Microbial fermentation process monitoring based on higher order cumulant statistics

Peng CHANG,Pu WANG,Xue-jin GAO
DOI: https://doi.org/10.13195/j.kzyjc.2016.1261
2017-01-01
Abstract:The essence of the traditional multiway kernel independent component analysis(MKICA) method is that the independent component analysis(ICA) whitened principal component analysis(PCA) is replaced with KPCA by using second order statistics of the monitoring and controlling process, not by the stage characteristic of process data and higher-order cumulant information. To solve this problem, the high order cumulant analysis(HCA) and multiway kernel entropy independent component analysis(MKECA) are combined, and the analysis of high order cumulant multiway kernel entropy independent component analysis(HCA-MKEICA) method is proposed. Firstyly, the kernel entropy independent component analysis(KECA) method is used for original data conversion to solve the problem of nonlinear. Then, in the high-dimensional kernel entropy space, the HCA technology is used to construct the new statistics for process monitoring. Finally, the proposed method is applied to the microbial fermentation process, and the comparison results with the traditional methods show that the proposed method can achieve a better detection, and verify its effectivess.
What problem does this paper attempt to address?