模糊神经网络在磨矿控制系统中的应用研究

发布时间：2018-03-04 12:08

本文选题：磨矿分级过程　切入点：粒度　出处：《河北工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：磨矿分级生产过程是整个选矿流程中的重要环节。通过磨矿分级生产过程对矿石进行研磨和分级,使有用矿物与脉石矿物进行单体解离,为后续的选别作业创造条件。本文以典型两段式磨矿分级生产流程为研究对象,采用补偿模糊神经网络对磨矿控制器进行设计,并在磨矿优化虚拟设备上进行仿真实验。本文的主要内容如下:首先对近年来国内外磨矿控制研究发展历程与应用现状进行了总结,对磨矿分级生产过程的工艺特点以及控制难点进行了分析。然后通过对磨矿流程中各参变量的研究,总结归纳出相关控制规律,确定以磨矿粒度为最终的控制指标,将给矿量、一段球磨机入口加水量、螺旋分级机加水量以及旋给流量作为其控制变量,以神经网络对磨矿系统进行辨识,建立数学模型。针对典型两段式磨矿系统中机理复杂、影响因素多等诸多难点问题,采用补偿模糊神经网络对磨矿控制器进行设计,通过利用模糊控制对模糊信息的处理能力和神经网络强大的学习能力很好地解决了磨矿控制中非线性、强耦合等问题。补偿模糊神经元的引入,能使网络从初始正确定义的模糊规则或者初始错误定义的模糊规则进行训练,使其具有更高的容错性,系统更稳定。通过MATLAB软件对其进行仿真验证,结果表明矿石粒度被很好地控制在设定值附近,证明了补偿模糊神经网络对磨矿控制的有效性和实用性。最后将补偿模糊神经网络控制算法加载到磨矿优化虚拟设备上进行半实物仿真验证,得到的磨矿粒度趋势图基本满意,取得了很好的控制效果。
[Abstract]:Grinding and grading production process is an important link in the whole process of mineral processing. By grinding and classifying ore, the useful minerals are separated from gangue minerals by monomer dissociation. In this paper, the typical two-stage grinding classification production process is taken as the research object, and the compensatory fuzzy neural network is used to design the grinding controller. The main contents of this paper are as follows: firstly, the history and application of grinding control research at home and abroad in recent years are summarized. The technological characteristics and control difficulties of grinding classification production process are analyzed. Then through the study of the parameters in the grinding process, the relevant control rules are summarized, and the final control index of grinding granularity is determined, and the ore quantity will be given. In this paper, the inlet water quantity of the first stage ball mill, the water quantity of the spiral classifier and the rotating feed flow rate are taken as the control variables. The neural network is used to identify the grinding system and the mathematical model is established. The mechanism of the typical two-stage grinding system is complex. Many difficult problems, such as many influencing factors, are used to design the grinding controller using compensatory fuzzy neural network. By using the ability of fuzzy control to deal with fuzzy information and the powerful learning ability of neural network, the nonlinear and strong coupling problems in grinding control are well solved. The network can be trained from the original correctly defined fuzzy rules or the initial misdefined fuzzy rules, which makes the network more fault-tolerant and more stable. The simulation results are verified by MATLAB software. The results show that the ore size is well controlled near the set value. The effectiveness and practicability of the compensatory fuzzy neural network for grinding control are proved. Finally, the compensatory fuzzy neural network control algorithm is loaded into the grinding optimization virtual equipment for hardware-in-the-loop simulation, and the obtained grinding particle size trend diagram is basically satisfactory. Good control effect has been achieved.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD453;TP273

【相似文献】