An advertisement material production scheduling method and system based on production working condition perception

Abstract

The application discloses a kind of based on production condition perception's advertisement material production scheduling method and system, it is related to advertisement material manufacturing management technical field.The kind of based on production condition perception's advertisement material production scheduling method and system, including S1, collection and pre-processing advertisement material production process in execution time related data and resource queue data, constructs standardization production condition dataset;S2, analysis rework influence in production process in the degree of conduction, adjustment execution arrangement in production process;S3, assesses the rhythm stability degree of production process, linkage constraint production execution arrangement in time window occupancy and task release rhythm;S4, assesses the adjustable degree of task time landing point, corrects task plan start time.It solves the problem that existing advertisement material production management relies on single task average duration scheduling, lacks the depiction of rework timing aggregation and cross-node conduction disturbance, leading to scheduling deviation accumulation and causing resource occupancy imbalance.

Description

Technical Field

[0001] This invention relates to the field of advertising material manufacturing management technology, specifically to an advertising material production scheduling method and system based on production status perception. Background Technology

[0002] With the continuous improvement of digital production and process collaboration capabilities, advertising material production has gradually evolved from a linear production model driven by human experience to a process-oriented production model with multi-node parallel collaboration. Personnel, equipment, and production tasks are highly coupled within the same time window, making the impact of production scheduling on production efficiency and delivery stability increasingly prominent. Currently, advertising material production management typically relies on project management tools or production schedules to break down and arrange task durations, using historical average processing times or standard production cycles as the primary scheduling basis. Deviations in production schedules are addressed through manual adjustments or simple rule corrections. In this process, the operational status of production nodes, resource usage, and task flow relationships are mostly recorded statically, lacking continuous perception and structured expression of real-time production conditions, making it difficult to fully reflect the dynamic changes in the production process.

[0003] For example, invention patent CN118798625B discloses an advertising review method, apparatus, device, and storage medium. The method includes: generating a smart contract based on preset advertising review rules and preset listing rules, and publishing the smart contract to the blockchain; performing automated review based on advertising material data and advertiser data to obtain automated review results; obtaining a review index and ranking advertising reviewers to obtain an advertising reviewer data list; determining target advertising reviewers based on the advertising reviewer data list, and allocating advertising materials to target advertising reviewers for manual review through the smart contract to obtain manual review results; publishing the automated review results and manual review results to the blockchain, and calculating the advertising review result through the smart contract. This method addresses the problems of low advertising review efficiency and the risk of human-caused conflicts of interest, improves review efficiency, avoids conflicts of interest between reviewers and advertisers, and ensures fair and impartial listing and ranking of compliant advertisements.

[0004] For example, the invention patent with announcement number CN115456529B relates to a method, system, computer device, and storage medium for early warning of advertising material inventory. The method includes: obtaining the types and quantities of advertising materials running daily within a first time period in chronological order, and obtaining the daily advertising expenditure value; inputting the obtained types and quantities of daily advertising materials running daily and the obtained daily advertising expenditure value into a preset machine learning classification model for training until the model converges; obtaining the relationship between the advertising expenditure value and the types and quantities of advertising materials based on the trained machine learning classification model, and calculating the types and quantities of advertising materials currently needed; comparing this with the types and quantities of unused materials in the current material library, and issuing an alarm message if there are insufficient unused materials in the current material library. This achieves automatic monitoring of the material library inventory and can accurately predict the required quantity of various types of advertisements, achieving the effect of reasonable management of material inventory.

[0005] In actual production scenarios, rework is a frequent occurrence in advertising material production. It is often influenced by factors such as design modifications, review feedback, and material adaptation, exhibiting a concentrated burst pattern on the timeline and forming a chain reaction across multiple production nodes. Existing production management methods typically record or statistically analyze rework as an independent event, failing to systematically analyze the degree of rework aggregation over time and its continuous disruption to subsequent nodes' schedules. This results in the impact of rework gradually accumulating during production, making it difficult to manifest in a timely manner. Furthermore, traditional schedule correction mechanisms often rely on single-task delays or partial postponements, lacking a comprehensive portrayal of parallel processing capacity, node schedule changes, and time window competition, easily leading to passive occupation of production resources and frequent shifts in delivery windows.

[0006] To address the above issues, there is an urgent need for a method and system for scheduling advertising material production based on production status awareness. Summary of the Invention

[0007] Technical problems to be solved

[0008] To address the shortcomings of existing technologies, this invention provides a method and system for scheduling advertising material production based on production status awareness. This solves the problem that existing advertising material production management relies on average single-task scheduling, lacks characterization of rework sequence aggregation and cross-node propagation disturbances, leading to accumulated scheduling deviations and resource imbalance.

[0009] Technical solution

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method and system for scheduling advertising material production based on production condition awareness, comprising: S1, collecting execution time-related data and resource queue data during the advertising material production process, preprocessing the execution time-related data and resource queue data, and constructing a standardized production condition dataset; S2, based on the standardized production condition dataset, analyzing the transmission degree of rework impact in the production process from the relationship between adjacent production node rhythm changes caused by rework, and adjusting the execution arrangement in the advertising material production process based on the transmission analysis results; S3, based on the standardized production condition dataset, evaluating the stability of the production process rhythm from the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution, and imposing linkage constraints on the time window occupancy and task release rhythm in the production execution arrangement based on the offset evaluation results; S4, using the transmission analysis results and offset evaluation results as input, evaluating the adjustability of task time landing points from the time and space margin of tasks in the execution calendar and the node scheduling impedance state, and performing continuous time correction on the task plan start time based on the scheduling evaluation results to reconstruct the execution calendar distribution.

[0011] Furthermore, the specific steps for collecting execution time-related data and resource queue data during the advertising material production process are as follows: Execution time-related data is collected through the node operation log and execution calendar records. This data includes: the queuing time experienced by the material before entering processing, the duration of continuous resource occupation at the node, the planned start time and actual start time of each advertising material task in the original production sequence within the node, and the continuous time interval from the current moment to the next execution structure change in the node's execution calendar is defined as a scheduling cycle. The available time window length of each production node within the scheduling cycle that has not yet been occupied by any task and the single... The actual processing time consumed by each advertising creative task to complete processing at the corresponding node; resource queue data related to parallel carrying and queuing status is collected through monitoring the task queue status and resource usage of production nodes. Resource queue data includes: the number of parallel carrying that each production node is allowed to run simultaneously, the number of executing tasks that are in the execution state at any given time, the number of waiting tasks that have not yet entered the execution state, the required processing time for a task to complete one production process at the production node, and the number of rework occurrences at each node at each time. The required processing time of the task is based on the task's own attributes or the task's required processing time as determined by the task declaration, and the actual processing time is the time consumed by the task to actually complete processing as recorded in the execution log.

[0012] Furthermore, the specific steps for preprocessing execution time-related data and resource queue data to construct a standardized production condition dataset are as follows: After completing the collection of execution time-related data, using the start of the scheduling cycle as a unified time benchmark, the queuing waiting time, occupied time length, planned start time, and actual start time are converted into relative time expressions relative to the start of the scheduling cycle, eliminating the influence of absolute time offset between different nodes and different record sources; the execution time-related data is segmented and pruned, and when the occupied time or available time window crosses the scheduling cycle boundary, only the portion falling within the scheduling cycle is retained; a planned start time sequence and an actual start time sequence are constructed based on the planned start time and the actual start time, and monotonicity checks are performed on the planned start time sequence and the actual start time sequence within the same node, respectively. When it is found that the time record order is inconsistent with the actual execution order, it is rearranged according to the timestamp size, only correcting the order without changing the value itself; After collecting resource queue data, the number of parallel tasks, the number of tasks being executed, the number of tasks waiting, and the number of reworks are filtered by time according to the scheduling cycle range, retaining only valid records occurring within the scheduling cycle. The number of tasks being executed, the number of tasks waiting, and the number of reworks are smoothed using a moving average algorithm along the time sequence to eliminate abnormal fluctuations caused by log jitter and instantaneous statistical errors. Duplicate count records generated by different logs at the same node at the same time are merged and summed. The consistency between the demand processing time and the actual processing time recorded in the execution log is checked; when there is a conflict between the demand processing time and the actual processing time, the actual processing time is used as the value for the task demand processing time in subsequent calculations. After standardizing the execution time-related data and resource queue data, the execution time-related data and resource queue data are normalized using a minimum-maximum value normalization algorithm to construct a standardized production condition dataset.

[0013] Furthermore, the specific steps for analyzing the transmission degree of the impact of rework on the production process based on the standardized production condition dataset and the relationship between the cycle time changes of adjacent production nodes caused by rework are as follows: Add the actual processing time, queuing time, and occupied time of the node at the current time to obtain the node's comprehensive cycle time; add the actual processing time, queuing time, and occupied time of the preceding node at the same time to obtain the preceding node's comprehensive cycle time; subtract the preceding node's comprehensive cycle time from the node's comprehensive cycle time to obtain the cycle time transition between adjacent nodes; calculate the relationship between the node and its preceding node. The absolute value of the difference in actual processing time is added to one to obtain the beat buffer constraint; the number of reworks occurring in the preceding node is used as the numerator, and the number of reworks occurring in the node is added to one as the denominator to obtain the rework propagation trigger; the beat transition amount of adjacent nodes is divided by the beat buffer constraint and then multiplied by the rework propagation trigger to obtain the rework disturbance propagation value; when there is no preceding node in the actual flow path of the advertising material, the current production node is regarded as the starting node of the rework disturbance propagation path, and the corresponding rework disturbance propagation value is determined only by the rework disturbance response result formed by the production node in the current scheduling cycle.

[0014] Furthermore, the specific steps for adjusting the execution schedule of advertising material production based on the transmission analysis results are as follows: The rework disturbance transmission values ​​calculated by each production node within the same scheduling cycle are uniformly summarized, and all production nodes are sorted according to the magnitude of the rework disturbance transmission values ​​to construct a rework disturbance transmission sequence. The top n production nodes in the rework disturbance transmission sequence are marked as time-constrained nodes. The number of parallel loads allowed to be simultaneously in execution state by the time-constrained nodes in the next scheduling cycle is reduced. Specifically, the original number of parallel loads is reduced according to the proportion of the position in the rework disturbance transmission sequence to the total number of nodes and rounded down to obtain the updated number of parallel loads. When the number of parallel bearers is zero, the corresponding production node will not receive new execution tasks in the next scheduling cycle. For the clock constraint node whose number of parallel bearers has been reduced, the advertising material tasks that exceed the updated number of parallel bearers will be removed from the bearer set of the production node. The original production flow order of the removed advertising material tasks will remain unchanged, and they will be assigned to other production nodes whose sorting position in the rework disturbance propagation sequence is after the clock constraint node for execution. For the production node whose sorting position in the rework disturbance propagation sequence is after the clock constraint node, it will receive the advertising material tasks that have been removed from the clock constraint node and are waiting to be executed within the range allowed by the updated number of parallel bearers, without changing the predetermined production process order of the advertising material tasks.

[0015] Furthermore, the specific steps for evaluating the stability of the production process rhythm based on the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution, using a standardized production workflow dataset, are as follows: Obtain the actual start time of the first advertising material task to enter the execution state within the current scheduling cycle and the actual start time of the currently executing advertising material task; subtract the actual start time of the first advertising material task from the actual start time of the currently executing advertising material task to obtain the actual execution time span; subtract the planned start time of the first task from the planned start time of the current task in the original production sequence to obtain the sequential execution time span; when there is only one advertising material task entering the execution state within the scheduling cycle... The sequential execution time span is taken as the required processing time of the task at the corresponding production node after consistency verification. Among them, the task processing time formed by the actual execution time span and the sequential execution time span is the required processing time of the task after consistency verification. The actual execution time span is divided by the sequential execution time span, one is added, and the natural logarithm is taken to obtain the cumulative clock speed stretch. The available time window length of the production node within the scheduling cycle is divided by the required processing time after consistency verification to obtain the number of tasks that can be accommodated. The number of tasks that are simultaneously running at the node at the current time is used as the numerator, and the sum of the number of tasks that are running and the number of tasks that can be accommodated is used as the denominator to obtain the resource squeezing ratio. The resource squeezing ratio is subjected to arctangent operation and then multiplied by the cumulative clock speed stretch to obtain the cumulative clock speed offset.

[0016] Furthermore, the specific steps for linking the time window occupancy and task release rhythm in the production execution arrangement based on the offset evaluation results are as follows: Based on the adjustment of the production node's carrying capacity according to the rework disturbance propagation value, the cumulative value of the beat offset corresponding to each production node is summarized within the same scheduling cycle, and all nodes are sorted by offset according to the magnitude of the cumulative beat offset value; the nodes with the top M cumulative beat offset values ​​in the offset sorting are recorded as time-occupying diffusion nodes. For time-occupying diffusion nodes, a reserved time segment is reserved in the available time window to absorb rework disturbances and handle rhythm rebound; the remaining time is used as an allocable time segment to participate in the execution calendar generation; the scheduling side locks the reserved time segment when generating the execution calendar, allowing only advertising material tasks to enter the allocable time segment for execution, limiting the continuous erosion of subsequent time windows by rework disturbances; regarding the task release method, a fixed release rhythm control is implemented for all waiting tasks of each node, only releasing tasks at a fixed time. Once all ad creative tasks currently in execution at a node are completed, and the next time boundary marked as an allocable time segment in the execution calendar is triggered, a new ad creative task is allowed to enter the execution state. The release time of subsequent tasks strictly depends on the completion time of the previous task and the boundary of the allocable time segment. When the allocable time segment of a node is insufficient to accommodate waiting tasks, waiting tasks that have not yet entered the execution state are migrated across nodes while maintaining their original process order. The migration target is selected from candidate nodes that meet the process constraints, choosing the node with the largest remaining available time window. The time point when a task enters the execution state is fixed as the start time of the next allocable time segment of the target node. For waiting tasks that have not yet entered the execution state, they are re-offset and sorted according to the sum of the cumulative beat offset values ​​formed by the waiting tasks in the completed production nodes. The sorting result only affects the order in which tasks enter the execution state and does not change the production path and node dependencies.

[0017] Furthermore, the specific steps for scheduling and evaluating the adjustability of task timing based on the time and space margin of the task in the execution calendar and the node scheduling impedance status, using the transmission analysis results and offset evaluation results as input, are as follows: Subtract the required processing time for the task to complete one production process from the available time window length of the current production node, to obtain the task time and space margin; where the required processing time is the effective required processing time determined after consistency verification; add the cumulative value of the cycle offset of the production node where the task is located to the rework disturbance transmission value and then add one to obtain the node scheduling impedance; divide the task time and space margin by the node scheduling impedance to obtain the task scheduling displacement value.

[0018] Furthermore, the specific steps for reconstructing the execution calendar distribution by continuously correcting the task plan start time based on the scheduling evaluation results are as follows: For each advertising material task, the planned start time of the task in the original production plan is read and recorded as the original start time. The task scheduling displacement value is subtracted from the original start time to obtain the corrected start time. After calculating the corrected start times for all advertising material tasks, the scheduling side writes the tasks into the execution calendar of the corresponding production node in the order of their corrected start times. When the execution time segments corresponding to different tasks overlap in the execution calendar, the task whose corrected start time was written earlier remains unchanged, and the corrected start time is adjusted accordingly. Tasks written later will have their execution times postponed by an amount equal to the required processing time of the previous task after consistency verification. When a task's execution segment exceeds the available time window of the current production node due to time postponement, the task will be removed from the current node's execution calendar. The next task will be selected from the execution calendars of other production nodes and found to be the first continuous idle time segment after the corrected start time that can fully accommodate the required processing time. The task's completion time is determined by the end time of the final execution segment. The completion time is not modified separately during scheduling. The delivery time is formed by the corrected start time and the required processing time after consistency verification.

[0019] The second aspect of this invention provides an advertising material production scheduling system based on production condition awareness, comprising: a rework timing aggregation awareness module, used to collect execution time-related data and resource queue data during the advertising material production process, preprocess the execution time-related data and resource queue data, and construct a standardized production condition dataset; a rework disturbance transmission analysis module, used to perform transmission analysis on the degree of transmission of rework impact in the production process based on the standardized production condition dataset and the relationship between the rhythm changes of adjacent production nodes caused by rework, and to make linkage adjustments to the execution arrangement in the advertising material production process based on the transmission analysis results; and a cumulative rhythm offset assessment module. The first module, based on a standardized production process dataset, assesses the stability of the production process rhythm by evaluating the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution. It then uses the offset assessment results to impose linkage constraints on the time window occupancy and task release rhythm in the production execution schedule. The second module, using the transmission analysis results and offset assessment results as input, assesses the adjustability of task time landing points by considering the time and space margin of tasks in the execution calendar and the node scheduling impedance status. Based on the scheduling assessment results, it performs continuous time correction on the task plan start time to reconstruct the execution calendar distribution.

[0020] Beneficial effects

[0021] The present invention has the following beneficial effects:

[0022] (1) The advertising material production scheduling method and system based on production condition perception continuously depicts the concentrated occurrence of rework on the time axis, so that rework is transformed from discrete records into working condition signals that can participate in scheduling calculation, thereby exposing the risk of rhythm fluctuation in the production process in advance.

[0023] (2) The advertising material production scheduling method and system based on production condition perception analyzes the transmission path of rework impact among multiple production nodes, so that the implicit disturbance of rework to subsequent nodes is expressed in a structured form, avoiding the disturbance being gradually covered up in the production process.

[0024] (3) The advertising material production scheduling method and system based on production condition perception can identify and participate in scheduling adjustment before the delivery node by cumulatively evaluating the rhythm deviation formed during the continuous task execution.

[0025] (4) This advertising material production scheduling method and system based on production condition perception reduces the probability of passive occupation of production resources and improves the overall stability and delivery controllability of the advertising material production process by guiding the natural migration and redistribution of tasks between production nodes.

[0026] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] Figure 1 This is a flowchart of an advertising material production scheduling method based on production status awareness according to the present invention.

[0028] Figure 2 This is a structural diagram of an advertising material production scheduling system based on production status perception according to the present invention;

[0029] Figure 3 This is a schematic diagram of the multi-layer mapping situation of rework disturbance involved in this invention;

[0030] Figure 4 This is a bar chart showing the task scheduling displacement values ​​involved in this invention. Detailed Implementation

[0031] 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.

[0032] Please see Figures 1-4This invention provides a technical solution: a method and system for scheduling advertising material production based on production condition awareness, comprising: S1, collecting execution time-related data and resource queue data during the advertising material production process, preprocessing the execution time-related data and resource queue data, and constructing a standardized production condition dataset; S2, based on the standardized production condition dataset, performing a transmission analysis on the transmission degree of rework impact in the production process from the relationship between adjacent production node rhythm changes caused by rework, and adjusting the execution arrangement in the advertising material production process based on the transmission analysis results; S3, based on the standardized production condition dataset, performing a deviation assessment on the stability of the production process rhythm from the deviation relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution, and imposing a linkage constraint on the time window occupancy and task release rhythm in the production execution arrangement based on the deviation assessment results; S4, using the transmission analysis results and deviation assessment results as input, performing a scheduling assessment on the adjustability of the task time landing point from the time space margin of the task in the execution calendar and the node scheduling impedance state, and performing continuous time correction on the task plan start time based on the scheduling assessment results to reconstruct the execution calendar distribution.

[0033] Specifically, the steps for collecting execution time-related data and resource queue data during the advertising material production process are as follows: Execution time-related data is collected through the production node's operation log and execution calendar records. This data includes the queuing time experienced by advertising material tasks before entering the corresponding production node for processing, the duration of continuous resource occupation at the production node, and the planned start time and actual start time of each advertising material task in its original production sequence. To ensure consistency in time reference during scheduling analysis, the moment when the scheduling side triggers scheduling calculations is taken as the unified current moment. This current moment is determined by the scheduling side based on the execution calendar generation or update operation and serves as the unified starting point for subsequent time interval divisions. Based on this, a scheduling cycle is defined as the continuous time interval starting from this current moment and ending at the time point in the execution calendar corresponding to the production node where the next change in execution structure occurs.

[0034] The execution structure change is used to characterize discrete switching events in the execution state of the production node. Specifically, it includes any of the following situations: an execution task in the production node has been completed and released, execution resources are released, a waiting task enters the execution state causing a change in the number of parallel executions, the boundary between the original allocable time segment and the reserved time segment in the execution calendar changes, or the scheduler updates the number of parallel loads on the node and writes it into the execution calendar. When any of these events occurs, it is considered that the execution structure of the production node has changed, and the corresponding time point is taken as the termination boundary of the current scheduling cycle.

[0035] Once the scheduling period is determined, the available time window length of each production node that has not yet been occupied by tasks within that scheduling period is collected, as well as the actual processing time consumed by a single advertising creative task to complete one production process on the corresponding production node. The actual processing time is derived from the task completion time and start time records in the production node's operation log, and is used to objectively reflect the actual execution consumption of the task on the node.

[0036] Simultaneously, resource queue data related to parallel processing and queuing status is collected by monitoring the task queue status and resource usage of production nodes. This resource queue data includes the number of parallel processing tasks that each production node is allowed to run simultaneously, the number of tasks in the execution state at any given time, the number of waiting tasks that have not yet entered the execution state, and the number of rework occurrences at each node at a given time. Furthermore, the task's required processing time is obtained based on the processing requirements defined in the advertising material task's own attributes or task declaration, and is used to describe the processing time required by the task under ideal execution conditions. The actual processing time of the task is obtained as objective time data from the execution log, reflecting the time consumed by the task to complete processing in the actual production environment.

[0037] This implementation plan establishes a unique, definitive, and repeatable time reference system for all subsequent time-based analysis and scheduling calculations, thereby avoiding inconsistencies in scheduling results due to inconsistent time bases. By using the moment when the scheduling calculation is triggered on the scheduling side as the unified current moment, and using the identifiable execution structure change events in the execution calendar as the boundary of the scheduling cycle, the scheduling cycle no longer depends on the local state of nodes or fuzzy time divisions, but is strictly bound to objective events in the execution calendar and runtime logs. This ensures that data collected within the same scheduling cycle, such as queuing, resource usage, parallel processing, and rework occurrences, have consistent temporal semantics, providing a stable and alignable data foundation for rework disturbance propagation analysis, cycle time offset cumulative assessment, and execution calendar reconstruction.

[0038] Specifically, the preprocessing of execution time-related data and resource queue data to construct a standardized production workflow dataset involves the following steps: After collecting execution time-related data, using the start of the scheduling cycle as a unified time benchmark, the queuing time, the duration of continuous resource occupation, the planned start time of each advertising material task, and the actual start time of entering the execution state are uniformly converted into a relative time expression form relative to the start of the corresponding scheduling cycle. This eliminates the time offset caused by differences in absolute timestamps between different production nodes and different log record sources. Based on this, the execution time-related data is segmented. When the occupied time or available time window crosses the scheduling cycle boundary, only the time segment falling within the current scheduling cycle is retained, ensuring that the data used in subsequent analysis strictly corresponds to the same scheduling cycle range.

[0039] Based on the planned start time and actual start time within each production node, a planned start time sequence and an actual start time sequence are constructed respectively, and monotonicity checks are performed on both types of time series within the node dimension. When a discrepancy is found between the time recording order and the actual execution order, the sequence is rearranged only according to the timestamp size to correct the recording order error, without modifying the time values ​​themselves, thereby ensuring that the time data has the correct temporal sequence relationship while maintaining the original measurement results.

[0040] After completing the resource queue data collection, the number of parallel tasks, the number of tasks being executed, the number of tasks waiting, and the number of rework attempts are also filtered by time according to the scheduling cycle range. Only valid records occurring within the current scheduling cycle are retained to avoid inconsistencies in statistical standards caused by cross-cycle data mixing. Subsequently, the number of tasks being executed, the number of tasks waiting, and the number of rework attempts are smoothed using a moving average algorithm in chronological order to reduce abnormal fluctuations caused by log sampling jitter, instantaneous statistical errors, and short-term state switching. For duplicate counts caused by multiple log records of the same production node at the same time, a merging and summing process is performed to ensure that the resource queue status of the node at any given time has a unique and definite count result.

[0041] At the same time, a consistency check is performed between the required processing time of the task and the actual processing time recorded in the execution log. When there is a discrepancy between the required processing time and the actual processing time, the actual processing time recorded in the execution log is used as the value of the required processing time of the task in subsequent calculations, thereby ensuring that the processing time data used in subsequent analysis all come from observable actual execution results.

[0042] After standardizing the execution time-related data and resource queue data, the execution time-related data and resource queue data are uniformly normalized using the minimum-maximum value normalization algorithm, with a single scheduling cycle as the normalization range and the data value range of each production node within the scheduling cycle as the normalization object. This ensures that the working condition data obtained by different production nodes within the same scheduling cycle are on a comparable numerical scale, thereby constructing a standardized production working condition dataset. This provides a stable and consistent data foundation for subsequent rework disturbance propagation analysis, cycle time offset accumulation calculation, and scheduling strategy updates.

[0043] This implementation scheme unifies the starting point of the scheduling cycle and limits the normalization range, ensuring that runtime data from different production nodes and log sources remain consistent in temporal semantics and numerical scale. This avoids data incomparability issues caused by absolute time offsets, cross-cycle mixing, and differences in node scales. Simultaneously, through time rearrangement, segment pruning, smoothing, and consistency checks, the interference of record jitter, sequence errors, and duplicate counts on subsequent analysis is eliminated. This ensures that rework disturbance propagation analysis, cycle offset accumulation calculation, and execution calendar reconstruction are all based on stable and reliable data, guaranteeing the consistency and reproducibility of scheduling decisions across different scheduling cycles.

[0044] Specifically, based on a standardized production condition dataset, the transmission analysis of the impact of rework on the production process is conducted by examining the relationship between the cycle time changes of adjacent production nodes caused by rework. The specific steps are as follows:

[0045] The node's overall beat rate is obtained by adding the actual processing time, queuing time, and occupied time at the current time. The reason for choosing to add the actual processing time, queuing time, and occupied time is that these three factors respectively reflect the direct impact of processing consumption, queue delays, and resource consumption on the execution rhythm. The sum of these factors forms a quantitative expression that can fully describe the overall beat rate of the node at the current time.

[0046] The total cycle time of the preceding node is obtained by adding the actual processing time, queuing time, and occupied time of the preceding node at the same time. The time dimension and calculation method are the same as those of the current node to ensure that the cycle time of the preceding and following nodes are comparable, so that the subsequent difference calculation only reflects the state difference between nodes and does not introduce the deviation of the calculation method.

[0047] Subtracting the total beat count of the preceding node from the total beat count of the node yields the beat transition of the adjacent node. The difference is used to characterize the change in beat state from the preceding node to the current node in the actual flow path, so that the beat change is expressed as a beat transition with a clear direction.

[0048] The absolute value of the difference between the actual processing time of the calculated node and the preceding node is added to obtain the beat buffer constraint. The absolute value of the difference in actual processing time is used to eliminate the directional influence, and only the difference in processing rhythm itself is retained. Adding one is used to avoid zero denominator in subsequent calculations, so as to build the basic buffer constraint for beat variation without introducing artificial parameters.

[0049] The rework propagation trigger value is obtained by using the number of reworks occurring in the preceding node as the numerator and the number of reworks occurring in the current node plus one as the denominator. The number of reworks occurring in the preceding node is used as the numerator to reflect the intensity of the rework source, and the number of reworks occurring in the current node plus one is used as the denominator to characterize the relative diffusion relationship of rework in the propagation process. At the same time, the addition of one avoids calculation failure when the number of reworks is zero.

[0050] The rework disturbance propagation value is obtained by dividing the adjacent node beat transition amount by the beat buffer constraint amount and then multiplying it by the rework propagation trigger amount. The beat transition amount is constrained within the buffer range allowed by the processing rhythm difference by division, and the rework propagation trigger relationship is introduced by multiplication, so that the beat change and rework propagation are coupled in the same quantification result.

[0051] When there is no preceding node in the actual flow path of the advertising material, the current production node is regarded as the starting node of the rework disturbance transmission path. The corresponding rework disturbance transmission value is determined only by the rework disturbance response result formed by the production node in the current scheduling cycle. By excluding the related calculations of the preceding node, false beat comparisons are avoided under the condition of no effective reference, so that the rework disturbance transmission value of the starting node only reflects the direct impact of its own rework behavior on the beat state.

[0052] The formula for calculating the rework disturbance transmission value is:

[0053] ;

[0054] In the formula, This represents the actual processing time consumed by node i to complete the processing of a unit of advertising material within time t. It is used to reflect the processing cycle level of the node under the current working conditions and is an important basic data for measuring changes in node execution efficiency. This represents the queuing time that node i experiences before the material enters processing within time t. It is used to characterize the change in the waiting state of the node due to upstream task backlog or mismatch in cycle time, and is an important basis for judging whether rework disturbances cause scheduling congestion. This represents the length of time that node i is continuously occupied within time t. It reflects the degree to which node resources are continuously occupied and is a direct indicator of resource crowding and release restrictions caused by rework. This represents the actual processing time consumed by the preceding production node of node i to complete the processing of a unit of advertising material within time t. It is used to depict the real processing rhythm of the rework source node in the corresponding time period and is an important reference for judging whether rework is accompanied by changes in processing efficiency and transmitted downstream. This represents the queuing time experienced by the preceding production node of node i before the material enters the processing within time t. It is used to reflect the change in the waiting state of the rework node within the corresponding time period due to task backlog or rhythm imbalance. It is a key basis for identifying whether rework causes process blockage and spreads to subsequent nodes. This indicates the length of time that the preceding production node of node i is in a continuously occupied state within time t. It is used to quantify the degree to which the resources of the rework node are continuously occupied during this time period. It is an important basic data for judging whether rework leads to limited resource release and creates transmission pressure. It represents the number of reworks that occur at node i within time t. It is used to characterize the internal digestion state of the rework disturbance at the current node. Its value changes reflect the node's own ability to absorb the disturbance and is an important constraint to limit the continued propagation of the rework impact. This represents the number of rework events that occurred in the preceding node of node i within time t. It reflects the concentration of rework in upstream nodes and is the direct source of events that triggers the propagation of rework disturbances to subsequent nodes.

[0055] In this implementation plan, the implicit impact of rework on adjacent production nodes in the advertising material production process is transformed into a calculable and comparable rework disturbance transmission value. This is achieved by quantifying the comprehensive changes in processing time, queuing time, and continuous occupation time of adjacent nodes, and by combining the distribution of rework in preceding and current nodes, characterizing whether and the intensity of the rework disturbance continuously spreads from upstream to downstream nodes. Using this rework disturbance transmission value, the effects of rework-induced rhythm imbalance, resource congestion, and execution sequence shifts can be expressed in a unified numerical form without introducing manually set boundaries or subjective judgment. This ensures that rework is no longer viewed merely as a single-node anomaly, but rather as a continuous disturbance propagating across multiple nodes. This provides a direct basis for subsequent adjustments to the parallel carrying capacity of production nodes, reordering of task initiation, and reorganization of delivery sequences, enabling the impact of rework to be promptly made explicit and continuously constrained during the production process.

[0056] Specifically, the steps for adjusting the execution arrangements in the advertising material production process based on the results of the transmission analysis are as follows:

[0057] The rework disturbance propagation values ​​calculated by each production node within the same scheduling cycle are aggregated and sorted according to their values ​​to construct a rework disturbance propagation sequence. This sequence reflects the degree to which different production nodes bear and spill over rework disturbances within the current scheduling cycle; nodes ranked earlier in the sequence have a more significant effect on the aggregation and propagation of rework disturbances. Based on this sequence, the first N production nodes are selected as cycle time constraint nodes, where N is the number of nodes requiring simultaneous cycle time constraints within the current scheduling cycle. The value of N is determined by the total number of nodes in the rework disturbance propagation sequence and the adjustable parallel carrying capacity of the scheduling side, thus limiting the scope of parallel carrying capacity compression.

[0058] For nodes with clock speed constraints, the number of parallel bearers allowed to be in execution simultaneously is reduced in the next scheduling cycle. The reduction in the number of parallel bearers is monotonically correlated with the node's ranking position in the rework disturbance propagation sequence; nodes with higher ranking positions experience larger reductions, while those with lower ranking positions experience smaller reductions. Specifically, the original number of parallel bearers for a node is mapped to the proportion of remaining bearers based on its ranking position in the rework disturbance propagation sequence, and the calculation result is rounded down to obtain the updated number of parallel bearers. This allows nodes with higher rework disturbance concentration to bear stricter execution clock speed constraints in subsequent scheduling cycles. When the updated number of parallel bearers is zero, the corresponding production node will no longer receive new execution tasks in the next scheduling cycle and will only be used to complete tasks already in the execution state.

[0059] For clock constraint nodes where the number of parallel loads is reduced, advertising material tasks that exceed the updated number of parallel loads are removed from the node's current load set. The removed tasks maintain their original production flow order to avoid introducing process sequence disturbances during scheduling adjustments. These tasks are then assigned to other production nodes whose sorting position in the rework disturbance propagation sequence is after the clock constraint node for execution.

[0060] For production nodes whose sorting position in the rework disturbance propagation sequence is after the beat constraint node, the pending tasks moved out from the beat constraint node are received within the allowable range of the updated parallel carrying capacity. The execution of the tasks on the new node still follows the established production process path and does not change the process sequence of the advertising materials in the production process, thereby realizing the orderly distribution and absorption of the rework disturbance at the node level.

[0061] like Figure 3The diagram illustrates a multi-layered mapping of rework disturbances provided in this application. This multi-layered mapping systematically expresses the overall situation of rework disturbances during the production of advertising materials, from their generation and transmission to their eventual absorption and scheduling, revealing the structural relationships and evolutionary logic within the production process. The diagram, from top to bottom, shows a rework disturbance potential layer, a disturbance transmission relationship layer, and a scheduling absorption and redistribution layer. These three layers are spatially parallel, and the same set of production nodes is repeated in each layer, allowing the state of the same node to be presented side-by-side from different analytical perspectives. In the rework disturbance potential layer, the production nodes are connected by multiple intersecting lines. The large number of lines and their wide coverage are used to depict the overall disturbance potential formed when rework occurs concentratedly within a short period. This layer emphasizes that rework is not an isolated event but may be active simultaneously among multiple production nodes, forming a fluctuating state with a large impact, thus providing a background for subsequent transmission analysis. The disturbance transmission relationship layer in the middle retains the same production nodes, but the connection structure between the nodes shows a significant selective difference compared to the upper layers. This layer focuses on the transmission relationships formed when rework disturbances propagate along the actual production flow path to subsequent nodes. The number and connection methods between different nodes vary, reflecting the path differentiation and structural differences that arise during the propagation of rework impacts between nodes. Through this layer, it can be visually observed that rework disturbances do not spread uniformly but are continuously propagated along specific node relationships. The bottom scheduling absorption and redistribution layer also includes all production nodes, but the connections between nodes are relatively sparse, and the overall structure is more convergent. This layer is used to express the state where rework disturbances are redistributed, dispersed, and gradually absorbed after the intervention of scheduling strategies, reflecting the constraint and mitigation effect of scheduling behavior on the scope of disturbance impact, and reflecting the stable outcome of disturbances under scheduling from an overall situational perspective. Vertical dashed lines connect the three layers with corresponding production nodes, indicating that nodes with the same position in different layers represent the state mapping of the same actual production node under different operational analysis dimensions. This vertical correspondence ensures the spatial continuity of the evolution process of rework disturbances from generation and propagation to absorption. By using a multi-layered parallel structure and node mapping relationship, the active state, transmission structure, and scheduling absorption effect of rework disturbances are integrated into the overall situational framework, which intuitively reveals the internal correlation and hierarchy of the production process, providing clear mechanistic support for understanding the production scheduling method based on the condition awareness.

[0062] In this implementation plan, by uniformly sorting the rework disturbance propagation degree of production nodes and applying differentiated parallel carrying constraints, production nodes with high rework disturbance concentration can actively release execution pressure in subsequent scheduling cycles. At the same time, tasks to be executed are guided to transfer along the rework disturbance propagation sequence to production nodes with stronger disturbance absorption capabilities. Thus, without changing the predetermined production process sequence of advertising materials, the execution load is redistributed in an orderly manner among nodes, suppressing the continuous accumulation of rework disturbance in local nodes and its spread along the time axis, stabilizing the overall execution rhythm, and reducing the risk of frequent distortion of scheduling results due to rework fluctuations.

[0063] Specifically, based on a standardized production workflow dataset, the stability of the production process rhythm is evaluated by the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution. The specific steps are as follows: obtain the actual start time of the first advertising material task to enter the execution state within the current scheduling cycle and the actual start time of the advertising material task currently entering the execution state. Subtract the actual start time of the first advertising material task from the actual start time of the current advertising material task to obtain the actual execution time span. By selecting the first task to enter the execution state as the time anchor point, the actual start behavior of all subsequent tasks is measured with the same execution starting point, thereby uniformly mapping the time unfolding caused by parallel and interleaved execution within the scheduling cycle into a single time span expression.

[0064] The sequential execution time span is obtained by subtracting the planned start time of the first task from the planned start time of the current task in the original production sequence. The sequential execution time span is constructed using the planned start time under the original production sequence to characterize the ideal execution rhythm baseline that the task should form under the condition of no rework disturbance and parallel adjustment, so that subsequent comparisons only reflect the offset impact caused by scheduling and rework.

[0065] When there is only one ad creative task in the execution state within the scheduling period, the sequential execution time span is taken as the requirement processing time of the task at the corresponding production node after consistency verification; in the absence of a multi-task sequence reference, the requirement processing time of a single task is used as the sequential execution time span to avoid the sequential baseline degenerating to zero or meaningless time difference due to insufficient samples.

[0066] The task processing time, which is formed by the actual execution time span and the sequential execution time span, adopts the task requirement processing time after consistency verification. By uniformly using the requirement processing time after consistency verification, the consistency between the actual execution and the sequential baseline in terms of time scale is ensured, and the implicit proportional deviation introduced by mixing processing times from different sources is avoided.

[0067] The cumulative beat stretch is obtained by dividing the actual execution time span by the sequential execution time span, adding one, and taking the natural logarithm. The ratio characterizes the stretch of the actual execution beat relative to the sequential baseline. Adding one is used to avoid the problem of the logarithm being undefinable when the ratio is less than or equal to zero. The natural logarithm is used to map the linearly growing time stretch relationship into a continuous quantity that is more sensitive to incremental changes and slows down the growth of extreme values, thus reflecting the cumulative trend of beat stretch rather than instantaneous jumps.

[0068] The number of tasks that can be accommodated is obtained by dividing the available time window length of the production node within the scheduling cycle by the required processing time after consistency verification. The division relationship between the available time window and the processing consumption of a single task directly describes the task capacity that the node can carry in the time dimension within the current scheduling cycle, so that resource evaluation is based on objective time constraints.

[0069] The resource squeeze ratio is obtained by taking the number of tasks currently running on a node as the numerator and the sum of the number of tasks being run and the number of tasks that can be accommodated as the denominator. By placing both the number of tasks being run and the number of tasks that can be accommodated in the denominator, the ratio always falls between zero and one. This ratio is used to express the degree of occupancy of the execution load relative to the node's time carrying capacity, and to avoid the ratio from becoming unbounded when the number of tasks being run approaches or exceeds the number of tasks that can be accommodated.

[0070] The resource crowding ratio is subjected to arctangent calculation, and then multiplied by the cumulative beat stretch to obtain the cumulative beat offset. The arctangent function is used to map the resource crowding ratio to a continuous response curve that gradually slows down as the load increases, simulating the objective characteristic that the impact of resource congestion on beat offset gradually saturates in the high occupancy range. By multiplying it by the cumulative beat stretch, the beat stretch in the time dimension and the congestion state in the resource dimension are coupled and expressed, thereby avoiding the simple weighted superposition that masks the synergistic amplification effect of the two factors on the beat offset at different stages.

[0071] The formula for calculating the cumulative value of beat offset is:

[0072] ;

[0073] In the formula, This represents the actual start time of the first advertising creative task to enter the execution state at node i, and is used to characterize the starting time reference of the actual execution sequence of this node in this period; This indicates the actual start time of the advertising material task that node i is currently in the execution state within the current scheduling period. It is used to depict the end point of the execution sequence of this node in the actual running state on the time axis. This represents the planned start time of the first corresponding task in the original production sequence of node i, and is used to reflect the starting time reference of the ideal execution sequence of this node under the condition of no rework disturbance; This indicates the planned start time of node i in the original production sequence corresponding to the current task, and is used to reflect the end point of the execution sequence of this node on the time axis under ideal conditions; It represents the length of the available time window that node i has not yet been occupied by tasks within the scheduling period corresponding to time t. It is used to quantify the remaining execution space of the node in the time dimension and is direct runtime data that characterizes the degree of time window crowding. This represents the number of tasks that node i is simultaneously executing at time t, reflecting the actual processing capacity that node can be used in parallel under the current operating conditions. This represents the processing time required for task i to complete one production process at the current production node. It is used to quantify the objective time resource requirements of the task and is the basic operational data for determining whether the task can be carried out by the current node.

[0074] In this implementation scheme, the impact of rework disturbances on production nodes during continuous task execution is transformed into a quantifiable cumulative value of beat offset. By performing a nonlinear mapping on the difference between the actual execution sequence and the original sequential execution sequence on the time axis, and combining the node waiting task status and available time window constraints, the coupling results of processing sequence shift, parallel resource constraints, and time window crowding in the execution process are characterized. This allows potential plan deviation risks in the production process to be made explicit in advance before the task is completed, providing a direct basis for subsequent execution control based on time window segmentation, task release rhythm control, and cross-node migration arrangements.

[0075] Specifically, the steps for linking time window occupancy and task release rhythm in production execution scheduling based on offset assessment results are as follows: After adjusting the capacity of production nodes based on rework disturbance propagation values, the cumulative values ​​of beat offsets for all production nodes are centrally summarized within the same scheduling cycle. Using the cumulative beat offset value as a unified unit, the execution offset status of production nodes in the time dimension is sorted to form an offset ranking result reflecting the distribution of time encroachment risks for each node. This ranking is used to characterize the comprehensive degree of continuous time window encroachment and rhythm rebound lag of different production nodes within the current scheduling cycle, providing a direct basis for subsequent time window management.

[0076] Based on the offset sorting results, production nodes with the highest cumulative beat offset values ​​(ranking in the top M positions) are identified as time-crowding diffusion nodes. M represents the number of nodes requiring time window constraints within the current scheduling cycle, and its value is derived from the correspondence between the offset sorting scale and the adjustable time window range on the scheduling side. For time-crowding diffusion nodes, a reserved time segment is specifically designated within the available time window of the node's scheduling cycle to absorb rework disturbances and execution rhythm rebounds. The remaining time windows are designated as allocable time segments for participation in execution calendar generation. During the execution calendar construction process, the scheduling side locks the reserved time segments, prohibiting any advertising material tasks from occupying them. Only tasks are allowed to enter the allocable time segments for execution, thus creating a buffer and isolation against rework disturbances in the time dimension and preventing time crowding from continuously spreading backward.

[0077] Regarding task release methods, a fixed delivery rhythm is adopted to control all waiting tasks at each production node. Tasks no longer enter the execution state continuously, but are strictly dependent on the completion events of existing tasks within the node and the time boundaries of the allocable time segments in the execution calendar. Only when all ad creative tasks currently in the execution state of a node are completed, and the next time boundary marked as an allocable time segment in the execution calendar is reached, are new ad creative tasks allowed to enter the execution state. This creates a discrete and controllable delivery structure for task release rhythm on the timeline.

[0078] When a production node's available time slot within the current scheduling period is insufficient to accommodate its waiting tasks, cross-node migration is performed on the waiting tasks that have not yet entered the execution state. During the migration, the original production process sequence of the advertising materials remains unchanged. The migration target is selected from candidate production nodes that meet the process constraints, prioritizing the production node with the largest remaining available time window length within the current scheduling period. The time when the task enters the execution state is fixed at the start time of the next available time slot of the target production node, thereby avoiding new time consumption caused by temporary insertion of execution.

[0079] For advertising material tasks that are still in the waiting state and have not yet entered the execution state, the offset is re-sorted based on the sum of the cumulative beat offset values ​​formed in the completed production nodes. The sorting result is only used to determine the order in which the tasks enter the execution state and does not affect the established production path and node dependency relationship of the advertising material. Thus, without breaking the process constraints, the risk of time squeezing is orderly distributed at the task level.

[0080] In this implementation scheme, by simultaneously introducing time-dimensional constraints and rearrangement mechanisms at the node and task levels, the time-dimensional offset caused by rework disturbances is transformed from a continuously accumulating state into a time structure that can be absorbed in segments and dispersed in a targeted manner. On the one hand, by identifying time-crowding nodes and reserving unusable time segments within their scheduling cycles, rework disturbances and execution rhythm rebounds have a clear time buffer space, preventing execution delays from continuously accumulating on the timeline. On the other hand, by fixing the task release rhythm, limiting the time boundaries for task entry into execution, and performing cross-node migration when necessary, the release behavior of waiting tasks is kept consistent with the actual available time window of the nodes, thereby avoiding new time crowding caused by excessive parallelism or continuous insertion. At the same time, by reordering waiting tasks according to the cumulative time-dimensional offset results, the order in which tasks enter execution matches their impact on the overall time-dimensional stability, achieving self-correction and stable convergence of scheduling results in the time dimension without changing the production path and process dependencies.

[0081] Specifically, using the results of conduction analysis and offset evaluation as input, the scheduling evaluation of the adjustability of task timing based on the task's time and space margin in the execution calendar and the node scheduling impedance status is carried out through the following steps:

[0082] The task time space margin is obtained by subtracting the required processing time for a task to complete one production process from the available time window length available for the current production node. This subtracts the required processing time from the available time window length, directly depicting the remaining time space that can be released after the task is completed without changing the scheduling cycle boundary. This allows the task's placeability in the time dimension to be expressed in absolute time form, avoiding the interpretation ambiguity caused by relative proportions. The required processing time is taken from the effective required processing time after consistency verification, to ensure that the time consumption reflects the actual execution requirements.

[0083] The node scheduling impedance is obtained by adding the cumulative value of the clock offset of the production node where the task is located to the rework disturbance propagation value and then adding one. The cumulative value of the clock offset and the rework disturbance propagation value are added to form a superimposed expression of the continuous offset effect of the node in the time dimension and the propagation effect of rework in the path dimension, thereby reflecting the comprehensive degree of obstruction of the node to the scheduling of new tasks. The addition of one is used to avoid zero denominators in subsequent calculations and to maintain the continuity of the scheduling impedance without introducing external parameters.

[0084] Dividing the task's time and space margin by the node scheduling impedance yields the task scheduling displacement value. The division relationship is used to offset the task's available space in the time dimension with the degree of node scheduling obstruction, so that the task can obtain a larger scheduling displacement result when it has more time and space and lower node impedance. This provides a directly comparable quantitative basis for subsequent adjustments to the task deployment order and execution position.

[0085] The formula for calculating task scheduling displacement is:

[0086] ;

[0087] In the formula, It represents the length of the available time window that task j can occupy at the current production node at time t. It is used to quantify the available space for the task in the time dimension and is a direct indicator reflecting the supply status of node time resources. This represents the processing time required for task j to complete one production process at the current production node. It is used to quantify the objective time resource requirements of the task and is the basic operational data for determining whether the task can be carried out by the current node. It represents the cumulative value of the beat offset of the production node where task j is located at time t. It is used to characterize the stretching state of the execution rhythm formed by the solidification of rework disturbance during the continuous execution of tasks. It is an important quantitative result reflecting the stability of the node rhythm. This represents the rework disturbance propagation value at time t for the production node where task j is located. It is used to characterize the continuous propagation state of the rework disturbance at that node and is a direct indicator reflecting whether the impact of rework is still spreading at the node level.

[0088] In this implementation example, the available time window length of task unit A is set to 18.0, the demand processing time is set to 6.0, the cumulative value of cycle offset is set to 0.8, and the rework disturbance propagation value is set to 0.6.

[0089] The available time window length for task unit B is set to 15.0, the requirement processing time is set to 7.0, the cumulative cycle offset value is set to 1.2, and the rework disturbance propagation value is set to 0.9.

[0090] The available time window length for task unit C is set to 20.0, the required processing time is set to 8.0, the cumulative cycle offset value is set to 0.5, and the rework disturbance propagation value is set to 0.4.

[0091] The available time window length for task unit D is set to 12.0, the requirement processing time is set to 6.5, the cumulative cycle offset value is set to 1.6, and the rework disturbance propagation value is set to 1.1.

[0092] The available time window length for task unit E is set to 22.0, the requirement processing time is set to 9.0, the cumulative cycle offset is set to 0.3, and the rework disturbance propagation value is set to 0.2. The task scheduling displacement values ​​for each task unit are calculated, as shown in Table 1.

[0093] Table 1 Task Scheduling Displacement Value Data Table Task Unit Number Available time window length Request processing time Beat offset cumulative value Rework disturbance transmission value Task scheduling displacement value Task Unit A 18.0 6.0 0.8 0.6 5.00 Task Unit B 15.0 7.0 1.2 0.9 2.58 Task Unit C 20.0 8.0 0.5 0.4 6.32 Task Unit D 12.0 6.5 1.6 1.1 1.49 Task Unit E 22.0 9.0 0.3 0.2 8.67

[0094] like Figure 4 The table shown is a bar chart of task scheduling displacement values ​​provided in this application example. (See Table 1 and...) Figure 4 It can be seen that task unit E has the highest task scheduling displacement value, indicating that under the combined effect of a longer available time window, a lower cumulative clock offset value, and a lower rework disturbance propagation value, task unit E has more time placement space and a lower degree of scheduling obstruction. Task unit D has the lowest task scheduling displacement value, reflecting that under the conditions of a shorter available time window and higher cumulative clock offset values ​​and rework disturbance propagation values, the scheduling flexibility of the task in the time dimension is significantly compressed. The task scheduling displacement value bar chart uses task units as the horizontal axis and task scheduling displacement value as the vertical axis, and uses orange bars to intuitively present the different distribution of each task unit in the scheduling space. The difference in bar height clearly reflects the placement capability gradient formed between task units under the superimposed influence of time resources and scheduling impedance. Tall bars correspond to tasks with larger scheduling displacement space, while low bars correspond to tasks that require stricter placement constraints in the execution calendar. Overall, the task scheduling displacement value bar chart transforms the abstract scheduling displacement calculation results into directly comparable visual results, which helps to quickly identify task units that are more suitable for priority release of execution resources and task units that need to be postponed during the scheduling decision-making stage.

[0095] Specifically, the steps for reconstructing the execution calendar distribution by continuously adjusting the task start time based on the scheduling evaluation results are as follows: For each advertising material task, the planned start time in the original production plan is read and recorded as the original start time. The task scheduling displacement value is used as the time adjustment amount, and the original start time is subtracted from the task scheduling displacement value to obtain the adjusted start time. This time rollback method causes the task's start position within the scheduling cycle to shift according to the available time space and node scheduling impedance state. This allows tasks with larger scheduling displacement space to enter the execution calendar earlier, while avoiding path inconsistencies caused by directly changing the production process sequence.

[0096] After calculating the start time for all advertising material tasks, the scheduling side writes the tasks sequentially into the execution calendar of the corresponding production node based on the order of their start times. This creates a time-based arrangement structure in the execution calendar, with the start time being the primary factor. When different tasks overlap in their corresponding execution time segments, tasks with earlier start times are left unchanged, while tasks with later start times have their execution time delayed. The delay is calculated as the processing time of the preceding task after consistency verification, ensuring that only a single task exists within any given time segment and preventing concurrent overlaps from disrupting the execution rhythm.

[0097] When a task's execution segment exceeds the available time window of the current production node due to delayed processing, the task is removed from the current node's execution calendar and reassigned to another production node's execution calendar. During reassignment, a search is performed backward from the correction start time to select the first continuous idle time segment that can fully accommodate the task's processing time requirements. This segment becomes the task's new execution segment, ensuring the task has executable time space without altering process constraints.

[0098] The task completion time is naturally determined by the final end time of the execution segment. The completion time is not set or modified separately during the scheduling process. The delivery time is directly determined by the modified start time and the requirement processing time after consistency verification, thereby ensuring that the time projection results in the execution calendar are consistent with the actual execution consumption.

[0099] In this implementation scheme, the task scheduling displacement results calculated in the preceding steps are directly mapped to the time adjustment behavior in the execution calendar, enabling the actual placement of tasks on the timeline to adaptively change according to the running status of production nodes and rework disturbances. By correcting the start time in the original production plan with task scheduling displacement values, tasks with more time resources and lower scheduling impedance are given priority to occupy the execution window, thereby reconstructing the execution order without disrupting the manufacturing process path. Simultaneously, through delayed processing and cross-node relocation mechanisms, conflicts caused by overlapping execution segments and time window overflows are eliminated, ensuring that the execution calendar always maintains non-overlapping time segments and allows for complete execution. Finally, the task completion time is determined by the natural end time of the execution segment, avoiding human intervention in completion nodes and ensuring consistency between the predicted delivery time and actual processing consumption, thereby improving the executability and time stability of the overall scheduling results.

[0100] The second aspect of this invention provides an advertising material production scheduling system based on production condition awareness, comprising: a rework time sequence aggregation awareness module, used to synchronously collect execution time-related data and resource queue data during the advertising material production process, and to perform time alignment, segment clipping, sequence verification and smoothing processing on the two types of data around a unified scheduling cycle, transforming the scattered and inconsistent granularity of the operation records into production condition expressions with consistent time semantics and comparable numerical scales, thereby forming a standardized production condition dataset that can reflect changes in rework density, fluctuations in execution rhythm and resource occupancy status, providing a stable data foundation for subsequent analysis.

[0101] The rework disturbance propagation analysis module is used to compare and analyze the rhythm change relationship between adjacent production nodes under the same time reference along the actual flow path of advertising materials based on a standardized production process dataset. It transforms the processing time stretching, waiting time superposition, and resource occupation extension caused by rework into quantifiable propagation results, thereby characterizing the transmission direction and diffusion intensity of rework impact between production nodes. Based on the propagation results, it implements linkage adjustment on the execution carrying relationship and task distribution status between different nodes to avoid the continuous accumulation of rework disturbances in local nodes.

[0102] The cumulative beat offset assessment module is used to compare and analyze the actual time unfolding formed during continuous task execution with the sequential time unfolding formed under the original production sequence based on a standardized production condition dataset. By characterizing the degree of offset accumulation of the two time unfoldings within the scheduling cycle, it reflects the stable state of the production process in the time dimension. Based on the offset assessment results, it implements synchronous constraints on the time window division method and task release rhythm in the production execution arrangement, so that the time crowding trend is suppressed in the early stage.

[0103] The adaptive update module for scheduling strategy is used to comprehensively utilize the analysis results of rework disturbance propagation and the evaluation results of cumulative cycle offset. Starting from the time and space margin that the task can obtain in the execution calendar and the current scheduling impedance state of the production node, it evaluates the adjustable range of the task on the time axis and performs continuous time correction on the original planned start time of the task accordingly. This allows the execution calendar to be adaptively reconstructed as the production conditions change without changing the production process path, thereby achieving continuous and stable evolution of scheduling results in complex rework scenarios.

[0104] In this implementation plan, the rework time-series aggregation perception module is used to collect and organize the operational data that is scattered in the execution log and resource monitoring during the production of advertising materials. Through time alignment and scale unification, the concentrated occurrence of rework behavior on the time axis is transformed from discrete records into continuous and perceptible operational characteristics, thereby providing a data foundation for subsequent analysis that can reflect the intensity of production rhythm fluctuations and rework aggregation trends.

[0105] The rework disturbance transmission analysis module is designed to analyze the rhythm change relationship between adjacent production nodes in the actual flow path of advertising materials based on standardized production process data. It transforms the time stretching and resource occupation extension caused by rework into measurable transmission results, clearly expressing the diffusion direction and intensity of the rework impact between nodes, and providing a basis for the linkage adjustment of execution arrangements at the node level.

[0106] The function of the cumulative beat offset assessment module is to evaluate the cumulative offset status of the production process in the time dimension by comparing the actual time unfolding formed during the continuous execution of tasks with the time unfolding under the original production sequence. This reflects whether the production rhythm is in a stable range, and accordingly imposes constraints on the time window occupation method and task release rhythm to suppress the continuous amplification of time squeezing within the scheduling cycle.

[0107] The adaptive update module of the scheduling strategy is to combine the rework disturbance propagation results and the cycle offset evaluation results, and determine the adjustable range of the task on the time axis based on the time and space margin of the task in the execution calendar and the scheduling impedance state presented by the production node. By continuously correcting the planned start time, the execution calendar distribution is reconstructed so that the scheduling results can continuously evolve with changes in the production conditions and maintain executability.

[0108] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for scheduling advertising material production based on production status awareness, characterized in that: include: S1: Collect execution time-related data and resource queue data during the production process of advertising materials, preprocess the execution time-related data and resource queue data, and construct a standardized production condition dataset; S2, based on a standardized production process dataset, analyzes the transmission degree of the impact of rework in the production process by examining the relationship between the rhythm changes of adjacent production nodes caused by rework, and makes coordinated adjustments to the execution arrangements in the production process of advertising materials based on the transmission analysis results. S3, based on a standardized production process dataset, evaluates the stability of the production process rhythm by assessing the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution, and sets linkage constraints on the time window occupancy and task release rhythm in the production execution arrangement based on the offset evaluation results. S4 takes the results of conduction analysis and offset evaluation as input, evaluates the adjustability of task time landing point based on the time and space margin of task in execution calendar and node scheduling impedance status, and performs continuous time correction on task plan start time based on scheduling evaluation results to reconstruct execution calendar distribution.

2. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for collecting execution time-related data and resource queue data during the advertising material production process are as follows: By creating node operation logs and execution calendar records, execution time-related data is collected. This data includes: the queuing time before the material enters the processing stage, the length of time the node resources are continuously occupied, the planned start time of each ad creative task in the original production sequence, and the actual start time of entering the execution stage. A scheduling cycle is defined as a continuous time interval starting from the current time and ending at the time point of the next change in the execution structure in the node execution calendar. The length of the available time window that each production node has not yet been occupied by tasks within the scheduling cycle and the actual processing time consumed by a single ad creative task to complete processing at the corresponding node are collected. By monitoring the status of the task queue and resource usage of the production nodes, resource queue data related to parallel carrying and queuing status is collected. The resource queue data includes: the number of parallel carrying that each production node is allowed to run at the same time, the number of executing tasks that are in the execution state at any given time, the number of waiting tasks that have not yet entered the execution state, the required processing time for a task to complete one production process at the production node, and the number of rework occurrences at each node at each time. The task processing time is based on the task's own attributes or the task declaration, while the actual processing time is the time consumed by the task to actually complete the processing, as recorded in the execution log.

3. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for preprocessing execution time-related data and resource queue data to construct a standardized production condition dataset are as follows: After completing the collection of execution time-related data, the starting point of the scheduling cycle is used as a unified time benchmark. The queuing waiting time, occupied time, planned start time and actual start time are converted into relative time expressions relative to the starting point of the scheduling cycle, eliminating the influence of absolute time offset between different nodes and different record sources. The execution time-related data is processed by segment pruning. When the occupied time or available time window crosses the boundary of the scheduling cycle, only the part that falls within the scheduling cycle is retained. Based on the planned start time and the actual start time, a planned start time sequence and an actual start time sequence are constructed. Monotonicity checks are performed on the planned start time sequence and the actual start time sequence within the same node. When the time record order is found to be inconsistent with the actual execution order, the order is rearranged according to the timestamp size, only correcting the order without changing the values ​​themselves. After collecting resource queue data, the number of parallel loads, the number of executed tasks, the number of waiting tasks, and the number of reworks are filtered by time according to the scheduling cycle range, retaining only valid records that occurred within the scheduling cycle. The number of executed tasks, the number of waiting tasks, and the number of reworks are smoothed by a moving average algorithm in chronological order to eliminate abnormal fluctuations caused by log jitter and instantaneous statistical errors. Duplicate count records generated by different logs at the same node at the same time are merged and summed. The consistency between the demand processing time and the actual processing time recorded in the execution log is checked. When there is a conflict between the demand processing time and the actual processing time, the actual processing time is used as the value of the task demand processing time for subsequent calculations. After standardizing the execution time-related data and resource queue data, the minimum and maximum value normalization algorithms are used to normalize the execution time-related data and resource queue data to construct a standardized production condition dataset.

4. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for analyzing the transmission degree of rework impact in the production process based on the standardized production condition dataset and the relationship between the cycle time changes of adjacent production nodes caused by rework are as follows: The node's overall cycle time is obtained by adding the node's actual processing time, queuing time, and occupied time at the current time. The total cycle time of the preceding node is obtained by adding the actual processing time, queuing time, and occupied time of the preceding node in the same time period. Subtracting the total clock speed of the preceding node from the total clock speed of the node yields the clock speed transition of the adjacent node. The absolute value of the difference between the actual processing time of the calculated node and the preceding node is added to one to obtain the tick buffer constraint. The number of reworks occurring within the preceding node is used as the numerator, and the number of reworks occurring within the node plus one is used as the denominator to obtain the rework propagation trigger quantity. Divide the clock transition amount of adjacent nodes by the clock buffer constraint amount and then multiply it by the rework propagation trigger amount to obtain the rework disturbance propagation value. When the production node does not have a preceding node in the actual flow path of the advertising material, the current production node is regarded as the starting node of the rework disturbance propagation path, and the corresponding rework disturbance propagation value is determined only by the rework disturbance response result formed by the production node in the current scheduling cycle.

5. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for adjusting the execution arrangements in the advertising material production process based on the transmission analysis results are as follows: The rework disturbance propagation values ​​calculated by each production node within the same scheduling cycle are aggregated and sorted according to the magnitude of the rework disturbance propagation values ​​to construct a rework disturbance propagation sequence. The production nodes in the first n positions of the rework disturbance propagation sequence are marked as clock constraint nodes. The number of parallel bearers that clock constraint nodes are allowed to be in execution state at the same time in the next scheduling cycle is reduced. Specifically, the original number of parallel bearers is reduced according to the proportion of the position in the rework disturbance propagation sequence to the total number of nodes and rounded down to obtain the updated number of parallel bearers. When the updated number of parallel bearers is zero, the corresponding production node will not receive new execution tasks in the next scheduling cycle. For the clock constraint node where the number of parallel loads is reduced, the advertising material tasks that exceed the updated number of parallel loads are removed from the load set of the production node. The original production flow order of the removed advertising material tasks remains unchanged, and they are assigned to other production nodes in the rework disturbance propagation sequence that are sorted after the clock constraint node. For production nodes whose sorting position in the rework disturbance propagation sequence is after the beat constraint node, they receive advertising material tasks that have been moved out of the beat constraint node and are waiting to be executed, within the range allowed by the updated parallel carrying capacity, without changing the predetermined production process order of the advertising material tasks.

6. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for evaluating the stability of the production process rhythm based on the offset relationship between the actual time unfolding and the original sequential time unfolding during continuous task execution, using a standardized production condition dataset, are as follows: Obtain the actual start time of the first ad creative task to enter the execution state within the current scheduling period and the actual start time of the ad creative task currently entering the execution state. Subtract the actual start time of the first ad creative task to enter the execution state from the actual start time of the ad creative task currently entering the execution state to obtain the actual execution time span. Subtract the planned start time of the first task from the planned start time of the current task in the original production sequence to obtain the sequential execution time span. When there is only one ad creative task in the execution state within the scheduling period, the sequential execution time span is taken as the requirement processing time of the task on the corresponding production node after consistency verification; Among them, the task processing time formed by the actual execution time span and the sequential execution time span adopts the task requirement processing time after consistency verification; Divide the actual execution time span by the sequential execution time span, add one, and take the natural logarithm to obtain the cumulative beat stretch. The number of tasks that can be accommodated is obtained by dividing the available time window length of the production node within the scheduling cycle by the required processing time after consistency verification. The resource squeezing ratio is obtained by taking the number of tasks that are currently running on a node as the numerator and the sum of the number of tasks being run and the number of tasks that can be accommodated as the denominator. The resource squeezing ratio is calculated by arctangent, and then multiplied by the cumulative beat stretch to obtain the cumulative beat offset value.

7. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for linking the time window occupancy and task release rhythm in the production execution schedule based on the offset evaluation results are as follows: Based on the adjustment of the carrying capacity of the production node according to the rework disturbance transmission value, the cumulative value of the cycle offset of each production node is summarized within the same scheduling cycle, and all nodes are sorted by offset according to the size of the cumulative value of the cycle offset. The nodes with the highest cumulative beat offset values ​​in the offset sorting are designated as time-occupying and diffusion nodes. For these nodes, a reserved time segment is set aside within the available time window to absorb rework disturbances and handle rhythm rebounds. The remaining time is used as an allocatable time segment to participate in the execution calendar generation. When generating the execution calendar, the scheduling side locks the reserved time segment, allowing only advertising material tasks to enter the allocatable time segment for execution, thus limiting the continuous erosion of subsequent time windows by rework disturbances. Regarding the task release method, a fixed delivery rhythm is implemented for all waiting tasks at each node. Only when all advertising material tasks currently in the execution state of the node are completed, and the next time boundary marked as an allocable time segment in the execution calendar is triggered, can a new advertising material task enter the execution state. The release time of subsequent tasks strictly depends on the completion time of the previous task and the boundary of the allocable time segment. When the available time interval of a certain node is insufficient to accommodate waiting tasks, the waiting tasks that have not yet entered the execution state will be migrated across nodes while maintaining the original process sequence. The migration target is selected from the candidate nodes that meet the process constraints, and the node with the largest remaining available time window is selected. The time point when the task enters the execution state is fixed as the start time of the next available time interval of the target node. For waiting tasks that have not yet entered the execution state, the offsets are reordered based on the sum of the cumulative beat offsets formed by the waiting tasks in the completed production nodes. The sorting result only affects the order in which the tasks enter the execution state and does not change the production path and node dependencies.

8. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for scheduling evaluation of the adjustability of task timing, using the results of conduction analysis and offset assessment as inputs, and considering the task's temporal and spatial margins in the execution calendar and the node scheduling impedance status, are as follows: Subtract the required processing time for a task to complete one production process from the available time window length of the current production node to obtain the task time space margin; where the required processing time is the effective required processing time determined after consistency verification. Add the cumulative value of the cycle offset of the production node where the task is located to the value of the rework disturbance propagation, and then add one to obtain the node scheduling impedance. Divide the task time and space margin by the node scheduling impedance to obtain the task scheduling displacement value.

9. The advertising material production scheduling method based on production status awareness according to claim 1, characterized in that: The specific steps for reconstructing the execution calendar distribution by performing continuous time correction on the task plan start time based on scheduling evaluation results are as follows: For each advertising material task, read the planned start time of the task in the original production plan and record it as the original start time. Subtract the task scheduling displacement value from the original start time to obtain the corrected start time. After calculating the corrected start time for all advertising material tasks, the scheduling side writes the tasks into the execution calendar of the corresponding production node in the order of their corrected start times. When the execution time segments of different tasks overlap in the execution calendar, the task with the earlier corrected start time remains unchanged, and the execution time of the task with the later corrected start time is postponed. The postponement amount is equal to the required processing time of the previous task after consistency verification. When a task's execution segment exceeds the available time window length of the current production node due to time delay, the task is removed from the current node's execution calendar. The first continuous idle time segment that can fully accommodate the required processing time after the corrected start time is found in the execution calendars of other production nodes and used as the new execution segment for the task. The completion time of a task is naturally determined by the end time of the final execution segment. The completion time is not modified separately during the scheduling process. The delivery time is formed by the correction of the start time and the requirement processing time after consistency verification.

10. An advertising material production scheduling system based on production status awareness, employing the advertising material production scheduling method based on production status awareness as described in any one of claims 1-9, characterized in that: include: The rework time sequence aggregation and perception module is used to collect execution time-related data and resource queue data during the production process of advertising materials, preprocess the execution time-related data and resource queue data, and construct a standardized production condition dataset. The rework disturbance transmission analysis module is used to analyze the transmission degree of the rework impact in the production process based on the relationship between the rhythm changes of adjacent production nodes caused by rework, using a standardized production condition dataset. Based on the transmission analysis results, the execution arrangements in the production process of advertising materials are adjusted accordingly. The cumulative beat offset evaluation module is used to evaluate the stability of the production process rhythm based on the offset relationship between the actual time unfolding and the original sequential time unfolding during the continuous task execution process, based on the standardized production work condition dataset. Based on the offset evaluation results, the module also sets linkage constraints on the time window occupation and task release rhythm in the production execution arrangement. The adaptive update module for scheduling strategy is used to take the results of transmission analysis and offset evaluation as input, evaluate the adjustability of task time landing points based on the time and space margin of tasks in the execution calendar and the scheduling impedance status of nodes, and perform continuous time correction on the task plan start time based on the scheduling evaluation results to reconstruct the execution calendar distribution.

Images