Dynamic Multi-Objective Evolutionary Algorithms And Its Research In Rolling Load Distribution

With the wide application of steel products in all aspects of national production and life,the demand for product quality is increasing,so the rolling accuracy requirements are also increasingly fine.In rolling production,load distribution is the main link of rolling process automation system,responsible for formulating the optimal rolling production plan.In this paper,the dynamic process of acceleration and deceleration in tandem cold rolling is taken as the research object,and the dynamic multi-objective algorithm is used to deal with this process to reduce energy consumption and improve product quality.Rolling process is a multi-variable,time-varying and strong coupling system.The objective function and constraints that meet production requirements are formulated according to rolling model,and an improved multi-objective evolutionary algorithm is used to quantitatively analyze the relationship between different objectives in high-speed section and low-speed section,respectively.The results show that there are great errors in the optimal Pareto front under different speeds,which further illustrates the non-linear characteristics of rolling speed change process,and provide decision-makers with intuitive rolling rules and scientific guidance.The problem of rolling load distribution in variable speed stage can be abstracted as a dynamic multi-objective problem with velocity as the environmental change factor.When the objective function changes with time(environment),a two stages prediction strategy for evolutionary dynamic multi-objective optimization is proposed to track the new Pareto frontier quickly and accurately.Each Pareto solution set is divided into two parts: the center point and the manifold,and the prediction accuracy of the center points directly affect the prediction accuracy of the population.In this paper,a two stages prediction strategy is used to improve the prediction accuracy of the center point.The simulation results show that the proposed algorithm has faster convergence speed and higher convergence accuracy for complex Pareto frontier changes and strong environmental changes.Taking sliping factor and equal power margin as objective functions,the dynamic process of rolling mill is optimized by the proposed algorithm,the traditional linear processing method,and static algorithm respectively.The simulation results show that the traditional processing method can meet the general production requirements,but the algorithm in this paper can obtain the optimal results of the whole rolling process,and further improve the setting accuracy of rolling process and the rolling process.The rolling compound load distribution scheme with higher precision can be obtained by product quality.