一种基于高斯过程的非线性PLS建模方法

提出了一种基于高斯过程(GP)和偏最小二乘法(PLS)的非线性PLS方法(GP-PLS),以更加有效地处理过程非线性、多输入和数据共线性等复杂特性,提高模型的推广能力和精度。该方法首先采用PLS进行特征提取,再用GP建立PLS的内部模型,因而具有GP与PLS的优点。对工业丙烯腈生产过程丙烯腈收率软测量建模的应用表明,采用该方法建立的软测量模型在模型精度、推广能力等方面明显优于一些传统软测量建模方法,满足工业现场应用要求。

A Nonlinear Partial Least Square Modeling Method Based on Gaussian Process

A new nonlinear partial least squares method(GP-PLS) based on Gaussian process is(proposed) to deal with complicated processes with nonlinearities and a large number of correlated inputs.The GP-PLS method,which has merits of both GP and PLS,is an integration of GP models and partial least squares.The PLS outer projection is used as a dimension reduction tool to remove collinearity and the GP models are trained to capture the nonlinearities in the projected latent space.Soft sensor modeling of acrylonitrile yield using GP-PLS method is established.It is found that the generalization ability and the accuracy of the soft sensor using the method proposed are superior to traditional methods,and the performance of the soft sensor meets the demands of industrial application in the field.