摘要
针对双进双出磨煤机料位检测难题,提出了一种基于多信息数据融合的双进双出磨煤机料位检测方法.该方法将粗糙集(RST)和径向基(RBF)神经网络相结合,利用粗糙集数据简约和规则抽取特性,有效地去除大量冗余数据.利用RBF神经网络函数逼近能力更强和收敛速度更快等优点,引入带遗忘因子的梯度下降算法来整定RBF神经网络参数,简化了神经网络结构,提高了神经网络的学习效率,同时拥有自学习和容错能力,从而有效地保证了数据融合的快速收敛性和稳定性.实验结果表明,在料位检测过程中,将两种智能算法相结合所构成的融合系统,能使双进双出磨煤机准确地完成复杂环境的料位检测任务.
Aiming at the difficulty in detecting the material level of BBD ball mill,a material level measurement method based on multi-information data fusion was proposed.The method combined Rough Set（RST） and Radial Basis Function（RBF） neural network and used the RST simple extraction characteristics to remove a large number of redundant data effectively.The RBF neural network function has stronger approximation ability and faster convergence character.The RBF neural network parameters can be tuned by introducing the gradient descent algorithm with forgetting factor.Thus,the structure of neural network gets simplified and the learning efficiency of neural network is improved.At the same time,the self-learning and fault-tolerant ability can be achieved,and the rapid convergence and stability of data fusion are ensured effectively.The experimental results show that the BBD ball mill can accurately complete the material level measurement in complex environment with the fusion system combined two intelligent algorithms.
出处
《沈阳工业大学学报》
EI
CAS
2010年第2期182-186,共5页
Journal of Shenyang University of Technology
基金
国家自然科学基金资助项目(60672078)
辽宁省教育厅基金资助项目(2006T102)
沈阳工业大学博士启动基金资助项目(521102302)
关键词
双进双出磨煤机
料位检测
运行参数分析
粗糙集
径向基神经网络
多信息数据融合
遗忘因子
梯度下降算法
BBD ball mill
material level measurement
operating parameters analysis
rough set
RBF neural network
multi-information data fusion
forgetting factor
gradient descent algorithm
