On-line monitoring of substrates and biomass using near-infrared spectroscopy and model-based state estimation for enzyme production by S. cerevisiae

摘要

In the early process development, feeding trajectories and cultivation conditions are altered in order to maximise the desired product. This is in contrast to industrial cultivations, where variations are to be eliminated. In order to fully understand and safely run a new process, real-time information are beneficial. Near-infrared spectrometer that are coupled to a fermenter offer the possibility to gather information on-line. The NIR data can be transformed via partial least squares modeling to estimate substrate and biomass concentrations. At low concentrations and under changed cultivation conditions during process development, however, the estimates may differ from reality. This uncertainty may be reduced by integrating biological and physico-chemical knowledge in the on-line estimation. In this contribution, we present a hybrid approach of near-infrared (NIR) spectroscopy and nonlinear model-based state estimation to enable an improved quality in the on-line estimation of substrates and biomass in a yeast cultivation. As only three cultivations are needed for calibration, on-line state estimation is available in the early stage of development for this process. This approach is compared to the use of both methods separately for estimation of biomass, ammonium, glucose, phosphate and ethanol in cultivations of S. cerevisiae.