摘要
针对LS-SVM算法中小波提取特征存在小波基函数选择和小波分解层次、系数选取的问题,提出了一种基于因子分析技术的故障特征识别方法;该方法通过构建采样数据的相关矩阵求出因子载荷和因子得分,按照累计贡献率自动提取出1~3个因子组成特征向量,从而降低了输入维度,提高了算法训练诊断效率,降低了收敛难度;四运放典型电路的仿真实验结果表明：文中算法的诊断正确率超过了同类方法,同时提高了训练时间和诊断效率。
This paper presents a fault feature recognition method based on factor analysis techniques for wavelet feature extraction in LS-SVM algorithm existing problem of wavelet bases function selection,wavelet decomposition level and coefficient selection.The method computes factor loadings and factor scores by constructing a correlation matrix of sample data,extract factors 1-3to compose feature vector automatically according to the cumulative contribution rate,thereby reduce the dimension of the input,improve the efficiency of training and diagnostic algorithm,reduce the convergence difficulty.The simulation results of four op-amp biquad high-pass filter show：The diagnostic accuracy of the algorithm in this paper is beyond similar methods,while increasing the training time and the efficiency of diagnosis.
出处
《计算机测量与控制》
北大核心
2014年第11期3470-3472,共3页
Computer Measurement &Control
关键词
因子分析
故障特征
因子得分
特征向量
factor analysis
fault feature
factor score
feature vector
