Computer Engineering and Applications ›› 2008, Vol. 44 ›› Issue (16): 157-159.

• 数据库、信号与信息处理 • Previous Articles     Next Articles

Trend prediction of non-stationary vibration signals based on Empirical Mode Decomposition and Least Square Support Vector Machine

WANG Hong-bo,ZHU Qi-bing   

  1. School of Communication and Control Engineering,Jiangnan University,Wuxi,Jiangsu 214122,China
  • Received:2007-09-11 Revised:2007-11-26 Online:2008-06-01 Published:2008-06-01
  • Contact: WANG Hong-bo

基于EMD和LS-SVM的非平稳振动信号趋势预测

王洪波,朱启兵   

  1. 江南大学 通信与控制工程学院,江苏 无锡 214122
  • 通讯作者: 王洪波

Abstract: The trend forecasting of vibration signals is an important content of condition monitoring and fault diagnosis.The old method of identification of machinery system is not practicable because the non-linear and non-stationary character is becoming more and more prominent.A prediction modelling method based on Empirical Mode Decomposition(EMD) and Least Square Support Vector Machine(LS-SVM) is proposed.Firstly,the trend time series is adaptively decomposed into a series of stationary Intrinsic Mode Functions(IMF) in different scale space using EMD.Then the right parameter and kernel functions are chosen to build different LS-SVM respectively to each and every IMF.Finally,these forecasting results of each IMF are combined to obtain final forecasting result.The simulation results show that the hybrid method has faster speed,higher precision and greater generalization ability than that of the single LS-SVM method.

Key words: trend prediction, non-linear and non-stationary time series, Least Squares Support Vector Machine(LS-SVM), Empirical Mode Decomposition(EMD)

摘要: 镇动信号的趋势预测是设备状态监测与故障诊断中的一个重要内容。随着运行设备的非线性、非平稳特点越来越明显,传统的数学建摸预报方法已不能满足设备的复杂化和现代化要求。提出了一种基于经验模式分解EMD(Empirical Mode Decomposition)和最小二乘支持向量机LS-SVM(Least Square Support Vector Machine)的新模型。首先,运用 EMD 将趋势时间序列自适应地分解成一系列不同尺度的本征模式分量IMF(intrinsic mode function);其次,对每个本征模式分量,采用合适的核函数和超参数构造不同的LS-SVM 进行预测;最后对各分量的预测值进行拟合得到最终的预测值。仿真实验表明,此方法与单一的LS-SVM预测法相比,具有较高的精度和较强的推广能力。

关键词: 趋势预测, 非线性、非平稳时间序列, 最小二乘支持向量机(LS-SVM), 经验模式分解(EMD)