Intelligent Control And Research Of Self-learning Model In Sinter Mixture Moisture

Sinter ore is one of the most significant raw materials of blast furnace ironmaking.The moisture content of the mixture can only be controlled within the optimum range to stabilize the sintering machine working conditions and to improve the sintering production efficiency.However,at present,most of the sintering plants still add water by people,which makes the mixture of water fluctuate in a great range and makes sintering production efficiency improve difficultly.The main reason is the delay of process equipment.It must use the feedforward control in automation,according to the weight of the feed flow to determine the amount of water.However,the water content of various materials sometimes fluctuate.It causes that parameter of feedforward can't follow the variety of materials.To solve this problem,the material balance is used as the theoretical basis to build a water model in this thesis.The water content of all kinds of materials is used by intelligent identification method for the self-learning modeling research.First of all,mission collecting data of flow rate,water flow rate and water content of the mixture is completed in the online ingredients.The consistency of the communication protocol between the on-site PLC and industrial PC collecting data,and the delay time from each ingredient steelyard to the inlet of mixing machine are taken into account.It achieves real-time receiving data and sending data besides it achieves storing in binary files to disk.Secondly,after attaining a large amount of data,the data is processed,including removing the break point data at the time of stoppage and removing the singular point of the remaining data by Patuta criterion.The simulation platform of graphical user interface of MATLAB is built.The research and simulation of water content identification,feedforward model correction,and feedforward water control are completed on it.Finally,it achieves that the feedforward model parameters track the variation of material moisture,timely corrects feedforward model and calculates the flow of water.Because of considering the water of materials fluctuating in a small range,material moisture is set the different degree of random fluctuations and is identified again.When the moisture measurement exceeds the setting value of ± 0.3%,the water content various materials is automatically identified and corrects the feedward model.The results of simulation compare with the effect of adding water by people from record data,which the range of fluctuation of mixture moisture narrows obviously.In the field test,the industrial computer according to the result of offline identification and flow of material compute the water for feedforward control,so as to realize adding water by feedforward control.The result of simulation demonstrates the feedforward control adding water with self-learning modeling,which is feasible and can make model parameters track the variation of material moisture,effectively restrain the interference caused by the flow of material and water content of material,solve the disadvantage effect caused by the lag of process for water control and make the mixture moisture fluctuate in small range.The achievement provides a method of self-learning modeling which can follow variation of material moisture and adapt parameter of feedforward,for modeling of feedforward of sinter mixture and consolidating the foundation in online application.