Biochemical processes often results in highly nonlinear concentration-time profiles during the fed-batch fermentation. Considering the inherent complexity associated with these processes, it becomes difficult to implement a model-based control scheme successfully. This particular lacuna has rushed a welter of studies in process controller area. Accounting for the recent advancements in Data Analytics, Voronoi Tessellation has been established as a well-proven theory, with huge application in fields like systems biology. We have developed a Voronoi Tessellated Neural Network (VTNN) based on a wide operating range of substrate and DO set points and modulated its use as a feed-forward process controller to track the dynamics of <i>Corynebacterium lilium</i> (M 128) fermentation.  The VTNN training dataset has been generated through simulation using a theoretical model. The performance of the VTNN controller has been compared with existing controllers based on Proportional Integral Derivative (PID) and Artificial Neural Network (ANN) based control schemes. Stability analysis for the controllers has also been done by introducing systematic and random perturbations in the system. It is observed that the VTNN controller is more robust than its counterparts and exhibits a stable tracking of the substrate and DO setpoint concentration trajectories in the presence of measurement noise. The controller has also been tested with highly variant offline datasets to test for its adaptation performance. The VTNN can adapt to the perturbations robustly. In conclusion, it can be perceived that the online implementation of VTNN for a general bioprocess is practicable.<br/>

Application of voronoi tessellated neural networks (VTNN) for bioprocess control

2019 Spring National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.