Department of Chemical and Biochemical Engineering, Technical University of Denmark, Building 229, 2800 Kgs. Lyngby, Denmark.
J Ind Microbiol Biotechnol. 2011 Oct;38(10):1679-90. doi: 10.1007/s10295-011-0957-0. Epub 2011 Apr 2.
The recent process analytical technology (PAT) initiative has put an increased focus on online sensors to generate process-relevant information in real time. Specifically for fermentation, however, introduction of online sensors is often far from straightforward, and online measurement of biomass is one of the best examples. The purpose of this study was therefore to compare the performance of various online biomass sensors, and secondly to demonstrate their use in early development of a filamentous cultivation process. Eight Streptomyces coelicolor fed-batch cultivations were run as part of process development in which the pH, the feeding strategy, and the medium composition were varied. The cultivations were monitored in situ using multi-wavelength fluorescence (MWF) spectroscopy, scanning dielectric (DE) spectroscopy, and turbidity measurements. In addition, we logged all of the classical cultivation data, such as the carbon dioxide evolution rate (CER) and the concentration of dissolved oxygen. Prediction models for the biomass concentrations were estimated on the basis of the individual sensors and on combinations of the sensors. The results showed that the more advanced sensors based on MWF and scanning DE spectroscopy did not offer any advantages over the simpler sensors based on dual frequency DE spectroscopy, turbidity, and CER measurements for prediction of biomass concentration. By combining CER, DE spectroscopy, and turbidity measurements, the prediction error was reduced to 1.5 g/l, corresponding to 6% of the covered biomass range. Moreover, by using multiple sensors it was possible to check the quality of the individual predictions and switch between the sensors in real time.
最近的过程分析技术（PAT）计划更加关注在线传感器，以实时生成与过程相关的信息。然而，特别是对于发酵而言，引入在线传感器远非易事，在线测量生物量就是一个很好的例子。因此，本研究的目的是比较各种在线生物量传感器的性能，其次是展示它们在丝状培养过程早期开发中的应用。作为工艺开发的一部分，共进行了 8 次链霉菌分批补料培养，其中 pH 值、进料策略和培养基成分都有所变化。使用多波长荧光（MWF）光谱、扫描介电（DE）光谱和浊度测量法对培养物进行原位监测。此外，我们还记录了所有经典的培养数据，例如二氧化碳释放率（CER）和溶解氧浓度。基于各个传感器以及传感器组合，对生物量浓度的预测模型进行了估计。结果表明，基于 MWF 和扫描 DE 光谱的更先进的传感器在预测生物量浓度方面并没有比基于双频 DE 光谱、浊度和 CER 测量的更简单的传感器具有优势。通过结合 CER、DE 光谱和浊度测量，预测误差降低到 1.5 g/L，相当于所涵盖的生物量范围的 6%。此外，通过使用多个传感器，可以检查各个预测的质量，并实时在传感器之间切换。
