AbstractFluorescence spectroscopy can provide high-level chemical characterization and quantification that is suitable for use in online process monitoring and control. However, the high-dimensionality of excitation–emission matrices and superposition of underlying signals is a major challenge to implementation. Herein the use of Convolutional Neural Networks (CNNs) is investigated to interpret fluorescence spectra and predict the formation of disinfection by-products during drinking water treatment. Using deep CNNs, mean absolute prediction error on a test set of data for total trihalomethanes, total haloacetic acids, and the major individual species were all < 6 µg/L and represent a significant difference improved by 39–62% compared to multi-layer perceptron type networks. Heat maps that identify spectral areas of importance for prediction showed unique humic-like and protein-like regions for individual disinfection by-product species that can be used to validate models and provide insight into precursor characteristics. The use of fluorescence spectroscopy coupled with deep CNNs shows promise to be used for rapid estimation of DBP formation potentials without the need for extensive data pre-processing or dimensionality reduction. Knowledge of DBP formation potentials in near real-time can enable tighter treatment controls and management efforts to minimize the exposure of the public to DBPs.
Fluorescence spectroscopy offers a cheap, simple, and fast approach to monitor poly(3-hydroxybutyrate) (PHB) formation, a biodegradable polymer belonging to the biodegradable polyester class polyhydroxyalkanoates. In the present study, a fluorescence and side scatter-based spectroscopic setup was developed to monitor in situ biomass, and PHB formation of biotechnological applied Cupriavidus necator strain. To establish PHB quantification of C. necator, the dyes 2,2-difluoro-4,6,8,10,12-pentamethyl-3-aza-1-azonia-2-boranuidatricyclo[7.3.0.03,7]dodeca-1(12),4,6,8,10-pentaene (BODIPY493/503), ethyl 5-methoxy-1,2-bis(3-methylbut-2-enyl)-3-oxoindole-2-carboxylate (LipidGreen2), and 9-(diethylamino)benzo[a]phenoxazin-5-one (Nile red) were compared with each other. Fluorescence staining efficacy was obtained through 3D-excitation-emission matrix and design of experiments. The coefficients of determination were ≥ 0.98 for all three dyes and linear to the high-pressure liquid chromatography obtained PHB content, and the side scatter to the biomass concentration. The fluorescence correlation models were further improved by the incorporation of the biomass-related side scatter. Afterward, the resulting regression fluorescence models were successfully applied to nitrogen-deficit, phosphor-deficit, and NaCl-stressed C. necator cultures. The highest transferability of the regression models was shown by using LipidGreen2. The novel approach opens a tailor-made way for a fast and simultaneous detection of the crucial biotechnological parameters biomass and PHB content during fermentation.
• Intracellular quantification of PHB and biomass using fluorescence spectroscopy.
• Optimizing fluorescence staining conditions and 3D-excitation-emission matrix.
• PHB was best obtained by LipidGreen2, followed by BODIPDY493/503 and Nile red.