摘要
针对核函数方法中单个核函数的局限性,以及PLS非线性处理能力差的特点,提出混合核函数PLS建模方法,以提高模型的推广能力和非线性处理能力。混合核函数集中了多个局部和全局核函数,兼具局部和全局特性,并可以通过参数调节局部和全局核函数对混合核函数的作用,将过程的先验知识融入到核函数的确定,进而适合具有不同数据特征的工业过程。工业丙烯腈收率软测量建模的应用表明,混合核函数PLS软测量模型具有较好的数据适应性和非线性特性,满足了工业应用要求。
With regard to the limitations of single kernel and poor performance of nonlinearity of partial least squares( PLS), mixtures of kernels PLS is proposed and this method is applied to the modeling of chemical process to improve the nonlinearity and generalization ability of the model. Mixtures of kernels are characterized by both global and local ability because it is combined by several local and global kernels. The prior knowledge of process is incorporated in the determination of the kernels by tuning the parameters which change the influence of local and global kernels on the mixtures of kernels, thus makes this method more suitable for applications with different data distribution. Soft sensing modeling using this method to the selectivity to aerylonitrile process have shown that such method is superior to other methods and meets the demands of field applications.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2007年第2期239-242,共4页
Computers and Applied Chemistry
关键词
核函数
部分最小二乘回归
推广能力
软测量
kernels, partial least squares(PLS), generalization ability, soft sensing