基于PCA-SVM的柴油凝点近红外光谱软测量法

PCA-SVM-based Near-infrared Spectroscopy(NIRS) Soft Sensing Method for Condensation Point of Diesel Fuel

提出一种基于PCA-SVM及近红外光谱(NIRS)分析技术的柴油凝点软测量方法。首先,采用多项式卷积对原始的柴油NIRS数据进行光谱平滑、基线校正和标准归一化;然后,利用主元分析(PCA)对近红外光谱数据集的高维特征进行组合并向低维空间投影;最后,利用SVM回归算法建立凝点的软测量模型。与BP、SVM及PCA-BP方法相比,实验结果表明所提方法具有更高的测量精度,且与标准方法测量的结果更为接近,因此适合柴油凝点的在线测量。

The soft sensing method based on PCA-SVM and NIRS analysis technology for condensation point of diesel fuel is presented. Firstly, the polynomial convolution algorithm was used in spectrogram smoothness, baseline correction and standardization for original NIRS data of diesel fuel. Then, using principal component analysis （PCA） , the NIRS data sets with higher dimensional feature were associated and projected to lower dimensional space. Finally, the soft sensing model for condensation point was built by adopting SVM regression algorithm. Experimental results show that comparing with BP, SVM and PCA-BP, the proposed method offers higher precision, and is closer to the standard measuring result, so it fits ouline measurement of condensation point.