An intelligent optimization strategy of operating parameters for stable-efficient operation in the iron ore sintering process

Abstract

To improve the stability and efficiency of the iron ore sintering process, an intelligent optimization strategy of operating parameters is proposed. The strategy integrates operating mode perception, finished sinter ratio prediction, and multi-objective optimization to address the strong nonlinearity, time delay, and multivariable coupling of the process. A feedforward neural network is developed to predict burn-through point location and temperature for operating mode perception. A hybrid model based on bidirectional long short-term memory networks and fuzzy c-means clustering is constructed to predict the finished sinter ratio under time-varying operating modes. A plant-constrained multi-objective optimization problem is then formulated to coordinate strand velocity, main flue pressure, and sintering bed height, and standard NSGA-II is adopted as its solution method. Offline evaluation using industrial historical data indicates that the integrated framework can generate candidate operating-parameter schemes that balance the predicted normal-mode proportion and the predicted finished sinter ratio. The contribution of this study lies in the domain-specific formulation and integration of operating-mode perception, time-varying production prediction, and constrained multi-objective decision making.