基于改进灰色神经网络的液压泵寿命预测

摘要/Abstract

摘要：

改进了GM(1，1)模型，提高了其精度和适应范围;将改进的GM(1，1)模型与神经网络预测模型相结合来构建灰色神经网络组合预测模型;提出了基于支持向量机的液压泵寿命特征启发式搜索策略，以液压泵寿命特征参数特征集的交叉验证错误率为评价指标，从液压泵的特征参数（振动、压力、流量、温度、油液信息等）中选取寿命特征因子;运用小波阈值降噪法进行降噪处理，提取典型的小波包能量特征作为模型的输入。以齿轮泵为例，将改进的灰色神经网络预测模型与原始GM(1，1)模型和改进GM(1，1)模型比较可知，灰色神经网络预测模型预测精度最高，达到98.42%。

关键词: 液压泵;寿命预测;GM(1, 1)模型;神经网络;支持向量机

Abstract:

A life prediction method of hydraulic pump based on improved grey neural network was presented for the shortcomings of low precision forecasting model. Firstly, a new model was proposed based on the combination of the initial condition and the background value to improve the precision of the grey forecasting model. A SVM-based hydraulic pump lifetime feature heuristic elimination strategy was put forward, and the evaluation criterion of cross-validating error rate was adopted to select life feature form hydraulic pump features(vibration, pressure, flux, temperature, oil, and so on). Feature signals were de-noised by wavelet threshold de-noising method. Then representative energy features were selected by wavelet pocket energy spectrum algorithm. Taking gear pump as an example, the improved grey neural network model has higher precision than original GM(1，1) model and improved GM(1，1) model. 

Key words: hydraulic pump, life prediction, GM(1, 1) model, neural network, support vector machine(SVM)

中图分类号:

何庆飞, 陈桂明, 陈小虎, 姚春江. 基于改进灰色神经网络的液压泵寿命预测[J]. 中国机械工程, 2013, 24(4): 500-506.

HE Qiang-Fei, CHEN Gui-Meng, CHEN Xiao-Hu, TAO Chun-Jiang. Life Prediction of Hydraulic Pump Based on an Improved Grey Neural Network[J]. China Mechanical Engineering, 2013, 24(4): 500-506.

参考文献

[1]党耀国[1],刘思峰[1],刘斌[1].以x^（1）（n）为初始条件的GM模型[J].中国管理科学,2005,13(1):132-135.
[2]魏勇[1],张怡[2].灰色模型的最优化及其参数的直接求法[J].数学的实践与认识,2006,36(12):203-207.
[3]何文章 [1],宋国乡 [2],吴爱弟 [3].估计GM（1，1）模型中参数的一族算法[J].系统工程理论与实践,2005,25(1):69-75.
[4]何庆飞[1],陈桂明[1],陈小虎[1],姚春江[1],杨庆[1].基于改进灰色预测模型的液压泵寿命预测[J].润滑与密封,2011,36(7):27-31.
[5]魏勇[1],胡大红[1].光滑性条件的缺陷及其弥补办法[J].系统工程理论与实践,2009(8):165-170.
[6]李源[1,2],韩旭[1],姜潮[1],王林军[1].一种基于灰色预测模型的预测桁架结构疲劳寿命的数值方法[J].中国机械工程,2011,22(6):710-714.
[7]刘威[1],赵选民[1],邓春龙[2],李文军[2].灰色神经网络模型在海水腐蚀预测中的应用[J].中国腐蚀与防护学报,2008,28(4):201-204.
[8]Harpham C,Dawson C W. The Effect of DifferentBasis Functions on a Radial Basis Function Networkfor Time Series Prediction: A Comparative Study[J], Neurocomputing,2006,69(16) :2161-2170.
[9]李军红[1],周天瑞[1],郑荣[1].基于神经网络的冷轧带肋钢筋机械性能预测[J].中国机械工程,2006,17(15):1580-1582.
[10]刘思峰,谢乃明等.灰色系统理论及其应用[M].4版.北京:科学出版社,2008.
[11]罗党,刘思峰,党耀国.灰色模型GM（1，1）优化[J].中国工程科学,2003,5(8):50-53.
[12]Yuan Y L, Zhong L, Jiang Q. A Study on GreyRBF Prediction Model[C]// Proceedings of 2005International Conference on Machine Learning andCybernetics. Guangzhou, 2005 :4140-4143.
[13]刘玉兵[1],张宗扬[2],吴国军[1].基于径向基函数神经网络的发动机磨损预测分析[J].润滑与密封,2009,34(1):71-72.
[14]许东,吴铮.基于MATLAB6.X的系统分析与设计一神经网络[M]. 2版.西安:西安电子科技大学出版社,2003.
[15]Chen Y W?Lin C J. Combining SVMs with Vari-ous Feature Selection Strategies [ J]. Studies inFuzziness and Soft Computing,2006,207 : 319-328.
[16]Kuo M,Sklansky J. Comparison of Algorithmsthat Select Features for Pattern Slassifiers [J].Pattern Recognition, 2000,33(1) : 25-41.
[17]Sun Z H, Bebis Z H. Object Detection Using Fea-ture Subset Selection [ J]. Pattern Recognition,2004,37(11) : 2165-2176.
[18]VapnikV N. An Overview of Statistical LearningTheory [J], IEEE Transactions on Neural Net-works,1999,10(5);988-999.
[19]Huang R Q,Xi L F,Li X L,et al. Residual LifePredictions for Ball Bearings Based on Self-organ-izing Map and Back Propagation Neural NetworkMethods[J]. Mechanical Systems and Signal Pro-cessing,2007,21 (1) : 193-207.
[20]Chen Zhixin,Xu Jinwu?Yang Debin. New Methodof Extracting Weak Failure Information in Gearboxby Complex Wavelet De- noising [J]. ChineseJournal of Mechanical Engineering, 2008 .21(4):87-91.
[21]王奉涛 [1],马孝江 [2],邹岩崑 [3],张志新 [4].基于小波包分解的频带局部能量特征提取方法[J].农业机械学报,2004,35(5):177-180.