Nonlinear multivariate filtering and bioprocess monitoring for supervising nonlinear biological processes

On-line monitoring of bioprocesses is crucial to the safe production of high-quality products. However, biological processes tend to have nonlinear behavior patterns that depend on the influent loads, temperature, microorganism activity and so on. Moreover, since biosystems are generally operated under process control systems, data from biosystems tend to be characterized by autocorrelation and dynamic patterns. Although several nonlinear principal component analysis techniques have been recently developed for bioprocess monitoring, no nonlinear monitoring research that considers the bioprocess dynamics has been developed. In order to better monitor bioprocesses, a new dynamic nonlinear monitoring method that combines a kernel principal component analysis (KPCA) and an exponentially weighted moving average (EWMA) is proposed in this research. The kernel functions of KPCA can capture the nonlinearity of bioprocesses and the filtering of EWMA can catch the dynamics of bioprocesses. The proposed method is applied to two case studies: a simple dynamic nonlinear process and a simulation benchmark of a biological treatment process. The simulation results clearly show that the proposed method outperforms other static and linear methods, especially for detecting small shifts in processes.