Abstract:Previous studies on batch microbial fermentation usually considered data maximization but lack of data cluster structure information. A Multi-way Kernel Entropy Component Analysis (MKECA) method was proposed to solve this problem, which overcome the drawbacks of traditional monitoring methods on high monitoring failure rates. The AT method was first used for historical data preprocessing and mapping data from low-dimensional space to high dimensional feature space to solve data nonlinearity. Data in the high dimensional feature space was moved to lower dimension based on the size of the data kernel entropy, in order to keep the original data distribution. Meanwhile, the proposed method was equivalent to the traditional method under certain conditions. Penicillin simulation data verifies that MKECA is more reliable and accurate which may have broad potential applications.

Batch Process Monitoring for Microbial Fermentation Based on Multi-Way Kernel Entropy Component Analysis