重介旋流器分选过程智能控制策略研究与实现  

Intelligent control strategy for the separation process of dense medium cyclone

作　　者：邓建军 匡亚莉 赵建章[1,2] 孙小路[1] 王传真 DENG Jianjun;KUANG Yali;ZHAO Jianzhang;SUN Xiaolu;WANG Chuanzhen(Key Laboratory of Coal Processing and Efficient Utilization of Ministry of Education,China University of Mining and Technology,Xuzhou,Jiangsu 221116,China;Department of Chemical and Environmental Engineering,Xinjiang Institute of Engineering,Urumqi,Xinjiang 830091,China)

机构地区：[1]煤炭加工与高效洁净利用教育部重点实验室,江苏徐州221116 [2]新疆工程学院化学与环境工程系,新疆乌鲁木齐830091

出　　处：《中国矿业大学学报》2019年第3期624-632,共9页Journal of China University of Mining & Technology

基　　金：国家自然科学基金项目(51174202)

摘　　要：为实现重介旋流器精煤灰分的稳定和分选数量效率的提高,以唐口选煤厂重介分选工艺为研究对象,针对其控制系统的多变量、非线性、强耦合、纯滞后等特点,提出了以PID反馈控制为基础、基于神经网络学习和带抗饱和积分控制为核心的前馈控制策略.针对该策略中的神经网络学习问题,构建了"多输入、两输出"的智能算法,即以精煤灰分和数量效率为输出目标的重介分选过程的多个输入变量BP神经网络算法,并用Python语言编制了针对该网络结构的BP算法程序;采集和分析了唐口选煤厂生产实时在线数据,并进行了相关训练和仿真模拟.结果表明:当输入层节点数为3,且在输入变量为入料灰分、入料压力、循环介质密度条件下,输出变量精煤灰分和数量效率的预测值与实际值间的误差平方和最小,为0.114.通过系统的工业化应用可以看出,采用智能控制的悬浮液密度偏差幅度为0.01 g/cm^3,仅为人工控制的1/6,同时密度曲线呈短周期(3～5 min)高频率波动,比人工控制速度提高了4倍左右.此外,密度智能控制系统的稳定运行时间可长达6 h,表明该智能策略在选煤厂具有较好的应用潜力.To obtain steady clean coal ash and higher quantity efficiency of the heavy medium cyclone, the control strategy combined the PID feedback and feedforward of the neural network learning as well as the anti-saturation integral were investigated, which focuses on the multivariable, nonlinear, strong coupling and hysteretic control system in Tangkou Coal Preparation Plant. The intelligent algorithms of multi-input and two output variables were studied, namely, the BP neural network algorithm with multiple input variables for the heavy medium separation process were developed to output the clean coal ash and the quantity efficiency, which was programmed by the Python. The training and simulation of data collected from Tangkou Coal preparation Plant were conducted. The results show that three input variables of feed ash, feed pressure and circulating medium density could result in the least error sum of squares of 0.114 between the predicted and actual values. The industrial application of control system indicates that the deviation amplitude of suspension density is 0.01 g/cm^3, which is only 1/6 of that in manual control. The density curve presents high-frequency fluctuation with short cycle(3-5 min), about 4 times faster than manual control. Moreover, the intelligent control system could run more than 6 h. The above results indicate the excellent application potential of the control strategy.

关 键 词：智能化 控制策略 重介质旋流器 BP神经网络 PYTHON 

分 类 号：TD94[矿业工程—选矿]