Search

Pixian broad bean paste (PBP) is a popular fermentation condiment known in home and abroad. Aroma is a significant index for evaluating PBP quality during fermentation process. Hence, in this study, electronic nose (E-nose) system combined machine learning methods were applied for PBP quality characterization. The machine learning methods including partial least squares discriminant analysis (PLS-DA), partial least squares regression (PLSR), support vector machine (SVM), random forest (RF), and artificial neural networks (ANN) were introduced for qualitatively discriminating fermentation time and quantitatively analyzing the contents of key aromas of PBP samples. The PLS-DA result indicated that it is feasible to identify the fermentation stages of PBP samples by E-nose and a classification accuracy of 99% could be achieved. As for the quantitative prediction modelling, ANN exhibited preferable performance than PLSR, SVM and RF for analyzing the contents of phenethyl alcohol (R2= 0.846, RMSE = 10.270), isoamyl alcohol (R2= 0.940, RMSE = 6.857), 3-methylthiopropanol (R2= 0.910, RMSE = 2.205), benzaldehyde (R2= 0.824, RMSE = 4.172), furfural (R2= 0.902, RMSE = 2.066), 4-ethyl guaiacol (R2= 0.877, RMSE = 11.249) and 4-ethylphenol (R2= 0.913, RMSE = 12.754).

Compared with current global navigation satellite system (GNSS), which consists of medium- and high-altitude orbit satellites, low Earth orbit (LEO) satellites move faster and can greatly improve the observing geometry. In this contribution, LEO constellation augmented GNSS (LeGNSS) troposphere estimation is investigated, and the impacts of relevant factors are analyzed in detail. When the temporal resolutions of zenith troposphere delay (ZTD) and horizontal gradients are 1 and 2 h, while standard deviations (STDs) of phase and pseudorange observations at the zenith direction are 0.005 and 0.5 m, the accuracies of ZTD, north gradient, and east gradient augmented by LEO constellation improve by 15.7%, 29.6%, and 16.4%, respectively, compared with GNSS solution. The results of troposphere estimation under obstructed environment and using low-cost dual frequency receivers also become more robust after adding LEO observations. Most importantly, analysis results during periods with rapid varying troposphere parameters suggest that the contribution of LEO constellation becomes even bigger with increasing temporal resolutions of ZTD and horizontal gradients, indicating that LeGNSS can be used to extract tropospheric parameters with improved accuracies at high temporal resolution. Thus, the capability in rapidly capturing severe weather events can be improved by LeGNSS observations, and the performance can become even better with an increasing number of LEO satellites. All these results suggest that LeGNSS might be an important tool in improving the performance of troposphere estimation, and would promote GNSS application in water vapor monitoring.

Near-infrared (NIR) spectroscopy coupled with chemometric methods were employed for quantitatively analyzing broad bean paste quality during fermentation. The quality parameters including the contents of moisture, total acid (TA), and amino acid nitrogen (AAN) were determined by traditional chemical methods as reference. Savitzky–Golay first derivative (FD), Savitzky–Golay second derivative, Savitzky–Golay smoothing, standard normal variate, and multiplicative scatter correction (MSC) were employed to preprocess the raw spectrums. Furthermore, the pretreated spectrums were selected by successive projections algorithm for dimension reducing. To investigate the quantitative relationship between quality parameters and spectrums, linear models of partial least squares regression and multiple linear regression, as well as non-linear models of random forest, gradient boosting decision tree, and adaptive boosting (AdaBoost) were introduced for regression modeling. The results revealed that the MSC was the best preprocessing method for the NIR spectrums of broad bean paste. Meanwhile, compared with linear modeling, non-linear models had better performances for the prediction of the parameters. In addition, among all the three non-linear methods, AdaBoost provided the optimal regression results with satisfactory prediction statistics for moisture (Rp2= 0.963, RMSEP = 0.416), TA (Rp2= 0.917, RMSEP = 0.030) and AAN (Rp2= 0.908, RMSEP = 0.011) contents. This study demonstrated that NIR spectroscopy could be applied as a rapid and effective tool for quantitatively analyzing the quality parameters of broad bean paste during the fermentation process.

Promising progress in the field of icephobicity has been made in the recent years. However, a majority of the reported icephobic surfaces rely on static mechanisms, and they maintain low ice adhesion on surfaces at extreme temperatures (as low as −60 °C), which is highly challenging. Dynamic anti-icing surfaces, which can melt ice or change the ice–substrate interfaces from the solid to liquid phase after the formation of ice, serve as a viable alternative. In this study, liquid layer generators (LLGs), which can release ethanol to the ice–solid interface and convert the ice–substrate contact from the solid–solid mode to the solid–liquid–solid mode, were introduced. Excellent icephobicity on surfaces with an ethanol lubricating layer was found to withstand extremely low temperatures (−60 °C), which was proven by both molecular dynamics simulations and experiments. Two prototypes of LLGs, one by packing ethanol inside and the other by storing replenishable ethanol below the substrate, were fabricated. These LLGs could constantly release ethanol for a maximum of 593 days without source replenishment. Both these prototypes exhibited super-low ice adhesion strengths of 1.0–4.6 kPa and 2.2–2.8 kPa at −18 °C. For select samples, by introducing an interfacial ethanol layer, the ice adhesion strength on the same surfaces decreased in an unprecedented manner from 709.2–760.9 kPa to 22.1–25.2 kPa at a low temperature of −60 °C.

It is anticipated that the performance of real‐time (RT) GNSS meteorology can be further improved by incorporating observations from multiple Global Navigation Satellite System (GNSS), including GPS, GLONASS, Galileo, and BeiDou. In this paper, an operational RT system for extracting zenith troposphere delay (ZTD) using a modified version of the Precise Point Positioning With Integer and Zero‐difference Ambiguity Resolution Demonstrator (PPP‐WIZARD) was established. GNSS, including GPS, GLONASS, and Galileo, observation streams were processed using RT Precise Point Positioning (PPP) strategy based on RT satellite orbit/clock products from the Centre National d'Etudes Spatiales. An experiment covering 30 days was conducted, in which the observation streams of 20 globally distributed stations were processed. The initialization time and accuracy of the RT troposphere results using single‐system and multisystem observations were evaluated. The effect of PPP ambiguity resolution was also evaluated. Results reveal that RT troposphere estimates based on single‐system observations can both be applied in weather nowcasting, in which the GPS‐only solution is better than the GLONASS‐only solution. The performance can also be improved by PPP ambiguity resolution and utilizing GNSS observations. Specifically, we notice that ambiguity resolution is more effective in improving the accuracy of ZTD, whereas the initialization process can be better accelerated by GNSS observations. Combining all techniques, the RT troposphere results with an average accuracy of about 8 mm in ZTD can be achieved after an initialization process of approximately 8.5 min, which demonstrates superior results for applying GNSS observations and ambiguity resolution for RT meteorological applications. Real‐time troposphere estimation based on actual real‐time GNSS observation streamsAnalysis of initialization period and accuracy of GNSS troposphere estimatesEvaluation of the impact of PPP ambiguity resolution on troposphere estimates

Pixian broad-bean paste (PXDB) is a traditional Chinese fermented condiment with a distinct flavor. The flavors stems from its strong relationship to local microorganisms. This study analyzed changes in key aroma compounds, organic acids, free amino acids (FAA), and the microbial community during different fermentation stages in order to study the correlation between flavors and microbes. Results indicated that the presence of key aroma compounds and organic acids initially increased, then decreased. Changes in total FAA content were divided into two stages due to the addition of chili pepper, however, both stages displayed an increase in FAA levels as fermentation time increased. High-throughput sequencing was used to study the microbial community and succession. Lactobacillus, Prevotella 9, Ruminococcaceae ucg-014, Lysinibacillus, Enterococcus, Weissella, Escherichia-Shigella, Sphingomonas, Aspergillus, Zygosaccharomyces, Candida,and Pichiapresented as the primary microorganism during fermentation process. Simultaneously, it was found that bacteria genera indicated a stronger potential correlation with flavors, versus fungi genera. Escherichia-Shigella, Leuconostoc, Acetobacter, Chromohalobacter, Candida, Mortierella, Plectosphaerella, Podoscypha,and Podosporadisplayed a significant correlation with flavors (isoamyl alcohol, 2,5-dimethylpyrazine, phenethyl alcohol, 2,3,5,6-tetramethylpyrazine, citric acid, serine, asparagine, glutamate, etc.). This study provides an improved understanding on formation mechanisms of flavors, and guidance for screening PXDB fermentation strains.