基于大数据DQN的烘丝料头预热温度在线调节与仿真模拟

On-Line Adjustment and Simulation of Preheating Temperature of Cut Tobacco Drying Based on DQN

卷烟制丝工艺流程中,保持烟丝含水率维持在一个稳定的范围是保证烟丝质量的一个重要因素。在调节烘丝机温度以保证烟丝水分达标时,烘丝筒壁温度根据挡车工经验值设定。对于不同的烟草来料情况,不同人员在烘前温度预热调整时往往会在生产阶段产生一定的偏差,而在烘后水分反馈调节时又会产生滞后性,造成产线上烟丝含水率不稳定的现象。为了解决烘丝过程中的诸多问题,以及工艺知识固化、少人化与质量一致化等一系列需求问题,实现工厂的一体化智能管控,提出了一种基于大数据的环境、应用深度强化学习算法模拟预测料头阶段筒壁温度设定值。根据大数据,构建了一个仿真模拟系统,并集成在烟厂MAS工业控制平台进行试车应用。模拟仿真和现场应用试验结果证明了强化方法的可行性,其能够替代人工在线控制并提高生产效率。

In the process of cut tobacco production,maintaining the moisture content of cut tobacco in a stable range is an important factor to ensure the quality of tobacco.When adjusting the temperature of the dryer to ensure the moisture level of cut tobacco,the temperature of the drying cylinder wall is always set according to the experience value of the machinists.For different tobacco materials,different personnel tend to produce certain deviation in the production stage when adjusting the preheating temperature before drying,and then lag behind when the moisture feedback adjustment is made after drying,resulting in the instability of the moisture content of the cut tobacco on the production line.In order to solve many problems in the process of cut tobacco drying,solve a series of demand problems such as solidification of process knowledge,reduction of manpower and consistency of quality,and realize the integrated intelligent management and control of the factory,a deep reinforcement learning algorithm based on big data was proposed to simulate and predict the temperature setting value of the cylinder wall in the feed head stage.Based on the big data,a simulation system was constructed and integrated into the MAS industrial control platform of the factory for trial operation.The simulation and field application test results showed that the proposed method has stable performance and can replace the manual on-line control and improve production efficiency.