基于自适应等距映射算法的软测量建模

Soft sensor modeling based on adaptive Isomap algorithm

针对等距映射(Isomap)算法中的邻域图构造问题,提出1种自适应确定邻域的方法。利用欧氏距离计算样本相似系数。基于各样本的局部密度和平均密度构造密度指数函数。根据密度指数函数自适应调整样本的近邻数,构造合理的邻域图。采用高斯过程回归(GPR)建立模型。将该方法应用于某双酚A生产装置的软测量建模中。仿真结果表明,基于自适应Isomap算法建立的GPR模型比Isomap-GPR模型具有更高的估计精度,均方根误差减小了约15%。

An adaptive neighborhood construction method is proposed for the neighborhood graph construction in the isometric mapping(Isomap)algorithm. The sample similarity coefficient is calculated by using the Euclidean distance. A density exponential function is constructed based on the local density and average density of each sample. The neighbor number of samples is adjusted adaptively according to the density exponential function to construct a reasonable neighborhood graph. A model is developed by using the Gaussian process regression(GPR). This method is applied to the soft sensor modeling of a Bisphenol A production device. The simulation results show that the GPR model based on the adaptive Isomap algorithm has higher estimation accuracy than the Isomap-GPR model,and the root mean square error(RMSE)of the model is reduced by about 15%.