Research on modelling and optimization of hot rolling scheduling

Multiple species and random change in batch order forms has become the core business of Steel Intelligent Production Enterprises, necessitating new requirements for scheduling real-time adaptability and precision in the steel scheduling model. In this paper, a digital twin intelligent agent with cyberspace physical space integration is proposed as the unified driving source. First, an Open Multiple Objective Travelling Salesmen Problem model was established. For constraints such as a minimum rolling unit plan, process specification and a minimum power consumption per ton of steel, the NSGA-II algorithm was used to obtain a Pareto front for these constraints. With the help of appropriate penalty coefficients for the Pareto front, the constraints were defined as 3D-coordinates among cities of the Travelling Salesmen Problem. Combining the simulated annealing (SA) algorithm and the multi-objective particle swarm optimization (PSO) algorithm, the PSO algorithm was redefined and modified by introducing the metropolis criterion of the SA algorithm twice. This was used to obtain two extremes of the particle swarm, including the individual optimal solution and the global optimal solution to avoid local extrema. A model based on SA-MOPSO was thus obtained. Then, with the help of real-time mapping between the steel production line and the information model, the double-flow digital twin agent was established. This agent can achieve production scheduling model dynamic development and maturation itself. Finally, taking the steel hot rolling production line as a case in point, training and trial application of the model were performed to verify its adaptive development ability. Simulation results show that the algorithm discussed in this paper can solve the problem of hot rolling scheduling and provide support for decision-makers.