Data-Driven Multi-Objective Optimization of Power Consumption and Quality for Finishing Rolling Process

This paper investigates the operational optimization in the context of the steel hot rolling production and finishing facilities. Optimizing the operational variables of the finishing rolling process can improve the rolling quality and reduce the unit power consumption. However, there are too many operational variables in the process and their interactions are strongly coupled, making it difficult to develop an effective model for in-depth study. Therefore, this paper adopts a data-driven approach, using neural network algorithms to construct the data prediction models between performance and set values respectively. Multi-objective optimization of the set values is performed by combining the actual mechanism and equipment constraints. In the obtained Pareto optimal solutions, the conflicting choices are further balanced among multiple objectives to obtain the most suitable operating parameters. Experimentally verified using production line production data, the method proposed in this paper significantly improves the performance indexes.