LS-SVM在基于小波变换的模态分析中端部效应的应用

系统地阐述了运用改进的Morlet小波进行模态参数识别的方法。运用小波熵对小波参数进行了优化选择从而可以进行密频模态的识别,针对小波分析时产生的端部效应问题,提出了运用最小二乘支持向量机（LS-SVM）对小波骨架进行预测延拓的方法,经预测分析后可获取较准确的模态参数。通过仿真及实验信号的验证分析,表明基于LS-SVM方法可以有效地消除端部效应,且其准确效果优于基于RBF的神经网络和时变自回归的预测方法。

Application of LS-SVM to the End Effect of Modal Analysis Based on Wavelet Transformation

A method that the improved Morlet wavelet was used to identify the modal parameter was described systematically, and the wavelet parameter was optimized by the wavelet entropy. In the meantime, in order to resolving the end effect of the wavelet analysis a method that the LS-SVM was used to predict the wavelet skeleton was put forward. The results of simulation analysis and experimental analysis demonstrate that the time series forecasting method based on LS-SVM is applied to resolve the end effect of modal analysis based on Morlet wavelet is effective to identify the modal parameter accurately and restrain the end effect. And the accuracy is the highest among methods based on radial basis function(RBF) neural network and based on time-varying autoregressive(TVAR).