Online detection method for glucose and ethanol in yeast fermentation process based on Raman spectrum

Abstract

The invention discloses a Raman spectrum-based online detection method for glucose and ethanol in a yeast fermentation process. The Raman spectrum-based online detection method specifically comprises the following steps: preparing a yeast seed solution, and inoculating the yeast seed solution into a fermentation tank for fermentation; in the fermentation process, an immersion type Raman optical fiber probe is used for collecting a Raman spectrum in real time, and meanwhile sampling is conducted at regular time to measure the concentration of glucose and ethanol; sequentially carrying out median filtering, adaptive iteration reweighted penalty least square method (airPLS) baseline correction, Savitz-golay (SG) smoothing processing and characteristic peak normalization on the acquired spectrum; respectively establishing concentration prediction models of glucose and ethanol by adopting a PLS1 algorithm; and applying the model to a subsequent batch fermentation process to realize online prediction of glucose and ethanol. According to the method, manual sampling is not needed, prediction is accurate, operation is easy and convenient, and in-situ online detection of glucose and ethyl alcohol in the yeast fermentation engineering can be achieved.

Description

[0001] This invention relates to the field of bio-fermentation, specifically to an online detection method for glucose and ethanol during yeast fermentation based on Raman spectroscopy, used for in-situ online monitoring of glucose and ethanol concentrations in yeast fermentation broth.

[0002] Yeast fermentation has wide applications in various industries, including food, beverages, alcohol, and biopharmaceuticals. In actual production, yeast typically uses glucose as its primary carbon source, and under suitable fermentation conditions, it produces various metabolites, including ethanol, through metabolic activity. Ethanol, as one of the main products, is widely used in the brewing of alcoholic beverages such as beer, wine, and spirits, and is also used in the production of fuel ethanol and chemicals, playing a vital role in both the food and energy sectors. As a renewable alternative energy source, increasing ethanol yield not only helps reduce production costs and improve the economic efficiency of end products, but also reduces dependence on petroleum resources, lowers carbon emissions, and promotes green manufacturing and environmentally friendly production, thus having significant environmental benefits.

[0003] Currently, high-performance liquid chromatography (HPLC) is the primary method for detecting the concentrations of glucose, the substrate, and ethanol, the main product, during yeast fermentation. This method utilizes a high-pressure pump to propel the mobile phase solvent into a chromatographic column packed with a stationary phase. Separation of substances is achieved based on the distribution differences between components, followed by quantitative analysis using appropriate detectors. HPLC detectors include ultraviolet-visible detectors, differential refractive index detectors, and fluorescence detectors, which can be flexibly selected based on the physicochemical properties of the target components. This method offers high accuracy and sensitivity, and can simultaneously detect multiple organic acids, sugars, and alcohols, thus it is widely used for component detection in fermentation processes. However, HPLC detection typically requires manual, timed sampling from the fermenter, sample pretreatment, and quantitative analysis. The overall operation is cumbersome, the detection cycle is long, and the data often lags behind the actual fermentation progress. Furthermore, the operation is difficult, and external contamination can be introduced during sampling, preservation, and analysis, increasing experimental errors.

[0004] In recent years, with the continuous enrichment of fermentation detection methods, Raman spectroscopy, as a spectroscopic technique based on molecular vibrational information to obtain the structure and content of substances, has been gradually applied to the online monitoring of fermentation processes. Compared with traditional detection methods, Raman spectroscopy has the advantages of being rapid, non-destructive, suitable for aqueous systems, and capable of in-situ detection within fermenters. Raman spectroscopy allows for in-situ online detection of components such as glucose and ethanol in fermentation broth, eliminating the need for manual sampling, simplifying the detection process, and reducing operational difficulty.

[0005] The technical problem to be solved by the present invention is to provide an online detection method for glucose and ethanol in yeast fermentation based on Raman spectroscopy. This method does not require manual sampling, can monitor the concentration of glucose and ethanol in yeast fermentation broth in situ online, reduce the difficulty of operation, and reduce the risk of external contamination during fermentation.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] An online detection method for glucose and ethanol during yeast fermentation based on Raman spectroscopy, specifically including the following steps:

[0008] (1) Select a single yeast colony with intact morphology and inoculate it into sterilized and cooled YPD liquid medium. Incubate it in a constant temperature shaking incubator at 30℃ and 150rpm for 12 hours to prepare seed liquid. Inoculate the prepared seed liquid into the fermenter at an inoculation ratio of 10% and ferment it in the fermenter at 30℃ and 150rpm.

[0009] (2) During the fermentation process, real-time spectral data were acquired using an immersion Raman fiber optic probe; at the same time, samples were taken periodically and the actual concentrations of glucose and ethanol were determined by high performance liquid chromatography.

[0010] (3) The original spectral data were subjected to median filtering, adaptive iterative reweighted penalized least squares (airPLS) baseline correction, Savitsky-Gorye (SG) smoothing and normalization with the stable characteristic peak at the sapphire window as the internal standard in sequence.

[0011] (4) The processed Raman spectral data were paired with the glucose and ethanol concentration values ​​measured by high performance liquid chromatography (HPLC) at the corresponding time points, and the PLS1 algorithm was used to establish glucose concentration prediction models and ethanol concentration prediction models respectively.

[0012] (5) The prediction model is applied to the actual monitoring process of subsequent fermentation batches to continuously predict the concentration of glucose and ethanol in the yeast fermentation broth online.

[0013] Further, the specific process of step (1) is as follows: First, yeast extract, peptone, and glucose are dissolved in deionized water to prepare YPD medium. The prepared YPD medium is then autoclaved and naturally cooled to room temperature for later use. Subsequently, a single, intact yeast colony is picked from a pre-prepared plate using an inoculation loop and inoculated into the YPD liquid medium. The inoculated shake flask is placed in a constant temperature shaking incubator, and the temperature and rotation speed are set for shaking culture for 12 hours to obtain the yeast seed culture. Before fermentation in the fermenter, the fermenter and its piping system need to be cleaned and sterilized. The fermenter is rinsed several times with hot water circulation, followed by high-temperature autoclaving to ensure that the tank and accessories are clean and free of bacteria. The fermentation medium is prepared by weighing and dissolving the raw materials according to the ratio, autoclaving, and then naturally cooling to room temperature for later use. Subsequently, the seed culture is inoculated into the fermenter at a ratio of 10%, and sterilized with sterile water to a final volume, and the fermentation conditions are set.

[0014] Further, the specific process of step (2) is as follows: A Raman fiber optic probe is installed in the fermenter and directly inserted below the surface of the fermentation broth for in-situ sampling. Subsequently, the laser power, integration time, and average number of scans of the Raman spectrometer are set. During fermentation, samples are taken from the fermenter at regular intervals, pretreated with a microporous membrane to remove suspended cells, and then detected by high-performance liquid chromatography (HPLC). For HPLC detection, a standard solution covering the sample concentration range needs to be prepared to plot a standard curve. After detection, the actual concentrations of glucose and ethanol in the sample can be calculated by using the peak areas of glucose and ethanol, combined with the pre-plotted standard curve.

[0015] Further, the specific process of step (3) is as follows: First, the original spectrum is smoothed using a 5-point median filter to remove isolated spikes caused by occasional interference or cosmic ray interference, reducing the interference of random noise on subsequent processing. Second, the baseline drift correction of the spectrum is performed using the adaptive iterative reweighted penalized least squares (airPLS) algorithm, and the smoothing factor and weight function parameters are set to achieve effective removal of fluorescence background. After baseline correction, the Savitzky-Gore (SG) smoothing algorithm is used to process the spectrum to improve the spectral signal-to-noise ratio and enhance the characteristic peak resolution. Finally, the stable characteristic peak at the sapphire window is used as the internal standard peak to normalize all spectra, thereby eliminating the overall intensity difference caused by laser power fluctuations, fiber coupling efficiency, or probe position deviation. After the above multi-step spectral preprocessing, the obtained spectral data can significantly improve the stability, accuracy, and generalization ability of subsequent PLS1 modeling.

[0016] Further, the specific process of step (4) is as follows: First, the glucose and ethanol concentration data are processed using a linear interpolation method to obtain a reference concentration value corresponding to each spectral data. Then, the Raman spectral data is paired and organized with the corresponding glucose and ethanol concentration data, and the optimal number of latent variables is determined using a 5-fold cross-validation method, thereby establishing a prediction model for glucose and ethanol concentrations during fermentation. In subsequent fermentation processes, the real-time collected Raman spectral data is input into the prediction model, enabling in-situ online detection of glucose and ethanol concentrations in the fermentation broth.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] The method described in this invention enables online detection of glucose and ethanol concentrations in yeast fermentation broth by acquiring spectral data online and combining it with modeling techniques. Compared to traditional methods that require frequent offline sampling, this invention eliminates the need for manual sampling and sample pretreatment, offers rapid response, and has a short detection cycle, effectively reducing operational difficulty and the risk of contamination.

[0019] To more clearly illustrate the technical solution of the present invention, the present invention will be further described below in conjunction with the accompanying drawings, but the present invention is not limited to the specific form shown in the drawings.

[0020] Figure 1 This is a schematic diagram of the operation process of the online detection method for glucose and ethanol in yeast fermentation based on Raman spectroscopy described in this invention.

[0021] Figure 2 This is a schematic diagram of the experimental apparatus used for online detection during yeast fermentation according to the present invention. Reference numerals: 1 Fermentation tank; 2 Immersion Raman fiber optic probe; 3 Raman spectrometer; 4 Notebook.

[0022] Figure 3 This is a graph comparing the glucose and ethanol concentrations predicted by Raman spectroscopy with the measured values ​​by high performance liquid chromatography during the fermentation process.

[0023] The technical solutions of the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings, but are not limited to the following embodiments.

[0024] like Figure 1 As shown, the online detection method for glucose and ethanol in yeast fermentation based on Raman spectroscopy involved in this invention includes the following steps: seed culture preparation, culture medium preparation, fermenter inoculation, fermentation and real-time acquisition of Raman spectra, offline detection, spectral preprocessing, data modeling and online prediction.

[0025] like Figure 2As shown, the online detection device used in this invention mainly includes a fermenter 1, an immersion Raman fiber optic probe 2, a Raman spectrometer 3, and a laptop 4. The fermenter 1 is equipped with a stirring shaft and stirring blades to maintain the uniform mixing of the fermentation broth. The immersion Raman fiber optic probe 2 is fixedly installed on the upper end of the fermenter through a flange interface. The Raman fiber optic probe 2 is connected to the Raman spectrometer 3 through an optical fiber to collect the raw Raman spectrum. The Raman spectrometer 3 is connected to the laptop 4 through a data transmission line to realize online detection of preprocessing and PLS1 prediction model.

[0026] First, the seed culture was prepared. 20.0g of yeast extract, 10.0g of peptone, and 20.0g of glucose were weighed into separate beakers, and the volume was adjusted to 1L with ultrapure water to obtain YPD medium. The prepared YPD medium was then dispensed into three 250mL shake flasks, 100mL per flask. The shake flasks were double-wrapped with non-woven fabric and autoclaved at 121°C for 15 minutes, then allowed to cool naturally to room temperature. After cooling, a single morphologically normal yeast colony was picked from the activated plate using an inoculation loop and inoculated into the shake flask. The flasks were then incubated at 30°C and 150rpm with shaking for 12 hours, at which point the culture was ready for use as the seed culture.

[0027] Before inoculation, the fermentation medium was prepared in fermenter 1. The fermentation medium formula was 100 g / L glucose, 5 g / L yeast extract, 1 g / L ammonium bicarbonate, and 0.3 g / L potassium dihydrogen phosphate. The above substances were weighed and added to ultrapure water, stirred until completely dissolved, ensuring no visible solid particles remained. The prepared medium was then slowly poured into fermenter 1 through a funnel. The condenser, air inlet, air outlet, and pressure relief valve were installed, and the fermenter was placed in an autoclave for high-pressure steam sterilization for 15 minutes. After sterilization, the autoclave was slowly vented. When the pressure in the autoclave dropped to 0, the autoclave was opened, and fermenter 1 containing the culture medium was removed and allowed to cool naturally for later use.

[0028] After cooling, the three bottles of seed culture were inoculated into fermenter 1, and then diluted to 3L with sterile water. Simultaneously, an immersion Raman fiber optic probe 2 was installed in fermenter 1. The Raman fiber optic probe 2 had an excitation wavelength of 785nm and a laser power of 300mW, and was connected to a Raman spectrometer 3 via fiber optic cable, which in turn was connected to a laptop 4. The Raman data acquisition software was set to an integration time of 1.5 seconds, and one spectral data point was output after averaging 10 scans. Simultaneously, during the fermentation process, 5mL of fermentation broth sample was collected every 1 hour through the sampling valve of fermenter 1, and immediately filtered through a 0.22μm PES microporous membrane. The filtrate was then analyzed by high-performance liquid chromatography (HPLC). The HPLC analysis used a mobile phase of 60% acetonitrile-water, a flow rate of 1.0mL / min, a column temperature of 30℃, and a differential refractive index detector. Before high-performance liquid chromatography (HPLC) detection, standard solutions of glucose (0, 20, 40, 60, 80, 100 g / L) and ethanol (0, 20, 40, 60 g / L) need to be prepared to plot standard curves for glucose and ethanol. The peak areas of the fermentation broth sample are then substituted into the standard curves to calculate the actual concentrations of glucose and ethanol in the sample.

[0029] After acquiring the spectral data, firstly, a 5-point median filter was used to remove isolated spikes and sharp peaks caused by cosmic rays, reducing noise interference. Then, an adaptive iterative reweighted penalized least squares (airPLS) method was used to correct the spectral baseline, setting a smoothing factor and weight update strategy that matched the spectral characteristics to eliminate baseline drift caused by background fluorescence. After baseline correction, the spectrum was smoothed using Savitzky-Gory (SG) smoothing with an 11-point window and third-order polynomial fitting to further suppress high-frequency noise. Finally, using the stable characteristic peak at the sapphire window as the internal standard peak, all spectra were normalized, and all spectra were scaled according to the peak intensity to enhance the model's generalization ability. After these steps, the obtained standardized spectral data can be used to construct a quantitative prediction model.

[0030] Before data modeling, linear interpolation was performed on the measured glucose and ethanol concentration data to obtain the concentration values ​​corresponding to each Raman spectrum acquisition time point. Then, the PLS1 method was used for model training, and the optimal number of latent variables was selected through 5-fold cross-validation. Finally, the glucose model determination coefficient (R²) was obtained. 2 =0.93, Root Mean Square Error (RMSEC) = 2.57 g / L, Ethanol Model Coefficient of Determination (R²) = 0.93, Root Mean Square Error (RMSEC) = 2.57 g / L, R² = 0.93, Root Mean Square Error (RMSEC) = 2.57 g / L, Ethanol Model Coefficient of Determination (R²) = 0.93, Root Mean Square Error (RMSEC) = 2. 2 The model's performance was assessed with a mean square error (RMSEC) of 0.91 and a root mean square error (RMSEC) of 1.53 g / L. After model construction and validation, it was applied to the next batch of fermentation processes to predict glucose and ethanol concentrations online. Figure 3As can be seen, the predicted results and the curve trends of the HPLC measured results are highly consistent, which fully verifies the accuracy and stability of the method of the present invention.

Claims

1. An online detection method for glucose and ethanol during yeast fermentation based on Raman spectroscopy, specifically comprising the following steps: (1) Select a single yeast colony with intact morphology and inoculate it into sterilized and cooled YPD liquid medium. Incubate it in a constant temperature shaking incubator at 30℃ and 150 rpm for 12 hours to prepare seed liquid. Inoculate the prepared seed liquid into the fermenter at an inoculation ratio of 10% and ferment it in the fermenter at 30℃ and 150 rpm. (2) During the fermentation process, real-time spectral data were acquired using an immersion Raman fiber optic probe; at the same time, samples were taken periodically and the actual concentrations of glucose and ethanol were determined by high performance liquid chromatography. (3) The original spectral data were subjected to median filtering, adaptive iterative reweighted penalized least squares (airPLS) baseline correction, Savitsky-Gorye (SG) smoothing and normalization with the stable characteristic peak at the sapphire window as the internal standard in sequence. (4) The processed Raman spectral data were paired with the glucose and ethanol concentration values ​​measured by high performance liquid chromatography (HPLC) at the corresponding time points, and the PLS1 algorithm was used to establish glucose concentration prediction models and ethanol concentration prediction models respectively. (5) The prediction model is applied to the actual monitoring process of subsequent fermentation batches to continuously predict the concentration of glucose and ethanol in yeast fermentation broth online.

2. The method according to claim 1, characterized in that: The fermentation medium formulation in step (1) is 100 g / L glucose, 5 g / L yeast extract, 1 g / L ammonium bicarbonate, and 0.3 g / L potassium dihydrogen phosphate.

3. The method according to claim 1, characterized in that: In step (2), the spectral acquisition time is 15 seconds each time, and the average value is taken after 10 acquisitions. Data acquisition is completed every 150 seconds.

4. The method according to claim 1, characterized in that: The Raman fiber optic probe mentioned in step (2) is an immersion probe, which is installed on the interface above the fermenter and inserted below the surface of the fermentation liquid for in-situ spectral acquisition.

5. The method according to claim 1, characterized in that: In step (2), the high-performance liquid chromatography (HPLC) method uses acetonitrile / water (60:40 v / v) as the mobile phase, a flow rate of 1 mL / min, and a column temperature of 30 °C. The injection volume is 10 μL, and the detector is a differential refractive index detector.

6. The method according to claim 1, characterized in that: The spectral preprocessing in step (3) includes median filtering, adaptive iterative reweighted penalized least squares (airPLS) baseline correction, Savitsky-Gorye (SG) smoothing, and characteristic peak normalization. The median filter window width is 5 points. Baseline correction employs the adaptive iterative reweighted penalized least squares (airPLS) algorithm to remove background fluorescence and baseline drift interference; Savitsky-Gorye (SG) smoothing has an order of 3 and a window width of 11 points; normalization is performed using the stable characteristic peak at the sapphire window as the internal standard peak.

7. The method according to claim 1, characterized in that: The model in step (4) is established using the PLS1 algorithm, and the number of latent variables is optimized and selected using the 5-fold cross-validation method to determine the optimal number of latent variables.

8. The method according to claim 1, characterized in that: The wavelength of the Raman laser used can be selected from 785 nm, 534 nm, and 1024 nm.

Images