Method and system for optimizing intelligent scheduling and productivity balance of liquid crystal panel production line

Abstract

The invention provides a liquid crystal panel production line intelligent scheduling and capacity balance optimization method and system, and relates to the technical field of liquid crystal panel production, and the method comprises the steps: calculating the capacity matching degree through the real-time collection of equipment production state data, determining the optimal control parameter through a time sequence difference and deep reinforcement learning algorithm, and obtaining the optimal control parameter; regulation sequence optimization is realized by combining ant colony optimization and dynamic planning, equipment operation is monitored in real time, and an emergency instruction is generated, so that productivity balance of each process link of the production line can be realized, the equipment utilization rate is improved, the production cost is reduced, and the response capability of the production line to abnormal conditions is enhanced.

Description

Technical Field

[0001] This invention relates to the field of liquid crystal panel manufacturing technology, and in particular to a method and system for intelligent scheduling and capacity balancing optimization of a liquid crystal panel production line. Background Technology

[0002] LCD panel manufacturing is a core industry in modern electronic display technology, involving multiple processes such as array fabrication and cell assembly. In LCD panel production lines, real-time monitoring and optimization of equipment operation at each process stage have a decisive impact on production line efficiency and product quality. Traditional LCD panel production lines typically rely on manual experience and semi-automated management for production scheduling. With the rapid development of the LCD panel industry and intensified market competition, intelligent and automated scheduling of production lines has become an industry trend. However, existing technologies still lack effective capacity matching assessment mechanisms, and capacity imbalances between array and cell assembly processes are common, leading to low overall production line efficiency. Furthermore, the lack of data-driven intelligent optimization methods makes it difficult to accurately assess equipment status and optimize parameter settings, hindering the achievement of theoretically maximum equipment operating efficiency. Additionally, the absence of real-time dynamic adjustment mechanisms prevents the rapid generation of emergency scheduling strategies, impacting the overall efficiency of the production line. Summary of the Invention

[0003] This invention provides a method and system for intelligent scheduling and capacity balancing optimization of a liquid crystal panel production line, which can at least solve some of the problems existing in the prior art.

[0004] A first aspect of this invention provides a method for intelligent scheduling and capacity balancing optimization of a liquid crystal panel production line, comprising: Real-time collection and digital processing of production status data of various process equipment in the LCD panel production line; combined with the material flow process of the production line to calculate the capacity matching degree of array process and cell assembly process. The optimization target of the process equipment is determined based on the capacity matching degree. The state evaluation of the process equipment is carried out by the time difference algorithm. The operation control parameters of each process equipment are updated by calculating the deviation between the equipment operation state and the target state until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The performance of the candidate control parameters is evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is generated and optimized by dynamic neighborhood search to obtain the global optimal adjustment sequence. The production capacity matching improvement path after executing the global optimal adjustment sequence is calculated in real time by dynamic programming method. The unexecuted adjustment sequence is dynamically adjusted according to the path matching result until the pre-set production target is met. The system monitors the operating status of each process equipment in real time. When an equipment malfunction or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

[0005] In one alternative implementation, Real-time acquisition and digital processing of production status data from various process equipment in the LCD panel production line; calculation of capacity matching between array manufacturing and cell assembly processes based on material flow process of the production line, including: Real-time acquisition of operating status data of various process equipment in the LCD panel production line, including processing time, number of workpieces, and equipment fault information; The operating status data is reorganized according to the material transfer sequence of the process equipment to establish the processing time relationship between process equipment and calculate the material flow status between adjacent process equipment in real time. Based on the material flow status of the process equipment, the equipment utilization rate, work-in-process quantity and processing cycle of the array process and the box process are calculated, and the parameters are weighted and summed to obtain the capacity matching degree between the processes.

[0006] In one alternative implementation, The optimization objectives for process equipment are determined based on the capacity matching degree. A time-difference algorithm is used to evaluate the state of the process equipment. The operating control parameters of each process equipment are updated by calculating the deviation between the equipment's operating state and the target state, until the preset maximum number of iterations is reached or convergence is achieved. Candidate control parameters include: The optimization target of the process equipment is determined based on the capacity matching degree, and the deviation between the capacity matching degree and the preset target state is weighted and summed to obtain the state evaluation result. The time-difference algorithm is used to evaluate the status of process equipment. The equipment operating parameters and material inventory information of the process equipment are collected to obtain the equipment operating status. Equipment rate adjustment parameters and batch size adjustment parameters are generated as control actions. The instant reward value is calculated based on the deviation between the equipment operating status and the target status, and the current operating status of the process equipment is determined based on the instant reward value. The gradient of the deviation between the operating state and the target state of the calculated equipment with respect to the operating control parameters is used to update the operating control parameters of each process equipment along the negative direction of the gradient. Determine whether the number of iterations has reached the preset maximum number of iterations or whether the deviation change of consecutive preset number of iterations is less than the preset threshold. When either condition is met, the current operation control parameter is determined as a candidate control parameter.

[0007] In one alternative implementation, The performance of candidate control parameters is evaluated using deep reinforcement learning algorithms, and the optimal control parameters are determined as follows: Construct a causal influence propagation diagram among candidate control parameters, calculate the direct change derivative and indirect transmission derivative between different control parameters to obtain the causal influence intensity matrix, calculate the parameter correlation degree based on the causal influence intensity matrix, and determine the parameters with a correlation degree greater than a preset threshold as the main control parameters; Collect the operating status data of the control system, calculate the performance evaluation index based on the causal influence intensity matrix to obtain the parameter performance score, and prioritize the main control parameters according to the parameter performance score; The main control parameters are subjected to disturbances. The change transmission path is predicted by the causal influence intensity matrix. The parameter sensitivity value is calculated as the parameter evaluation weight. The independent influence coefficient and the synergistic influence coefficient are calculated based on the causal influence intensity matrix. The independent effect score and the synergistic effect score are obtained by multiplying them with the parameter evaluation weight. The parameter comprehensive score is obtained by weighted summing of the independent action score and the synergistic action score. The parameter combination with the highest comprehensive score is selected as the current optimal parameter combination. The parameter exploration frequency is adjusted according to the changing trend of the comprehensive score until the comprehensive score converges and the optimal control parameter is determined.

[0008] In one alternative implementation, Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is then generated and optimized using dynamic neighborhood search to obtain the globally optimal adjustment sequence, including: An initial adjustment sequence is generated using the optimal control parameters, and the adjustment constraints between adjacent parameters are calculated based on the parameter performance scores of the optimal control parameters to obtain the initial sequence priority. Based on the pre-acquired historical scheduling data and equipment operating characteristics, the initial value of the pheromone volatilization coefficient and the pheromone increment update rule are set. The pheromone concentration and heuristic information value between different control parameters are calculated to obtain the adjustment value selection probability. The adjustment value selection probability is used to generate an optimized adjustment sequence. The pheromone concentration is updated based on the pheromone volatilization coefficient and the pheromone increment. The visited regulation sequences are recorded to form a tabu list. The optimized regulation sequence is searched locally using exchange and reverse shifts to obtain candidate regulation sequences. The penalty factor of the candidate regulation sequence is determined by combining the parameter performance score and the tabu list. The initial regulation sequence is optimized according to the regulation value selection probability to obtain the optimized regulation sequence. The parameter performance score of the optimized regulation sequence is calculated. The regulation sequence with the highest parameter performance score is updated as the global optimum. The global optimal solution is optimized by dynamic neighborhood search. The search direction is updated based on adaptive step size. The target value is calculated by combining parameter performance score and penalty factor. The tabu list and the global optimal solution are updated according to the target value. The process is iterated until convergence to obtain the global optimal adjustment sequence.

[0009] In one alternative implementation, The production capacity matching improvement path is calculated in real time using dynamic programming methods after executing the globally optimal adjustment sequence. Based on the path matching results, the unexecuted adjustment sequences are dynamically adjusted until the pre-set production target is met, including: Calculate the capacity matching degree of the process equipment during the execution of the global optimal adjustment sequence, divide the capacity matching degree into multiple decision stages, obtain the state transition probability based on the capacity matching degree of each decision stage, calculate the state transition cost by combining the equipment adjustment cost coefficient and the capacity matching degree weight coefficient, and obtain the capacity matching degree improvement path by reverse recursion. Calculate the deviation between the current capacity status of the process equipment and the production target, obtain the adjustment amount using a pre-set proportional adjustment matrix and differential adjustment coefficient, and dynamically adjust the unexecuted adjustment sequence based on the product of the adjustment amount and the time-varying weight function; The adjustment range, adjustment rate, and capacity balance constraints of the process equipment are transformed into a barrier function with a penalty factor. The unexecuted adjustment sequence is dynamically adjusted according to the barrier function until the pre-set production target is met.

[0010] In one alternative implementation, Real-time monitoring of the operating status of each process equipment; when equipment malfunction or capacity imbalance is detected, alarm signals are triggered and emergency dispatch instructions are generated, including: Real-time acquisition of operational status data of process equipment; calculation of capacity balance of each process equipment and probability of equipment malfunction based on operational status data. When the capacity balance is lower than a preset threshold or the probability of equipment malfunction is higher than a preset threshold, an alarm signal is sent to the control system. Based on the current operating status of the process equipment and historical scheduling experience, an emergency scheduling instruction sequence is generated, which includes equipment power adjustment instructions and equipment start-up and shutdown sequence.

[0011] A second aspect of this invention provides an intelligent scheduling and capacity balancing optimization system for a liquid crystal panel production line, comprising: The first unit is used to collect production status data of each process equipment in the LCD panel production line in real time and perform digital processing, and calculate the capacity matching degree of array process and cell assembly process in combination with the material flow process of the production line. The second unit is used to determine the optimization target of the process equipment based on the capacity matching degree. It uses the time difference algorithm to evaluate the state of the process equipment, updates the operation control parameters of each process equipment by calculating the deviation between the equipment operation state and the target state, until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The candidate control parameters are then evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. The third unit is used to calculate the process equipment adjustment sequence based on the ant colony optimization algorithm and the optimal control parameters and perform sequence optimization, generate an optimized adjustment sequence and perform dynamic neighborhood search optimization to obtain the global optimal adjustment sequence, calculate the capacity matching improvement path after executing the global optimal adjustment sequence in real time through dynamic programming method, and dynamically adjust the unexecuted adjustment sequence in combination with the path matching result until the pre-set production target is met. The fourth unit is used to monitor the operating status of each process equipment in real time. When an equipment abnormality or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

[0012] A third aspect of the present invention provides an electronic device, comprising: A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0014] This invention achieves intelligent scheduling and capacity balancing optimization of the production line by real-time acquisition of data from various process equipment on the LCD panel production line and calculation of capacity matching degree in conjunction with the material flow process. Based on a data-driven decision model, it effectively solves the problems of strong reliance on human experience, slow response speed, and difficulty in achieving precise control in traditional scheduling methods, significantly improving the overall operating efficiency of the production line. By combining temporal difference algorithm and deep reinforcement learning, the state evaluation and parameter optimization of process equipment are performed, realizing adaptive adjustment of equipment operating parameters and keeping equipment performance at its best. The combined application of ant colony optimization algorithm and dynamic neighborhood search provides an effective solution for optimizing the process equipment adjustment sequence, greatly improving the capacity balancing of the production line. The application of dynamic programming method realizes real-time optimization of capacity matching degree, ensuring the continuity and stability of the LCD panel production process, and has a significant effect on improving product yield and reducing production costs. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the intelligent scheduling and capacity balancing optimization method for a liquid crystal panel production line according to an embodiment of the present invention. Figure 2 This is a simulation diagram of the capacity matching degree optimization and flow balance of the intelligent scheduling and capacity balancing optimization method for LCD panel production lines according to an embodiment of the present invention. Figure 3This is an optimization diagram of the adjustment trajectory under the obstacle function constraint of the intelligent scheduling and capacity balancing optimization method for LCD panel production lines according to an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0017] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0018] Figure 1 This is a flowchart illustrating the intelligent scheduling and capacity balancing optimization method for a liquid crystal panel production line according to an embodiment of the present invention. Figure 1 As shown, the method includes: Real-time collection and digital processing of production status data of various process equipment in the LCD panel production line; combined with the material flow process of the production line to calculate the capacity matching degree of array process and cell assembly process. The optimization target of the process equipment is determined based on the capacity matching degree. The state evaluation of the process equipment is carried out by the time difference algorithm. The operation control parameters of each process equipment are updated by calculating the deviation between the equipment operation state and the target state until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The performance of the candidate control parameters is evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is generated and optimized by dynamic neighborhood search to obtain the global optimal adjustment sequence. The production capacity matching improvement path after executing the global optimal adjustment sequence is calculated in real time by dynamic programming method. The unexecuted adjustment sequence is dynamically adjusted according to the path matching result until the pre-set production target is met. The system monitors the operating status of each process equipment in real time. When an equipment malfunction or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

[0019] In one alternative implementation, Real-time acquisition and digital processing of production status data from various process equipment in the LCD panel production line; calculation of capacity matching between array manufacturing and cell assembly processes based on material flow process of the production line, including: Real-time acquisition of operating status data of various process equipment in the LCD panel production line, including processing time, number of workpieces, and equipment fault information; The operating status data is reorganized according to the material transfer sequence of the process equipment to establish the processing time relationship between process equipment and calculate the material flow status between adjacent process equipment in real time. Based on the material flow status of the process equipment, the equipment utilization rate, work-in-process quantity and processing cycle of the array process and the box process are calculated, and the parameters are weighted and summed to obtain the capacity matching degree between the processes.

[0020] Data acquisition terminals are installed on each process piece of equipment in the LCD panel production line to collect real-time operating status data, including processing time, number of workpieces, and equipment fault information. The data acquisition terminals can be industrial-grade PLC controllers or embedded computing devices, connected to the equipment via industrial communication interfaces such as RS485, PROFIBUS, or industrial Ethernet. The acquisition frequency can be set to 5 seconds per acquisition to ensure real-time data accuracy.

[0021] In the array fabrication process, the main equipment includes cleaning machines, coating machines, exposure machines, developing machines, etching machines, and stripping machines. Taking the coating machine as an example, its operating status data includes: current batch number A20230512001, single substrate processing time 23.5 seconds, number of processed workpieces 216, equipment temperature 56.2℃, and equipment fault code NULL (indicating normal operation). This data is acquired in real time by the acquisition terminal and transmitted to the central data processing server via the industrial network.

[0022] In the cell assembly process, the main equipment includes alignment film coating machines, liquid crystal dispensing machines, laminating machines, and cutting machines. Taking the liquid crystal dispensing machine as an example, the operating status data includes: current batch number B20230512001, dispensing time per panel 28.3 seconds, number of panels dispensed 195, dispensing accuracy 99.7%, and equipment fault code E00215 (indicating abnormal pressure in valve number 2). The fault information is recorded and an early warning mechanism is immediately triggered, while the relevant data is transmitted to the central server.

[0023] After receiving the operating status data of each process equipment, the data processing server reorganizes the data according to the flow sequence of materials on the production line and pre-establishes a process route model for the material flow of the production line. For example, the process route for array manufacturing is: cleaning machine → coating machine → exposure machine → developing machine → etching machine → peeling machine, while the process route for cell assembly manufacturing is: alignment film coating machine → liquid crystal dispensing machine → laminating machine → cutting machine.

[0024] Based on the process route model, the operating status data of different equipment are sorted according to timestamps to construct the processing time sequence relationship between process equipment. For example, at a specific time point (May 12, 2023, 10:15:30), it was recorded that the coating machine completed the processing of the 157th substrate of batch number A20230512001, while at the same time point, the upstream cleaning machine had completed the processing of the 175th substrate, and the downstream exposure machine had completed the processing of the 142nd substrate. Using this data, the work-in-process quantity between the coating machine and the cleaning machine is calculated to be 18 pieces, and the work-in-process quantity between the coating machine and the exposure machine is 15 pieces.

[0025] The real-time calculation of material flow status between adjacent process equipment mainly includes the following indicators: material transfer rate (pieces / hour), material accumulation between equipment (pieces), and material transfer time (minutes). Taking the coating machine and the exposure machine as an example, based on the collected data, the calculated material transfer rate between the two machines is 152 pieces / hour, the material accumulation is 15 pieces, and the average material transfer time is 5.9 minutes.

[0026] Based on the material flow status of the process equipment, key production parameters for the array process and the box-forming process are further calculated. For the array process, the parameters calculated by the system include: average equipment utilization rate of 88.5% (calculated by dividing the actual working time of each piece of equipment by the total available time), total work-in-process inventory of 125 pieces (the sum of work-in-process inventory across all processes), and average processing cycle time of 153 pieces / hour (calculated based on the actual output of the bottleneck process). For the box-forming process, the calculated parameters include: average equipment utilization rate of 92.3%, total work-in-process inventory of 98 pieces, and average processing cycle time of 145 pieces / hour.

[0027] To calculate the capacity matching degree between processes, the above parameters are weighted and summed. Equipment utilization rate has a weight of 0.4, work-in-process inventory has a weight of 0.2, and cycle time has a weight of 0.4. The weighted score for the array process is: 88.5×0.4+(1-125 / 200)×0.2+(153 / 160)×0.4=85.6 points (where 200 is the upper limit of the work-in-process target, and 160 is the target value of the cycle time). The weighted score for the box-forming process is: 92.3×0.4+(1-98 / 200)×0.2+(145 / 160)×0.4=87.1 points. The capacity matching degree between processes is calculated as: (1-|85.6-87.1| / 100)×100%=98.5%, indicating a very high capacity matching degree between the two processes and balanced production line operation.

[0028] When the capacity matching rate is detected to be lower than a set threshold (such as 90%), an early warning mechanism will be triggered and adjustment suggestions will be generated. For example, when the equipment utilization rate of the array process drops to 75%, causing the capacity matching rate to drop to 85%, it is recommended to increase the working time of the array process or adjust the equipment maintenance plan to improve equipment utilization and capacity output.

[0029] In this embodiment, by collecting real-time operating status data of each process equipment on the production line and combining it with equipment fault information, the real-time operating status of the production line can be fully grasped, providing a complete data foundation for subsequent capacity matching degree calculation. By establishing a time-series correlation based on the material transfer sequence of the process equipment, the material flow status between adjacent equipment can be intuitively reflected, which helps to promptly identify capacity bottlenecks and material backlog issues between processes. By using equipment utilization rate, work-in-process quantity, and processing cycle as key parameters for weighted calculation, the capacity matching degree can accurately reflect the capacity balance between the array process and the cell assembly process, providing a quantitative basis for capacity adjustment.

[0030] In one alternative implementation, The optimization objectives for process equipment are determined based on the capacity matching degree. A time-difference algorithm is used to evaluate the state of the process equipment. The operating control parameters of each process equipment are updated by calculating the deviation between the equipment's operating state and the target state, until the preset maximum number of iterations is reached or convergence is achieved. Candidate control parameters include: The optimization target of the process equipment is determined based on the capacity matching degree, and the deviation between the capacity matching degree and the preset target state is weighted and summed to obtain the state evaluation result. The time-difference algorithm is used to evaluate the status of process equipment. The equipment operating parameters and material inventory information of the process equipment are collected to obtain the equipment operating status. Equipment rate adjustment parameters and batch size adjustment parameters are generated as control actions. The instant reward value is calculated based on the deviation between the equipment operating status and the target status, and the current operating status of the process equipment is determined based on the instant reward value. The gradient of the deviation between the operating state and the target state of the calculated equipment with respect to the operating control parameters is used to update the operating control parameters of each process equipment along the negative direction of the gradient. Determine whether the number of iterations has reached the preset maximum number of iterations or whether the deviation change of consecutive preset number of iterations is less than the preset threshold. When either condition is met, the current operation control parameter is determined as a candidate control parameter.

[0031] Obtain the initial operating parameters and process capacity data for each piece of equipment. Initial operating parameters include equipment speed, batch size, and operating time; process capacity data includes the theoretical and actual capacity of each process. For example, on a coating production line, the initial operating speed of the coating equipment is 15 pieces / hour, the batch size is 30 pieces / batch, the theoretical capacity is 480 pieces / day, and the actual capacity is 420 pieces / day.

[0032] Based on the acquired data, the capacity matching degree is calculated. The capacity matching degree represents the degree of coordination between the capacities of adjacent processes, and is obtained by calculating the ratio of the actual capacities of adjacent processes. Taking a production line containing three processes—pre-treatment, gluing, and drying—as an example, if the actual capacity of the pre-treatment process is 450 pieces / day, the actual capacity of the gluing process is 420 pieces / day, and the actual capacity of the drying process is 460 pieces / day, then the capacity matching degree between the pre-treatment and gluing processes is 450 / 420 = 1.07, and the capacity matching degree between the gluing and drying processes is 420 / 460 = 0.91.

[0033] The optimization objectives for process equipment are determined based on the capacity matching degree. An ideal capacity matching degree of 1 indicates that the capacities of adjacent processes are perfectly matched. When the capacity matching degree is greater than 1, it indicates that the upstream process has excess capacity, and the optimization objective should be to reduce the capacity of the upstream process or increase the capacity of the downstream process. When the capacity matching degree is less than 1, it indicates that the upstream process has insufficient capacity, and the optimization objective should be to increase the capacity of the upstream process or decrease the capacity of the downstream process.

[0034] The state assessment result is obtained by weighted summing the deviations between the capacity matching degree and the preset target state. Assuming the production line has n processes, there are n-1 capacity matching degree values. A weight coefficient is assigned to each capacity matching degree to reflect the degree of influence of that matching relationship on the overall production line efficiency. For example, if the glue coating process is the bottleneck process of the entire production line, the weight of the capacity matching degree related to the glue coating process can be set to 0.6, and the weight of other matching degrees can be 0.4. The deviation between each capacity matching degree and the target value of 1 is calculated, multiplied by the corresponding weight, and then summed to obtain the state assessment result. In the above example, the state assessment result is 0.6 × |1.07-1| + 0.4 × |0.91-1| = 0.062.

[0035] A time-series differential algorithm is used to assess the status of process equipment. Equipment operating parameters and material inventory information are collected, including the current equipment speed, current batch size, equipment status code, and material buffer amounts for upstream and downstream processes. Taking a coating equipment as an example, the current speed is 15 pieces / hour, the current batch size is 30 pieces / batch, the equipment is operating normally, the upstream process has a material buffer of 50 pieces, and the downstream process has a material buffer of 25 pieces.

[0036] The system generates equipment speed adjustment parameters and batch size adjustment parameters as control actions. The equipment speed adjustment parameters can be adjusted within ±10%, and the batch size adjustment parameters can be adjusted within ±5 pieces / batch. For the coating equipment, control actions can be generated to adjust the speed to 16.5 pieces / hour (an increase of 10%) and the batch size to 32 pieces / batch (an increase of 2 pieces / batch).

[0037] The immediate reward value is calculated based on the deviation between the equipment's operating status and the target status. The target status is set as a capacity matching degree of 1 for all processes, with actual capacity reaching over 95% of theoretical capacity. The smaller the deviation between the current status and the target status, the larger the immediate reward value. In this example, the deviation value is 0.062, resulting in an immediate reward value of 0.8 (out of 1).

[0038] The current operating status of the process equipment is determined based on the immediate reward value. If the immediate reward value is greater than 0.9, it indicates that the equipment is operating well; if it is between 0.7 and 0.9, it indicates that the status is average; if it is less than 0.7, it indicates that the status needs improvement. In the example, the reward value is 0.8, indicating that the equipment is operating in an average state and needs further optimization.

[0039] The gradient of the deviation between the equipment's operating state and the target state with respect to the operating control parameters is calculated. By making small adjustments to each control parameter, the changes in the state assessment results are observed, and the gradient is estimated. For example, adjusting the coating equipment rate from 15 pieces / hour to 15.5 pieces / hour changes the state assessment result from 0.062 to 0.058, with a gradient estimate of -0.008; adjusting the batch size from 30 pieces / batch to 31 pieces / batch changes the state assessment result from 0.062 to 0.060, with a gradient estimate of -0.002.

[0040] Update the operating control parameters of each process device along the negative direction of the gradient. A negative gradient indicates that increasing the parameter can reduce the deviation, while a positive gradient indicates that decreasing the parameter can reduce the deviation. Calculate the parameter adjustment amount based on the gradient magnitude and a preset learning rate (e.g., 0.1). For example, the glue coating equipment speed is adjusted to 15 + 0.1 × 0.008 × 15 = 15.12 pieces / hour, and the batch size is adjusted to 30 + 0.1 × 0.002 × 30 = 30.006 ≈ 30 pieces / batch.

[0041] Determine whether the number of iterations has reached the preset maximum number of iterations (e.g., 100 times) or whether the deviation change after a consecutive preset number of iterations (e.g., 10 times) is less than a preset threshold (e.g., 0.001). If either condition is met, the current operating control parameter is determined as a candidate control parameter; otherwise, the iterative optimization process continues.

[0042] After multiple iterations, assuming that in the 58th iteration, the deviation change is less than 0.001 for 10 consecutive iterations, the operating control parameters of the coating equipment are: rate 16.8 pieces / hour, batch size 32 pieces / batch, capacity matching evaluation result 0.998, and immediate reward value 0.95, achieving a good operating state. These parameters are selected as candidate control parameters and applied to actual production. In this embodiment, a comprehensive state assessment mechanism is established by weighted summing of the deviations between the capacity matching degree and the preset target state. This mechanism can accurately reflect the gap between the operating state of the process equipment and the optimization target. The state assessment is performed using a time-series difference algorithm, which transforms equipment operating parameters and material inventory information into real-time reward values. This enables a quantitative evaluation of the operating state of the process equipment, giving the optimization process a clear evaluation standard. The operating control parameters are dynamically updated based on the gradient descent method, ensuring the convergence of the optimization process. At the same time, the update method in the negative gradient direction ensures that the parameter adjustment always moves in the direction of reducing the deviation.

[0043] Figure 2 This diagram illustrates the simulation results of capacity matching optimization and flow balancing in the intelligent scheduling and capacity balancing optimization method for LCD panel production lines according to an embodiment of the present invention. It showcases the process of optimizing the capacity matching of process equipment based on a time-difference algorithm. Initially, the capacity matching degree for the pre-treatment-coating process is 1.20, the coating process rate is 1.50 pieces / minute, the drying process rate is 1.80 pieces / minute, and the overall capacity matching degree index is 0.60. As the optimization iterations proceed, the system dynamically adjusts the equipment rate and batch size of each process, gradually bringing the parameters closer to the ideal matching state (1.0). In the 40th iteration (the initial convergence stage), the pre-treatment-coating matching degree drops to 0.84, the coating process rate is adjusted to 0.93 pieces / minute, the drying process rate drops to 1.28 pieces / minute, and the capacity matching degree index improves to 0.80. After the 60th iteration (stabilization optimization phase), the system entered the fine-tuning phase. Finally, in the 70th iteration, the pretreatment-adhesive matching degree was 0.76, the adhesive coating process rate was 0.98 pieces / minute, the drying process rate was 1.01 pieces / minute, and the overall capacity matching degree reached 0.99. This achieved a high degree of coordination and matching between the various processes, significantly improving the overall efficiency of the production line.

[0044] In one alternative implementation, The performance of candidate control parameters is evaluated using deep reinforcement learning algorithms, and the optimal control parameters are determined as follows: Construct a causal influence propagation diagram among candidate control parameters, calculate the direct change derivative and indirect transmission derivative between different control parameters to obtain the causal influence intensity matrix, calculate the parameter correlation degree based on the causal influence intensity matrix, and determine the parameters with a correlation degree greater than a preset threshold as the main control parameters; Collect the operating status data of the control system, calculate the performance evaluation index based on the causal influence intensity matrix to obtain the parameter performance score, and prioritize the main control parameters according to the parameter performance score; The main control parameters are subjected to disturbances. The change transmission path is predicted by the causal influence intensity matrix. The parameter sensitivity value is calculated as the parameter evaluation weight. The independent influence coefficient and the synergistic influence coefficient are calculated based on the causal influence intensity matrix. The independent effect score and the synergistic effect score are obtained by multiplying them with the parameter evaluation weight. The parameter comprehensive score is obtained by weighted summing of the independent action score and the synergistic action score. The parameter combination with the highest comprehensive score is selected as the current optimal parameter combination. The parameter exploration frequency is adjusted according to the changing trend of the comprehensive score until the comprehensive score converges and the optimal control parameter is determined.

[0045] A causal propagation diagram of the relationships between candidate control parameters in a liquid crystal panel production line was constructed. Historical data for each control parameter was collected, including exposure energy, development time, and etching temperature in the array fabrication process, and bonding pressure, curing temperature, and alignment accuracy in the cell assembly process, with at least 1000 sets of sample data collected. Based on the collected data, the direct derivatives of the parameters were calculated, such as the rate of change of development time when exposure energy changes, with values ​​ranging from [-1, 1]. Simultaneously, the indirect derivatives, i.e., the transfer effect of exposure energy on etching quality through development time, were also calculated, with values ​​also ranging from [-1, 1]. For example, if the exposure energy increases by 5 mJ / cm², and the development time is subsequently extended by 10 seconds, the direct derivative is 2; if this change in development time leads to a 0.3 μm decrease in linewidth, the indirect derivative of exposure energy on linewidth is -0.06. The direct and indirect derivatives of each parameter were combined to form a causal influence strength matrix.

[0046] The correlation degree of each parameter is calculated based on the causal influence strength matrix. The parameter correlation degree is obtained by summing and normalizing the absolute values ​​of the row and column elements of a parameter in the matrix. For example, for 10 key control parameters on an LCD panel production line, if the absolute value sum of the row and column elements of the exposure energy parameter in the matrix is ​​8.5, and the maximum value among all parameters is 10, then the correlation degree of the exposure energy parameter is 0.85. Parameters with a correlation degree greater than 0.6 are identified as primary control parameters. For example, if the correlation degrees of exposure energy, development time, and bonding pressure are 0.85, 0.72, and 0.58, respectively, then exposure energy and development time are identified as primary control parameters.

[0047] Real-time data collection of the LCD panel production line's operational status, including current values ​​of various parameters and product quality indicators, is performed. Based on a causal influence strength matrix, the contribution of each key control parameter to the performance evaluation indicators is calculated, resulting in a parameter performance score. Performance evaluation indicators include panel yield, production cycle time, and product accuracy, with weights assigned according to production requirements. For example, if exposure energy has weights of 0.5, 0.2, and 0.3 for yield, cycle time, and accuracy, respectively, and current scores of 92, 88, and 90, then the performance score for the exposure energy parameter is 90.6. The key control parameters are then prioritized based on their performance scores.

[0048] A disturbance is applied to the main control parameters, typically 5%-10% of the normal range of parameter variation. The causal influence strength matrix is ​​used to predict the propagation path of the change, and the parameter sensitivity value is calculated as an evaluation weight. For example, if an increase in exposure energy of 1 mJ / cm² leads to a 0.5% improvement in yield, then its sensitivity value is 0.5. Based on the causal influence strength matrix, independent influence coefficients and synergistic influence coefficients are calculated separately. The independent influence coefficient reflects the direct impact of the parameter on product quality, while the synergistic influence coefficient reflects the cooperative effect between parameters.

[0049] The independent effect coefficient and the synergistic effect coefficient are multiplied by the parameter evaluation weights respectively to obtain the independent effect score and the synergistic effect score. For example, if the independent effect coefficient of exposure energy is 0.75, the synergistic effect coefficient is 0.65, and the sensitivity value is 0.5, then its independent effect score is 0.375 and its synergistic effect score is 0.325. The independent effect score and the synergistic effect score are then weighted and summed with a weight of 0.6:0.4 to obtain the comprehensive parameter score. The parameter combination with the highest comprehensive parameter score is selected as the current optimal parameter combination.

[0050] The parameter exploration frequency is adjusted based on the trend of the overall parameter score. A higher exploration frequency is maintained when the score continues to improve, and the exploration frequency is reduced when the score tends to stabilize. The exploration frequency is adjusted during production shift changes, ranging from once every 1 to 8 hours. When the score change rate is less than 0.5% over 10 consecutive evaluations, the current parameter combination is determined to be the optimal control parameter and applied to actual production.

[0051] In this embodiment, by constructing a causal influence propagation diagram and calculating the causal influence intensity matrix, a quantitative expression of the complex relationships between control parameters is achieved, providing a scientific theoretical basis for parameter optimization. The introduction of calculation methods for direct change derivatives and indirect transmission derivatives not only considers the direct influence between parameters but also the indirect influence generated through parameter transmission, making parameter correlation analysis more comprehensive and accurate. Based on the causal influence intensity matrix, a parameter evaluation system is established, combining parameter performance scores, independent influence coefficients, and synergistic influence coefficients to achieve a multi-dimensional evaluation of parameter importance, avoiding the one-sidedness that may be caused by a single evaluation index.

[0052] In one alternative implementation, Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is then generated and optimized using dynamic neighborhood search to obtain the globally optimal adjustment sequence, including: An initial adjustment sequence is generated using the optimal control parameters, and the adjustment constraints between adjacent parameters are calculated based on the parameter performance scores of the optimal control parameters to obtain the initial sequence priority. Based on the pre-acquired historical scheduling data and equipment operating characteristics, the initial value of the pheromone volatilization coefficient and the pheromone increment update rule are set. The pheromone concentration and heuristic information value between different control parameters are calculated to obtain the adjustment value selection probability. The adjustment value selection probability is used to generate an optimized adjustment sequence. The pheromone concentration is updated based on the pheromone volatilization coefficient and the pheromone increment. The visited regulation sequences are recorded to form a tabu list. The optimized regulation sequence is searched locally using exchange and reverse shifts to obtain candidate regulation sequences. The penalty factor of the candidate regulation sequence is determined by combining the parameter performance score and the tabu list. The initial regulation sequence is optimized according to the regulation value selection probability to obtain the optimized regulation sequence. The parameter performance score of the optimized regulation sequence is calculated. The regulation sequence with the highest parameter performance score is updated as the global optimum. The global optimal solution is optimized by dynamic neighborhood search. The search direction is updated based on adaptive step size. The target value is calculated by combining parameter performance score and penalty factor. The tabu list and the global optimal solution are updated according to the target value. The process is iterated until convergence to obtain the global optimal adjustment sequence.

[0053] An initial adjustment sequence is generated based on the optimal control parameters. Assuming the optimal control parameters for the LCD panel production line are: exposure energy 280 mJ / cm², development time 35 seconds, and etching temperature 75℃, adjustment constraints between adjacent parameters are calculated based on their performance scores. The parameter performance scores are derived from historical operating data analysis: exposure energy score is 89 points, development time score is 85 points, and etching temperature score is 82 points. The adjustment constraints consider the degree of mutual influence between parameters; for example, the influence coefficient of exposure energy change on development time is 0.82, the influence coefficient of development time change on etching quality is 0.75, and the influence coefficient of exposure energy on etching quality is 0.45. Based on these scores and constraints, the initial sequence priority is generated as follows: exposure energy → development time → etching temperature.

[0054] The pheromone evaporation coefficient was initially set to 0.5 based on historical scheduling data from the LCD panel production line. Pheromone increment calculations were based on yield improvements resulting from parameter adjustments; the greater the yield improvement, the more pheromones were added. Analysis of panel production data from the past six months summarized the effectiveness evaluations of different parameter adjustment sequences, which were then converted into initial pheromone distributions. For example, the initial pheromone value for the exposure energy-development time-etching temperature sequence was 0.85, for the development time-exposure energy-etching temperature sequence it was 0.65, and for the etching temperature-development time-exposure energy sequence it was 0.45. Heuristic pheromone values ​​were calculated based on parameter scores and adjustment constraints: the heuristic pheromone value for exposure energy was 3.5, for development time it was 3.2, and for etching temperature it was 2.8.

[0055] The probability of selecting the adjustment value is obtained by weighting the pheromone concentration and the heuristic information value. Taking the exposure energy parameter as an example, with a current pheromone concentration of 0.85 and a heuristic information value of 3.5, setting the pheromone weight to 0.6 and the heuristic information weight to 0.4, the probability of selecting the exposure energy parameter is 0.85 × 0.6 + 3.5 × 0.4 = 2.11. Based on this, multiple candidate adjustment sequences are generated and their performance is evaluated. During each iteration, the pheromone evaporation coefficient is gradually reduced from the initial value of 0.5 to 0.3 to ensure the convergence ability of the algorithm in the later stages.

[0056] A taboo list is formed by recording visited adjustment sequences, with an initial taboo length of 5 and a taboo period of 3 iterations. Local searches are performed on the obtained optimized adjustment sequences using swap-shift and reverse-shift techniques to generate candidate adjustment sequences. For example, for the sequence "exposure energy-development time-etching temperature", a swap-shift might generate "development time-exposure energy-etching temperature". A penalty factor is determined by combining parameter performance scores with the taboo list. If a candidate sequence is in the taboo list and does not meet the amnesty criterion, its penalty factor is set to 1.5; otherwise, it is 1.0. The amnesty criterion states that the yield improvement of the candidate sequence exceeds 108% of the current global optimum.

[0057] The initial adjustment sequence is optimized based on the selection probability of the adjustment value, and the parameter performance score of the optimized sequence is calculated. The performance score comprehensively considers parameter adjustment speed, product yield, and registration accuracy. Taking the "exposure energy-development time-etching temperature" sequence as an example, its total adjustment time is 3.5 minutes, the yield is 98.5%, the registration accuracy is ±0.3μm, and the comprehensive score is 91 points. The adjustment sequence with the highest parameter performance score is updated as the global optimal solution.

[0058] A dynamic neighborhood search optimization is performed on the global optimal solution, and an adaptive step size adjustment strategy is used to update the search direction. The initial search step size is set to 2.0. As the iteration progresses, if no better solution is found for 3 consecutive iterations, the step size is reduced to 0.8 times the original size; if a better solution is found, the step size is increased to 1.2 times the original size, but does not exceed the maximum step size of 3.0.

[0059] The process was iterated until the convergence condition was met, i.e., the global optimal solution remained unchanged for 10 consecutive iterations or the maximum number of iterations (100) was reached. After 31 iterations, convergence was achieved, and the globally optimal adjustment sequence was determined to be "exposure energy - development time - etching temperature". The parameter performance score of this sequence was 93 points, which was 4 points higher than the initial sequence. The product yield was improved by 1.2%, the registration accuracy was improved by 15%, and the adjustment time was shortened by 10%.

[0060] In this embodiment, an initial adjustment sequence is generated based on the optimal control parameters, and the adjustment constraints between adjacent parameters are calculated by combining parameter performance scores, ensuring the feasibility and rationality of the adjustment sequence. The pheromone mechanism of the ant colony algorithm is introduced to optimize the adjustment sequence. Pheromones are set using historical scheduling data, allowing the optimization process to draw on historical experience. A taboo list is established to record the adjustment sequences that have been tried, and local searches are performed using exchange moves and reverse moves, which can effectively escape local optima and find better adjustment schemes. A dynamic neighborhood search method is used to further optimize the global optimal solution, and the search efficiency is improved by adaptive step size adjustment. The parameter search strategy can be dynamically adjusted according to the actual production situation, ensuring both optimization effect and production stability.

[0061] In one alternative implementation, The production capacity matching improvement path is calculated in real time using dynamic programming methods after executing the globally optimal adjustment sequence. Based on the path matching results, the unexecuted adjustment sequences are dynamically adjusted until the pre-set production target is met, including: Calculate the capacity matching degree of the process equipment during the execution of the global optimal adjustment sequence, divide the capacity matching degree into multiple decision stages, obtain the state transition probability based on the capacity matching degree of each decision stage, calculate the state transition cost by combining the equipment adjustment cost coefficient and the capacity matching degree weight coefficient, and obtain the capacity matching degree improvement path by reverse recursion. Calculate the deviation between the current capacity status of the process equipment and the production target, obtain the adjustment amount using a pre-set proportional adjustment matrix and differential adjustment coefficient, and dynamically adjust the unexecuted adjustment sequence based on the product of the adjustment amount and the time-varying weight function; The adjustment range, adjustment rate, and capacity balance constraints of the process equipment are transformed into a barrier function with a penalty factor. The unexecuted adjustment sequence is dynamically adjusted according to the barrier function until the pre-set production target is met.

[0062] Real-time capacity data of all process equipment on the LCD panel production line is collected. Data is acquired via sensors on each process device, collected every 5 seconds, and uploaded to the central processing system. The collected data includes capacity output values ​​for key processes such as array fabrication and cell assembly; for example, the exposure process capacity is 150 panels / hour, and the developing process capacity is 142 panels / hour. Capacity matching degree is defined as the degree to which the ratio of the capacity of adjacent process equipment is close to 1. Specifically, the capacity of each process device is divided by the capacity of the downstream process device to obtain a series of matching degree values.

[0063] Using a 15-minute decision cycle, the 8-hour production shift is divided into 32 decision stages. At the end of each decision stage, the capacity matching degree of each process is recorded. For example, the matching degree of the exposure-development process in the first stage is 0.93, which improves to 0.95 in the second stage. Based on historical data analysis, a state transition probability matrix is ​​established to describe the probability of transitioning from the current matching degree state to other possible states. For example, from a matching degree of 0.93, there is a 0.70 probability of transitioning to 0.95, a 0.20 probability of transitioning to 0.91, and a 0.10 probability that remains unchanged.

[0064] The state transition cost calculation combines the equipment adjustment cost coefficient and the capacity matching degree weighting coefficient. The adjustment cost coefficient for the exposure machine is 0.85, and for the developing machine it is 0.70, reflecting the difficulty and cost of adjusting the capacity of different equipment. The capacity matching degree weighting coefficient is set to 0.8, reflecting the degree of influence of matching degree on panel yield. The state transition cost is calculated by combining the two. For example, the transition cost from a matching degree of 0.93 to 0.95 is 0.85×0.8+0.2×(1-0.95)=0.69.

[0065] Starting from the production endpoint, a backward recursion is performed. The optimal path is calculated using a dynamic programming algorithm. The target capacity matching degree is set to 0.98. From this target state, the minimum cumulative cost to reach each possible preceding state is calculated. For example, the cost to reach 0.98 from a matching degree of 0.96 is 0.65; the cost to reach 0.98 from 0.94 is 1.20. The path with the minimum cumulative cost is selected as the capacity matching degree improvement path, such as 0.93→0.95→0.96→0.98.

[0066] The deviation value is calculated by comparing the real-time capacity with the preset target capacity. For example, if the target capacity for the exposure process is 160 pieces / hour, and the actual capacity is 150 pieces / hour, the deviation is 10 pieces / hour. Similarly, if the target capacity for the developing process is 158 pieces / hour, and the actual capacity is 142 pieces / hour, the deviation is 16 pieces / hour.

[0067] A pre-set proportional adjustment matrix describes the capacity impact relationship between each process. For the exposure-development process, the proportional adjustment matrix is ​​set to [[1.0, 0.4], [0.3, 1.0]], indicating that the impact coefficient of adjusting 1 unit in the exposure process on the development process is 0.4. The differential adjustment coefficient is set to 0.45, representing the sensitivity to the rate of change of deviation. The calculated adjustment amount for the exposure process is 12 pieces / hour, and the adjustment amount for the development process is 18 pieces / hour.

[0068] The time-varying weight function adopts an exponential decay form, with an initial value of 1.0, decaying by 4% in each decision cycle to smooth the adjustment process and avoid product quality fluctuations. The dynamic adjustment sequence uses the product of the adjustment amount and the time-varying weight function as the actual adjustment command executed. When the current time-varying weight is 0.88, the actual adjustment amount for the exposure process is 12 × 0.88 = 10.56 pieces / hour.

[0069] The adjustment constraints of the process equipment are achieved through a barrier function. The capacity range for the exposure process is 120-180 pieces / hour, and for the developing process, it is 110-170 pieces / hour. Outside these ranges, the barrier function value increases sharply. Adjustment rate constraints limit the maximum adjustment range within 15 minutes; for example, the maximum adjustment range for the exposure process is 15 pieces / hour. Capacity balance constraints ensure that the capacity difference between adjacent processes does not exceed 12 pieces / hour.

[0070] The barrier function is logarithmic, and its value increases rapidly as the constraint approaches the boundary. A penalty factor of 12 is set to effectively eliminate constraint-violating adjustments during the optimization process. Unexecuted adjustment sequences are corrected based on the barrier function; for example, the original plan to adjust the exposure process to 185 pieces / hour is corrected to 178 pieces / hour because it exceeds the upper limit.

[0071] In this embodiment, a dynamic programming framework is established by dividing the capacity matching degree into multiple decision stages and calculating the state transition probability. This makes the parameter adjustment process of the LCD panel production line more systematic and controllable. By introducing the equipment adjustment cost coefficient and the capacity matching degree weight coefficient, a comprehensive consideration of adjustment cost and capacity balance is achieved. This can minimize the production fluctuations caused by parameter adjustment while ensuring product quality. It is particularly suitable for products with extremely high requirements for process stability. The combined control strategy of proportional adjustment matrix and differential adjustment coefficient provides a more accurate method for calculating adjustment amount, making the adjustment of key parameters such as exposure energy and development time more precise and avoiding the problems of over-adjustment or under-adjustment.

[0072] Figure 3 This is an optimization diagram of the adjustment trajectory under the obstacle function constraint of the intelligent scheduling and capacity balancing optimization method for LCD panel production lines according to an embodiment of the present invention. It shows the trajectory optimization effect of different adjustment methods under the obstacle function constraint.

[0073] In the figure, the horizontal axis represents the adjustment range of the LCD panel production line equipment (unit: panels / hour), the vertical axis represents the barrier function value, and the shaded area represents the constraint boundary area (adjustment range exceeding 15 panels / hour). The barrier function designed in this technical solution exhibits a significant "sharp rise" characteristic when the adjustment range approaches the constraint boundary of 15 panels / hour. The barrier function value rapidly rises from 14.2 at an adjustment range of 14.2 panels / hour to 21.5 at 14.8 panels / hour, and then sharply climbs to 37.6 when approaching the 15 panels / hour boundary, forming a mathematical "energy barrier". This barrier can accurately sense the boundary and adjust the strategy in time when approaching the constraint conditions, ensuring that the adjustment range is strictly controlled within a safe range.

[0074] In contrast, the theoretically optimal trajectory (an ideal state without considering actual equipment constraints) has a barrier function value of only 26.6 when the adjustment range reaches 20 pieces / hour, and even at 30 pieces / hour, it only reaches 38.7. This inevitably leads to equipment overload and product quality fluctuations in actual production. Traditional penalty factor adjustment methods lack sensitivity at constraint boundaries; their barrier function is only 23.1 when the adjustment range is 15 pieces / hour, and only rises to 35.9 after the adjustment range exceeds the constraint to 21 pieces / hour, failing to effectively prevent constraint violations. This technical solution achieves a precise mathematical description and efficient implementation of the constraint conditions through a logarithmic barrier function (with a penalty factor set to 12), ensuring the safe operation of production equipment while improving capacity matching to the optimal level.

[0075] In one alternative implementation, Real-time monitoring of the operating status of each process equipment; when equipment malfunction or capacity imbalance is detected, alarm signals are triggered and emergency dispatch instructions are generated, including: Real-time acquisition of operational status data of process equipment; calculation of capacity balance of each process equipment and probability of equipment malfunction based on operational status data. When the capacity balance is lower than a preset threshold or the probability of equipment malfunction is higher than a preset threshold, an alarm signal is sent to the control system. Based on the current operating status of the process equipment and historical scheduling experience, an emergency scheduling instruction sequence is generated, which includes equipment power adjustment instructions and equipment start-up and shutdown sequence.

[0076] A sensor network is installed on various process equipment on the LCD panel production line to collect operational status data. The sensor network includes temperature sensors, light intensity sensors, air pressure sensors, and speed sensors, which collect multi-dimensional data such as temperature changes, light intensity stability, air pressure fluctuations, and transmission speed. These sensors are connected to a central monitoring system via an industrial fieldbus for real-time data transmission. The data acquisition frequency is set according to the importance of the process: 50ms / time for critical equipment such as exposure machines and developing machines, 200ms / time for general processes such as etching and cleaning, and 500ms / time for inspection processes.

[0077] Capacity balance calculation is based on the ratio of actual capacity to designed capacity for each process. In practical applications, actual capacity data for each process on the panel production line is collected, such as an exposure machine with a capacity of 155 pieces / hour, a developing machine with 148 pieces / hour, and an etching machine with 160 pieces / hour. Capacity balance is calculated based on the capacity ratio of adjacent processes, specifically by dividing the minimum capacity of the preceding and following processes by the maximum capacity. In the example above, the capacity balance between the exposure machine and the developing machine is 148 / 155 = 0.955, and the capacity balance between the developing machine and the etching machine is 148 / 160 = 0.925. The capacity balance warning threshold is set at 0.95; values ​​below this are considered a capacity imbalance.

[0078] The probability of equipment malfunction is calculated by the deviation of real-time operating parameters from the normal parameter range. A normal value model for operating parameters is established for each process, including an exposure machine light intensity stability of ±2%, a developing machine temperature range of 23℃±1℃, an etching solution concentration fluctuation range of ±3%, and a substrate conveying speed deviation of no more than ±5mm / s. When the real-time collected operating parameters deviate from the normal range, the degree of malfunction is calculated and converted into an malfunction probability. For example, when the developing machine temperature reaches 24.5℃, exceeding the normal upper limit by 0.5℃, the temperature malfunction degree is 50%; when the developing solution concentration deviation reaches 5%, the concentration malfunction degree is 80%. Considering the malfunction degrees of all parameters, the overall probability of equipment malfunction is 65%. The equipment malfunction probability warning threshold is set at 60%.

[0079] The alarm signal triggering system employs a multi-level early warning mechanism. When the capacity balance is below 0.95 or the probability of equipment malfunction exceeds 60%, a yellow warning message is displayed on the monitoring interface. When the capacity balance is below 0.9 or the probability of malfunction exceeds 80%, a red alarm is displayed and an alarm message is sent to the on-site engineer. When the capacity balance is below 0.85 or the probability of malfunction exceeds 90%, an audible and visual alarm is triggered, and alarm messages are sent to the engineer and supervisor simultaneously. For example, when the capacity balance between the developing machine and the etching equipment is detected to drop to 0.87, a red alarm is triggered, and an alarm message is sent to the engineer's mobile phone stating, "The capacity of the developing machine and the etching equipment is severely mismatched; it is recommended to adjust the developing machine parameters or activate the backup developing machine."

[0080] Emergency dispatch instructions are generated based on the current operating status of process equipment and a historical dispatch experience database. A database containing 15,000 historical dispatch cases is maintained, recording dispatch strategies and effectiveness scores under different abnormal conditions. Emergency dispatch instructions include equipment parameter adjustment instructions and equipment start-up / shutdown sequences. Parameter adjustment instructions specify in detail the process parameters that need to be adjusted for each piece of equipment, such as "adjust the developer temperature to 22.5℃, increase the developer concentration by 2%; reduce the conveyor speed to 380mm / s," etc. Equipment start-up / shutdown sequences specify the precise order and time intervals for equipment switching.

[0081] For example, when the developer temperature reaches 24.8℃ (exceeding the normal upper limit by 1.8℃) and the developer concentration deviation reaches 6%, the probability of equipment malfunction is calculated to be 85%, triggering a red alarm. Simultaneously, the actual capacity of the developer drops to 135 pieces / h, and the balance with the exposure machine's capacity of 155 pieces / h drops to 0.87. Based on the abnormal state analysis and historical experience, an emergency dispatch command is generated: "1. Immediately cool the developer to a target temperature of 22.5℃, and increase the cooling system power to 95%; 2. Adjust the developer concentration online, and replenish 2.5L of stock solution; 3. Temporarily reduce the substrate conveying speed to 350mm / s; 4. Synchronously adjust the exposure machine parameters, reducing the energy to 270mJ / cm²; 5. Prepare a backup developer; if the main developer parameters remain abnormal, execute the equipment switching procedure; 6. Equipment switching sequence: Backup developer preheats to operating temperature → Backup developer starts and stabilizes → Main developer decelerates by 30% → Switch substrate conveying path → Completely stop the main developer → Enter maintenance procedure." In this embodiment, a dual early warning mechanism for capacity balance and anomaly probability is established by real-time monitoring of the operating status of the LCD panel production line process equipment. This mechanism can promptly detect capacity imbalances or equipment malfunctions, avoiding batch product quality issues. An alarm mechanism based on preset thresholds enables early warning of production anomalies, allowing sufficient response time for process adjustments. Historical scheduling experience is introduced to guide the generation of emergency scheduling instructions, improving the accuracy and effectiveness of emergency response. The combined control strategy of equipment power adjustment instructions and start-stop sequence ensures process stability during emergency scheduling, enabling the adjustment of exposure machine power or the implementation of equipment switching according to a reasonable sequence, avoiding large-scale production line shutdowns.

[0082] A second aspect of this invention provides an intelligent scheduling and capacity balancing optimization system for a liquid crystal panel production line, comprising: The first unit is used to collect production status data of each process equipment in the LCD panel production line in real time and perform digital processing, and calculate the capacity matching degree of array process and cell assembly process in combination with the material flow process of the production line. The second unit is used to determine the optimization target of the process equipment based on the capacity matching degree. It uses the time difference algorithm to evaluate the state of the process equipment, updates the operation control parameters of each process equipment by calculating the deviation between the equipment operation state and the target state, until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The candidate control parameters are then evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. The third unit is used to calculate the process equipment adjustment sequence based on the ant colony optimization algorithm and the optimal control parameters and perform sequence optimization, generate an optimized adjustment sequence and perform dynamic neighborhood search optimization to obtain the global optimal adjustment sequence, calculate the capacity matching improvement path after executing the global optimal adjustment sequence in real time through dynamic programming method, and dynamically adjust the unexecuted adjustment sequence in combination with the path matching result until the pre-set production target is met. The fourth unit is used to monitor the operating status of each process equipment in real time. When an equipment abnormality or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

[0083] A third aspect of the present invention provides an electronic device, comprising: A processor and a memory for storing processor-executable instructions, wherein the processor is configured to invoke instructions stored in the memory to perform the aforementioned method.

[0084] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0085] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent scheduling and capacity balancing optimization of LCD panel production lines, characterized in that, include: Real-time collection and digital processing of production status data of various process equipment in the LCD panel production line; combined with the material flow process of the production line to calculate the capacity matching degree of array process and cell assembly process. The optimization target of the process equipment is determined based on the capacity matching degree. The state evaluation of the process equipment is carried out by the time difference algorithm. The operation control parameters of each process equipment are updated by calculating the deviation between the equipment operation state and the target state until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The performance of the candidate control parameters is evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is generated and optimized by dynamic neighborhood search to obtain the global optimal adjustment sequence. The production capacity matching improvement path after executing the global optimal adjustment sequence is calculated in real time by dynamic programming method. The unexecuted adjustment sequence is dynamically adjusted according to the path matching result until the pre-set production target is met. The system monitors the operating status of each process equipment in real time. When an equipment malfunction or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

2. The method according to claim 1, characterized in that, Real-time acquisition and digital processing of production status data from various process equipment in the LCD panel production line; calculation of capacity matching between array manufacturing and cell assembly processes based on material flow process of the production line, including: Real-time acquisition of operating status data of various process equipment in the LCD panel production line, including processing time, number of workpieces, and equipment fault information; The operating status data is reorganized according to the material transfer sequence of the process equipment to establish the processing time relationship between process equipment and calculate the material flow status between adjacent process equipment in real time. Based on the material flow status of the process equipment, the equipment utilization rate, work-in-process quantity and processing cycle of the array process and the box process are calculated, and the parameters are weighted and summed to obtain the capacity matching degree between the processes.

3. The method according to claim 1, characterized in that, The optimization objectives for process equipment are determined based on the capacity matching degree. A time-difference algorithm is used to evaluate the state of the process equipment. The operating control parameters of each process equipment are updated by calculating the deviation between the equipment's operating state and the target state, until the preset maximum number of iterations is reached or convergence is achieved. Candidate control parameters include: The optimization target of the process equipment is determined based on the capacity matching degree, and the deviation between the capacity matching degree and the preset target state is weighted and summed to obtain the state evaluation result. The time-difference algorithm is used to evaluate the status of process equipment. The equipment operating parameters and material inventory information of the process equipment are collected to obtain the equipment operating status. Equipment rate adjustment parameters and batch size adjustment parameters are generated as control actions. The instant reward value is calculated based on the deviation between the equipment operating status and the target status, and the current operating status of the process equipment is determined based on the instant reward value. The gradient of the deviation between the operating state and the target state of the calculated equipment with respect to the operating control parameters is used to update the operating control parameters of each process equipment along the negative direction of the gradient. Determine whether the number of iterations has reached the preset maximum number of iterations or whether the deviation change of consecutive preset number of iterations is less than the preset threshold. When either condition is met, the current operation control parameter is determined as a candidate control parameter.

4. The method according to claim 1, characterized in that, The performance of candidate control parameters is evaluated using deep reinforcement learning algorithms, and the optimal control parameters are determined as follows: Construct a causal influence propagation diagram among candidate control parameters, calculate the direct change derivative and indirect transmission derivative between different control parameters to obtain the causal influence intensity matrix, calculate the parameter correlation degree based on the causal influence intensity matrix, and determine the parameters with a correlation degree greater than a preset threshold as the main control parameters; Collect the operating status data of the control system, calculate the performance evaluation index based on the causal influence intensity matrix to obtain the parameter performance score, and prioritize the main control parameters according to the parameter performance score; The main control parameters are subjected to disturbances. The change transmission path is predicted by the causal influence intensity matrix. The parameter sensitivity value is calculated as the parameter evaluation weight. The independent influence coefficient and the synergistic influence coefficient are calculated based on the causal influence intensity matrix. The independent effect score and the synergistic effect score are obtained by multiplying them with the parameter evaluation weight. The parameter comprehensive score is obtained by weighted summing of the independent action score and the synergistic action score. The parameter combination with the highest comprehensive score is selected as the current optimal parameter combination. The parameter exploration frequency is adjusted according to the changing trend of the comprehensive score until the comprehensive score converges and the optimal control parameter is determined.

5. The method according to claim 1, characterized in that, Based on the ant colony optimization algorithm and the optimal control parameters, the process equipment adjustment sequence is calculated and optimized. An optimized adjustment sequence is then generated and optimized using dynamic neighborhood search to obtain the globally optimal adjustment sequence, including: An initial adjustment sequence is generated using the optimal control parameters, and the adjustment constraints between adjacent parameters are calculated based on the parameter performance scores of the optimal control parameters to obtain the initial sequence priority. Based on the pre-acquired historical scheduling data and equipment operating characteristics, the initial value of the pheromone volatilization coefficient and the pheromone increment update rule are set. The pheromone concentration and heuristic information value between different control parameters are calculated to obtain the adjustment value selection probability. The adjustment value selection probability is used to generate an optimized adjustment sequence. The pheromone concentration is updated based on the pheromone volatilization coefficient and the pheromone increment. The visited regulation sequences are recorded to form a tabu list. The optimized regulation sequence is searched locally using exchange and reverse shifts to obtain candidate regulation sequences. The penalty factor of the candidate regulation sequence is determined by combining the parameter performance score and the tabu list. The initial regulation sequence is optimized according to the regulation value selection probability to obtain the optimized regulation sequence. The parameter performance score of the optimized regulation sequence is calculated. The regulation sequence with the highest parameter performance score is updated as the global optimum. The global optimal solution is optimized by dynamic neighborhood search. The search direction is updated based on adaptive step size. The target value is calculated by combining parameter performance score and penalty factor. The tabu list and the global optimal solution are updated according to the target value. The process is iterated until convergence to obtain the global optimal adjustment sequence.

6. The method according to claim 1, characterized in that, The production capacity matching improvement path is calculated in real time using dynamic programming methods after executing the globally optimal adjustment sequence. Based on the path matching results, the unexecuted adjustment sequences are dynamically adjusted until the pre-set production target is met, including: Calculate the capacity matching degree of the process equipment during the execution of the global optimal adjustment sequence, divide the capacity matching degree into multiple decision stages, obtain the state transition probability based on the capacity matching degree of each decision stage, calculate the state transition cost by combining the equipment adjustment cost coefficient and the capacity matching degree weight coefficient, and obtain the capacity matching degree improvement path by reverse recursion. Calculate the deviation between the current capacity status of the process equipment and the production target, obtain the adjustment amount using a pre-set proportional adjustment matrix and differential adjustment coefficient, and dynamically adjust the unexecuted adjustment sequence based on the product of the adjustment amount and the time-varying weight function; The adjustment range, adjustment rate, and capacity balance constraints of the process equipment are transformed into a barrier function with a penalty factor. The unexecuted adjustment sequence is dynamically adjusted according to the barrier function until the pre-set production target is met.

7. The method according to claim 1, characterized in that, Real-time monitoring of the operating status of each process equipment; when equipment malfunction or capacity imbalance is detected, alarm signals are triggered and emergency dispatch instructions are generated, including: Real-time acquisition of operational status data of process equipment; calculation of capacity balance of each process equipment and probability of equipment malfunction based on operational status data. When the capacity balance is lower than a preset threshold or the probability of equipment malfunction is higher than a preset threshold, an alarm signal is sent to the control system. Based on the current operating status of the process equipment and historical scheduling experience, an emergency scheduling instruction sequence is generated, which includes equipment power adjustment instructions and equipment start-up and shutdown sequence.

8. A smart scheduling and capacity balancing optimization system for a liquid crystal panel production line, used to implement the method described in any one of claims 1-7, characterized in that, include: The first unit is used to collect production status data of each process equipment in the LCD panel production line in real time and perform digital processing, and calculate the capacity matching degree of array process and cell assembly process in combination with the material flow process of the production line. The second unit is used to determine the optimization target of the process equipment based on the capacity matching degree. It uses the time difference algorithm to evaluate the state of the process equipment, updates the operation control parameters of each process equipment by calculating the deviation between the equipment operation state and the target state, until the preset maximum number of iterations is reached or the candidate control parameters are obtained through convergence. The candidate control parameters are then evaluated by the deep reinforcement learning algorithm to determine the optimal control parameters. The third unit is used to calculate the process equipment adjustment sequence based on the ant colony optimization algorithm and the optimal control parameters and perform sequence optimization, generate an optimized adjustment sequence and perform dynamic neighborhood search optimization to obtain the global optimal adjustment sequence, calculate the capacity matching improvement path after executing the global optimal adjustment sequence in real time through dynamic programming method, and dynamically adjust the unexecuted adjustment sequence in combination with the path matching result until the pre-set production target is met. The fourth unit is used to monitor the operating status of each process equipment in real time. When an equipment abnormality or production capacity imbalance is detected, an alarm signal is triggered and an emergency dispatch command is generated.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images