Estimation of operating blast furnace reactor invisible interior surface using Differential Evolution

A blast furnace reactor that transforms processed minerals into liquid iron is the most important link in a steel making process chain. Estimating the condition of refractory shield in a running furnace is important for production campaign planning and hazardous leakage prevention. A method coupling Differential Evolution with heat transfer simulations is developed for estimating refractory inner surface within process real times. The fitness value that drives this evolutionary algorithm is calculated from the differences between temperatures measured from thermocouples embedded in the refractory, and those obtained from numerical solution of the heat transfer equations. A population of different refractory inner surfaces constitutes the candidate solution set. These evolving surfaces tend to become craggy which impact variably on accuracy of numerical solution with significant consequence on the final synthesized solution. These inter-candidate accuracy variations are neutralized by transforming each physical refractory grid configuration onto identical rectangular computational domains using a set of continuity-ensuring elliptic partial differential equations. The heat transfer equation is correspondingly transformed into the computational space and solved within a finite difference framework. A set of practical constraints are imposed on the Differential Evolution algorithm to guide evolution toward the optimal surface that best represents the measured temperatures. Rigorous verification of the various stages of development facilitating industrial acceptance of the method are presented. This simulation-embedded evolutionary inverse-design approach can be used generically for online estimation of unobservable critical parameters of infrastructure, equipment and processes in diverse domains.