k-means-RBF集成神经网络在工业尾气检测中的应用
PDF下载 (221)刘 萍,龚雪飞,简家文*,张 帆,陈志芸.k-means-RBF集成神经网络在工业尾气检测中的应用[J].宁波大学学报(理工版),2017,30(1):116-120.DOI:
LIU Ping,GONG Xue-fei,JIAN Jia-wen*,ZHANG Fan,CHEN Zhi-yun.k-means-RBF integrated neural network and its application in industrial exhaust emission detection[J].Journal of Ningbo University(Natural Science & Engineering Edition),2017,30(1):116-120.DOI:
| Title: | k-means-RBF integrated neural network and its application in industrial exhaust emission detection |
| 作者: | 刘 萍, 龚雪飞, 简家文*, 张 帆, 陈志芸 |
| Author(s): | LIU Ping, GONG Xue-fei, JIAN Jia-wen*, ZHANG Fan, CHEN Zhi-yun |
| 关键词: | 工业尾气检测; 径向基函数神经网络; k-means算法 |
| Keywords: | industrial exhaust emission detection; RBF neural network; k-means algorithm |
| 分类号: | TP183 |
| 文献标识码: | A |
| 摘要: | 神经网络是工业尾气检测系统的一个重要组成部分. 为提高神经网络的预测精度和收敛速度, 建立k-means-RBF集成神经网络模型. 首先, 通过选取不同的径向基函数神经网络参数, 得到一组RBF神经网络; 然后, 利用k-means算法对生成的RBF神经网络进行聚类, 并筛选出各类中精度较高的神经网络; 最后, 通过简单平均法对筛选出的神经网络进行集成, 得到高性能的k-means-RBF集成神经网络模型. 为验证模型有效性, 搭建基于k-means-RBF集成神经网络模型的工业尾气检测系统进行验证. 结果表明, 与粒子群算法优化后的Back Propagation (PSO-BP)神经网络模型相比, k-means-RBF集成神经网络模型的平均预测精度提高78.27%, 收敛时间节省99.65% |
| Abstract: | Neural network is very important in detection of industrial exhaust emission. In order to improve stability and prediction accuracy of neural network and reduce the training time of neural network, k-means-RBF integrated neural network model is established based on k-means algorithm. Firstly, a set of RBF neural networks are obtained by choosing different RBF network function parameters. Secondly, the neural network with higher precision is selected from each type. Finally, high performance integration RBF neural network is obtained through integrating RBF neural network using the simple average method. In order to verify the validity of the model, the developed neural network model is used on the industrial exhaust emission detection system. The results show that, compared with the Back Propagation (BP) neural network method optimized by particle swarm optimization (PSO), the average precision predicted by the proposed neural network increases by 78.27%, and the processing time is saved by 99.65% |
| 参考文献 /References: | [1].郑杰. 混合有毒气体检测算法研究[J]. 计算机仿真, 2011, 28(8): 413-416. [2].张志宇, 赵丹国, 侯晓宇. PSO-RBF 神经网络在城市需水量预测中的应用[J]. 水电能源科学, 2013, 31(6): 55-57. [3].ZHOU Z H, WU J, TANG W. Ensembling neural networks: many could be better than all[J]. Artificial ntelligence, 2002, 137(1):239-263. [4].李凯, 黄厚宽. 一种基于聚类技术的选择性神经网络集成方法[J]. 计算机研究与发展, 2005, 42(4):594-598.龚雪飞, 徐景, 孙寿通等. PSO-BP 神经网络在多元有害气体检测中的应用[J]. 传感器与微系统, 2015, (2):154-160. [5].韩红桂, 乔俊飞, 薄迎春. 基于信息强度的RBF神经网络结构设计研究[J]. 自动化学报, 2012, 38(7):1083- 1090. [6].庞振, 徐蔚鸿. 一种基于改进 k-means的RBF神经网络学习方法[J]. 计算机工程与应用, 2012, 48(11):161- 163. [7].谢娟英, 王艳娥. 最小方差优化初始聚类中心的K- means算法[J]. 计算机工程, 2014, 40(8):205-211. [8].张靖, 段富. 优化初始聚类中心的改进k-means算法[J]. 计算机工程与设计, 2013, 34(5):1691-1694. [9].MINKU L L, WHITE A P, YAO X. The impact of diversity on online ensemble learning in the presence of concept drift[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(5):730-742. |
| 备注/Memo: | 收稿日期: 2015?05?21. 宁波大学学报(理工版)网址: http://journallg.nbu.edu.cn/ 基金项目: 国家自然科学基金( 61471210 ) ; 宁波市自然科学基金( 2013A 610002 ) . 第一作者: 刘萍(1989-), 女, 河北保定人, 在读硕士研究生, 主要研究方向: 新型气体传感器及其应用. E-mail: 710610290@qq.com * 通信作者: 简家文( 1967 -) , 男 , 河南淮滨人 , 博导 / 教授 , 主要研究方向 : 敏感材料与传感器制备 . E-mail: jianjiawen@nbu.edu.cn 宁波大学学报(理工版)网址:http://journallg.nbu.edu.cn/ |