Intelligent control system and control method for capsule production line

Abstract

The invention discloses an intelligent control system for a capsule production line and a control method thereof, the system comprises a mechanical module, a control module, a sensor module and a scheduling module, and the scheduling module combines a rule priority scheduling algorithm, a dynamic allocation scheduling algorithm, a deep learning algorithm and a reinforcement learning algorithm. And task priority calculation, task allocation optimization and process parameter dynamic adjustment are realized. The system determines the task priority through a priority calculation formula # imgabs0 #, optimizes the task allocation proportion and operation parameters in combination with a reinforcement learning algorithm, and forms a real-time feedback closed-loop control mechanism. According to the invention, the production efficiency, the product quality consistency and the system intelligence level of the capsule production line are obviously improved, and meanwhile, the energy consumption and the maintenance cost are reduced. The method is suitable for the field of capsule manufacturing with high requirements for production efficiency and quality.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing and industrial automation, and in particular to an intelligent control system and a control method for a capsule production line. Background Art

[0002] As a common pharmaceutical dosage form, the production process of capsules involves multiple processes such as mixing, dipping, drying, demoulding, cutting, fitting and collecting. These processes have strict requirements on production parameters (such as glue viscosity, drying temperature, wind pressure) and product quality (such as particle weight difference and wall thickness uniformity). However, in the existing technology, capsule production lines still face the following challenges: Low production efficiency: The scheduling mechanism of existing capsule production lines mostly relies on fixed rules and cannot dynamically adjust the task sequence and resource allocation according to the real-time production status, which easily leads to production bottlenecks and uneven utilization of module resources, reducing overall efficiency.

[0003] Product quality control is difficult: Differences in particle weight and uneven wall thickness are key issues affecting capsule quality. Existing technologies usually rely on subsequent testing to detect quality problems and lack real-time monitoring and adjustment mechanisms, resulting in a high defective rate.

[0004] The system lacks intelligence: Current production scheduling and parameter adjustments mainly rely on manual experience, with slow response speed and limited accuracy, making it difficult to adapt to complex dynamic environments such as order fluctuations, process changes or equipment abnormalities.

[0005] High energy consumption and maintenance costs: Due to the lack of real-time monitoring of module loads and task allocation optimization, some equipment is in an overloaded or inefficient state for a long time, resulting in increased energy consumption and equipment wear.

[0006] Although some studies have attempted to improve production efficiency by introducing simple automated control algorithms (such as rule-based scheduling), these solutions are usually only applicable to static environments and lack adaptive capabilities and quality feedback mechanisms. Therefore, the existing technology urgently needs a more intelligent capsule production control system that can combine real-time data and historical production data to dynamically optimize production scheduling, adjust process parameters, and improve product quality and production efficiency in closed-loop control. Summary of the invention

[0007] The present invention provides an intelligent control system and control method for a capsule production line, aiming to solve the problems of low production efficiency, difficult product quality control, insufficient intelligence level and high energy consumption in the existing capsule production line. The present invention combines rule priority scheduling algorithm, dynamic allocation scheduling algorithm, deep learning algorithm and reinforcement learning algorithm to construct a set of efficient and intelligent production scheduling and control systems to achieve dynamic optimization and real-time quality control of the capsule shell production process.

[0008] The technical solution of the capsule production line intelligent control system of the present invention is as follows: Mechanical module: including automatic mold pin oiling section, automatic glue mixing and dipping system section, automatic capsule mold conveying and drying section, automatic mold platform pushing section, automatic capsule demoulding, cutting and fitting section, and automatic capsule collection section outside the machine. Each section is interconnected and coordinated. Control module: including industrial Ethernet interconnected operation control system, PLC controller, touch screen operation terminal and multiple electrical components, forming an intelligent control hardware system; Sensor module: including viscosity sensor, temperature sensor, wind pressure sensor and quality detection module, used to collect real-time product quality data such as glue viscosity, drying kiln temperature, wind pressure, particle weight difference, wall thickness uniformity, etc. Scheduling module: Based on rule-based priority scheduling algorithm, dynamic allocation scheduling algorithm, deep learning algorithm and reinforcement learning algorithm, it dynamically optimizes production line task scheduling by collecting load status, task status and quality inspection data of each module in real time, including: Priority calculation: The task priority is calculated using the following formula: ; Among them, the weight parameter , , Dynamic optimization through deep learning models, with inputs including historical production data and real-time feedback data; : The priority of task i, used to sort tasks; : The timeliness of the task (e.g., whether it is approaching the deadline); : The importance of the task (such as whether it is a critical process); : The load status of the module (such as whether the drying kiln is close to full load).

[0009] Task allocation optimization: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following task allocation formula:

[0010] ; in, Dynamic evaluation through reinforcement learning model, combined with task priority and module load status optimization; : The proportion of tasks assigned to module j; : Current production line status, including module load, number of tasks, etc.

[0011] : Assign tasks to module j; :state Next, action The value of is dynamically optimized by the reinforcement learning model: Real-time feedback and closed-loop control: Use product quality test data (grain weight difference, wall thickness uniformity) as reinforcement learning reward signals , update the Q value and optimize the scheduling strategy through the following formula:

[0012] ) ; Through the synergistic effect of the modules, efficient operation of task scheduling of the capsule production line and real-time optimization of product quality are achieved.

[0013] Preferably, the weight optimization target of the rule-first scheduling algorithm is achieved by the following formula: ; in: w=[α,β,γ] represents the priority weight vector; Calculate the result for the current priority level; It is a target priority, generated based on actual production performance.

[0014] Preferably, the reward function of the reinforcement learning algorithm is Combined with the following formula: ; in: , , is a weight parameter used to balance product quality and production efficiency; The quality parameter in the reward function comes from real-time detection data.

[0015] Preferably, the dynamic task allocation algorithm adjusts the module task ratio by the following formula: ; in: represents the real-time available resources of module j; represents the task allocation ratio of module j; Indicates the total number of modules in the production line.

[0016] The control method of the intelligent control system for a capsule production line of the present invention comprises the following steps: A. Data collection: The sensor module collects the task status of each process, module load data and product quality inspection data in real time; B. Priority calculation: The task priority is calculated using the rule-based priority scheduling algorithm according to the following formula:

[0017] Among them, the weight parameters α, β, and γ are dynamically optimized by the deep learning model; C. Dynamic task allocation: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following formula: ; D. Real-time feedback: Optimizing the reward function through quality inspection data , and update the reinforcement learning model: ; E. Parameter adjustment: Dynamically adjust the production line operating parameters according to the optimization results, including conveying speed, drying temperature and glue viscosity.

[0018] Through the above steps, dynamic scheduling of the production line and closed-loop optimization of product quality can be achieved.

[0019] Technical Effects

[0020] The present invention constructs a set of efficient and intelligent capsule production line control system through the organic combination of rule priority scheduling algorithm, dynamic allocation scheduling algorithm, deep learning algorithm and reinforcement learning algorithm, which can significantly improve production efficiency, optimize product quality and consistency, and reduce energy consumption and maintenance costs. The specific technical effects are as follows: 1. Improve production efficiency Priority scheduling: Through the rule-based priority scheduling algorithm, the priority of tasks is dynamically calculated to ensure that key tasks (such as drying and demoulding) are processed first and reduce the waiting time between processes.

[0021] Dynamic task allocation: Combined with the dynamic allocation scheduling algorithm, tasks are reasonably allocated to modules with lower loads, eliminating production bottlenecks and balancing resource utilization.

[0022] Actual effect: The overall efficiency of the production line is increased by 10%~20%, and the cycle time is significantly shortened.

[0023] 2. Optimize product quality Real-time monitoring and closed-loop control: The sensor module monitors key quality parameters such as particle weight difference and wall thickness uniformity of the product in real time, and uses the detection results as input to the reinforcement learning algorithm to dynamically adjust process parameters (such as drying temperature, viscosity, and conveying speed).

[0024] Adaptive adjustment: The system can reduce the defective rate and ensure the stability of product quality through real-time feedback closed-loop control.

[0025] Actual effect: Particle weight variation and wall thickness non-uniformity are reduced by 15% to 30%, and product consistency is significantly enhanced.

[0026] 3. Improve the level of system intelligence Dynamic scheduling rule optimization: Use deep learning algorithms to optimize the weight parameters in the scheduling rules (such as the weight of task timeliness, importance, and module load status) in real time to ensure that the scheduling logic can adapt to changes in production line status.

[0027] Reinforcement learning strategy optimization: Continuously optimize the task allocation strategy through the reward mechanism (combined with indicators such as quality and latency) to enhance the system's adaptability in a dynamic production environment.

[0028] Actual effect: The system response speed is increased by 30%~50%, and it can quickly adapt to complex dynamic scenarios such as order fluctuations and equipment abnormalities.

[0029] 4. Reduce energy consumption and maintenance costs Load balancing: Through a dynamic allocation algorithm, tasks are assigned to modules with lower loads to prevent some modules from overloading or being idle inefficiently.

[0030] Parameter optimization and energy-saving control: Dynamically adjust the module operation speed and process parameters (such as drying kiln air pressure and conveying speed) according to the actual task volume to reduce the energy consumption and wear of the equipment.

[0031] Actual effect: System energy consumption is reduced by 10%~15%, equipment service life is extended, and maintenance costs are significantly reduced.

[0032] 5. Promote visualization and automation of production processes Visual monitoring: Through industrial Ethernet and touch screen terminals, the priority of production tasks, task allocation ratio and product quality status are displayed in real time, realizing comprehensive visual management of the production line.

[0033] Automated control: From task allocation to parameter optimization, the system operates automatically throughout the process without manual intervention, reducing human operation errors.

[0034] Actual effect: The frequency of manual intervention is reduced by 50%, and operational efficiency is significantly improved.

[0035] Comprehensive technical advantages The technical solution of the present invention significantly improves the overall performance of the capsule production line, which is specifically reflected in the following aspects: Improved production efficiency: process coordination is smoother and bottleneck problems are reduced.

[0036] Product quality optimization: product consistency is stronger and the defective rate is significantly reduced.

[0037] Enhanced system flexibility: Scheduling rules and allocation strategies are adaptive and can respond quickly to complex dynamic environments.

[0038] Energy saving and consumption reduction: higher resource utilization, lower energy consumption and equipment loss.

[0039] Improved automation level: improved operational efficiency and reduced manual intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a structural block diagram of the present invention.

[0041] Figure 2 It is a step diagram of the operating method of the present invention. DETAILED DESCRIPTION

[0042] Embodiment 1: A technical solution for an intelligent control system for a capsule production line is as follows: Mechanical module: including automatic mold pin oiling section, automatic glue mixing and dipping system section, automatic capsule mold conveying and drying section, automatic mold platform pushing section, automatic capsule demoulding, cutting and fitting section, and automatic capsule collection section outside the machine. Each section is interconnected and coordinated. Control module: including industrial Ethernet interconnected operation control system, PLC controller, touch screen operation terminal and multiple electrical components, forming an intelligent control hardware system; Sensor module: including viscosity sensor, temperature sensor, wind pressure sensor and quality detection module, used to collect real-time product quality data such as glue viscosity, drying kiln temperature, wind pressure, particle weight difference, wall thickness uniformity, etc. Scheduling module: Based on rule-based priority scheduling algorithm, dynamic allocation scheduling algorithm, deep learning algorithm and reinforcement learning algorithm, it dynamically optimizes production line task scheduling by collecting load status, task status and quality inspection data of each module in real time, including: Priority calculation: The task priority is calculated using the following formula: ; Among them, the weight parameter , , Dynamic optimization through deep learning models, with inputs including historical production data and real-time feedback data; : The priority of task i, used to sort tasks; : The timeliness of the task (e.g., whether it is approaching the deadline); : The importance of the task (such as whether it is a critical process); : The load status of the module (such as whether the drying kiln is close to full load).

[0043] Task allocation optimization: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following task allocation formula:

[0044] ; in, Dynamic evaluation through reinforcement learning model, combined with task priority and module load status optimization; : The proportion of tasks assigned to module j; : Current production line status, including module load, number of tasks, etc.

[0045] : Assign tasks to module j; :state Next, action The value of is dynamically optimized by the reinforcement learning model; Real-time feedback and closed-loop control: Use product quality test data (grain weight difference, wall thickness uniformity) as reinforcement learning reward signals , update the Q value and optimize the scheduling strategy through the following formula:

[0046] ; Through the synergistic effect of the modules, efficient operation of task scheduling of the capsule production line and real-time optimization of product quality are achieved.

[0047] Embodiment 2: The difference between this embodiment and embodiment 1 is that the weight optimization target of the rule priority scheduling algorithm described in this embodiment is achieved by the following formula: ; in: w=[α,β,γ] represents the priority weight vector; Calculate the result for the current priority level; It is a target priority, generated based on actual production performance.

[0048] Preferably, the reward function of the reinforcement learning algorithm is Combined with the following formula: ; in: , , is a weight parameter used to balance product quality and production efficiency; The quality parameter in the reward function comes from real-time detection data.

[0049] Preferably, the dynamic task allocation algorithm adjusts the module task ratio by the following formula:

[0050] in: represents the real-time available resources of module j; represents the task allocation ratio of module j; Indicates the total number of modules in the production line.

[0051] Embodiment 3: The control method of the intelligent control system for a capsule production line of the present invention comprises the following steps: A. Data collection: The sensor module collects the task status of each process, module load data and product quality inspection data in real time; B. Priority calculation: The task priority is calculated using the rule-based priority scheduling algorithm according to the following formula:

[0052] Among them, the weight parameters α, β, and γ are dynamically optimized by the deep learning model; C. Dynamic task allocation: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following formula: ; D. Real-time feedback: Optimizing the reward function through quality inspection data , and update the reinforcement learning model: ; E. Parameter adjustment: Dynamically adjust the production line operating parameters according to the optimization results, including conveying speed, drying temperature and glue viscosity.

[0053] The details are as follows: 1. Data Collection

[0054] The production line collects the following data in real time: 1.1 Task status:

[0055] The timeliness of the current process : The remaining time of the dipping process is 10 seconds, which is highly time-effective. =0.9W.

[0056] The importance of process :The drying process is a key process, with high importance and value =0.8.

[0057] 1.2 Module load status :

[0058] The load of drying kiln 1 is 80%, which is a high state.

[0059] The load of drying kiln 2 is 60%, which is a low state.

[0060] Example value: Assigning the load status of the drying kiln module =0.6.

[0061] 1.3 Product quality data: Grain weight difference: The current grain weight difference is 0.05 g, which affects the mass term in the reward function.

[0062] Wall thickness unevenness: The current wall thickness unevenness is 3%, which affects the uniformity term in the reward function.

[0063] 2. Priority calculation The system calculates the task priority based on the rule-based priority scheduling algorithm, using the following formula: ; Input parameters: · =0.9.

[0064] · =0.8.

[0065] · =0.6.

[0066] · , , ,: The result after weight optimization, assigning α=0.4, β=0.4, γ=0.2.

[0067] Calculation results: Priority of dipping process: =0.4⋅0.9+0.4⋅0.8+0.2⋅0.6=0.78; Drying process priority: =0.4⋅0.7+0.4⋅0.9+0.2⋅0.5=0.74; Result: Dipping process priority Higher than drying process .

[0068] 3. Dynamic task allocation The system calculates the task allocation ratio based on the priority results and module load using the following formula: ; Input parameters: Task value of drying kiln 1: collecting load status =0.6, calculate the task value =0.6.

[0069] Task value of drying kiln 2: collecting load status =0.8, calculate the task value =0.8.

[0070] Calculation results: Task distribution ratio of drying kiln 1:

[0071] 0.43; Task distribution ratio of drying kiln 2:

[0072] 0.57; Results: 43% of the tasks were assigned to Dry Kiln 1 and 57% of the tasks were assigned to Dry Kiln 2.

[0073] 4. Real-time feedback The system optimizes the task allocation strategy based on the reward function of reinforcement learning and the product quality inspection results: ; Input parameters: Grain weight difference: The collected data is 0.05 g, substitute 1-grain weight difference = 1-0.05 = 0.95.

[0074] Wall thickness unevenness: The collected data is 3%, substitute 1-wall thickness unevenness = 1-0.03 = 0.97.

[0075] Task delay: Data collection is 5 seconds and is directly substituted.

[0076] Reward weights (λ1,λ2,λ3): Set λ1=0.5, λ2=0.3, λ3=0.2.

[0077] Calculation results: =0.5⋅0.95+0.3⋅0.97−0.2⋅5=0.475+0.291−1=−0.234, Result: The reward value is low, and the system determines that the current strategy needs to be optimized.

[0078] 5. Parameter Adjustment According to the reward results and scheduling strategy optimization, the system adjusts the operating parameters: Drying kiln temperature: increased from 90°C to 95°C.

[0079] Glue viscosity: reduced from 200 mPa·s to 180 mPa·s.

[0080] Mould conveying speed: increased from 0.8 m / s to 1.0 m / s.

[0081] Implementation Results After the adjustment, the operation of the production line has improved significantly: 1. Difference in grain weight: reduced from 0.05 g to 0.03 g; 2. Wall thickness unevenness: reduced from 3% to 2%; 3. Production cycle time: reduced from 60 seconds to 55 seconds; 4. Defective rate: decreased by 15%.

Claims

1. An intelligent control system for a capsule production line, comprising: Mechanical module: including automatic mold pin oiling section, automatic glue mixing and dipping system section, automatic capsule mold conveying and drying section, automatic mold platform pushing section, automatic capsule demoulding, cutting and fitting section, and automatic capsule collection section outside the machine. Each section is interconnected and coordinated. Control module: including industrial Ethernet interconnected operation control system, PLC controller, touch screen operation terminal and multiple electrical components, forming an intelligent control hardware system; Sensor module: including viscosity sensor, temperature sensor, wind pressure sensor and quality detection module, used to collect real-time product quality data such as glue viscosity, drying kiln temperature, wind pressure, particle weight difference, wall thickness uniformity, etc. Scheduling module: Based on rule-based priority scheduling algorithm, dynamic allocation scheduling algorithm, deep learning algorithm and reinforcement learning algorithm, it dynamically optimizes production line task scheduling by collecting load status, task status and quality inspection data of each module in real time, including: Priority calculation: The task priority is calculated using the following formula: ； Among them, the weight parameter , , Dynamic optimization through deep learning models, with inputs including historical production data and real-time feedback data; : The priority of task i, used to sort tasks; : The timeliness of the task (e.g., whether it is approaching the deadline); : The importance of the task (such as whether it is a critical process); : The load status of the module (e.g. whether the drying kiln is close to full load) Task allocation optimization: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following task allocation formula: ； in, Dynamic evaluation through reinforcement learning model, combined with task priority and module load status optimization; : The proportion of tasks assigned to module j; : Current production line status, including module load, number of tasks, etc.; : Assign tasks to module j; :state Next, action The value of is dynamically optimized by the reinforcement learning model; Real-time feedback and closed-loop control: Use product quality test data (grain weight difference, wall thickness uniformity) as reinforcement learning reward signals , update the Q value and optimize the scheduling strategy through the following formula: ； Through the synergistic effect of the modules, efficient operation of task scheduling of the capsule production line and real-time optimization of product quality are achieved.

2. The intelligent control system for capsule production line according to claim 1, wherein: The weight optimization target of the rule priority scheduling algorithm is achieved by the following formula: ； in: w=[α,β,γ] represents the priority weight vector; · Calculate the result for the current priority level; · It is a target priority, generated based on actual production performance.

3. The intelligent control system for capsule production line according to claim 1, wherein: The reward function of the reinforcement learning algorithm Combined with the following formula: ； in: , , is a weight parameter used to balance product quality and production efficiency; The quality parameter in the reward function comes from real-time detection data.

4. The intelligent control system for a capsule production line according to claim 1, wherein: The dynamic task allocation algorithm adjusts the module task ratio through the following formula: ； in: represents the real-time available resources of module j; represents the task allocation ratio of module j; Indicates the total number of modules in the production line.

5. The intelligent control system for capsule production line according to claim 1, wherein: The system adjusts the module operating speed based on the following formula: ； in: is the running speed of module j; is the current task volume of the module; It is the cycle time of the production line.

6. A control method for an intelligent control system for a capsule production line according to claims 1 to 5, comprising the following steps: A. Data collection: The sensor module collects the task status of each process, module load data and product quality inspection data in real time; B. Priority calculation: The task priority is calculated using the rule-based priority scheduling algorithm according to the following formula: ; Among them, the weight parameters α, β, and γ are dynamically optimized by the deep learning model; C. Dynamic task allocation: Based on the reinforcement learning algorithm, the module task allocation ratio is dynamically adjusted according to the following formula: ; D. Real-time feedback: Optimizing the reward function through quality inspection data , and update the reinforcement learning model: ； E. Parameter adjustment: Dynamically adjust the production line operating parameters according to the optimization results, including conveying speed, drying temperature and glue viscosity; Through the above steps, dynamic scheduling of the production line and closed-loop optimization of product quality can be achieved.

Images