基于ESN混沌时间序列的RBF神经网络对浮选经济指标的预测分析

发布时间：2018-12-19 08:08

【摘要】：选矿企业作为典型的连续流程型企业,其最关键生产指标通常指精矿品位及作业回收率,对于浮选过程而言,精矿品位的稳定与否更是对企业的经济效益起着决定性的作用。传统的选矿厂通常参考要达到的经济目标根据选矿流程机理和工厂生产积累的经验把要达到的生产指标分解为对应的工艺指标,例如给矿粒度、矿浆浓度和浮选剂添加量等,车间工作人员的工作则是将这些工艺指标控制在在规定的范围内,而生产管理人员则按照工艺指标是否在规定范围内来判断生产操作的好坏。本文的所研究的主要是通过对浮选生产过程关键工艺指标进行优化从而实现对浮选生产过程中浮选经济指标最优控制的目的。本文以某选矿厂采集的浮选经济指标数据为基础,选取给矿品位、给矿粒度、给矿浓度、给矿流量四种数据作为RBF神经系统的输入量,精矿品位以及作业回收率两种指标作为神经网络的输出量,利用MATLAB软件中的simulink工具箱进行编译程序、模拟仿真。对比实际曲线与期望曲线之间的误差,观察实际曲线是否平滑,调节神经网络与混沌时间序列结合后的系统精度,考察调整参数后系统的拟合程度是否已达到要求值。实验结果分析表明:利用RBF神经网络可对混沌系统进行预测分析,算法简便,响应速度快,省去许多繁琐的步骤,提高运算效率。同时通过对Mackey-Glass和Lorenz混沌系统的模拟仿真也直接表明了利用神经网络对混沌系统的建模及分析可以有效提高系统的的精度。同时,对于浮选过程的建模及仿真也充分说明了RBF神经网络可对混沌时间序列进行有效预测,也是可以应用与生产实践的有效方法,也为未来的研究打好了基础。
[Abstract]:As a typical continuous process type enterprise, the most key production index usually refers to concentrate grade and operation recovery rate. For flotation process, the stability of concentrate grade plays a decisive role in the economic benefit of the enterprise. The traditional concentrator usually refers to the economic objectives to be achieved. According to the mechanism of the processing process and the experience accumulated in the production of the plant, the production indexes to be achieved are decomposed into corresponding technological indicators, such as the ore size, pulp concentration and the amount of flotation agent, etc. The work of the workshop staff is to control these process indicators within a specified range, while the production management personnel judge the quality of the production operation according to whether the process indicators are within the specified range. The main purpose of this paper is to optimize the key technological parameters of flotation production process so as to achieve the purpose of optimal control of flotation economic indexes in flotation production process. Based on the economic index data of flotation collected from a concentrator, four kinds of data, such as feed grade, ore size, ore concentration and ore flux, are selected as the input amount of RBF neural system. The two indexes of concentrate grade and job recovery are used as the output of neural network. The simulink toolbox in MATLAB software is used to compile the program, and the simulation is carried out. By comparing the error between the actual curve and the expected curve, the paper observes whether the actual curve is smooth, adjusts the precision of the system after the combination of neural network and chaotic time series, and investigates whether the fitting degree of the system has reached the required value after adjusting the parameters. The experimental results show that the chaotic system can be predicted and analyzed by RBF neural network. The algorithm is simple, the response speed is fast, and many tedious steps are eliminated, and the operation efficiency is improved. At the same time, the simulation of Mackey-Glass and Lorenz chaotic system also shows that the modeling and analysis of chaotic system using neural network can effectively improve the accuracy of the system. At the same time, the modeling and simulation of flotation process fully show that RBF neural network can effectively predict chaotic time series, is also an effective method for application and production practice, and provides a good foundation for future research.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F426.1;TP183

【参考文献】