Abstract:

As widely used process monitoring techniques，principal component analysis（PCA）and partial least squares（PLS）are limited to the application in linear and time-invariant systems.To handle the nonlinear and time-varying characteristics of real processes，a recursive kernel PCA（RKPCA）algorithm was proposed for adaptive monitoring of nonlinear processes.By extending the incremental singular value decomposition（SVD）to the kernel space，the kernel formulation of incremental kernel PCA，which possessed much lower computational complexity and was suitable for online model updating，was obtained.Finally，the proposed algorithm was applied to the Alstom gasifier for adaptive monitoring and RKPCA could efficiently capture the time-varying and nonlinear relationship in process variables.