作者：张文涛1,2, 钟文晶2, 胡伯勇2, 陆豪强2, 马永光1, 董子健1
作者单位：1. 华北电力大学控制与计算机工程学院, 河北 保定 071003;
2. 浙江浙能技术研究院有限公司, 浙江 杭州 310000
关键词：制粉系统;邻域粗糙集;相关向量机;故障诊断
摘要：
针对火电厂制粉系统的故障征兆参数复杂、不易诊断的特点，提出一种基于邻域粗糙集（NRS）与相关向量机（RVM）的故障诊断方法。该方法首先利用邻域粗糙集约简输入的特征向量，并将约简得到的最优决策表作为RVM的输入，采用组合核函数代替传统的单一核函数，利用网格搜索和交叉验证的方法确定最佳的核函数参数和组合核系数，建立二叉树RVM多分类模型，从而进行制粉系统故障识别和诊断。实验结果表明，该方法故障诊断准确率可达95%，且泛化能力强。
Fault diagnosis based on neighborhood rough set and RVM for pulverizing system
ZHANG Wentao1,2, ZHONG Wenjing2, HU Boyong2, LU Haoqiang2, MA Yongguang1, DONG Zijian1
1. School of Control and Computer Engineering, North China Electric Power University, Baoding 071003, China;
2. Zhejiang Energy Group Research Institute, Hangzhou 310000, China
Abstract: On account of the complex and difficult diagnosis characteristic of the fault sign parameters of the power plant, a fault diagnosis method based on the neighborhood rough set (NRS) and the relevance vector machine (RVM) is proposed. Firstly, the feature vectors of input are reduced by neighborhood rough set, and the optimal decision table is used as the input of RVM. Then the combination kernel function is used instead of the traditional single kernel function. The grid search and cross validation are used to determine the optimal parameters of the kernel function and the combination kernel coefficients. Finally, the two fork tree RVM multi-classification model is established, therefore it can be applied for the fault recognition and diagnosis of pulverizing system. The experimental results show that the accuracy of fault diagnosis can reach 95%, and the generalization ability is strong.
Keywords: pulverizing system;neighborhood rough set (NRS);relevance vector machine (RVM);fault diagnosis
2019, 45(8):151-155 收稿日期: 2018-05-08;收到修改稿日期: 2018-06-20
基金项目: 中央高校基本科研业务费专项资金（9160316004）
作者简介: 张文涛(1992-),男,山东泰安市人,硕士研究生,专业方向为发电设备的故障诊断
参考文献
[1] 钱虹,茅大钧,叶建华,等.火电厂磨煤机出口温度异常分析和故障诊断系统[J].自动化仪表, 2009(7):22-24
[2] 王禹新,牛玉广.基于专家系统的制粉系统故障诊断的研究[J].仪器仪表与分析测, 2010(3):1-3
[3] 刘定平,叶向荣,陈斌源,等.基于核主元分析和最小二乘支持向量机的中速磨煤机故障诊断[J].动力工程, 2009(2):155-158
[4] 陈斌源,朱军.基于径向基函数神经网络的中速磨煤机故障诊断[J].发电设备, 2011, 25(5):323-326
[5] 费树岷,李延红,柴琳.基于RSPNN的制粉系统故障诊断[J].控制工程, 2012, 19(3):412-415
[6] SHEN L, TAY F, QU L, et al. Fault diagnosis using rough sets theory[J]. Compute Ind, 2000, 43(1):61-72
[7] 王国胤,姚一豫,于洪.粗糙集理论与应用研究综述[J].计算机报, 2009, 32(7):1229-1245
[8] TIPPING M E. Sparse bayesian learing and the relevance vector machine[J]. Journal of Machine Learing Research, 2001, 1(3):211-244
[9] 张亚男,杨慧中.基于RVM组合核优化的软测量模型研究[J].系统仿真学报, 2018(1):272-277
[10] 胡清华,于达仁,谢宗霞.基于邻域粒化和粗糙逼近的数值属性约简[J].软件学报, 2008, 19(3):640-649
[11] 江峰,王莎莎,杜军威,等.基于近似决策熵的属性约简[J].控制与决策, 2015, 30(1):65-70
[12] 卢锦玲,绳菲菲,赵洪山.基于相关向量机的风机齿轮箱故障诊断方法[J].华北电力大学学报(自然科学版), 2017, 44(2):91-96
[13] PSORAKIS I, DAMOULAS T, GIROLAMI M A. Multiclass relevance vector machines:an evaluation of sparsity and accuracy[J]. IEEE Transactions on Neural Networks, 2010, 21(10):1588-1598
[14] 薛宁静.多类支持向量机分类器对比研究[J].计算机工程与设计, 2011, 32(5):1792-1795
[15] 樊帅,肖军.锅炉制粉系统故障诊断方法[J].热力发电, 2015, 44(2):13-17,23
