Data‐Driven Bi‐Objective Genetic Algorithms EvoNN Applied to Optimize Dephosphorization Process during Secondary Steel Making Operation for Producing LPG (Liquid Petroleum Gas Cylinder) Grade of Steel

摘要

>Though the potential scope of multi‐objective genetic algorithm in the field of secondary steel making (SSM) is enormous, the useful utilization of such evolutionary techniques in secondary steel making process is yet to be done. In this work, data driven multi‐objective optimization is implemented to lower the phosphorous content in the steel bath after completion of secondary steel making process by using minimum lime. The input variables considered for this investigation are process route, minimum vacuum level, process time, ladle inlet temperature, ladle outlet temperature, SSM carbon input, SSM manganese input, SSM phosphorus input, SSM sulfur input, SSM silicon input, Al addition, and tundish sulfur. A data driven Evolutionary Neural Network (EvoNN) algorithm is used to form the meta‐models by training of a dataset consisting of 76 data entries obtained from a steel plant. Analysis of Pareto front gives an effective and beneficial guideline so that one can achieve low tundish phosphorous content avoiding higher amount of slag generation due to extra lime addition during secondary steel making process. This work helps to select the optimum secondary steel making process route for producing LPG grade steel depending upon composition of liquid steel input to secondary steel making units.