Evaluation method of liquor aging association degree based on mid-infrared spectroscopy detection technology

Abstract

The invention provides a method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology, and realizes rapid detection of the degree of association of liquor, belonging to the field of detection of the degree of association of liquor storage. The method comprises: configuring samples treated with different storage times and different temperatures; acquiring sample spectrum data through a Fourier infrared spectroscopy system; selecting a band of hydroxyl (4000cm ‑1 ~2600cm ‑1 ); normalization processing to eliminate the differences caused by different sampling amounts; multi-peak fitting to process samples of different storage times and classify the degree of association; define labels for all spectral data according to the classification and create a data set; pre-process the spectral data in different ways; train and predict through different models, compare the effects of different combinations, and provide a reference for the rapid detection of the degree of association of liquor.

Description

Technical Field

[0001] The invention relates to a method for evaluating the aging association degree of liquor based on mid-infrared spectrum detection technology, belonging to the technical field of special quality detection for stored liquor. Background Art

[0002] The traditional production and processing of liquor includes raw material grinding, koji preparation, fermentation, distillation, aging, and blending. Aging is the process of storing liquor in a container to develop flavor and reduce irritation. The degree of liquor aging is closely related to storage time, storage container, temperature, and air circulation. In order to establish a standardized and sound controllable aging process standard, after years of research, five main mechanism theories have been summarized: "association theory", "esterification theory", "oxidation theory", "dissolution theory", and "volatilization theory", which explain the main changes in the liquor aging process from the perspectives of hydrogen bond association, liquor internal reaction, external environment, and storage container. The hydrogen bond association of liquor is an internal physical change with the ethanol-water association system as the main body. It can form stable clusters and play an important role in improving the taste of liquor and reducing irritation. Studies have shown that the association stability period of ethanol-water solution is much shorter than the required storage time of liquor. During the aging process, the stable association clusters in the liquor will also be destroyed due to changes in internal components and the external environment, thus forming an association-destruction-reassociation process. Therefore, if the degree of association can be detected in real time during the storage of liquor, the external environment can be changed to cope with the internal changes of the liquor, thus accelerating the aging process. The quality of the liquor can also be monitored and leaks can be found, thereby increasing production, ensuring quality, and greatly improving the economic benefits of the liquor industry.

[0003] At present, the evaluation of the degree of association is mainly applied to hydrophobic organic solutions, and the degree of association is evaluated by viscosity. For example, in patent No. CN102564899B, by comparing the zero shear viscosity before and after the destruction of the intermolecular association in the hydrophobic associating polyacrylamide solution, the contribution of the intermolecular association to the supramolecular solution structure is quantitatively calculated. In patent No. CN102967534B, by measuring the viscosity of adding different concentrations of β-cyclodextrin in a water-soluble hydrophobic associating polymer thickener aqueous solution, a quantitative degree of association comparison curve is established, and the degree of association of the water-soluble associating polymer thickener in a porous medium is obtained. However, it is obviously not suitable to use these methods to evaluate the degree of association for ethanol-water solutions with small viscosity changes.

[0004] Infrared spectrum is also called infrared absorption spectrum. It is a characteristic absorption spectrum curve produced by the resonance of infrared photons and the quantized energy levels of molecular vibration and rotation. Infrared spectrum is obtained when photons interact with molecules through the electric dipole moment transition. In other words, the structural characteristics of the highly polar group such as hydroxyl can be better obtained through infrared spectrum. Therefore, it is a feasible method to obtain the spectral information of ethanol-water solution through Fourier infrared spectrometer detection, and then extract the hydrogen bond structure information from the spectral information to evaluate the degree of association. Summary of the invention

[0005] The present invention provides a method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology. The method is to perform infrared spectroscopy detection on ethanol-water solutions stored for different times under the same environment, normalize the obtained spectral data to eliminate the differences caused by different sampling amounts, and then perform infrared spectroscopy on the hydroxyl spectral band (4000-2600cm -1 ) was used for multi-peak processing, and the spectral bands of hydroxyl were divided into four peaks: free hydroxyl, single-bridge hydroxyl, polyhydroxyl and weak hydrogen bond. According to their peak parameters, the association equilibrium constant K of each spectral data was calculated, and the degree of association of the solution was graded in combination with the storage time of the sample. Then, through heating or cooling treatment, spectral data of different degrees of association were obtained, and the rating was performed in the same way, and all the data were integrated to establish a data set. The training set and the test set were divided into a ratio of 4:1, and regression prediction was performed using a variety of preprocessing methods and modeling methods such as one-dimensional convolutional networks, partial least squares, random deep forests and support vector machines. The prediction results were compared, which provided a new reference for the evaluation of liquor association and the rapid detection of the degree of association, and also provided new technical guidance for the brewing industry to improve the quality of liquor.

[0006] In order to achieve the above invention object, the specific technical solution of the present invention is:

[0007] The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology comprises the following steps:

[0008] (1) The infrared spectra (4000 cm) of 60% ethanol-water (V / V) solutions treated with different storage times and temperatures were collected using a Fourier transform infrared spectrometer. -1 ~400cm -1 )data;

[0009] (2) Select the hydroxyl band (4000cm -1 ~2600cm -1 );

[0010] (3) Analyze the data differences caused by different sampling methods and sampling volumes, and select capillary sampling and normalization processing;

[0011] (4) Quantify the data characteristics using multi-peak fitting and calculate the corresponding association equilibrium constant K;

[0012] (5) Associating and classifying the samples according to their storage time to create a data set;

[0013] (6) Perform regression prediction on the original data and preprocessed data.

[0014] The Fourier infrared spectroscopy detection system is composed of a Frontier Fourier transform infrared instrument and a dedicated computer software Spectrum; the Frontier Fourier transform infrared instrument collects infrared spectral data by transmitting light through a KBr tablet with a solution, and the parameters for spectrum collection are: resolution 4cm -1 , scan number 8, spectral range 4000cm -1 ~400cm -1 The spectral curve was preliminarily processed by Spectrum software, the baseline of the spectrum was corrected using interactive baseline correction, and then the noise was smoothed using the Boxcar function with a smoothing factor of 50.

[0015] As a better implementation method in the present application, the hydroxyl band is selected in step (2) to eliminate the influence of other data.

[0016] As a better implementation method of the present application, the method of capillary sampling and normalization processing is selected by comparing the differences in spectral curves of different sampling methods and different sampling amounts in step (3), so that the detection results of the samples can be repeated.

[0017] As a better implementation method of the present application, the multi-peak fitting described in step (4) is used to divide the entire hydroxyl peak into four peaks: free hydroxyl, single bridge hydroxyl, polyhydroxyl and weak hydrogen bond, and the free molecule M is calculated by the parameters of each peak. free and aggregate cluster M assoc The concentration of the complex is then calculated using the equilibrium formula to determine the association equilibrium constant K.

[0018] As a better implementation method of the present application, in step (5), the association level is divided and a data set is prepared according to the association equilibrium constant K at different storage times.

[0019] As a better implementation method in the present application, the data set is preprocessed by standardization, mean centering and other methods as described in step (6), and the original data set and the preprocessed data set are input into modeling methods such as one-dimensional convolutional network, partial least squares, random forest and support vector machine for regression prediction.

[0020] The above method is used to evaluate the degree of association of liquor and realize rapid detection

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] (I) The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology disclosed in the present invention evaluates the degree of association of a hydrophilic solution such as ethanol-water solution for the first time. The present invention takes into account the influence of external factors such as temperature, sampling method, and sampling volume, and takes storage time as a standard to achieve the repeatability of association detection. The characteristics of the infrared spectrum curve are quantified by multi-peak fitting, so that the degree of association can be presented in a standard manner, providing a new reference for the quality assessment of liquor and a new direction for the development of liquor aging technology.

[0023] (II) In terms of technical field, the present invention combines infrared spectrum detection technology with multi-peak fitting processing, divides the entire hydroxyl peak into four peaks: free hydroxyl, single bridge hydroxyl, polyhydroxyl and weak hydrogen bond, analyzes the internal cluster structure of the aqueous solution from the perspective of microscopic molecular groups, and realizes in-depth evaluation of the degree of association. According to the equilibrium formula, each spectral curve is quantified, which can intuitively reflect the degree of association of the solution.

[0024] (III) In data processing, the present invention uses normalization to reduce the differences caused by different sampling amounts and improve repeatability. According to the change characteristics of the solution spectrum curves at different storage times, the parameters of the four peaks are divided into the associated part and the non-associated part, so as to calculate the concentration of free molecules and polymer clusters in the solution, and then calculate the association equilibrium constant K. This method realizes the quantification of the spectral data characteristics based on the actual storage time and the theory of hydroxyl hydrogen bonding, so it is very convincing for the determination of the degree of association.

[0025] (IV) In terms of prediction, the present invention adopts a variety of preprocessing methods and prediction models, adjusts parameters to obtain the optimal model results, and provides a reference for the realization of rapid detection of the degree of association of liquor.

[0026] (V) The present invention provides a method for evaluating the degree of association of liquor from the microscopic perspective of hydrogen bond structure. This method can be used not only for the detection of liquor but also for the evaluation of the association of other hydrophilic organic aqueous solutions, providing new technical guidance for hydrogen bond detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic diagram of the process of the method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology in the present invention;

[0028] Figure 2 General graph of infrared spectrum curve;

[0029] Figure 3aInfrared spectrum curves of the same sample with different sampling amounts;

[0030] Figure 3b Spectral curve after normalization;

[0031] Figure 4 Spectral curves of different sampling methods;

[0032] Figure 5 Spectral curves of different storage times;

[0033] Figure 6 Schematic diagram of hydrogen bond structure peak separation in infrared spectrum;

[0034] Figure 7 Schematic diagram of the changes of each peak at different storage times;

[0035] Figure 8a Standardized processing data graph;

[0036] Figure 8b Standard normal variable transformation processing data diagram;

[0037] Figure 8c Multivariate scatter correction processing data graph;

[0038] Figure 8d Sliding average filter processing data graph;

[0039] Figure 8e Mean-centered data graph;

[0040] Figure 8f Vector normalization processing data graph;

[0041] Figure 8g Wavelet transform processing data graph;

[0042] Figure 8h First-order difference processing data plot. DETAILED DESCRIPTION

[0043] The following is an explanation of the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0044] It should be noted that in order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.

[0045] Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Unless otherwise specified, the raw materials, equipment and methods used in the present invention are commonly used in the art.

[0047] Example:

[0048] The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology comprises the following steps:

[0049] 1. Ethanol-water solution sample preparation

[0050] Mix anhydrous ethanol and high-purity water (conductivity as low as 0.04μs / cm) to prepare a 60% ethanol-water (V / V) solution, and put it into a sealed container to prevent the volatilization of ethanol. Store it at room temperature for 2 days, 5 days, 30 days, 50 days and 100 days, respectively, and avoid severe vibration during the period. Some samples are heated and cooled before testing to change the degree of association of the solution. The infrared spectrum is collected by the KBr tablet method. KBr tablets need to be prepared before collection, and KBr crystals of spectrally pure grade are ground into tablets. Use a capillary or syringe to place the solution on the KBr tablet for easy spectrum collection.

[0051] 2. Infrared spectrum acquisition and preliminary data processing

[0052] The transmission infrared spectrum of the sample was collected using a Frontier Fourier transform infrared instrument under constant temperature conditions. Two magnetic iron blocks with holes pressed the KBr tablet containing the solution in the middle of the hole so that light can directly pass through the KBr tablet. The parameters for spectrum collection are: resolution 4cm -1 , scan number 8, spectral range 4000-400cm -1 .

[0053] The preliminary spectrum curve processing is carried out in the Spectrum software. The baseline of the spectrum is corrected using interactive baseline correction, and then the noise is smoothed using the Boxcar function with a smoothing factor of 50. The Boxcar averaging function eliminates high-frequency noise and makes the collected information clearer. When collecting weak signals, it can be used to increase vertical resolution, reduce noise, improve dynamic characteristics such as signal-to-noise ratio and spurious-free dynamic range, and improve vertical accuracy by using digital signal processing functions and averaging adjacent collected data.

[0054] 3. Eliminate the influence of different sampling volumes and choose the sampling method

[0055] The infrared spectrum curves obtained after the capillary dripping of solutions with obvious differences in quantity on the KBr tablet are mainly reflected in the difference in absorbance. In order to eliminate the influence of absorbance on the calculation of association value, the data needs to be normalized. Considering that the association evaluation only requires the spectral data of the hydroxyl part (4000-2600cm -1 ), extract this part and then normalize it.

[0056] There are three sampling methods for infrared spectroscopy detection: capillary sampling, syringe-intrusion sampling and syringe-drop sampling. The capillary contacts the solution alcohol to allow it to automatically adsorb, and then one end of the capillary is placed on the KBr tablet, and the solution slides naturally. This treatment has the least impact on the solution. Use a 1ml syringe to absorb the solution, and place a drop of solution on the KBr tablet by close intrusion and long-distance dripping. Compared with the use of capillaries, both methods of using syringes have greater damage to the association state of the solution, so capillary sampling is selected to maintain the original association state of the solution.

[0057] 4. Determine the degree of association based on storage time

[0058] The storage time of liquor is an important indicator for determining its taste, and one of the important factors is the degree of association. The degree of association of the spectral curve of ethanol-water solution is distinguished according to the storage time, and then the spectral data of different degrees of association are obtained by heating and cooling. Capillary sampling and detection of ethanol-water solution stored for 2 days, 5 days, 30 days, 50 days and 100 days are carried out, and the obtained spectral curves are compared to analyze the characteristics in the associative and non-associated directions.

[0059] In infrared spectroscopy, there are three types of hydroxyl groups under the action of hydrogen bonds: free hydroxyl groups (3650-3590cm -1 )、Single bridged hydroxyl (3550-3450cm -1 ) and polyhydroxy (3400-3200cm -1 ). Free hydroxyl refers to hydroxyl that does not form hydrogen bonds, which represents the main component of non-association. Single-bridged hydroxyl refers to hydroxyl that forms only one hydrogen bond per atom, which represents the main component of association. Polyhydroxyl refers to hydroxyl that forms two hydrogen bonds per atom. In a stable ring cluster, only a double ring structure will have a polyhydroxyl. The stretching vibration wavenumbers of methyl and methylene are both below 3000cm -1, the spectral curve sampled with a syringe has obvious bulges in the methyl and methylene regions. When the association state is destroyed and the degree of association is not high, the area occupied by the middle part of the alkane and the hydroxyl group is large, so this part is a weak hydrogen bond (HO...HC) connecting the hydroxyl group and the methyl group. When the degree of association of the solution is low, there are fewer stable clusters, and the methyl and methylene groups are prone to interact with the surrounding hydroxyl groups to form unstable weak hydrogen bonds, and are also prone to form small clusters with multiple hydroxyl groups. For solutions that have been recently destroyed by external forces, the methyl and methylene groups will also vibrate violently, which is manifested as a prominent peak group in the spectral curve. Therefore, the higher the degree of association of the spectral curve, the narrower and more concentrated its peak shape.

[0060] 5. Multi-peak fitting and association determination

[0061] The infrared spectrum curve (4000-2600cm -1 ) is divided into four sub-peaks: single-bridge hydroxyl, free hydroxyl, polyhydroxyl and weak hydrogen bond. The alkane part is not fitted. The free molecule concentration [M free ] can be stably expressed. A spectral curve can be fitted in many ways. To present the degree of association stably, restrictions must be set. The half-width height of the single-bridged hydroxyl peak with the largest peak area is determined to be dx1=260, and the weak hydrogen bond peak is close to the alkane part, and its half-width height is limited to dx4=180 to achieve the standardization of multi-peak fitting. Thus, each infrared spectral curve is quantified to calculate the association value that meets the spectral characteristics.

[0062] The basis for the study of hydrogen bonding is the cluster aggregation equation:

[0063]

[0064] Therefore, the calculation formula of the association equilibrium constant K is

[0065]

[0066] Where n is the association number, that is, the average number of molecules of all clusters in the solution; K is the association equilibrium constant, which is used to determine the degree of association of the solution in this study; M free and M assoc represent free molecules and aggregated clusters in the solution, respectively. free ] and [M assoc ] are the concentrations of free molecules and aggregated clusters, respectively.

[0067] Based on the molecular dynamics simulation, the cluster ratio of 60% (V / V) ethanol-water solution in the ideal state was obtained, and the total association number n in the ideal state was calculated to be 6.75. free ] can be obtained from the infrared spectrum by multi-peak fitting. The total molar concentration C0 of 60% (V / V) ethanol-water solution is 32.5 mol / L. The spectra obtained by multiple tests of the same sample have certain differences, and the fitted parameters are also different. The calculation of the association value needs to reduce this difference and increase the difference between different types. The free hydroxyl peak is used to determine the free molecule concentration [M free ] is the main factor. The previous analysis shows that the peak area of ​​polyhydroxyl and weak hydrogen bonds reflects the free direction, but in the stable association system, it also represents the stability of the cluster. Therefore, the free molecule concentration [M free ] The peak area of ​​free hydroxyl groups is taken as the main component, and the peak areas of polyhydroxyl groups and weak hydrogen bonds are multiplied by the peak value y and added. Because the peak value y is less than 1 and increases with the increase of the peak area, this reduces the difference caused by a small change in the peak area, and can also highlight the difference when the peak area difference is large. Therefore, the free molecule concentration [M free The calculation formula of ] is:

[0068]

[0069] The concentration of aggregate clusters [M assoc The calculation formula is as follows:

[0070]

[0071] Among them, C0 is the total molar concentration, S1, S2, S3 and S4 are the peak areas of the four standard peaks of single-bridge hydroxyl, free hydroxyl, polyhydroxyl and weak hydrogen bond, y3 and y4 are the peak values ​​of polyhydroxyl and weak hydrogen bond, respectively, and n is the association number.

[0072] Please refer to Table 1, which shows the parameters obtained by multi-peak fitting of infrared spectra at different storage times and the calculated association equilibrium constant K.

[0073] Table 1 Multi-peak fitting parameters and association equilibrium constant K at different storage times

[0074]

[0075]

[0076]

[0077] The K value of the same storage time fluctuates within a certain range, and all spectral curves are divided into five levels: 0 (K < 0.000002), 1 (0.000002 ≤ K < 0.000008), 2 (0.000008 ≤ K < 0.00002), 3 (0.00002 ≤ K < 0.00007) and 4 (K ≥ 0.00007).

[0078] 6. Build a detection model

[0079] After removing the abnormal sample data, a data set was established that divided the 114 sample data into 5 categories, and the training set and test set were allocated in a ratio of 4:1. All sample data were baseline corrected, smoothed, and normalized. Standardization, standard normal variable transformation, multivariate scattering correction, sliding average filtering, mean centering, vector normalization, wavelet transform, first-order difference, and second-order difference were used for preprocessing, and the original data were input into the modeling algorithms such as one-dimensional convolutional network (1D CNN), partial least squares (PLS), random forest (RF), and support vector machine (SVM) for regression prediction.

[0080] 1D CNN uses rectified linear unit (ReLU) as the activation function and mean squared error (MSE) as the loss function of the output layer. It uses 8 convolutional layers to extract features, and adds a maximum pooling layer after every two convolutional layers to retain the main features. The last deep layer outputs the combined prediction value. Under the approximation of the MSE loss function, the combined prediction value will tend more and more to the true value.

[0081] Before using PLS modeling and prediction, principal component analysis (PCA) was used to reduce the data dimension. The best result was obtained after the dimension was reduced to 25 dimensions. The best result was obtained when the number of trees in the RF algorithm was set to 200. The SVM algorithm used the linear kernel function (linear), the polynomial kernel function (poly), and the radial basis kernel function (rbf) for prediction.

[0082] Please refer to Table 2, which shows the prediction results obtained by combining different preprocessing methods and modeling algorithms.

[0083] Table 2 Prediction results (each grid data is training set accuracy, test set accuracy and prediction root mean square error RMSE)

[0084]

[0085]

[0086] From the prediction results, we can see that the prediction result of the support vector machine model using the "linear" kernel function after the sliding average filter preprocessing is the best, and the prediction accuracy of the other two kernel functions of the support vector is also higher than that of other models. Therefore, the sliding average filter combined with the support vector machine has a good advantage in the prediction of liquor association. The 1DCNN model has the best prediction effect for the original data.

[0087] The above descriptions are only typical embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology, comprising the following steps: (1) The degree of association was determined by infrared spectral data of 60% ethanol-water (V / V) solutions at different storage times and after heating or cooling. The infrared spectra (4000 cm) of 60% ethanol-water (V / V) solutions at different storage times and after heating or cooling were collected using a Fourier transform infrared spectrometer. -1 ~400cm -1 ) data to obtain spectral data of different association degrees; (2) Select the hydroxyl band (4000cm) from the infrared spectrum collected in step 1 -1 ~2600cm -1 ), eliminating the influence of the remaining data; (3) Analyze the data differences caused by different sampling methods and sampling volumes, and select capillary sampling and normalization processing; (4) The data characteristics were quantified using multi-peak fitting. The entire hydroxyl peak was divided into four standard peaks: free hydroxyl, single-bridge hydroxyl, polyhydroxyl and weak hydrogen bond. The free molecular concentration [M free ], cluster concentration [M assoc ] and the association equilibrium constant K: Wherein, C0 is the total molar concentration, S1, S2, S3 and S4 are the peak areas of the four standard peaks of single bridge hydroxyl, free hydroxyl, polyhydroxyl and weak hydrogen bond, y3 and y4 are the peak values ​​of polyhydroxyl and weak hydrogen bond, respectively, and n is the association number; (5) The association levels were divided according to the association equilibrium constant K of samples with different storage times. The K value of the same storage time fluctuated within a certain range. All spectral curves were divided into five levels: 0: K < 0.000002, 1: 0.000002 ≤ K < 0.000008, 2: 0.000008 ≤ K < 0.00002, 3: 0.00002 ≤ K < 0.00007 and 4: K ≥ 0.00007, so as to make a data set; (6) Perform regression prediction on the original data and preprocessed data.

2. The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology according to claim 1 is characterized in that: In step (3), by comparing the differences in spectral curves of different sampling methods and different sampling amounts, the capillary sampling and normalization processing methods are selected to make the detection results of the samples repeatable.

3. The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology according to claim 1 is characterized in that: In step (6), the data set is preprocessed by methods such as standardization and mean centering, and the original data set and the preprocessed data set are input into modeling methods such as one-dimensional convolutional network, partial least squares, random forest and support vector machine for regression prediction.

4. The method for evaluating the degree of association of liquor aging based on mid-infrared spectroscopy detection technology according to any one of claims 1 to 3, characterized in that: This method is used for association assessment and rapid detection during the aging process of liquor.

Images