基于KIsomap特征提取的软测量建模方法

A soft sensor modeling method nased on KIsomap feature extraction

来自化工生产过程的数据大多具有非线性和高维性,对数据进行特征提取是软测量建模过程的必要环节。流形学习作为一种非线性维数约简方法,可以获得高维数据在低维空间的嵌入。针对流形学习中的等距映射法(Isomap)鲁棒性差、拓扑稳定性不好等缺点,通过常数偏移的方法构造核矩阵,形成核等距映射法(KIsomap),提高了Isomap算法的鲁棒性和拓扑稳定性。运用一种将K近邻与ε-半径法相结合的方法构造邻域图,基于核等距映射法(KIsomap)对数据进行特征提取,并建立高斯过程回归软测量模型,提高了模型的泛化能力与学习效率。将该方法应用于某双酚A装置的软测量建模,仿真结果表明相比于其他特征提取的软测量建模方法,该方法具有更高的估计精度和学习效率。

Data from the process of chemical engineering is normally nonlinear and high dimension, so the feature extraction of data is an indispensable step in soft sensor modeling. Manifold learning as a nonlinear dimension reduction method can take high-dimensional data embedded to low-dimensional space. Aiming at the poor robustness, poor topology and other shortcomings of isometric mapping(Isomap) in manifold learning,a construction method of kernel matrix is proposed with constant translation to improve the robustness and topological stability of Isomap algorithm. In addition, a neighborhood graph is constructed by K-nearest neighbor algorithm with the -radius method, so that the data is extracted by kernel isometric mapping(KIsomap), and then a soft sensor model is built by Gaussian process regression so as to improve the generalization ability and learning efficiency of the model. Finally, the method is applied to the soft sensor modeling for a Bisphenol A production plant. The simulation results show that the proposed method has higher estimation accuracy and learning efficiency compared with other methods of feature extraction.