Optimization of iron ore sintering process based on ELM model and multi-criteria evaluation

Optimization of iron ore sintering process is to maximize the productivity and the sinter ore quality while minimizing the energy consumption. However, these economic and technical criteria are sometimes contradictory and the improvement of one criterion generally leads to the deterioration of other criteria. This paper aims to seek the optimal manipulated parameters from the production data by considering this contraction. Several typical groups of manipulated parameters were firstly obtained by clustering the production data, and then one new artificial neural network model—online sequential extreme learning machine (OS-ELM) was established to predict FeO content and tumble strength of sinter ores for each group. Validation results showed that the OS-ELM model established possessed a higher accuracy than conventional BP (back propagation) model. Finally, a multi-criteria evaluation method—VIKOR was applied to obtain a set of compromise optimal solutions. Taking into account both economic and technical criteria, the final solutions acquired are instructive, and the optimal solution can directly reduce fuel consumption by about 0.5 kg per ton sinter ore without degrading sinter ore quality seriously.