基于神经网络的烧结工艺优化

发布时间：2019-01-26 16:23

【摘要】：中国的钢铁行业经过高速发展,到现阶段出现了严重产能过剩导致的价格下降问题,造成钢铁企业的利润减少,同行业间的竞争日益加剧。因此,为了降低生产成本,提高产品质量,钢铁企业越来越期望通过先进生产工艺控制手段提高生产效率与效益。高炉炼铁以烧结生产的产出品作为原料,烧结成品产量,化学物理性能是否合格,直接限制和影响后续炼铁工序生产。烧结矿生产工艺过程中最能影响上述要求的参数是烧结终点位置。在分析国内外有关研究现状的过程中,首先分析了烧结生产工艺的基本流程和工艺参数控制方法,通过深入分析生产工艺过程,最终确定了四个影响烧结终点位置的关键参数:大烟道负压力值、辊道送风量、辊道前进速度和点火起始温度。然后研究了神经网络的基本理论,依据神经网络映射抽象控制参数与结果间非线性关系的能力,提出使用BP神经网络算法建立烧结工艺终点预测模型解决烧结终点位置预测问题。在研究烧结工艺特点与BP神经网络算法的基础上,本文为烧结终点位置预测问题建立一个基于BP神经网络的预测模型,然后使用仿真手段对网络进行验证。仿真过程首先是训练网络,训练样本使用生产现场收集到的实际数据,在网络输出达到可接受的误差后,使用另一组真实生产环境采集的工艺数据对神经网络进行检验,通过分析输出结果得出结论,BP神经网络预测模型可以在较短的时间内完成训练,预测结果与期望值误差小。仿真结果表明该方法可以用于解决预测烧结终点的问题上,应用效果明显,对烧结生产过程有良好的指导意义。
[Abstract]:With the rapid development of the steel industry in China, there appears the problem of price decline caused by serious overcapacity at the present stage, which results in the reduction of the profit of the iron and steel enterprises, and the competition between the same industries is intensifying day by day. Therefore, in order to reduce production cost and improve product quality, iron and steel enterprises are more and more eager to improve production efficiency and benefit through advanced production process control. In blast furnace ironmaking, the output of sintered products is used as raw material, the output of sintered products and the chemical and physical properties are up to standard, which directly limits and affects the subsequent ironmaking process. The point of sintering is the most effective parameter in the process of sinter production. In the process of analyzing the current research situation at home and abroad, the basic process of sintering production process and the control method of process parameters are first analyzed, and through in-depth analysis of the production process, Finally, four key parameters affecting the end point of sintering are determined: negative pressure value of the flue, air supply rate of the roller table, moving speed of the roller table and starting temperature of ignition. Then, the basic theory of neural network is studied, and the ability of abstracting the nonlinear relation between control parameters and results is analyzed according to the mapping of neural network. In this paper, BP neural network algorithm is used to set up the prediction model of sintering process end point to solve the problem of sintering end point prediction. On the basis of studying the characteristics of sintering process and the algorithm of BP neural network, this paper establishes a prediction model based on BP neural network for the prediction of sintering endpoint, and then verifies the network by means of simulation. The first step of the simulation is to train the network. The training sample uses the actual data collected from the production site. After the network output reaches the acceptable error, another set of process data collected in the real production environment is used to test the neural network. By analyzing the output results, it is concluded that the BP neural network prediction model can complete the training in a relatively short time, and the error between the prediction result and the expected value is small. The simulation results show that the method can be used to predict the sintering end point, and the application effect is obvious, and it has a good guiding significance for the sintering production process.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TF046.4;TP183

【相似文献】