Development and testing of mid-infrared sensors for in-line process monitoring in biotechnology

作者：Bogomolov Andrey^*; Hessling Martin; Wenzel Ulla; Princz Sascha; Hellmuth Thomas; Bernal Maria J Barraza; Sakharova Tatiana; Usenov Iskander; Artyushenko Viacheslav; Meyer Hans

来源：Sensors and Actuators B: Chemical , 2015, 221: 1601-1610.

DOI：10.1016/j.snb.2015.07.118

• Mid-infrared sensor
• Pyroelectric detection
• Fabry-Perot interferometer
• Attenuated total reflection
• Polycrystalline infrared fiber
• Saccharomyces cerevisiae fermentation

Three prototypes of mid-infrared (MIR) spectrometric sensor systems for simultaneous monitoring of ethanol and carbohydrates (in the present case glucose and fructose) in the course of biotechnological processes have been constructed based on recent developments in pyroelectric detection and fiber photonics. The sensors utilized were a grating spectrometer or a Fabry-Perot interferometer adjusted for the detection of analytes' characteristic absorbance bands in the spectral region of "fingerprints" between 1050 and 950 cm(-1). The measurements were performed with an attenuated total reflection (ATR) probe connected to the spectrometer by a polycrystalline infrared fiber (PIR). Two probes with different ATR elements were tested: with a diamond crystal (for both spectrometers) and with a detachable PIR loop head (for grating spectrometer). The sensor performances were assessed and compared using partial least-squares (PIS) regression modeling and prediction statistics for two designed sample sets of binary ethanol-glucose and glucose-fructose aqueous solutions. The models based on the FT-IR spectroscopic analysis of the same designed samples using a diamond ATR probe (a "gold standard" method) were used as a benchmark. The system based on a grating spectrometer connected to an ATR probe with a PIR loop head was additionally tested under the process conditions of Saccharomyces cerevisiae fermentation. The resulting root mean-square errors of prediction were 4.74 and 13.33 g/L, for ethanol and glucose models, respectively. Simultaneously, NIR spectroscopy in the range 1100-2100 nm was used both for the analysis of designed samples and for the fermentation process monitoring. In the latter case a biomass content prediction model has been built along with those for ethanol and glucose. All tested full-spectroscopic and sensor-based methods of analysis have been compared and their practical applications discussed.

• 出版日期2015-12-31