review library. The expert review library is a relational database that stores historical abnormal scheduling features, manual review judgment labels and corresponding scheduling action parameters, which is used to provide a comparative reference system for unknown abnormal forms. A cardboard box factory was producing two types of orders during the day shift. One type was fresh food e-commerce express boxes, with a delivery time margin of less than the preset time threshold and a processing quantity of more than the preset scale threshold. The other type was heavy-duty cardboard boxes for home appliance exports, with printing registration tolerance and die-cutting dimension tolerance needing to be less than the preset precision threshold. Around 10 a.m., the manufacturing execution system interface showed a sudden delay in the feedback of the completion of the printing process, and the work-in-process quantity distribution curve showed a local accumulation pattern. The system does not immediately rearrange the schedule. Instead, it first compares the actual time-of-flight data collected in real time with the ideal scheduling result, and then couples the result with the theoretically damaged scheduling result caused by network jitter. If the two are found to be highly consistent, it is determined that the fluctuation is mainly caused by non-physical deviations in communication transmission, and a freeze scheduling command is output to avoid transferring orders that are still running normally to other devices. Around 2 PM, the die-cutting section experienced a continuous decrease in speed, accompanied by changes in die-cutting pressure compensation and an increase in the downstream buffer. At this point, the actual residual was closer to the theoretical residual of physical attenuation, and the system output an adaptive rearrangement instruction to avoid this bottleneck workstation for subsequent switchable ordinary orders. The purpose of this step is to distinguish between data representation anomalies and physical entity anomalies, thereby enabling differentiated handling of network disturbances and physical bottlenecks, reducing frequent scheduling oscillations, and improving the consistency between scheduling decisions and the actual production line status.

[0018] In a preferred embodiment of the present invention, the order data includes at least a dynamic order pool, a bill of materials, delivery date data, product specification data, and process type data; the actual operating status data includes at least equipment cycle time data, work-in-process inventory data, and network service quality data.

[0019] This embodiment provides a data organization mechanism for scheduling of packaging cartons; specifically, in the aforementioned continuous production scenario, in order to make scheduling reference and anomaly identification have a feasible industrial basis, the order processing module outputs not only the order number, but also a set of data that can determine the production line route, capacity utilization, and order change costs. The edge acquisition module collects not only the device's on / off status, but also multi-dimensional operational status data that can reflect the sources of throughput changes; Specifically, the dynamic order pool reflects the process of order entry, insertion, cancellation, and priority changes. For carton manufacturing, regular orders, rush orders, and rework orders often exist simultaneously in the same shift. Without a dynamic order pool, the system cannot determine whether the current schedule fluctuation is abnormal or actively adjusted by the business side. The bill of materials (BOM) represents the requirements for raw paper, ink, printing rollers, die-cutting dies, and auxiliary materials for each order. Different combinations of materials will change the range of available equipment and preparation time. Delivery date data is used to define the latest completion window. Product specification data is used to determine the box type, size, number of layers, and pressure rating. Process type data determines whether printing, slotting, die-cutting, gluing, and other processes are required, as well as the equipment capacity requirements for each process. Correspondingly, actual operating status data includes at least equipment cycle time data, work-in-process inventory data, and network service quality data. Equipment cycle time data directly reflects the actual processing speed of the equipment per unit time and is the most direct source of physical speed reduction; work-in-process buffer data reflects the balance of logistics between processes. If the upstream is faster than the downstream, it will accumulate in the buffer. Work-in-process buffer data is generated in real time by photoelectric sensors, radio frequency identification acquisition equipment, or machine vision inspection devices deployed in buffer zones between each process. Network service quality data reflects the reliability of status reporting. For example, increased latency, jitter, or packet loss can cause the production line progress obtained by the system to lag behind the actual production line. These three types of data need to be used together to avoid making misjudgments based on a single indicator. As an exception handling mechanism, if the order data is temporarily missing complete process type information, the system can provide a temporary placeholder based on the approved process template of the same specification in the past, but process confirmation is required before the formal schedule is issued. If network service quality data is missing, but the equipment cycle time and cache size are stable, the system will prioritize maintaining the schedule based on physical flow continuity and will not expand the anomaly judgment based on missing network data. If the work-in-process cache size is interrupted due to manual data entry, the edge computing gateway can smoothly connect adjacent sampling windows, but will not overwrite the original records. In the aforementioned cardboard box factories, fresh food e-commerce express box orders typically use a high-speed corrugation, ordinary printing, and fast gluing process, while heavy-duty cardboard box orders for home appliance exports may require high-strength base paper, deeper creases, and higher die-cutting precision; when a batch of high-priority expedited orders is inserted in the morning, the dynamic order pool will update the priority. If the system only sees the increase in the back-end cache but does not know that the new order uses different die-cutting molds and inks, it may misjudge normal order change losses as equipment failures. By reading the order pool, bill of materials and edge takt simultaneously, the system can identify that the stage is a short-term fluctuation driven by business rather than an abnormal stagnation. The purpose of this step is to provide a unified, complete, and interpretable data foundation for subsequent ideal scheduling construction and anomaly source identification, thereby enabling correlation analysis between business flows, physical flows, and network flows.

[0020] In a preferred embodiment of the present invention, the data hub stores a mixed-integer linear programming solution model; the data hub is used to construct an ideal scheduling optimization model based on the dynamic order pool, bill of materials, delivery date data, product specification data and process type data, and solves the model using the mixed-integer linear programming solution model to obtain the planned start time, planned end time and equipment allocation relationship of each order process on each equipment, as well as the baseline time series data with time as the horizontal axis and work-in-process quantity or process throughput rate as the vertical axis; The ideal scheduling result is composed of the plan start time, plan end time, equipment allocation relationship, and baseline timing data.

[0021] This embodiment provides an ideal scheduling benchmark construction mechanism; specifically, order and equipment data alone are insufficient to support anomaly stripping, because without a unified ideal scheduling result, all subsequent deviations lack comparative coordinates. Therefore, the data center stores a mixed-integer linear programming solution model to generate an executable ideal scheduling result under the premise of satisfying process sequence, equipment occupancy constraints and delivery constraints; The mixed-integer linear programming solution model takes minimizing the maximum completion time and minimizing the total delivery delay time as joint optimization objectives, and performs iterative calculations by setting corresponding decision variables; Specifically, the input to the model is not the historical average working hours, but a structured production constraint oriented towards the current order pool and equipment capacity; for each order, the system first determines its necessary processes and the range of optional equipment, and then identifies which equipment can handle the process and which equipment needs to be changed or the die-cutting mold needs to be changed in combination with product specifications and process type. The system is solved under production constraints such as each process cannot be processed repeatedly by multiple devices at the same time, each device can only process one process at a time, and subsequent processes cannot be completed before the preceding processes; the output results include at least three types of information: the planned start time and the planned end time of each process; Second, which machine is assigned to this process; third, the baseline time series data derived from the entire process; after obtaining the specific time interval of each process, the system discretizes the time axis according to a unified time window scale, and counts the planned processing volume of each machine and the theoretical queuing volume between processes within each time window, thereby generating the baseline time series data, such as how many work-in-process should theoretically be in the post-printing buffer at a certain moment, or what level the theoretical throughput of the die-cutting section should reach in a certain time period; the baseline time series data here plays an important role. Gantt charts describe who processes what and when, while baseline time-series data describes how logistics and capacity should flow throughout the entire production cycle. In subsequent judgments, if only the start and end times of the plan are compared, only point-like deviations can often be observed; if the baseline throughput curve is also available, it is possible to observe whether the deviation occurs locally in a certain piece of equipment or spreads along the entire process chain. As an exception tolerance mechanism, if there is a delivery conflict that prevents all orders from being fulfilled at the same time, the system can prioritize high-priority orders according to the preset business strategy, and then output the expected delay mark for low-priority orders. However, the ideal scheduling result still needs to be generated completely to ensure that the residual comparison has a continuous baseline. If a process has no available equipment in the current shift, the system can mark the process as a standby process and postpone its entry into the formal schedule, but will not mistakenly merge it into other process segments; if the order pool changes frequently beyond the preset refresh frequency, the data hub can use a rolling window to periodically rebuild the ideal schedule to avoid the schedule becoming outdated before it is issued; In the aforementioned cardboard box factory, the system simultaneously models e-commerce express boxes, home appliance export boxes, and rework orders; express boxes are mostly processed on high-speed corrugated lines and standard printing machines, while export boxes are preferentially assigned to downstream equipment with higher pressure resistance and more stable die-cutting accuracy. After the solution is completed, the system obtains a shift-level ideal Gantt chart and simultaneously generates theoretical work-in-process curves for several key buffer areas; For example, if under ideal conditions the printed buffer area should maintain a low level of fluctuation between 11:00 and 11:20, but subsequent actual tests show a sudden high level of accumulation, the system can determine that the deviation is abnormal rather than a simple order switch. The purpose of this step is to establish a comparable, traceable, and retrospective absolute reference system to enable the targeted identification of subsequent network disturbances and physical attenuation.

[0022] In a preferred embodiment of the present invention, the data hub is further configured to map the network jitter parameters, packet loss parameters, and out-of-order parameters in the network disturbance parameters to an edge state reporting delay model, map the preset raw material anomaly parameters and equipment wear parameters to equipment operating speed attenuation parameters, and introduce the edge state reporting delay model and equipment operating speed attenuation parameters into the ideal scheduling optimization model to correct the model constraint parameters in order to generate theoretically damaged scheduling results.

[0023] This embodiment provides a theoretical damaged schedule generation mechanism; specifically, it is still difficult to explain the cause of the deviation by simply comparing the ideal schedule with real-time sampling, because the same late completion may come from network-reported lateness or from the equipment actually slowing down. Therefore, in addition to the ideal benchmark, this embodiment further constructs theoretically damaged scheduling results with single or combined perturbation characteristics as a control sample for cause identification. Specifically, network jitter, packet loss, and out-of-order delivery do not directly change the actual processing speed of cardboard on the equipment, but they do change the time sequence in which the data center obtains production progress; therefore, the system maps these network phenomena to an edge state reporting latency model. The industrial implications of this model are that it is not the production line itself that slows down, but rather that the time it takes for the completion message of a certain process to reach the scheduling system is delayed, scattered, or misaligned; conversely, raw material abnormalities and equipment wear and tear will directly affect the physical processing. For example, abnormal moisture content in raw paper increases the probability of paper breakage and deviation, and wear on the die-cutting die increases the pressure compensation requirement and lengthens the cycle time; therefore, these factors are mapped to equipment operating speed attenuation parameters to correct the available capacity of the equipment and the duration of the process. The specific correction logic is as follows: For the edge state reporting delay model, the system does not change the actual physical processing time and equipment capacity constraints of each process in the ideal scheduling optimization model, but only adds a time bias vector to the theoretical completion time node derived by the model to simulate the lag of state data transmission in the network node. The theoretical completion time node is set as The time bias vector is The observation completion time is affected by network disturbances. Satisfy the equation: in, It follows a non-negative probability distribution determined by the network perturbation parameters; For the equipment operating speed attenuation parameter, the system directly uses it as a reduction factor to apply to the equipment unit throughput constraint in the ideal scheduling optimization model, thereby stretching the theoretical processing time of each process solved by the model accordingly; in implementation, the data center can generate multiple sets of theoretically damaged scheduling results; For example, simulation states such as network jitter only, packet loss only, die-cutting wear only, and material abnormality superimposed on subsequent speed reduction can be generated respectively; each simulation state distorts the ideal scheduling result in a specific direction, thereby obtaining the theoretical deviation form corresponding to a certain type of abnormality mechanism; The significance of doing this is that the subsequent real residuals are no longer isolated data lacking theoretical reference, but can be compared item by item with theoretical samples of specific mechanisms; In addition, if both network disturbance parameters and equipment attenuation parameters are significant, the system can generate combined theoretical damage scheduling results instead of forcibly attributing them to a single cause. If there is a temporary lack of raw material quality testing input on site, the system can first use the vehicle speed decay and process compensation trend monitored by the equipment side over a long period of time to approximate the physical decay, but the theoretical sample must be marked as a low confidence combination. If a certain parameter exceeds the equipment's safety boundary, such as when the die-cutting mold is so worn that production cannot continue, the system will no longer treat it as a normal speed-down simulation, but will directly trigger a shutdown risk warning and suspend the equipment from participating in the subsequent ideal comparison; the aforementioned cardboard box factory experienced two different situations at noon; The first scenario is that the factory's wireless network is temporarily congested, and the completion records uploaded by the edge gateway arrive in batches with a delay. The theoretical sample of the reporting delay constructed by the system will show a pattern where the cycle time of the front-end equipment is basically normal, but the completion curve of the central side is shifted backward as a whole. The second scenario is that the cutting edge of the die-cutting die becomes dull after continuous use. In order to ensure the molding quality, the equipment automatically increases the pressure and reduces the operating speed. The theoretical sample of equipment attenuation constructed by the system will show a pattern of continuously decreasing throughput of the subsequent process and gradual accumulation of the upstream buffer. Although both will cause deviation from the plan, their residual profiles are completely different. The purpose of this step is to transform empirical sources of anomalies into comparable theoretical deviation samples, thereby enabling a structured breakdown of network-side and device-side anomalies.

[0024] In a preferred embodiment of the present invention, the data hub is further configured to compare the actual work-in-process throughput data characterized by equipment cycle time data and work-in-process buffer quantity data with the baseline time-series data in the ideal scheduling result at corresponding times according to a preset sampling period or a unified time window, so as to obtain the actual throughput deviation data. The theoretical work-in-process throughput data corresponding to the theoretical damage scheduling results is compared with the baseline time series data at the corresponding time to obtain the theoretical throughput deviation data. Among them, the actual throughput deviation data constitutes the actual residual vector, and the theoretical throughput deviation data constitutes the theoretical residual vector. The actual residual vector and the theoretical residual vector are either one-dimensional time series vectors formed based on unified time series sampling points or multi-dimensional vectors formed based on device dimension and time dimension expansion.

[0025] This embodiment provides a dual-track differential extraction mechanism; specifically, after generating the ideal scheduling result and the theoretically damaged scheduling result, the system also needs to uniformly convert the multi-source heterogeneous workshop operating status into a comparable deviation expression, otherwise the subsequent similarity judgment will lack a consistent input caliber. Specifically, the data center aligns and compares the actual work-in-process throughput data with the ideal baseline time-series data according to a preset sampling period or a unified time window to obtain the actual throughput deviation data; the actual work-in-process throughput data here is jointly represented by the equipment cycle time and buffer quantity; looking only at the equipment cycle time may ignore the logistics blockage between processes; Looking only at the buffer quantity might lead to misinterpreting proactive material preparation as an anomaly; therefore, the actual residual formed by combining the two can better reflect how deviations propagate on the production line. The system then performs the same comparison on the theoretically damaged scheduling results to obtain theoretical throughput deviation data; the two types of deviations are unified to form a residual vector, which facilitates comparison with the same caliber; In terms of expression, if only the total throughput of the entire line is considered, a one-dimensional time-series vector can be used; for example, five consecutive sampling windows can be used to describe whether the current shift is lower, close to, higher, or lagging behind the ideal state in these five windows respectively. If it is necessary to locate a specific device, it can be expanded into a multi-dimensional vector. For example, the deviation status of corrugation, printing, slotting, die-cutting, and gluing in each time window can be arranged sequentially along the device dimension. The advantage of multi-dimensional expression is that it can not only see whether the deviation exists, but also see where the deviation first appears and where it spreads. When forming multidimensional vectors, since the physical dimensions of the process throughput represented by the equipment cycle data and the work-in-process buffer data are different, the data center will pre-divide the calculated actual throughput deviation data and theoretical throughput deviation data by the ideal reference value of the corresponding equipment or buffer, and uniformly convert the absolute deviation into a dimensionless relative deviation magnitude to ensure that the subsequent calculations will not cause the feature extraction weights to be distorted due to the difference in data magnitude. To facilitate understanding, a simplified example can be used for verification: Suppose we observe three time windows T1, T2, and T3. The ideal benchmark shows that the die-cutting section should maintain continuous output within these three windows. However, the actual sampling shows that T1 is basically normal, T2 starts to decrease, and T3 continues to decrease. At the same time, the buffer amount after printing increases in T2 and T3. Therefore, the actual residual will show a profile of slowing down in the later stage and back pressure upstream. If a theoretical sample happens to show the same profile, then the two are highly comparable; conversely, if the theoretical sample reported by the network shows that the three data segments are all late, and the cache does not actually accumulate, then its residual profile is different from the above situation. As an anomaly tolerance mechanism, if the actual sampling period is slightly uneven due to network fluctuations, the edge computing gateway can first resample according to a unified time window and then send it to the data hub to avoid direct misalignment of data from different sources; if a device lacks effective tick data in a certain window, while the upstream and downstream buffers are continuous, the system can first mark the window as a local missing measurement and not interrupt the entire residual vector with a single missing measurement point. If the time scale of the theoretical damage scheduling result is inconsistent with the actual sampling scale, then the time axis of the ideal benchmark should be used for unified mapping first. When the die-cutting section of the aforementioned carton factory gradually slows down in the afternoon, the actual residual will be as follows: the throughput of the die-cutting process decreases in several consecutive windows, the pre-die-cutting buffer gradually increases, and the gluing section will experience insufficient material supply later. If the congestion is only reported online on the morning of the day, the actual residual is more likely to be the concentrated reporting of the completion records of multiple processes at the same time, while the equipment cycle time and cache do not form a real blockage. By using this dual-track differential extraction on a unified time axis, the system can convert late data arrival and slowed-down equipment into two distinguishable deviation profiles. The purpose of this step is to project the complex and heterogeneous workshop state into a comparable residual space, thereby standardizing the basic input for subsequent coupled decisions.

[0026] In a preferred embodiment of the present invention, the data center stores a similarity decision algorithm, which includes a dynamic time warping algorithm or a multidimensional space cosine similarity algorithm. The data center is also used to calculate the normalized matching degree between the actual residual vector and the theoretical residual vector using a similarity decision algorithm. The normalized matching degree ranges from 0 to 1, with a larger value indicating a higher degree of matching. The data hub is also used to generate coupling decision results based on the normalized matching degree, which includes network perturbation matching degree and physical attenuation matching degree.

[0027] This embodiment provides a residual coupling decision mechanism; specifically, it is not enough to simply obtain the actual residual and the theoretical residual, the system also needs to determine the degree of matching between the actual deviation and various theoretical causes; To this end, a data central storage similarity decision algorithm is used to quantify the degree of agreement between two types of residuals into a normalized matching degree. Specifically, the dynamic time warping algorithm is suitable for handling deviation profiles that are similar in shape but have slight time scaling. For example, if a network congestion lasts for a slightly longer time than the theoretical sample, but its behavior is still typical of delayed reporting, the algorithm can identify that the two belong to the same type of pattern. It should be noted that the original output of the dynamic time warping algorithm is the cumulative distance where smaller values ​​indicate greater similarity. Therefore, the system converts it into a normalized matching degree through a negative exponential transformation. For example, using computational logic: in, To control the preset coefficient of the mapping curvature, thereby smoothly mapping non-negative distance values ​​to... Interval; multidimensional spatial cosine similarity is more suitable for comparing whether the deviation directions are consistent after unfolding multiple devices and multiple time windows; For example, if the deviation is mainly concentrated between die-cutting and gluing, and the direction of upstream diffusion is consistent with the theoretical wear sample, then the matching degree will be higher; Due to the original value of cosine similarity Between -1 and 1, the system uses a linear mapping to scale the range to eliminate the negative value interval, for example, by employing computational logic. ; After the above transformation, the normalized matching degree The value of is strictly uniformly between 0 and 1. Its physical meaning is the degree of fit between the actual deviation and a certain theoretical mechanism, rather than the size of the deviation. The larger the value, the higher the degree of matching, thus laying the foundation for the uniform high and low threshold judgment in the future. To make it easier to understand, a simplified explanation can be given: If the system holds two sets of theoretical residuals, namely network-reported delay samples and device speed-down samples, when the actual residuals are shifted backward on the time axis but the actual cycle time of each device does not change much, the matching degree with the former will be higher than that with the latter. When the actual residual manifests as a device starting to slow down and the upstream and downstream cache relationship changes, the matching degree with the latter will be higher; the data center will form network disturbance matching degree and physical attenuation matching degree respectively based on this, as the basis for subsequent decision-making. As an anomaly tolerance mechanism, if both types of matching degrees are high at the same time, it indicates that there may be a superposition of network disturbance and physical attenuation on site. In this case, the system does not directly make a single attribution, but outputs a composite anomaly label or a label to be reviewed. If both types of matching degrees are low, it means that the existing theoretical samples cannot explain the current bias. In this case, the system will maintain the current schedule and record the residual shape for subsequent supplementation of the theoretical sample library. If the actual residual vector length is insufficient, for example, if the sampling window is too short, the system temporarily suspends the coupling decision to avoid amplifying occasional fluctuations due to insufficient observation. When the network was unstable in the aforementioned cardboard box factory in the morning, the matching degree between the actual residual and the network jitter sample was significantly higher than that of the die-cutting wear sample, because the actual cycle time did not decrease synchronously, and the deviation was mainly reflected in the disorder of the data arrival time. In the afternoon, the die-cutting pressure compensation increased, the cycle time decreased, and insufficient material supply in the later stages occurred continuously. The actual residuals were more consistent with the physical attenuation samples, and the network disturbance matching degree was relatively reduced. Based on this, the data center output the network disturbance matching degree and the physical attenuation matching degree respectively, as the basis for scheduling action selection. The purpose of this step is to transform ambiguous sources of anomalies into sortable and comparable coupled results, thereby enabling the transition from identifying biases to interpreting them.

[0028] In a preferred embodiment of the present invention, the data hub is further configured to generate a freeze scheduling instruction when the network disturbance matching degree is greater than or equal to a preset high threshold and the physical attenuation matching degree is less than a high threshold; generate an adaptive rearrangement instruction when the physical attenuation matching degree is greater than or equal to a high threshold and the network disturbance matching degree is less than a high threshold; generate a composite anomaly pending review flag when both the network disturbance matching degree and the physical attenuation matching degree are greater than or equal to a high threshold; generate a schedule hold instruction when both the network disturbance matching degree and the physical attenuation matching degree are less than or equal to a preset low threshold; and generate a pending review flag when either the network disturbance matching degree or the physical attenuation matching degree is greater than a low threshold and less than a high threshold; wherein, the high threshold is greater than the low threshold.

[0029] This embodiment provides a hierarchical decision output mechanism; specifically, although the aforementioned coupled decision can identify which type of cause the anomaly is closer to, without clear action hierarchy, the system may still frequently change the schedule under slight fluctuations. Therefore, this embodiment introduces a hierarchical decision interval consisting of a high threshold and a low threshold, enabling the data center to select different responses such as freezing, rearranging, keeping, or waiting for review based on the strength of the matching degree. Specifically, when the network disturbance matching degree reaches a level greater than or equal to the preset high threshold, it indicates that the current deviation is more likely to come from temporary distortion of the data link rather than a significant reduction in physical capacity. If the schedule is rearranged rashly at this time, it will disrupt the equipment that was originally operating normally. Therefore, the system generates a freeze schedule instruction, that is, the current plan will not be changed for the time being, and only monitoring will continue. Conversely, when the physical attenuation matching degree reaches a level greater than or equal to the preset high threshold, it indicates that the bottleneck is more likely to occur on the actual equipment or material side. At this time, continuing to adhere to the original schedule will cause subsequent orders to accumulate at the bottleneck workstation. Therefore, the system generates an adaptive rescheduling instruction. If both types of matching degrees are less than or equal to the preset low threshold, it means that the deviation has not yet formed a clear abnormal profile. The system generates a schedule hold instruction to avoid over-response. If the matching degree is between the high and low thresholds, it means that the confidence level is not yet sufficient or the anomaly is in the early stage. The system generates a check mark, which can be handed over to the on-duty dispatcher for confirmation in combination with the on-site process status. This high and low threshold stratification essentially changes workshop scheduling from single-point triggering to confidence-driven scheduling. The high threshold corresponds to the high confidence level that triggers the scheduling action, while the low threshold corresponds to the low confidence level that falls within the normal fluctuation tolerance range. The buffer between the two is used to absorb common unstable factors such as order changes, short-term manual intervention, and high-frequency abnormal pulses from sensors. As an anomaly tolerance mechanism, if the network disturbance matching degree and the physical attenuation matching degree both enter the high threshold range, the system should prioritize outputting a check mark or a combined alarm, rather than freezing and rearranging at the same time; if the check state continues for several cycles and the matching degree gradually increases, the system can automatically upgrade to freeze scheduling or adaptive rearrangement. If the high and low thresholds need to be adjusted due to different factories and shifts, they can be configured by the data center based on the deviation distribution of historical stable operating periods, but the threshold relationship should be maintained so that the high threshold is greater than the low threshold. Specifically, the high threshold and low threshold are obtained by statistically calculating the feature matching degree data within the historical fault-free operation cycle. The system sets the value of the cumulative distribution function of matching degree reaching the preset quantile as the corresponding judgment boundary. In the aforementioned network congestion incident at the cardboard box factory in the morning, the actual residual closely matched the theoretical sample of network disturbance, but differed significantly from the physical attenuation sample. Based on this, the system generated a freeze scheduling instruction and did not reallocate the already scheduled order of express boxes and export boxes. During the brief midday rescheduling, the deviation did not significantly match any theoretical samples, and the system maintained the original schedule. In the afternoon, when the die-cutting section slowed down and continued to propagate to the downstream processes, the physical attenuation matching degree increased to a level sufficient to affect delivery schedule assurance, and the system switched to generating adaptive rescheduling instructions. If both are in the middle range at a certain time, a check mark is displayed on the workshop Kanban board, prompting the dispatcher to check the raw paper quality and network switch status on-site. The purpose of this step is to configure different intensity scheduling actions for abnormal matching results with different confidence levels, so as to achieve a balanced control that avoids erroneous rescheduling without missing the real bottleneck.

[0030] In a preferred embodiment of the present invention, the data hub is further configured to, after generating the adaptive rescheduling instruction, reallocate the process execution order, equipment load and process flow path of subsequent orders among the selectable equipment based on the dynamic order pool, equipment cycle time data, work-in-process buffer data and coupling decision results, and output the updated scheduling result.

[0031] This embodiment provides an adaptive reordering mechanism; specifically, merely determining the existence of a real physical bottleneck is insufficient to solve the problem, because the scheduling system ultimately needs to redirect subsequent orders to a more feasible execution path. Therefore, after generating the adaptive reordering instruction, the data hub combines the current order pool, device takt time, cache status and coupling decision results to dynamically reallocate subsequent orders; Specifically, the rescheduling does not simply postpone all orders. Instead, it takes the current work-in-process inventory and the latest available equipment capacity as the new initial state, retains the completed and uninterruptible process paths, and inputs the remaining optional subsequent orders into the scheduling optimization model for solution, or reconstructs the local production route through heuristic rules. Specifically, if the bottleneck is located in the die-cutting section, the system identifies which subsequent orders have alternative die-cutting equipment, which orders can proceed with printing and slotting first and postpone die-cutting, and which orders must wait for the original equipment to be restored due to their highly specialized processes. Then, based on the current work-in-process inventory, it determines whether it is necessary to reduce the upstream release speed to prevent the upstream from continuing to push materials to the bottleneck section. At the same time, the equipment cycle time data is used to identify the remaining capacity of other optional equipment to avoid transferring the problem from one piece of equipment to another piece of equipment that is about to be saturated. In terms of process execution sequence, the system can prioritize orders with tighter delivery dates, lower changeover costs, or those that can be completed quickly, in order to reduce the overall delay spread during bottleneck periods; in terms of equipment load, the system can distribute tasks that were originally concentrated on one piece of equipment to multiple optional equipment. Regarding the process flow path, the system can adjust the process flow path of certain orders. For example, the previous process can be completed first and put into a controllable cache, and the subsequent process can be released in a concentrated manner after the bottleneck workstation is restored. The updated scheduling results include not only the new start and end times, but also the new equipment allocation and cache release rhythm. In addition, if there is no alternative to the bottleneck equipment, the system will not perform a formally invalid rearrangement, but will switch to a delay minimization strategy, such as prioritizing urgent orders, suspending material feeding for low-priority orders, and simultaneously outputting delivery date warnings. If multiple devices experience degradation simultaneously, the system can perform local reordering based on the priority of the critical path affecting delivery time, rather than overturning all plans globally at once; if an order has entered an uninterruptible process, its current execution path will be maintained to avoid the risk of scrapping work-in-process and semi-finished products due to forced switching; After the die-cutting dies at the aforementioned cardboard box factory wore out in the afternoon, the system identified a batch of ordinary express boxes that could be transferred to the backup die-cutting equipment. However, due to the dedicated die-cutting dies and high precision requirements of export heavy-duty boxes, it was not advisable to switch them at the moment. Therefore, the system first directed the ordinary orders to the backup equipment to relieve the burden on the main die-cutting machine. At the same time, the feeding rhythm of heavy-duty boxes for export in the upstream printing section is reduced to prevent the buffer before die-cutting from continuously exceeding the capacity limit; for fresh food boxes with the tightest delivery time, the system prioritizes them to take alternative routes and compresses the waiting time; after the rescheduling is completed, the data center outputs the updated scheduling results and synchronizes them to the manufacturing execution system and the workshop dashboard; The purpose of this step is to truly transform the anomaly identification results into actionable production adjustments, thereby achieving targeted relief of real bottlenecks.

[0032] In a preferred embodiment of the present invention, the production equipment includes a corrugating machine, a printing machine, a slotting machine, a die-cutting machine, and a gluing machine; the actual operating status data includes equipment cycle time data and equipment process status data, wherein the equipment cycle time data includes corrugating machine speed data, printing cycle time data, slotting cycle time data, and gluing cycle time data, and the equipment process status data includes die-cutting pressure data; the process type data includes at least one of the following: corrugation type data, printing process parameters, slotting process parameters, die-cutting process parameters, and gluing process parameters.

[0033] This embodiment provides a mechanism for implementing equipment and processes for a specific carton production line. Specifically, in order to enable the aforementioned scheduling, residual and coupling decisions to be directly applied to the carton manufacturing site, this embodiment further defines the typical equipment configuration and the corresponding key data collection and process parameters. Specifically, the corrugating machine is responsible for cardboard forming, and its speed directly affects the supply capacity of the entire production line; the printing machine determines the graphic processing rhythm; and the slotting machine affects the geometric accuracy of folding and forming. The die-cutting machine determines the cutting efficiency of irregularly shaped boxes or high-precision structural parts; the gluing machine is the final cycle control point for the subsequent packaging into boxes. For these devices, simply collecting start-up and shutdown status data is far from sufficient; corrugating machine speed can reflect the adaptability of raw paper and the stability of front-end feeding; printing cycle time, slotting cycle time, and gluing cycle time are used to identify whether each process follows the plan; and die-cutting pressure data has unique value because die wear, changes in paperboard strength, and die-cutting accuracy compensation are usually reflected in pressure adjustment first, and then further transmitted as a decrease in cycle time; process type data also needs to correspond to the equipment capacity. The flute profile determines the thickness and compressive strength of the cardboard; different flute profiles affect the difficulty of corrugation and subsequent folding. Printing process parameters determine the difficulty of plate changing and overprinting. The grooving process parameters affect the tool load and machining stability; the die-cutting process parameters determine the die-cutting pressure window and cycle time limit; the gluing process parameters are related to the amount of glue, pressing time and box exit stability. After these parameters are introduced into the scheduling system, the system can then identify the process sensitivity of each order to specific equipment and potential bottleneck processes. As an anomaly tolerance mechanism, if a certain piece of equipment does not yet have complete process status sensors, such as an old grooving machine that cannot upload fine-grained status in real time, it can be temporarily replaced by cycle time data and upstream and downstream buffer quantities, but the confidence level of the conclusions related to the equipment should be reduced when attributing the anomaly. If sensor drift occurs in the die-cutting pressure data, the system can perform cross-verification by combining the die-cutting cycle time and finished product sampling results to avoid misjudging sensor failure as die wear. If certain order process parameters are not entered electronically, the process engineer can fill in the key fields before the order is issued; otherwise, the system will only allow them to enter the manual review and scheduling pool. In the aforementioned continuous shift production of the cardboard box factory, fresh express boxes mostly use conventional corrugated type and high-speed gluing process. The focus is on whether the speed of the corrugating machine and the gluing cycle are coordinated. For export boxes of home appliances, which involve higher strength cardboard and stricter die-cutting precision, the system will pay special attention to whether changes in die-cutting pressure occur before the decrease in cycle time. When the die-cutting pressure continues to rise in the afternoon while the printing and slotting cycle times remain stable, the system will focus the abnormality on the die-cutting section, rather than mistakenly believing that the entire production line has lost speed. Meanwhile, if only out-of-order reports occur on the network in the morning, while the cycle times of corrugating, printing, slotting, and gluing are close to the ideal baseline, the system can more confidently determine that it is a communication layer problem. The purpose of this step is to anchor the abstract scheduling and decision logic to specific equipment, specific processes, and specific sensor data, thereby enabling the solution to be implemented in the engineering of packaging carton production scenarios.

[0034] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A packaging carton production scheduling system based on cloud-network convergence, characterized in that, include: The order processing module communicates with the business system and is used to provide order data; The edge acquisition module communicates with production equipment and network nodes to collect actual operating status data; Edge computing gateway; Data Hub; the order processing module and the edge acquisition module are both connected to the data hub via the edge computing gateway; the data hub is used to construct an ideal scheduling optimization model based on the order data and generate an ideal scheduling result, modify the ideal scheduling optimization model based on preset network disturbance parameters and preset device attenuation parameters to generate a theoretically damaged scheduling result, generate a real residual vector based on the comparison between the actual operating status data and the ideal scheduling result, generate a theoretical residual vector based on the comparison between the theoretically damaged scheduling result and the ideal scheduling result, perform a coupling decision based on the similarity between the real residual vector and the theoretical residual vector, and output a freeze scheduling instruction, an adaptive rescheduling instruction, or a schedule hold instruction to the edge computing gateway or business system according to the coupling decision result; The network disturbance parameters include at least one of network latency, jitter, packet loss, and out-of-order delivery; the equipment attenuation parameters include at least one of raw material abnormality, equipment wear, and equipment operating speed attenuation.

2. The cloud-network convergence-based packaging carton production scheduling system according to claim 1, characterized in that, The order data includes at least a dynamic order pool, bill of materials, delivery date data, product specification data, and process type data; the actual operating status data includes at least equipment cycle time data, work-in-process inventory data, and network service quality data.

3. The cloud-network convergence-based packaging carton production scheduling system according to claim 2, characterized in that, The data hub stores a mixed-integer linear programming solution model; The data hub is used to construct the ideal scheduling optimization model based on the dynamic order pool, the bill of materials, the delivery date data, the product specification data, and the process type data, and to solve it using the mixed integer linear programming solution model to obtain the planned start time, planned end time, and equipment allocation relationship of each order process on each equipment, as well as the baseline time series data with time as the horizontal axis and work-in-process quantity or process throughput rate as the vertical axis; the planned start time, planned end time, equipment allocation relationship, and the baseline time series data constitute the ideal scheduling result.

4. The cloud-network convergence-based packaging carton production scheduling system according to claim 3, characterized in that, The data center is also used to map the network jitter parameters, packet loss parameters, and out-of-order parameters in the network disturbance parameters to an edge state reporting delay model, map the preset raw material anomaly parameters and equipment wear parameters to equipment operating speed attenuation parameters, and introduce the edge state reporting delay model and the equipment operating speed attenuation parameters into the ideal scheduling optimization model to correct the model constraint parameters in order to generate the theoretically damaged scheduling result.

5. The cloud-network convergence-based packaging carton production scheduling system according to claim 4, characterized in that, The data hub is also used to compare the actual work-in-process throughput data, characterized by the equipment cycle data and the work-in-process buffer amount data, with the baseline time-series data in the ideal scheduling result at the corresponding time according to a preset sampling period or a unified time window, to obtain the actual throughput deviation data. The theoretical work-in-process throughput data corresponding to the theoretical damage scheduling result is compared with the baseline time series data at the corresponding time to obtain the theoretical throughput deviation data; The actual throughput deviation data constitutes the actual residual vector, and the theoretical throughput deviation data constitutes the theoretical residual vector. The actual residual vector and the theoretical residual vector are either one-dimensional time-series vectors formed based on unified time-series sampling points or multi-dimensional vectors formed based on device dimension and time dimension expansion.

6. The cloud-network convergence-based packaging carton production scheduling system according to claim 5, characterized in that, The data center stores a similarity decision algorithm, which includes a dynamic time warping algorithm or a multidimensional spatial cosine similarity algorithm. The data center is also used to calculate the normalized matching degree between the actual residual vector and the theoretical residual vector using the similarity decision algorithm. The normalized matching degree ranges from 0 to 1, with a larger value indicating a higher degree of matching. The data center is also used to generate a coupling decision result based on the normalized matching degree. The coupling decision result includes network perturbation matching degree and physical attenuation matching degree.

7. The cloud-network convergence-based packaging carton production scheduling system according to claim 6, characterized in that, The data hub is further configured to generate the freeze scheduling instruction when the network disturbance matching degree is greater than or equal to a preset high threshold and the physical attenuation matching degree is less than the high threshold; generate the adaptive rearrangement instruction when the physical attenuation matching degree is greater than or equal to the high threshold and the network disturbance matching degree is less than the high threshold; generate a composite anomaly to be reviewed flag when both the network disturbance matching degree and the physical attenuation matching degree are greater than or equal to the high threshold; generate the schedule hold instruction when both the network disturbance matching degree and the physical attenuation matching degree are less than or equal to a preset low threshold; and generate a review flag when either the network disturbance matching degree or the physical attenuation matching degree is greater than the low threshold and less than the high threshold; wherein the high threshold is greater than the low threshold.

8. The cloud-network convergence-based packaging carton production scheduling system according to claim 7, characterized in that, The data hub is also used to, after generating the adaptive rescheduling instruction, reallocate the process execution order, equipment load, and process flow path of subsequent orders among the selectable equipment based on the dynamic order pool, the equipment cycle time data, the work-in-process cache data, and the coupling decision result, and output the updated scheduling result.

9. The cloud-network convergence-based packaging carton production scheduling system according to any one of claims 1 to 8, characterized in that, The production equipment includes a corrugating machine, a printing machine, a slotting machine, a die-cutting machine, and a gluing machine; the actual operating status data includes equipment cycle time data and equipment process status data, wherein the equipment cycle time data includes corrugating machine speed data, printing cycle time data, slotting cycle time data, and gluing cycle time data, and the equipment process status data includes die-cutting pressure data; the process type data includes at least one of the following: corrugation type data, printing process parameters, slotting process parameters, die-cutting process parameters, and gluing process parameters.

Images