Application of Deep Learning in Parameter Optimization of Automatic Production Process in Hot Rolling Production Line

Application of Deep Learning in Parameter Optimization of Automatic Production Process in Hot Rolling Production Line

China is a major producer of steel and a pillar industry of the national economy, with huge coal consumption. With the increasingly prominent problem of energy shortage, traditional industrial models are constantly being transformed and upgraded using information technology, and a comprehensive energy information management system is being constructed. This article focuses on the production scheduling optimization problem of steel hot rolling production process. Firstly, based on the hot rolling process flow, the operation and maintenance time consumption of hot rolling equipment and the conversion time between hot rolling equipment are fully considered. The mathematical model of production scheduling for the hot rolling production process is established with the goals of minimizing work order completion time and balancing equipment working hours. Then, the classic NSGA-II algorithm is used as the basis for multi-objective solving. To solve the problems of the algorithm being prone to falling into local optima, insufficient distribution, and long solving time, the algorithm is improved by combining deep reinforcement learning ideas. Finally, through simulation experiments, the superiority of the improved algorithm in the convergence process is verified. At the same time, real hot rolling cases are used as scheduling objects to complete scheduling optimization and provide scheduling solutions.