主成分分析法与核主成分分析法在机械噪声数据降维中的应用比较

摘要

According to the classification principle of linear and non-linear dimensional reduction,this paper dealt with mechanical noise data under different working-modes through PCA and KPCA. Lastly,the paper computed the right recognition percentage of noise data, including had been reduced and not, by NN method and SVM method, and compared the excellence for PCA and KPCA in dimensional reduction. Consequently, a better method of dimensional reduction is selected for ribbed cylindrical double-shells according to the results.%依据线性降维与非线性降维的分类原则,分别选择主成分分析法和核主成分分析法对某双层圆柱壳体在不同工况下的机械噪声数据进行降维;然后使用神经网络和支持向量机两种方法分别计算噪声数据在降维前后的正确识别率,以比较不同降维方法的降维效果,从而确定适合于某双层圆柱壳体机械噪声数据的降维方法.