A data-driven decision-making framework for online control of vertical roller mill. (May 2020)

• Record Type: Journal Article
• Authors: Zhu, Mingrui; Ji, Yangjian; Zhang, Zhen; Sun, Yuanyi
• Is Part Of: Computers & industrial engineering. Volume 143(2020)
• Publication Date: 2020-05
• Subjects: Vertical roller mill -- Data mining -- Running state data -- Control decision-making; Engineering -- Data processing -- Periodicals; Industrial engineering -- Periodicals; 620.00285
• DOI: 10.1016/j.cie.2020.106441
• Languages: English
• ISSNs: 0360-8352

Abstract

Highlights: An online control decision-making framework based on the running state data is proposed. The real-time rolling prediction and identification of VRM working condition based on time window are carried out. The generation of control strategy is studied based on the correlation of stability indexes. The real-time stability control strategy can be given for unstable working conditions. A real industrial application example is presented. Abstract: Vertical roller mill (VRM) is a large-scale grinding equipment, which is used to grind raw materials from block/granule into powder. Due to harsh production environment and inconsistent raw material quality, VRM requires timely regulation. Currently, the regulation of VRM is manually conducted; operators make decisions based on their observation and experience, therefore the timeliness and accuracy of regulation cannot be guaranteed. This study presents a data-driven online control decision-making approach; it extracts several key indicators for state judgment from the historical running state data, constructs a stable mode library based on clustering the running state, mines the association rules among variables, and establishes the rolling prediction model to predict the changes in the key indicators. In real-time operation, the target state is obtained by comparing the real-time state and stable mode library, and then the corresponding control strategy, composed of key indicators, controllable parameters and target state, is auto-generated to support the management of VRM operation. In this way, a closed-loop framework is formed based on offline data mining and online decision-making, supporting the operation optimization of VRM. This approach is applied in a cement plant as a case study in Jiangsu, China. The results show that the control strategy is effective in actual working conditions; the continuous operation of the equipment with vibration reduction is achieved.