The Research On Sinter Iron Protoxide Content Of Real-time Detection System Based On RBFNN

At present, with the economic construction ,the need of steel is accelerated, it objectively promotes our steel industry to booming development, and the raw material of ironmaking is iron ore, and it determines the requirement of iron ore growing. We know that there are two kinds of basic iron ore, the first is natural iron ore, and the second is artificial sinter. The method of getting artificial sinter is sintering inferior iron ore. Natural iron of high quality ore are non-renewable resources, can be directly used as charging, but with exploiting ,is less and less, yield reduces year by year. And iron ore of poor inferior quality, cannot be directly used as charging, firstly, the method of sintering is used to made block, first, improve the quality of sinter, then it can be used as charging. So now, the iron ore sinter is used as charging mostly.Sinter quality affects each index of the blast furnace ironmaking directly, and sinter quality is decided by the most important parameters which is FeO content of sinter. Using chemical testing method to detect the content of sinter FeO in sinter method refers to finish production, and after cooling sinter, analysis the sampling of chemical, and then detect FeO content, and then adjust process parameters based on the content of the sintering FeO, so as to achieve the purpose of adjusting FeO content. This method detection FeO content is the most accurate, but it is the regulation for sintering process, we know, from the hysteresis of sinter mixture burned form, again, the sinter cooling and chemical analysis, probably need 2 to 3 hours, so one moment FeO content is detected before the 2 to 3 hours of sintering process parameters decision. Now there is using artificial in sintering the tail place, directly observes the sinter cross-section image, based on the experiences, gets the characteristics of rough FeO content, and feedback to FeO sintering starting point, to adjust the sintering process parameters, achieve the purpose of controlling FeO content. But this method for man's subjective and objective factors, often deviation is bigger.Therefore, it is necessary to develop a set of automatic real-time detection system of sinter FeO. This article USES the method of the radial basis function neural network (RBFNN) and fuzzy c-mean combined to forecast the sinter FeO content real-time. The characteristic of the RBFNN network structure is simple and with fast convergence rate, and the characteristic of c-means clustering is that the cluster precision is high. Firstly, the characteristic parameters of sintering cross-section image are clustered, then the FeO content are divided into four types, for four different types of samples, with corresponding sintering process parameters to train RBFNN networks, thereby we obtain RBFNN network model of fourvariables spreading rate relatively small. The forecast of sinter FeO content as follows:firstly, we uses c-means clustering to judge the level of section FeO content preliminary.secondly, we enter the sintering process parameters into the corresponding RBFNN network model, and the FeO content will be given.