Development Of The Sintering Expert System

During the past decade, the price of iron ore kept a dramatic growth rate due to two main facts: one was the increasing need for ores with the development of iron and steel industry in China, and the other was the decrease of high-quality ores all over the world. Thus, the cost of raw materials has been taken up a high fraction in the economic benefit of enterprises. As a result, varieties of low-grade or high harmful elements containing ores were used in the ironmaking process which led to a great fluctuation in chemical compositions and physical properties of the mixture.For the complex materials structure in the sintering production, in which the ore is used in poor performance, this paper developed an expert system for sintering process are consist of raw material management system modules, burdening optimization modules, granulation prediction modules and sintering prediction modules. The raw material management module give the ore physical chemistry information, resources information and pictures information; Blending optimizing module mainly using linear planning method, under the constraint of the chemical composition and particle size distribution of the mixture, calculate the best burden with lowest cost; Granulating prediction module, mainly introducing the concept and application of moisture capacity, establish the optimum moisture prediction model, and is based on BP neural network established granulating effect, prediction model predict the fraction of mixture 3-8 mm and the permeability of the particle bed; Sintering prediction module, through the sintering raw material and thermal equilibrium calculation, get the carbon dosage during the sintering in theory, prediction model based on BP algorithm and establish the sinter performance neural network forecasting model, and forecast the sinter strength performance of physics, sintering utilization coefficient and yield, etc. After the application of this system in sintering plant, the conclusion can be summarized as follows:①The moisture capacity of the mixture and have a linear relationship with the optimum moisture, the expression is y = 0.365×x + 1.962, which is used to predict the optimum moisture;②Granulating effect neural network model with a shooting rate of 92%, has a good prediction accuracy, robust, and the high ability of recognition to new sample, which can give a good guide to granulation process. ③The sinter performance model based on neural network, with the shooting rate of 88.5%, have good prediction accuracy, can give a good guidance to the sintering process.④According to the property of the information transferring in modern steel enterprise, the expert system of sintering process is divided into three layers: the operation layer, model layer, and decision-making layer. The function is different to different users, realizing the configuration of information resources among the layered, and optimization of the sintering process for the modern enterprise.