Many-objective optimization of hot-rolling process of steel: A hybrid approach

In this study, many-objective optimization is carried out in search of combinations of process and chemistry parameters that can lead to simultaneous maximization of three conflicting mechanical properties of hot-rolled steel process. A novel approach combining evolutionary (MOEA/DD) and classical (Normalized Normal Constraint, NNC) algorithms has been proposed to perform the optimization. Through this hybrid approach, the known ability of evolutionary optimizers to escape a locally optimal basin is amalgamated with the strong local search ability of classical optimizers to quickly find better solutions. The efficacy of the proposed approach has been demonstrated using realistic industrial case studies as compared to the optimizers considered alone. Further, mechanical properties and the processing parameters corresponding to multiple Pareto optimal solutions have been correlated for identifying operators' rules to run the plant in near optimal fashion.