Optimized scheduling method for production and assembly process of liquid crystal television complete machine

Abstract

The application discloses a kind of optimization scheduling methods of liquid crystal television complete machine production assembly process, comprising the order comprehensive characteristic data of target work order and the time series data of the running state of multiple assembly lines;Based on the time series data of the running state, whether the pre-defined production abnormal event occurs is monitored;If yes, then calculate the scheduling risk influence factor of the affected work order under the current abnormal scene according to the order comprehensive characteristic data;Scheduling risk influence factor, order comprehensive characteristic data and time series data of the running state are input into pre-trained scheduling decision model, and adaptive dynamic adjustment strategy is output;Control assembly line to execute adjustment strategy, and calculate scheduling effect evaluation value after execution, and the weight parameter of scheduling decision model is updated using rolling horizon optimization algorithm in combination with digital twin simulation.The application realizes adaptive scheduling and online continuous optimization under abnormal scene, significantly improves delivery guarantee capacity and line running stability.

Description

Technical Field

[0001] This invention relates to the field of LCD TV technology, and in particular to an optimized scheduling method for the production and assembly process of an LCD TV. Background Technology

[0002] Currently, LCD TV assembly lines generally use a combination of MES (Manufacturing Execution System) and human experience for production scheduling. Common practices include setting static queuing rules based on work order priorities or performing simple capacity allocation based on the load of critical equipment. Some companies have introduced advanced planning and scheduling systems, which generate initial scheduling plans by establishing linear programming or heuristic algorithm models to minimize completion time or maximize equipment utilization.

[0003] However, existing technologies have significant limitations. First, static scheduling rules struggle to respond to real-time changes in production line status. When a production line experiences a sudden equipment stoppage or material supply delay, schedulers often rely on experience for emergency adjustments, resulting in slow response times and inconsistent decision-making quality, easily leading to the spread of abnormal impacts across production lines. Second, traditional scheduling models typically only consider equipment capacity and process constraints, lacking a multi-dimensional integrated assessment of order attributes and real-time production line operating status. This leads to a disconnect between scheduling plans and actual on-site execution capabilities, resulting in a "good-looking schedule, but ineffective execution" dilemma. Furthermore, existing systems lack effective feedback and model self-optimization mechanisms after each scheduling decision, preventing the accumulation of scheduling knowledge. Repeated occurrences of similar abnormal events necessitate recalculation, and scheduling efficiency gradually declines as production line status drifts. Therefore, there is an urgent need for an assembly scheduling method that can integrate multi-source order characteristics with dynamic production line status, possess adaptive response capabilities to abnormalities, and continuously optimize online, in order to improve the delivery assurance capabilities and operational stability of the LCD TV manufacturing process. Summary of the Invention

[0004] To address at least one of the aforementioned technical problems, this invention provides an optimized scheduling method for the production and assembly process of a complete LCD TV.

[0005] In a first aspect, the present invention provides an optimized scheduling method for the production and assembly process of an LCD TV, comprising: Obtain comprehensive order feature data for the target work order and time-series data of production line operation status for multiple assembly lines; the comprehensive order feature data includes at least delivery urgency, material availability rate, and process complexity; the time-series data of production line operation status includes at least equipment health, work-in-process inventory rate, and personnel skill suitability. Based on production line operation status time-series data, monitor whether predefined production anomaly events occur; if a production anomaly event is detected, calculate the scheduling risk impact factor of the affected work order under the current anomaly scenario based on the comprehensive characteristic data of the order; The scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time series data are input into the pre-trained scheduling decision model, and an adaptive dynamic adjustment strategy is output. The assembly line is controlled to execute an adaptive dynamic adjustment strategy, and the scheduling effect evaluation value is calculated after execution. Based on digital twin simulation and scheduling effect evaluation value, the weight parameters of the scheduling decision model are updated using a rolling time domain optimization algorithm.

[0006] Secondly, the present invention also provides an optimized scheduling system for the production and assembly process of an LCD TV, comprising: The operation data acquisition module is used to acquire the comprehensive order feature data of the target work order and the production line operation status time sequence data of multiple assembly production lines; the comprehensive order feature data includes at least the delivery urgency, material availability rate and process complexity; the production line operation status time sequence data includes at least the equipment health status, work-in-process backlog rate and personnel skill suitability. The scheduling risk calculation module is used to monitor whether predefined production anomaly events occur based on the production line operation status time-series data; if a production anomaly event is detected, the scheduling risk impact factor of the affected work order under the current anomaly scenario is calculated based on the comprehensive characteristic data of the order. The strategy output module is used to input scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time series data into the pre-trained scheduling decision model and output an adaptive dynamic adjustment strategy. The scheduling model update module is used to control the assembly line to execute adaptive dynamic adjustment strategies and calculate the scheduling effect evaluation value after execution. Based on digital twin simulation and scheduling effect evaluation value, the weight parameters of the scheduling decision model are updated using a rolling time domain optimization algorithm.

[0007] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) This invention constructs a multi-dimensional information perception foundation covering "order attributes - production line status" by acquiring comprehensive order characteristic data of the target work order, including delivery urgency, material availability rate, process complexity, and time-series data of production line operation status of multiple assembly lines, including equipment health, work-in-process inventory rate, and personnel skill suitability. Compared with the scheduling methods in the prior art that rely solely on static scheduling rules or single capacity indicators, this invention enables the scheduling system to comprehensively capture key factors affecting production execution, providing sufficient data support for subsequent anomaly identification and strategy generation, effectively avoiding decision-making biases caused by missing information, and laying the foundation for improving scheduling quality from an information perspective.

[0008] 2) Upon detecting a production anomaly, this invention calculates a delivery risk baseline, material risk gain coefficient, and process risk gain coefficient based on delivery urgency, material availability rate, and process complexity, respectively, and then weights and fuses them to obtain a scheduling risk impact factor. This factor transforms qualitative or discrete order characteristics into a continuously comparable unified risk value, enabling the system to objectively quantify the impact degree of different work orders under the same anomaly scenario. Based on this, this invention further determines the risk protection priority according to the scheduling risk impact factor and matches the corresponding strategy combination template from a predefined strategy combination mapping table. This template explicitly specifies the activation combination and parameter range of three strategies: work order cross-line diversion, line-side buffer replenishment, and process cycle time adjustment. This mode significantly compresses the time window from anomaly occurrence to strategy generation. Simultaneously, the three strategies provide complementary intervention methods from three dimensions: production line load transfer, material buffer reinforcement, and workstation cycle time adaptation, respectively, allowing for flexible combinations based on anomaly type and avoiding the limitations of a single strategy.

[0009] 3) This invention calculates a scheduling effect evaluation value after strategy execution. This evaluation value is not a simple achievement rate statistic, but rather incorporates a dynamic loss coefficient to include line switching losses in the effect evaluation system. The higher the line switching loss, the greater the gradient of the loss coefficient decrease, thus guiding the system to consider production line switching costs while pursuing delivery achievement, avoiding efficiency losses caused by "over-scheduling." Based on this, this invention constructs a multi-objective optimization cost function in a digital twin simulation environment, consisting of a weighted sum of a delivery deviation penalty term, a production line efficiency fluctuation penalty term, and a scheduling improvement reward term. A rolling time-domain optimization algorithm is used to iteratively update the weight parameters of the scheduling decision model. The core value of this mechanism lies in the fact that the digital twin environment provides a zero-risk trial-and-error space for parameter optimization, while rolling time-domain optimization ensures that the model parameters can continuously and adaptively adjust as the production line state drifts. Scheduling risk influencing factors participate in both strategy generation and the dynamic determination of cost function weights, causing the weight of the delivery deviation penalty term to adaptively increase in high-risk scenarios. This ensures that the model prioritizes delivery in emergency situations while considering efficiency under normal circumstances, achieving differentiated self-adjustment of the scheduling strategy under different risk levels.

[0010] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the background art, the accompanying drawings used in the embodiments of the present invention or the background art will be described below.

[0012] Figure 1 A flowchart illustrating an optimized scheduling method for the production and assembly process of a liquid crystal television set, provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of an optimized scheduling system for the production and assembly process of an LCD TV, provided as an embodiment of the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the present invention, 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.

[0014] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0015] Example 1: Please see Figure 1 , Figure 1 This is a flowchart illustrating an optimized scheduling method for the production and assembly process of a liquid crystal television set, as provided by the present invention. Figure 1 As shown, the method includes: S10. Obtain comprehensive order feature data of the target work order and production line operation status time sequence data of multiple assembly production lines; the comprehensive order feature data shall at least include delivery urgency, material availability rate and process complexity; the production line operation status time sequence data shall at least include equipment health, work-in-process inventory rate and personnel skill suitability. S20. Based on the production line operation status time sequence data, monitor whether a predefined production abnormal event occurs; if a production abnormal event is detected, calculate the scheduling risk impact factor of the affected work order under the current abnormal scenario based on the comprehensive characteristic data of the order; S30. Input the scheduling risk impact factors, comprehensive order characteristic data and production line operation status time series data into the pre-trained scheduling decision model, and output an adaptive dynamic adjustment strategy; S40. Control the assembly line to execute an adaptive dynamic adjustment strategy and calculate the scheduling effect evaluation value after execution; based on digital twin simulation and scheduling effect evaluation value, use a rolling time domain optimization algorithm to update the weight parameters of the scheduling decision model.

[0016] Existing LCD TV assembly scheduling relies heavily on static scheduling rules or manual experience, making it difficult to respond in real-time to fluctuations in production line status. Furthermore, it lacks multi-dimensional integrated assessment of order attributes and production line dynamics, leading to delayed decision-making, a disconnect between strategies and the actual situation, and an inability to accumulate and optimize scheduling knowledge when anomalies occur. To address this issue, this embodiment designs an adaptive scheduling scheme for anomaly scenarios based on a closed-loop control logic of "perception—judgment—decision—evolution." The overall approach is as follows: First, a multi-dimensional information perception foundation covering order attributes and production line status is constructed. Then, anomaly events are proactively identified based on real-time production line data, and the risk level of affected work orders is quantified. Subsequently, the risk quantification results, along with order and production line status data, are input into the scheduling decision model, outputting executable multi-dimensional adjustment strategies. Finally, the effectiveness is evaluated after execution, and model parameters are continuously iterated and optimized using a digital twin simulation environment, enabling scheduling capabilities to increase over time.

[0017] Specifically, step S10 aims to establish a comprehensive information perception layer. Existing systems typically focus only on equipment load or work order priority, resulting in a single information dimension. This step, by collecting order characteristics including delivery urgency, material availability rate, and process complexity, as well as production line time-series data including equipment health, work-in-process inventory rate, and personnel skill suitability, provides a sufficient data foundation for subsequent anomaly identification and strategy generation, avoiding decision-making biases caused by missing information. Step S20 differs from the traditional passive waiting for manual fault reporting by automatically monitoring predefined abnormal events such as sudden drops in equipment health, materials falling below safety stock, and personnel skill mismatch based on production line status time-series data. Once an anomaly is detected, a unified scheduling risk impact factor is immediately calculated based on the order's delivery urgency, material shortage rate, and process difficulty, transforming multi-dimensional order attributes into comparable risk quantification values. This allows the system to objectively assess the impact of different work orders under the same anomaly, providing a basis for differentiated strategies. Existing systems lack a structured strategy generation mechanism after anomalies occur, often relying on manual "case-by-case" handling. Step S30 inputs risk impact factors, order characteristics, and production line status into a pre-trained scheduling decision model. The model matches preset strategy combination templates according to the risk level and comprehensively considers the acceptance capacity scores of candidate production lines to output a comprehensive adjustment strategy that includes work order cross-line diversion, line-side buffer replenishment, and process cycle adjustment. This achieves end-to-end automated decision-making from anomaly identification to strategy output. Finally, in step S40, execution feedback and model evolution are completed. First, the production line execution strategy is controlled, and a scheduling effect evaluation value with a dynamic loss coefficient is calculated. This evaluation focuses on both delivery achievement and execution costs such as line changeover losses. Subsequently, the evaluation value is input into a digital twin simulation environment, and the model weight parameters are iteratively updated using a rolling time-domain optimization method. This allows the model to continuously and adaptively adjust as the production line status drifts, achieving the accumulation of scheduling knowledge and a spiral improvement in scheduling quality.

[0018] This embodiment, through the organic coordination of the above four steps, forms a complete closed loop of "real-time perception → anomaly warning → risk quantification → strategy matching → execution feedback → model evolution", which can significantly improve the scheduling response speed, strategy accuracy and long-term operating efficiency in abnormal scenarios, and effectively ensure the timely delivery of LCD TV assembly and the stable operation of the production line.

[0019] In one specific implementation, in step S10, the comprehensive order feature data of the target work order and the time sequence data of the production line operation status of multiple assembly lines are acquired in real time through the Manufacturing Execution System (MES) interface, Enterprise Resource Planning (ERP) system and industrial Internet of Things sensor network of the production line.

[0020] The comprehensive feature data of an order includes the following parameters: Delivery urgency: Determined based on the difference between the customer's contract delivery date and the current time. The smaller the difference, the higher the urgency. It can also be weighted and adjusted based on the customer's level to quantify the risk of delivery default in the future.

[0021] Material availability rate: refers to the percentage of all materials required for the target work order that have been fully supplied to the line-side warehouse. It focuses on tracking the arrival of key materials such as LCD panels, backlight modules and motherboards, and is used to assess the material support capability for production start-up or continuous operation.

[0022] Process complexity: Determined by the product model corresponding to the work order, it is quantitatively graded by referring to the number of processes in the whole machine assembly, special process requirements (such as precision fitting tolerance, high temperature aging time) and historical average assembly time, and is used to measure the requirements of the work order on production line skills and equipment precision.

[0023] Production line operation status timing data includes the following parameters: Equipment operational health: By collecting multi-dimensional sensor signals such as vibration, temperature, and current from key equipment (such as pick-and-place machines and laminators), and combining them with historical statistics of equipment failures, a continuous value of 0 to 100% is calculated by a health index model to reflect the degree of degradation and failure probability of the equipment at the current moment.

[0024] Work-in-process inventory ratio: The ratio of the number of products waiting to be processed in the buffer area of ​​each process to the rated buffer capacity of that process, used to determine whether there are signs of bottlenecks or decreased flow in the production line.

[0025] Personnel skill fit: Based on the skill level certificates and historical operation quality data of on-the-job personnel, combined with the process complexity requirements of the current work order, it is calculated through a skill-job matching matrix, reflecting the degree of matching between personnel capabilities and current production tasks.

[0026] In one specific implementation, the predefined production anomaly event in step S20 includes at least one of the following types: Equipment malfunction event: When the rate of decline of the equipment health timing signal of any assembly line exceeds the preset rate threshold within a predetermined time window, an equipment malfunction event is triggered. Material supply anomaly event: When the real-time inventory level in the line-side buffer is lower than the safety stock limit associated with the material availability rate in the order comprehensive feature data, a material supply anomaly event is triggered. Personnel skill mismatch event: A personnel skill mismatch event is triggered when the skill level of the personnel currently on the production line does not meet the minimum skill suitability required by the process complexity in the comprehensive feature data of the order. Emergency order insertion event: When an emergency order insertion signal is received, and the delivery urgency of the inserted work order is higher than the average delivery urgency of the current work orders in progress, an emergency order insertion event is triggered.

[0027] In step S20, a combined monitoring mechanism of "threshold triggering + status comparison" is typically used to scan the production line operation status time-series data in real time. For equipment malfunctions, the system extracts the linear regression slope within a predetermined time window from the equipment health time-series signal. If the slope is negative and the absolute value exceeds a preset slope threshold, the system determines that the equipment is at risk of rapid degradation. For material supply anomalies, the system dynamically calculates the lower limit of the line-side safety stock corresponding to the current work order based on the material availability rate in the order's comprehensive feature data (if the availability rate is higher, the lower limit can be appropriately lowered), compares the actual inventory in real time, and triggers an alert if it falls below the lower limit. For personnel skill mismatches, the system reads the skill codes of on-duty personnel from the human resources database and matches them with the minimum skill level required by the process complexity of the current work order. For urgent orders, the system continuously monitors order insertion signals from the ERP or sales department. Once a new work order is received, its delivery urgency is calculated and compared with the average value of work orders in production.

[0028] The selection of the above four types of abnormal events is specifically targeted: equipment malfunctions directly affect the available production capacity and processing quality of the production line, and are the most uncontrollable source of disturbance in scheduling; material supply abnormalities are a high-frequency problem in LCD TV assembly, especially the large fluctuations in the arrival of key materials such as panels and backlight modules, which directly determine whether work orders can be produced continuously; personnel skill mismatch targets processes such as LCD TV whole-set testing and precision bonding, which require high operational proficiency, and insufficient personnel skills can lead to decreased yield and hidden line stoppages; emergency order insertion reflects the uncertainty of market demand and is an external disturbance that must be addressed in mixed-flow production mode. These four types of events cover the four core dimensions of "equipment-materials-personnel-orders," and can comprehensively capture the main sources of abnormalities on the production floor. Through the above real-time monitoring mechanism, the system can proactively identify abnormalities in their early stages, rather than waiting for manual reporting or post-event statistics, thus gaining valuable response time for subsequent risk quantification and strategy generation, and effectively preventing the impact of abnormalities from spreading between production lines.

[0029] Furthermore, if a production anomaly is detected, the scheduling risk impact factor of the affected work order under the current anomaly scenario is calculated based on the comprehensive characteristic data of the order, including: Based on the urgency of delivery, query the preset urgency-benchmark value mapping table to obtain the delivery risk benchmark value; The critical material shortage rate is calculated based on the material availability rate, and the material risk gain coefficient is determined based on the critical material shortage rate; wherein, the critical material includes at least one of the following: LCD panel, backlight module or motherboard; Based on the complexity of the process, determine the assembly difficulty level corresponding to the target work order, and determine the process risk gain coefficient based on the assembly difficulty level. The delivery risk benchmark value is weighted and fused with the material risk gain coefficient and the process risk gain coefficient to obtain the scheduling risk impact factor.

[0030] In this implementation, after detecting a production anomaly, the system calculates the scheduling risk impact factor based on the comprehensive order characteristic data of the affected work orders in the following manner: 1) Determine the delivery risk benchmark value based on the urgency of the work order. The delivery urgency is calculated by converting the customer's contract delivery date into the current time; the smaller the difference, the higher the urgency. The system has a pre-set urgency-benchmark value mapping table, corresponding to different numerical benchmarks for different urgency levels. For example, work orders to be delivered on the same day correspond to a higher benchmark value, while work orders to be delivered several days later correspond to a lower benchmark value. This benchmark value reflects the inherent default risk of the work order in the time dimension and is the basic component of the risk factor.

[0031] 2) Calculate the critical material shortage rate based on the material availability rate, and determine the material risk gain coefficient accordingly. Specifically, the lower the material availability rate of a work order, the higher the shortage rate of its critical materials (such as LCD panels, backlight modules, motherboards, etc.). When abnormal events cause production line switching or production interruption, work orders with incomplete material availability will face a longer waiting time for recovery, thus further amplifying their risk. The system achieves this amplification effect through the material risk gain coefficient; the lower the availability rate of a work order, the higher the gain coefficient, allowing the risk factor to reflect the additional execution difficulty caused by material shortages.

[0032] 3) The assembly difficulty level is determined based on the process complexity, and a corresponding process risk gain coefficient is obtained. Work orders with high process complexity involve more precision bonding, high-temperature aging, or complex testing procedures, requiring higher equipment precision and personnel skills. In abnormal scenarios, such work orders face greater risks due to their less adjustable flexibility and higher barriers to production recovery. The system assigns higher risk weights to high-complexity work orders through the process risk gain coefficient.

[0033] 4) The delivery risk benchmark value is weighted and integrated with the material risk gain coefficient and the process risk gain coefficient to obtain a unified scheduling risk impact factor.

[0034] It's important to note that delivery urgency reflects the uncompromising hard constraint of "time pressure," serving as a fundamental component. Material shortages and process complexity, on the other hand, are moderating factors for "executability." They don't independently generate risk but rather amplify or mitigate the underlying risk. The three factors are multiplied rather than added because material shortages or high process complexity don't directly cause default, but they significantly exacerbate the vulnerability of urgent work orders to abnormal shocks. Through this implementation method, the originally discrete and heterogeneous order attributes can be transformed into uniform and comparable risk values. This allows the system to objectively quantify the different impact levels experienced by different work orders under the same abnormal scenario, providing a quantitative basis for subsequent priority prioritization and differentiated resource allocation, avoiding a "one-size-fits-all" approach to all work orders during emergency response.

[0035] In one specific implementation, step S30, which involves inputting scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time-series data into a pre-trained scheduling decision model and outputting an adaptive dynamic adjustment strategy, includes: Based on the scheduling risk impact factors, the risk protection priority of affected work orders is determined. Then, based on the risk protection priority, the corresponding strategy combination template is matched from a predefined strategy combination mapping table. The strategy combination template specifies the activation combination and parameter range of three strategies: work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. Based on the time-series data of production line operation status, the current equipment health, work-in-process inventory rate and personnel skill suitability of each candidate backup production line are obtained, and the ability score of each candidate backup production line to accept affected work orders is calculated. The strategy combination template, the acceptance capacity score of each candidate backup production line, and the scheduling risk impact factor are used as input features of the model. Through the reasoning of the scheduling decision model, an adaptive dynamic adjustment strategy is generated.

[0036] Specifically, in order to output an adaptive dynamic adjustment strategy, the main steps are as follows: 1) Strategy Combination Template Matching: Based on the numerical value of the scheduling risk impact factor, affected work orders are divided into different risk protection priority levels. For example, work orders with risk factors higher than the first threshold are classified as the highest priority, while those lower than the second threshold are classified as ordinary priority. The system has a pre-set strategy combination mapping table, which pre-defines a corresponding strategy combination template for each priority level. Each template explicitly specifies the activation combination method and adjustable parameter range of three strategies: work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. For example, the template corresponding to the highest priority work order may simultaneously enable all three strategies, limiting the candidate range for cross-line diversion to the two production lines with the highest equipment health; while the template for ordinary priority work orders may only enable the line-side buffer replenishment strategy, with a correspondingly narrower parameter adjustment range. This mapping mechanism directly associates the discretized risk level with the structured strategy action space, avoiding the need to search for strategy combinations from scratch every time an anomaly occurs, significantly reducing decision-making delay.

[0037] 2) Candidate Production Line Capability Scoring: Based on real-time acquisition of production line operation status time-series data for each candidate backup production line, three indicators are extracted for each production line: current equipment health, work-in-process inventory (WIP) rate, and personnel skill suitability. The system then performs a comprehensive weighted calculation of these three indicators to obtain a capability score for each backup production line to handle affected work orders. During the calculation, production lines with higher equipment health, lower WIP rates, and higher personnel skill suitability receive higher capability scores. If a production line's personnel skill suitability cannot meet the process complexity requirements of the work order, the system will automatically remove that production line from the candidate list, ensuring that the hard constraints of feasibility are met.

[0038] 3) Joint Inference: The matched strategy combination template, the capacity scores of each candidate production line, and the scheduling risk impact factors are used as input feature vectors and fed into a pre-trained scheduling decision model. This model is trained on a large number of historical scheduling samples, learning how to output specific strategy parameter values ​​(such as the number of work orders diverted across lines, the increase in safety stock for buffer replenishment, and the time offset for cycle time adjustment) under given risk levels, available production line resources, and strategy template constraints. Through multi-objective trade-offs, the model balances production line switching costs and efficiency fluctuations while ensuring the delivery of high-risk work orders, ultimately outputting a complete adaptive dynamic adjustment strategy.

[0039] By employing a priority-driven template matching mechanism, the infinitely possible strategy search space is reduced to a finite set of structured combinations, significantly decreasing the inference difficulty and computation time of the decision-making model. By introducing a capacity score, the generated strategies are ensured to be executable under actual production line conditions, avoiding the dilemma of "looking good but not being able to be implemented." The input features, composed of these three elements, take into account the triple constraints of risk urgency, available resources, and strategy boundaries, making the model's output strategies targeted and operable, and enabling end-to-end automated generation from anomaly identification to executable instructions.

[0040] Preferably, the adaptive dynamic adjustment strategy includes at least one or more combinations of work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy; The work order cross-line diversion strategy is to determine the target receiving production line based on the production line operation status time sequence data of each candidate backup production line and the scheduling risk impact factor of the affected work order, and generate a work order switching instruction to migrate the affected work order from the current production line to the target receiving production line. The line-side buffer replenishment strategy dynamically adjusts the safety stock threshold of the line-side buffer based on the material availability rate and scheduling risk impact factors in the comprehensive feature data of the order, and generates material pull instructions based on the adjusted safety stock threshold. The process cycle adjustment strategy involves recalculating the standard operating time or equipment operating frequency of the target workstation based on the process complexity in the comprehensive feature data of the order and the current equipment operating health of the production line, and generating process control instructions containing the adjusted cycle parameters.

[0041] This preferred embodiment defines three core adjustment strategies and their specific implementation logic, namely: Work order cross-line diversion strategy: When a production line is unable to continue processing the current work order due to equipment malfunction or overload, the system comprehensively evaluates the equipment health, work-in-process inventory rate, and personnel skill suitability of each candidate backup production line, and combines this with the scheduling risk impact factors of the affected work order to calculate a comprehensive acceptance score for each candidate production line. Based on this, the system selects the production line with the highest score as the target acceptance production line and automatically generates a work order switching instruction to migrate the work order from the current production line to the target production line for continued production. The purpose of this strategy is to enable high-risk work orders to be promptly removed from the abnormal situation through dynamic load rebalancing between production lines, avoiding delays in the entire order due to a single point of failure.

[0042] Line-side buffer replenishment strategy: The system dynamically adjusts the safety stock threshold of the line-side buffer based on the material availability rate and scheduling risk impact factor of the work order; the lower the material availability rate or the higher the risk factor, the larger the increase. Subsequently, the system automatically generates a material pull instruction based on the difference between the adjusted new threshold and the actual inventory, triggering the warehouse to replenish the line-side. The purpose of this strategy is to proactively increase the material buffer to offset the risk of supply interruption caused by abnormal events, ensuring continuous operation of the production line after recovery.

[0043] Process cycle time adjustment strategy: The system recalculates the standard operating time or equipment operating frequency for the target workstation based on the process complexity of the work order and the current equipment operating health of the production line. When the process complexity is high or the equipment health is low, the system can appropriately extend the standard operating time or reduce the equipment frequency to reduce quality degradation or accelerated equipment wear caused by forced speed increases. The purpose of this strategy is to maintain the basic output of the production line while ensuring quality through flexible adjustment of the cycle time.

[0044] The three strategies mentioned above can be flexibly combined or activated individually according to abnormal scenarios, providing intervention measures from three dimensions: "spatial transfer", "material replenishment" and "rhythm adaptation". They cover the most important control points on the production site, enabling the system to select the most suitable set of strategies for different types of abnormal events, and achieve accurate and low-cost emergency response.

[0045] When the adaptive dynamic adjustment strategy is output, it is sent to the assembly line, and in step S40, the assembly line is controlled to execute the adaptive dynamic adjustment strategy. Further, in one specific embodiment, after executing the strategy, a scheduling effect evaluation value is calculated, specifically including: Compare the on-time delivery rate of orders with a preset delivery effectiveness threshold; When the on-time delivery rate of an order is less than or equal to the delivery effectiveness threshold, the scheduling effect evaluation value will be set to zero or a preset baseline penalty value. When the on-time delivery rate of an order is greater than the delivery effectiveness threshold, the output achievement rate is used as the basic performance value, and the dynamic loss coefficient is determined based on the line change loss. The dynamic loss coefficient is negatively correlated with the line change loss, and when the line change loss exceeds the preset loss warning threshold, the decreasing gradient of the dynamic loss coefficient increases. The scheduling effect evaluation value is obtained by multiplying the basic performance value by the dynamic loss coefficient.

[0046] The core design idea of ​​this implementation is that scheduling effectiveness evaluation should not simply use delivery achievement as the sole criterion, but should simultaneously consider both output results and the costs incurred during the scheduling process. In LCD TV assembly, although emergency scheduling aims to ensure delivery, frequent work order cross-line migrations can lead to line changeover losses, including production line downtime for switching, the cost of reheating and parameter verification of new production line equipment, and the cost of secondary material handling. If line changeover losses are too high, even if delivery is ultimately on time, the overall efficiency of the production line will have been severely eroded, and this loss will continue to propagate in subsequent batches. Therefore, this solution introduces line changeover losses into the evaluation system in the form of a dynamic loss coefficient, enabling the system to weigh "delivery achievement" and "scheduling costs" simultaneously when evaluating scheduling plans, guiding decisions towards a direction that balances both.

[0047] Specifically, after executing the adaptive dynamic adjustment strategy, the system first calculates the on-time delivery rate of the target work orders, which is the proportion of work orders completed and delivered on time out of the total number of affected work orders. Then, the system compares this on-time delivery rate with a preset delivery effectiveness threshold, which is typically set as the lower limit of acceptable delivery assurance for the enterprise. If the on-time delivery rate is lower than or equal to this threshold, it indicates that the scheduling strategy has failed to effectively guarantee basic delivery requirements, and the scheduling plan is actually ineffective. The system directly sets the scheduling effect evaluation value to zero or a preset baseline penalty value, placing the evaluation result of this scheduling at a low level and providing a negative sample for subsequent model parameter updates. If the on-time delivery rate is higher than the threshold, the system considers that the delivery level has met the basic requirements, and then enters the refined evaluation stage. The system first calculates the output achievement rate, which is the ratio of the actual number of completed work orders to the planned number, using this as the basic performance value. Based on this, the system calculates the actual line changeover losses incurred during this scheduling process, which comprehensively considers factors such as production line changeover time, equipment recalibration costs, and material transfer workload. The system calculates a dynamic loss coefficient based on line switching losses. The relationship between the two is set as follows: the lower the line switching loss, the closer the loss coefficient is to one, meaning there is almost no reduction in basic performance; as line switching losses increase, the loss coefficient decreases non-linearly. Especially when line switching losses exceed a preset loss warning threshold, the descent gradient of the loss coefficient further increases, meaning that excessive scheduling beyond a reasonable range will be subject to more severe evaluation penalties. Finally, the system multiplies the basic performance value by the dynamic loss coefficient to obtain the final scheduling effect evaluation value.

[0048] Through the aforementioned evaluation mechanism, the system can provide a comprehensive evaluation of each scheduling decision, taking into account both "effectiveness" and "cost." Compared to simply using delivery rate as an evaluation indicator, this embodiment effectively suppresses the aggressive scheduling tendency of "ensuring delivery regardless of cost," guiding the model to proactively select the scheme with lower line-switching losses during strategy generation. Simultaneously, the nonlinear loss mechanism generates strong negative feedback on high-loss scheduling schemes exceeding the warning threshold, prompting the model to gradually converge to a reasonable balance between delivery achievement and efficiency loss in subsequent iterations. This evaluation value, as the core feedback signal for model parameter updates in step S40, directly determines the quality of the model optimization direction, providing a reliable cost measurement basis for rolling time-domain optimization.

[0049] Furthermore, in one embodiment, updating the weight parameters of the scheduling decision model using a rolling time-domain optimization algorithm based on digital twin simulation and scheduling effect evaluation values ​​includes: In a digital twin simulation environment, a multi-objective optimization cost function corresponding to the scheduling decision model is constructed; the cost function is composed of a weighted sum of a delivery deviation penalty term, a production line efficiency fluctuation penalty term, and a scheduling improvement reward term; Based on the scheduling risk impact factors corresponding to the historically implemented adaptive dynamic adjustment strategies, the weight coefficients of the delivery deviation penalty, production line efficiency fluctuation penalty, and scheduling improvement reward are dynamically determined; among them, the scheduling risk impact factor is positively correlated with the weight coefficient of the delivery deviation penalty. Input the currently collected scheduling effect evaluation value into the multi-objective optimization cost function to calculate the comprehensive cost value of this scheduling. With the goal of minimizing the cumulative comprehensive cost value within a future preset time domain, multiple sets of simulated scheduling trajectories are generated using a digital twin simulation environment, and the weight parameters of the scheduling decision model are iteratively optimized and updated based on the simulated scheduling trajectories.

[0050] Traditional scheduling decision-making models typically face two challenges. First, production line conditions are time-varying; for example, equipment aging, personnel turnover, and product process iterations cause the mapping relationships learned from historical data to drift over time. Without continuous updates, the quality of model decisions will gradually decline. Second, scheduling decisions have a lag effect; an inappropriate cross-line diversion may lead to new work-in-process inventory buildup or equipment overload within hours, and these long-term costs are difficult to detect in real time through single-step evaluation. Therefore, traditional offline training methods cannot address these issues. To resolve this, this implementation method combines digital twin simulation and rolling time-domain optimization algorithms.

[0051] Specifically, firstly, a multi-objective optimization cost function corresponding to the scheduling decision model is constructed in the digital twin simulation environment. This cost function consists of three components: a delivery deviation penalty, used to quantify the severity of the deviation between the actual completion time of the work order and the planned delivery date; a production line efficiency fluctuation penalty, used to measure the fluctuation range of production line utilization before and after the execution of the scheduling strategy; and a scheduling improvement reward, used to encourage the model to proactively take preventative adjustments rather than passive responses. The three components are weighted and summed to form a comprehensive cost value. In the weight determination stage, the system dynamically adjusts the weights of the three components according to the scheduling risk impact factors corresponding to this scheduling. When the risk factor is high, the weight of the delivery deviation penalty is adaptively increased, enabling the model to prioritize ensuring delivery dates in high-risk scenarios; when the risk factor is low, the weight of the efficiency fluctuation penalty is appropriately increased, guiding the model to focus on maintaining stable production line operation. Subsequently, the system inputs the scheduling effect evaluation value calculated in step S40 into the cost function to calculate the comprehensive cost value of this scheduling, serving as a quantitative evaluation of the merits of the current strategy. In the optimization and update phase, the system aims to minimize the cumulative comprehensive cost over a predetermined future time domain. It generates multiple candidate simulation scheduling trajectories in a digital twin simulation environment, with each trajectory corresponding to a different combination of model weight parameters. The system iterates through the cumulative costs generated by each set of parameters, selects the parameter combination with the lowest cost, and feeds it back to the scheduling decision model as the updated weight value.

[0052] Digital twin simulation environments provide a platform for risk-free trial and error, enabling systems to explore the long-term effects of various weight configurations in a virtual space without incurring trial and error costs on the actual production line. Rolling time-domain optimization allows model parameters to continuously and adaptively adjust as production line states drift, overcoming the limitations of traditional offline models that are "deployed and then fixed." A dynamic weighting mechanism ensures the model exhibits differentiated decision preferences under different risk scenarios, prioritizing delivery in high-risk situations and efficiency in low-risk situations, achieving an adaptive multi-objective balance. Through this closed-loop iteration, the scheduling decision model is optimized after each anomaly handling, the quality of handling similar events increases with runtime, and the overall scheduling capability of the system continuously evolves.

[0053] Example 2: See Figure 2 The present invention also provides an optimized scheduling system for the production and assembly process of an LCD TV, comprising: The operation data acquisition module 100 is used to acquire the comprehensive order feature data of the target work order and the time sequence data of the production line operation status of multiple assembly lines; the comprehensive order feature data includes at least the delivery urgency, material availability rate and process complexity; the time sequence data of the production line operation status includes at least the equipment health, work-in-process backlog rate and personnel skill suitability. The scheduling risk calculation module 200 is used to monitor whether a predefined production anomaly event occurs based on the production line operation status time sequence data; if a production anomaly event is detected, the scheduling risk impact factor of the affected work order under the current anomaly scenario is calculated based on the comprehensive characteristic data of the order. The adjustment strategy output module 300 is used to input scheduling risk impact factors, comprehensive order characteristic data and production line operation status time series data into the pre-trained scheduling decision model and output an adaptive dynamic adjustment strategy. The scheduling model update module 400 is used to control the assembly line to execute an adaptive dynamic adjustment strategy and calculate the scheduling effect evaluation value after execution. Based on the digital twin simulation and the scheduling effect evaluation value, the weight parameters of the scheduling decision model are updated using a rolling time-domain optimization algorithm.

[0054] Preferably, the adjustment strategy output module 300 is further configured to: Based on the scheduling risk impact factors, the risk protection priority of affected work orders is determined. Then, based on the risk protection priority, the corresponding strategy combination template is matched from a predefined strategy combination mapping table. The strategy combination template specifies the activation combination and parameter range of three strategies: work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. Based on the time-series data of production line operation status, the current equipment health, work-in-process inventory rate and personnel skill suitability of each candidate backup production line are obtained, and the ability score of each candidate backup production line to accept affected work orders is calculated. The strategy combination template, the acceptance capacity score of each candidate backup production line, and the scheduling risk impact factor are used as input features of the model. Through the reasoning of the scheduling decision model, an adaptive dynamic adjustment strategy is generated.

[0055] Preferably, the scheduling model update module 400 is further configured to: In a digital twin simulation environment, a multi-objective optimization cost function corresponding to the scheduling decision model is constructed; the cost function is composed of a weighted sum of a delivery deviation penalty term, a production line efficiency fluctuation penalty term, and a scheduling improvement reward term; Based on the scheduling risk impact factors corresponding to the historically implemented adaptive dynamic adjustment strategies, the weight coefficients of the delivery deviation penalty, production line efficiency fluctuation penalty, and scheduling improvement reward are dynamically determined; among them, the scheduling risk impact factor is positively correlated with the weight coefficient of the delivery deviation penalty. Input the currently collected scheduling effect evaluation value into the multi-objective optimization cost function to calculate the comprehensive cost value of this scheduling. With the goal of minimizing the cumulative comprehensive cost value within a future preset time domain, multiple sets of simulated scheduling trajectories are generated using a digital twin simulation environment, and the weight parameters of the scheduling decision model are iteratively optimized and updated based on the simulated scheduling trajectories. It is understood that the system provided in this embodiment has functions or includes modules that can be used to execute the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0056] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

Claims

1. An optimized scheduling method for the production and assembly process of an LCD TV, characterized in that, include: Obtain comprehensive order feature data for the target work order and time-series data of production line operation status for multiple assembly lines; The comprehensive order feature data includes at least delivery urgency, material availability rate, and process complexity; the production line operation status time-series data includes at least equipment health, work-in-process inventory rate, and personnel skill suitability. Based on production line operation status time-series data, monitor whether predefined production anomaly events occur; If a production anomaly is detected, the scheduling risk impact factor of the affected work order under the current anomaly scenario is calculated based on the comprehensive characteristic data of the order. The scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time series data are input into the pre-trained scheduling decision model, and an adaptive dynamic adjustment strategy is output. Control the assembly line to execute an adaptive dynamic adjustment strategy, and calculate the scheduling effect evaluation value after execution; Based on digital twin simulation and scheduling effect evaluation, a rolling time-domain optimization algorithm is used to update the weight parameters of the scheduling decision model.

2. The optimized scheduling method for the production and assembly process of a liquid crystal television set according to claim 1, characterized in that, The adaptive dynamic adjustment strategy includes at least one or more combinations of work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. The work order cross-line diversion strategy is to determine the target receiving production line based on the production line operation status time sequence data of each candidate backup production line and the scheduling risk impact factor of the affected work order, and generate a work order switching instruction to migrate the affected work order from the current production line to the target receiving production line. The line-side buffer replenishment strategy dynamically adjusts the safety stock threshold of the line-side buffer based on the material availability rate and scheduling risk impact factors in the comprehensive feature data of the order, and generates material pull instructions based on the adjusted safety stock threshold. The process cycle adjustment strategy involves recalculating the standard operating time or equipment operating frequency of the target workstation based on the process complexity in the comprehensive feature data of the order and the current equipment operating health of the production line, and generating process control instructions containing the adjusted cycle parameters.

3. The optimized scheduling method for the production and assembly process of a complete LCD TV as described in claim 1, characterized in that, The calculated scheduling effect evaluation value includes: Compare the on-time delivery rate of orders with a preset delivery effectiveness threshold; When the on-time delivery rate of an order is less than or equal to the delivery effectiveness threshold, the scheduling effect evaluation value will be set to zero or a preset baseline penalty value. When the on-time delivery rate of an order is greater than the delivery effectiveness threshold, the output achievement rate is used as the basic performance value, and the dynamic loss coefficient is determined based on the line change loss. The dynamic loss coefficient is negatively correlated with the line change loss, and when the line change loss exceeds the preset loss warning threshold, the decreasing gradient of the dynamic loss coefficient increases. The scheduling effect evaluation value is obtained by multiplying the basic performance value by the dynamic loss coefficient.

4. The optimized scheduling method for the production and assembly process of a complete LCD TV as described in claim 1, characterized in that, The calculation of the scheduling risk impact factor of affected work orders under the current abnormal scenario based on comprehensive order characteristic data includes: Based on the urgency of delivery, query the preset urgency-benchmark value mapping table to obtain the delivery risk benchmark value; The critical material shortage rate is calculated based on the material availability rate, and the material risk gain coefficient is determined based on the critical material shortage rate; wherein, the critical material includes at least one of the following: LCD panel, backlight module or motherboard; Based on the complexity of the process, determine the assembly difficulty level corresponding to the target work order, and determine the process risk gain coefficient based on the assembly difficulty level. The delivery risk benchmark value is weighted and fused with the material risk gain coefficient and the process risk gain coefficient to obtain the scheduling risk impact factor.

5. The optimized scheduling method for the production and assembly process of a liquid crystal television set according to claim 1, characterized in that, The predefined production anomaly events include at least one of the following types: Equipment malfunction event: When the rate of decline of the equipment health timing signal of any assembly line exceeds the preset rate threshold within a predetermined time window, an equipment malfunction event is triggered. Material supply anomaly event: When the real-time inventory level in the line-side buffer is lower than the safety stock limit associated with the material availability rate in the order comprehensive feature data, a material supply anomaly event is triggered. Personnel skill mismatch event: A personnel skill mismatch event is triggered when the skill level of the personnel currently on the production line does not meet the minimum skill suitability required by the process complexity in the comprehensive feature data of the order. Emergency order insertion event: When an emergency order insertion signal is received, and the delivery urgency of the inserted work order is higher than the average delivery urgency of the current work orders in progress, an emergency order insertion event is triggered.

6. The optimized scheduling method for the production and assembly process of a liquid crystal television set according to claim 1, characterized in that, The process of inputting scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time-series data into a pre-trained scheduling decision model, and outputting an adaptive dynamic adjustment strategy, includes: Based on the scheduling risk impact factors, the risk protection priority of affected work orders is determined. Then, based on the risk protection priority, the corresponding strategy combination template is matched from a predefined strategy combination mapping table. The strategy combination template specifies the activation combination and parameter range of three strategies: work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. Based on the time-series data of production line operation status, the current equipment health, work-in-process inventory rate and personnel skill suitability of each candidate backup production line are obtained, and the ability score of each candidate backup production line to accept affected work orders is calculated. The strategy combination template, the acceptance capacity score of each candidate backup production line, and the scheduling risk impact factor are used as input features of the model. Through the reasoning of the scheduling decision model, an adaptive dynamic adjustment strategy is generated.

7. The optimized scheduling method for the production and assembly process of a liquid crystal television set according to claim 1, characterized in that, The method for updating the weight parameters of the scheduling decision model based on digital twin simulation and scheduling effect evaluation values ​​using a rolling time-domain optimization algorithm includes: In a digital twin simulation environment, a multi-objective optimization cost function corresponding to the scheduling decision model is constructed; the cost function is composed of a weighted sum of a delivery deviation penalty term, a production line efficiency fluctuation penalty term, and a scheduling improvement reward term; Based on the scheduling risk impact factors corresponding to the historically implemented adaptive dynamic adjustment strategies, the weight coefficients of the delivery deviation penalty, production line efficiency fluctuation penalty, and scheduling improvement reward are dynamically determined; among them, the scheduling risk impact factor is positively correlated with the weight coefficient of the delivery deviation penalty. Input the currently collected scheduling effect evaluation value into the multi-objective optimization cost function to calculate the comprehensive cost value of this scheduling. With the goal of minimizing the cumulative comprehensive cost value within a future preset time domain, multiple sets of simulated scheduling trajectories are generated using a digital twin simulation environment, and the weight parameters of the scheduling decision model are iteratively optimized and updated based on the simulated scheduling trajectories.

8. An optimized scheduling system for the production and assembly process of an LCD TV, characterized in that, include: The data acquisition module is used to acquire comprehensive order feature data of the target work order and time-series data of production line operation status of multiple assembly lines; The comprehensive order feature data includes at least delivery urgency, material availability rate, and process complexity; the production line operation status time-series data includes at least equipment health, work-in-process inventory rate, and personnel skill suitability. The scheduling risk calculation module is used to monitor whether predefined production anomaly events occur based on the time-series data of the production line operation status. If a production anomaly is detected, the scheduling risk impact factor of the affected work order under the current anomaly scenario is calculated based on the comprehensive characteristic data of the order. The strategy output module is used to input scheduling risk impact factors, comprehensive order characteristic data, and production line operation status time series data into the pre-trained scheduling decision model and output an adaptive dynamic adjustment strategy. The scheduling model update module is used to control the assembly line to execute adaptive dynamic adjustment strategies and calculate the scheduling effect evaluation value after execution. Based on digital twin simulation and scheduling effect evaluation, a rolling time-domain optimization algorithm is used to update the weight parameters of the scheduling decision model.

9. The optimized scheduling system for the production and assembly process of a liquid crystal television set according to claim 8, characterized in that, The adjustment strategy output module is also used for: Based on the scheduling risk impact factors, the risk protection priority of affected work orders is determined. Then, based on the risk protection priority, the corresponding strategy combination template is matched from a predefined strategy combination mapping table. The strategy combination template specifies the activation combination and parameter range of three strategies: work order cross-line diversion strategy, line-side buffer replenishment strategy, and process cycle adjustment strategy. Based on the time-series data of production line operation status, the current equipment health, work-in-process inventory rate and personnel skill suitability of each candidate backup production line are obtained, and the ability score of each candidate backup production line to accept affected work orders is calculated. The strategy combination template, the acceptance capacity score of each candidate backup production line, and the scheduling risk impact factor are used as input features of the model. Through the reasoning of the scheduling decision model, an adaptive dynamic adjustment strategy is generated.

10. The optimized scheduling method for the production and assembly process of a liquid crystal television set according to claim 8, characterized in that, The scheduling model update module is also used for: In a digital twin simulation environment, a multi-objective optimization cost function corresponding to the scheduling decision model is constructed; the cost function is composed of a weighted sum of a delivery deviation penalty term, a production line efficiency fluctuation penalty term, and a scheduling improvement reward term; Based on the scheduling risk impact factors corresponding to the historically implemented adaptive dynamic adjustment strategies, the weight coefficients of the delivery deviation penalty, production line efficiency fluctuation penalty, and scheduling improvement reward are dynamically determined; among them, the scheduling risk impact factor is positively correlated with the weight coefficient of the delivery deviation penalty. Input the currently collected scheduling effect evaluation value into the multi-objective optimization cost function to calculate the comprehensive cost value of this scheduling. With the goal of minimizing the cumulative comprehensive cost value within a future preset time domain, multiple sets of simulated scheduling trajectories are generated using a digital twin simulation environment, and the weight parameters of the scheduling decision model are iteratively optimized and updated based on the simulated scheduling trajectories.

Images