Intelligent Control Strategy for Sintering Ignition Temperature Based on Working-Condition Recognition

Ignition temperature in a sintering process is an important factor for the quality and yield of sinter ore. Due to the diversity of the calorific value and the flow rate in the sintering process, there are many different steady states in the combustion process. Under different steady states, the change interval and fluctuation range of calorific value and flow rate are completely different which shows the sintering data is one kind of typical imperfect data. In this case, if these data are used to model and control in conventional machine learning methods, it will lead to inaccurate results. This paper presents an intelligent control strategy for ignition temperature based on working-condition recognition to stabilize the ignition temperature. First, this paper describes the structure of the intelligent control system for ignition temperature, including the working-condition recognition module and the different fuzzy prediction controllers. Then, this paper explains a fuzzy C-means clustering algorithm to recognize the working-condition of an ignition process. Next, this paper presents the detailed design of the fuzzy prediction controller for each specific working-condition, including an Elman prediction model optimized by the particle swarm optimization and a fuzzy controller. Finally, some experimental results based on actual production data show that the method controls the ignition temperature accurately in the complex production environment. Note to Practitioners—The motivation of this paper is to control the ignition temperature in the sintering process to improve ignition efficiency. The existing methods for ignition temperature control mainly focus on the design of the model and controller, and rarely consider the influence of ignition conditions on the system. This paper presents an intelligent control method of ignition temperature by recognizing different working conditions to accurately control the ignition parameters. The working condition recognization effectively solves the problem of the large difference between different data and states. By applying the proposed control strategy, the control error of the sintering plant is kept in a small range to meet the actual control requirements of the sintering plant. Future work is to extend the proposed method to more industrial application scenarios that have imperfect datasets characterized.