Intelligent Integrated Optimization And Control Strategy And Its Application For Mixture Preparation In Sintering Blending Process

Sintering is one of the important loops in the iron ores smelting process. The burden and mixing granulation belong to Iron Ores Mixture Preparation (IOMP). In actual industrial process, IOMP process has not been optimized enough. In the burden process, the proportions are with low accuracy leading the fluctuation of sintered ore quality, and the high sulfur content and cost cause a low economical and environmental benefit. In mixing granulation process, irrational granularity distribution may cause high energy consumption in sinttering. To deal with the problems, the thesis focuses on intelligent integrated optimization and control strategy in IOMP. Main researches and innovations of the thesis are given below:(1) An intelligent integrated modeling method for burden process based on mechanism analysis and data-driven methodTo deal with the quality fluctuation problem caused by the complexity of sintering quality prediction and low accuracy of ores proportions, a kind of intelligent integrated models with cascade structure for sinter qualities prediction is put forward. Firstly, after analyzing the qualitative and quantitative changes of chemicals in sintering process, the mechanism models of sinter qualities are worked out. On the basis of grey relational analysis, key parameters affecting the quality of sintered ore are found, and then they are separated into two parts:the known proportion information and the unkown sintering states parameters. According to the sintering stability requirements, Integrated T-S fuzzy fusion models, which are consisted by Gray Model (GM(1,1)) and Least Squares Support Vector Machine (LS-SVM), are worked out to give rational prediction values of sintering states papameters. On the above basis, BP Neural Network (BPNN) with global approximation capbility and LS-SVM with local generalization capability are seperatly worked out to predict the qualites of sintered ore. After that, from the viewpoint of information theory, an information entropy fusion method weitghting outputs by the variation degree of prediction error sequence is designed to integrate the outputs of mechanism model, BPNN and LS-SVM, so that the qualites of sintered ore can be predicted correctly. The simulation and industrial applications show that the accuracy of intelligent integrated model with cascade structure is higher than single-stage prediction model or single model with cascade structure. It can predict qualites of sintered ore correctly and satisfy data completeness requirements during burden optimization.After analyzing the numerical relationship between coke ratio and qualites of sintered ore, a coke proportion optimization model constrainted by qualites of sintered ore is proposed. Then according to the analysis of the main physical and chemical changes in sintering process, a computing model for lower limit of coke proportion based on heat balance is worked out to narrow the searching range of coke in the optimization model. The model can overcome the difficulty of conventional experience-based proportioning method that it is hard to realize coke proportion optimization, and establishes the foundation of energy saving optimization.(2) A multi-objective optimization model of burden process for cost and sulfur content reducingIn the burdening process, the traditional proportion methods usually is with high cost and sulfur content, and the ordinary optimization models can neither appoarch to the burden process precisely nor consider the cost, energy consumption and environment benifit comprehensively. To deal with those problems, linear weighted multi-objective optimization models of burden process (including primary and secondary proportion optimization models) using inventory and sintered ore chemical constraints, which is based on the different characteristics of primary and secondary proportions as well as the economy and key chemical components analysises of different materials, are proposed in the thesis. The optimization models could realize the comprehensive optimization of reducing cost and sulfur content compared with conventional proportion cost optimization model.(3) Multi-objective comprehensive optimization algorithm of burden processBased on the linear weighted multi-object optimization models of burden process, a multi-objective comprehensive optimization algorithm, which is integrated Linear Program (LP) with Genetic Algorithm and Particle Swarm Optimization (GA-PSO), is proposed to search the optimal solutions of primary and secondary proportions in the thesis. Firstly, LP method is applied to solve the linear weighted multi-object optimization models. If it does not work, GA-PSO algorithm is implemented. In this algorithm, Particle Swarm Optimization (PSO) algorithm is used during the early period. When PSO stops converging, the operations of crossing and mutations from Genetic Algorithm (GA) are executed with a certain probability to increase diversity of particle group, which avoids the fluctuation in late convergence and increase the convergence rate. The multi-objective comprehensive optimization algorithm is realized in actual industrial application to optimize the burden process.(4) Intelligent integrated optimal control algorithm for mixing granulation based on evaluation of granularity distributionThe factors effecting granularity distributions of iron ores mixture are various, while the process index of granularity distributions is quantitative. Hence, the granularity distribution optimal control lacks of accurate mathmatics models and clear targets, the process can hardly be optimized by using traditional optimal control strategies. Aimed at those problems, the thesis puts forward a control algorithm based on evaluation and optimization of granularity distribution. Firstly, the conception of granularity parameters is proposed to describe the continuous granularity distribution. Secondly, by analyzing the screening experiment datas and the corresponding sintering states, the height of material layer and the average permeability index are chosen to build the fuzzy evaluating functions, and then the sample set of granularity parameters and evaluation values can be obtained and BPNN can be trained to build the granularity parameters evaluating model. Thirdly, the optimal granularity parameters could be calculated using PSO algorithm, by taking the granularity parameters evaluating model, granularity parameters and process boundaries as objective function, decision variables and constraints, respectively. Finally, humidity setting model is built to covert the granularity parameters into real-time optimal humidity control setting value in mixing granulation process。The algorithm overcomes the limitation that the conventional humidity control is difficult to realize the granularity distribution optimal control. The simulation shows that the algorithm improves the effection of granulity distribution.(5) Industrial application of optimal control strategy in IOMPBased on industrial reality, software and data stream are designed, and then the burden optimization and decision support system is built for a360m2sintering production line to realize the comprehensive optimization of cost, energy consumption and sulfur content reducing, which bring obvious economical and environmental benefits.