Assessment of reinforcement learning applications for industrial control based on complexity measures

Machine learning and particularly reinforcement learning methods may be applied to control tasks ranging from single control loops to the operation of whole production plants. However, their utilization in industrial contexts lacks understandability and requires suitable levels of operability and maintainability. In order to asses different application scenarios a simple measure for their complexity is proposed and evaluated on four examples in a simulated palette transport system of a cold rolling mill. The measure is based on the size of controller input and output space determined by different granularity levels in a hierarchical process control model. The impact of these decomposition strategies on system characteristics, especially operability and maintainability, are discussed, assuming solvability and a suitable quality of the reinforcement learning solution is provided.

Improving Efficiency of Rolling Production by Predicting Negative Technological Events

Improving the production efficiency is the task with the increasing difficulty. Therefore, it is important to constantly expand the set of tools and perfect the methodology for improving the processes. Some of the losses associated with the negative technological events (breaks, drifting, etc.) are difficult to eliminate completely due to the complexity of making changes to the basic technology. But if you know in advance that the event will occur, you can significantly reduce the probability and consequences, thereby significantly improve the efficiency of the production process. Therefore, it is important to develop and introduce the applied approaches to forecasting the negative technological events in the production processes. This paper presents the method of the forecast-event statistical analysis of the cause-and-effect relationships. The technique was tested on the events of strip’s breakage during rolling at the cold rolling mill 1400 and strip’s drifting in the input storage of the continuous etching unit (CEU). Based on the presented methodology, the specialized digital service was developed and introduced in the production processes of the dynamo steels shop.

Collaboration Strategy Based on Conflict Resolution for Flatness Actuator Group

During the flatness control process, there are frequently some uncoordinated regulating behaviors in the flatness actuator group. This has a bad influence on the flatness control accuracy and the flatness control efficiency. Therefore, a collaboration strategy based on conflict resolution for the flatness actuator group has been proposed in this paper. First of all, the feature of flatness measurement value is extracted through establishing the actual flatness condition discriminating factor. After that, the coordination cooperation that is appropriate to the actual flatness condition is developed for the flatness actuator group. Finally, the optimal adjustment of the actuator population is solved by the coordinated algorithm of Topkis-Veinott and genetic algorithm collaborative optimization. The collaboration strategy proposed in this paper has been successfully applied to a flatness control system of a 1450 mm five-stand cold rolling mill.