Method for predicting the quality of butter or estimating the attributes of butter

Abstract

To provide a method by which test results equivalent to those of a sensory evaluator who is a skilled technician can be obtained by a simple method, and the flavor change due to the change in storage days of butter can be quickly predicted and the attributes of butter can be estimated.SOLUTION: A method of predicting the quality of butter or estimating the attributes of butter, includes the step of obtaining a quality prediction model of butter or an attribute estimation model of butter by subjecting an analytical sample prepared from the quality- or attribute-known butter to instrumental analysis to obtain first instrumental analysis data and a multivariate analysis of the relation between the first instrumental analysis data and the quality or attributes. In the method, second instrumental analysis data obtained by subjecting an analysis sample prepared from the quality- or attribute-unknown butter to instrumental analysis is compared with the quality prediction model or the attribute estimation model to predict the quality or attribute of the quality- or attribute-unknown butter.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a method for predicting the quality of butters or estimating the attributes of butters. [Background technology]

[0002] Numerous studies have been conducted on the quantitative measurement and changes in specific components of butter. For example, there is a study on a simple quantitative analysis of lactones, which are contained in large amounts in butter and cheese and are important aroma components that form the flavor of milk (Non-Patent Document 1), and a study on evaluating the changes in aroma components of sweet butter during storage and the effects of different packaging materials (Non-Patent Document 2). In addition, although not a type of butter, a method for predicting the quality of semi-hard cheese and hard cheese has been reported (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2015-55494 A [Non-patent literature]

[0004] [Non-Patent Document 1] Analytical Chemistry Vol. 66 No. 10 751-756 2017 [Non-Patent Document 2] JOURNAL Agricultural And Food Chemistry, Vol. 55, 7840-7846 2007 Summary of the Invention [Problem to be solved by the invention]

[0005] In general, butters, which are made up of fats and oils, including margarine and fat spreads, are said to be more delicious the fresher they are, i.e., the fewer days that have passed since the processing date (the number of storage days), the less the loss of volatile compounds with milk flavor derived from raw milk, the less chemical changes in aroma components, and the less oxidation of fatty acids. If it were possible to accurately predict the transition of flavor changes in butters according to the number of storage days, it would be possible to classify the factors that cause the flavor changes in butters and contribute to providing butters of stable quality that are less likely to change. Furthermore, if it were possible to predict the changes in flavor during the shelf life, it would lead to more rapid product development.

[0006] Conventionally, the quality control of butter products has been carried out by chemical tests such as oxidation value (AV) and peroxide value (POV), and by experienced technicians who perform sensory tests based on the taste and aroma of butter products. However, chemical tests such as AV and POV have the disadvantage that clear differences in numerical values ​​are difficult to detect in the test results until about three months after production. Sensory tests also have the disadvantage that it takes years of experience to master the techniques and become proficient, and that objectivity, reproducibility, and quantitation are unclear. In addition, securing sensory evaluators and carrying out sensory evaluations requires huge costs in terms of time and expenses. Furthermore, if the flavor cannot be measured quickly and future changes in flavor cannot be predicted, rapid product development may be hindered.

[0007] In addition, in conducting research and development of products that appeal to deliciousness, it is important to estimate the origin of raw milk from butter and predict the subsequent flavor trends of butter made from milk collected from each region. It is also known that the flavor and secretion amount of raw milk, which is the raw material for butter, changes depending on the feed (e.g. roughage, concentrated feed, pasture) fed to the dairy cows that secrete the raw milk. In addition, dairy farmers may take measures such as increasing or decreasing specific samples that change the composition, taking into account the rising price of feed, in order to keep the quality of raw milk, such as milk volume and fat percentage, constant throughout the year. If raw milk with different flavors is used, the flavor of not only the raw milk but also the final product, butter, will differ. Therefore, in order to maintain and improve the quality of butter as a product, as well as to manage the manufacturing process and for rapid research and development, it is necessary to estimate the attributes of butter ingredients (such as the collection area of ​​raw milk, the feed fed, and the collection season) using a simple and mechanized method.

[0008] The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide a method that makes it possible to obtain test results equivalent to those obtained by a sensory evaluator who is a skilled technician, using a simple method, and that can quickly predict changes in flavor of butter due to changes in the storage period of butter, and estimate the attributes of butter. [Means for solving the problem]

[0009] In order to solve these problems, the present inventors have conducted intensive research and found a correlation between the instrumental analysis data of butters and the quality or attributes of butters by analyzing the data and the quality or attributes of butters using a multivariate analysis method.Then, by utilizing the correlation, the present inventors have found a method for predicting the quality of butters or estimating the attributes of butters.

[0010] The method for predicting the quality of butter or estimating the attributes of butter according to the present invention comprises: A step of subjecting an analytical sample prepared from butters of known quality or attributes to instrumental analysis to obtain first instrumental analysis data; performing a multivariate analysis of the relationship between the first instrumental analysis data and the quality or the attribute; The method includes a step of obtaining a quality prediction model or an attribute estimation model of butters by This method predicts the quality or estimates the attributes of butters of unknown quality or attributes by subjecting an analytical sample prepared from butters of unknown quality or attributes to instrumental analysis and comparing the second instrumental analysis data obtained with the quality prediction model or the attribute estimation model. Effect of the Invention

[0011] According to the present invention, it is possible to obtain test results equivalent to those obtained by a sensory evaluator who is a skilled technician, using a simple method, and it is also possible to provide a method that can quickly predict changes in flavor of butter due to changes in the storage period of butter, and estimate the attributes of butter. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram showing the relationship between the actual measured values ​​and the predicted values ​​of the sensory evaluation scores of the butters analyzed in the Examples. [Diagram 2] FIG. 1 is a diagram showing the relationship between the actual measured values ​​and the predicted values ​​for the storage period of butters analyzed in the examples. [Diagram 3] FIG. 1 is a diagram showing the distribution of milk collection areas for the raw materials of butters analyzed in the Examples. [Figure 4] FIG. 2 is a diagram showing the distribution of milk collection seasons for the raw materials of butters analyzed in the Examples. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] A method for predicting the quality of butters or estimating the attributes of butters according to an embodiment of the present invention will be described in detail below. The method includes a step of obtaining first instrumental analysis data by subjecting an analytical sample prepared from butters of known quality or attributes to instrumental analysis, and a step of performing multivariate analysis of the relationship between the first instrumental analysis data and the quality or attributes, thereby obtaining a quality prediction model or an attribute estimation model for butters. The method is a method for predicting the quality or estimating the attributes of butters of unknown quality or attributes by comparing second instrumental analysis data obtained by subjecting an analytical sample prepared from butters of unknown quality or attributes to instrumental analysis with the quality prediction model or attribute estimation model.

[0014] <Butter> The butters in the present embodiment include not only butter but also margarine, fat spreads, etc. The butters contain fats and oils as a main component. In the present embodiment, the "quality" of butters refers to the evaluation value by sensory evaluation, the number of days of storage from the date of production, etc. The evaluation value by sensory evaluation of butters refers to the score evaluated by a sensory evaluator based on the flavor, texture, color tone, etc. of the butters. As a specific method of sensory evaluation, for example, the method in the examples described below can be adopted. The number of days of storage from the date of production refers to the number of days that the produced butters are stored in a specified environment.

[0015] In this embodiment, the "attributes" of butter refer to the region where raw milk, which is the raw material of butter, is collected, the feed fed to the dairy cows that secrete the raw milk, the season in which the raw milk is collected, etc. The region where raw milk is collected is an attribute related to the growing environment of the dairy cows and the type of feed fed to the dairy cows. Examples of milk collection regions include grassland regions, field crop regions, regions where grassland and field crops are mixed, and regions where grazing is the main focus. In terms of the feed fed, examples include regions where dairy cows are mainly fed with pasture grass that is high in sugar, regions where feed corn is fed, and regions where dairy cows are fed with feed (Total Mixed Rations) where concentrated feed such as corn is adjusted to match the nutrient requirements of the dairy cows.

[0016] Examples of milk collection seasons include spring, summer, autumn, and winter, which correspond to the four seasons in Japan. However, these seasons can be appropriately selected in consideration of the climate of the producing country of raw milk, which is the raw material for butter products, not limited to Japan. Furthermore, since the climate differs from region to region in Japan, the periods of each of the four seasons do not have to be equal. For example, for raw milk ordered from a production area in Hokkaido, spring may be set to April to May, summer to June to August, autumn to September to October, and winter to November to March. However, milk collection seasons may be appropriately set depending on the type of butter being produced.

[0017] <Butter quality prediction model or attribute estimation model> The quality prediction model or attribute estimation model for butters in this embodiment is obtained by subjecting an analytical sample prepared from butters of known quality or attributes to instrumental analysis to obtain first instrumental analysis data, and then performing multivariate analysis on the relationship between the obtained first instrumental analysis data and the quality or attributes of the butters.

[0018] The analysis sample is appropriately prepared from butters according to the instrumental analysis described below. In the present embodiment, the instrumental analysis includes gas chromatography (GC), liquid chromatography (LC) (e.g., high performance liquid chromatography (HPLC), ultra performance liquid chromatography (UPLC)), mass spectrometry (MS), infrared spectrometry (IR), and the like. These instrumental analyses may be combined. There are no particular limitations on the device used for the instrumental analysis, and any commonly used device may be used as long as it is capable of measuring substances contained in butters, such as fatty acids, lactones, alcohols, and aldehydes. Furthermore, the measurement conditions may be appropriately set so that these substances can be separated without any problems.

[0019] The first instrumental analysis data is preferably data on substances contained in butters, such as aroma components, obtained by the above-mentioned instrumental analysis. The first instrumental analysis data can be, for example, spectral data such as mass spectrometry signal intensity, absorption wavelength, and absorbance. Examples of the first instrumental analysis data include the content of aroma components such as fatty acids, lactones, alcohols, and aldehydes contained in butters. The first instrumental analysis data is preferably data on aroma components of butters. In this case, the data on aroma components of butters preferably includes fatty acids, lactones, alcohols, and aldehydes contained in butters.

[0020] Examples of the fatty acids include 9-decenoic acid, n-decanoic acid, 1-octanoic acid, butanoic acid, caproic acid, and butyric acid. It is preferable that the first instrumental analysis data includes data on 9-decenoic acid, n-decanoic acid, 1-octanoic acid, and butanoic acid. Examples of the lactones include δ-lactone (δ-octalactone, δ-decalactone, δ-dodecalactone), γ-lactone (γ-decalactone, γ-undecalactone), and the like. It is preferable that the first instrumental analysis data includes data on δ-dodecalactone, δ-decalactone, and δ-octalactone. Examples of the alcohols include ethanol and 1-octanol. It is preferable that the first instrumental analysis data includes data on 1-octanol. Examples of the aldehydes include 2-nonenal, hexanal, and nonanal. It is preferable that the first instrumental analysis data includes data on 2-nonenal.

[0021] The instrumental analysis for obtaining the first instrumental analysis data is preferably GC analysis. When extracting aroma components from an analysis sample prepared from butters, the extraction of the aroma components is preferably performed by a method selected from the group consisting of headspace-SPME, dynamic headspace, steam distillation extraction, SAFE-SPME, and solvent extraction using a solvent such as methanol, diethyl ether, or hexane. When detecting aroma components from an analysis sample prepared from butters, the detection of the aroma components is preferably performed by a method selected from the group consisting of mass spectrometry, sniffing mass spectrometry, and hydrogen flame ionization detection.

[0022] A dataset is prepared by preparing multiple pieces of data including first instrumental analysis data of butters of known quality or attributes obtained by the above-mentioned instrumental analysis and the quality or attributes linked to the first instrumental analysis data, and the dataset is used to perform multivariate analysis of the relationship between the first instrumental analysis data and the quality or attributes, thereby creating a quality prediction model or attribute estimation model for butters.

[0023] Multivariate analysis employs analytical tools commonly used in the field of chemometrics, such as PCA (Principal Component Analysis), PLS (Partial Least Squares Analysis), regression analysis, discriminant analysis, and HCA (Hierarchical Cluster Analysis).

[0024] The PLS regression analysis preferably employed in this embodiment is an effective method for creating a calibration curve from spectral data having correlation between variables. It is preferable to employ the PLS regression analysis in creating a quality prediction model for butter. In this embodiment, the term "predict" refers to quantitatively evaluating the quality using a calibration curve or the like.

[0025] PCA is a method for maximizing the variance of a group and displaying the subsequent principal components orthogonal to the principal components previously determined, thereby making it easier to visualize data, and is used for mapping. It is preferable to employ PCA in the attribute estimation model for butter. In this embodiment, the term "estimating" refers to qualitatively evaluating attributes by mapping or the like.

[0026] Normally, if there is a correlation between variables (for example, temperature and pressure), the accuracy of regression drops dramatically, but this method orthogonally transforms the variables in the data and uses those variables to perform regression analysis. The cumulative Q^2 in PLS regression analysis is a dimensionless index that measures the predictive ability of the model. Additionally, the PRESS root mean square is an abbreviation for Prediction Error Sum Squares, and indicates the average square root of the prediction errors, with smaller values ​​being more desirable. Multivariate analysis can be performed using commercially available software, and manuals are included with commercially available software.

[0027] The explanatory variables in the multivariate analysis are preferably data on aroma components of butters, more preferably data on fatty acids, lactones, alcohols, and aldehydes contained in butters, and more preferably MS signal intensity data.

[0028] <Prediction of quality or estimation of attributes of butter products with unknown quality or attributes> The quality of butters of unknown quality or attributes is predicted or the attributes are estimated by comparing the quality prediction model or attribute estimation model created as described above with second instrumental analysis data obtained by subjecting an analytical sample prepared from butters of unknown quality or attributes to instrumental analysis. The "second instrumental analysis data" of butters of unknown quality or attributes is data corresponding to the above-mentioned "first instrumental analysis data" of butters of known quality or attributes, and can take the same form as the above-mentioned first instrumental analysis data, so a detailed description will be omitted. Furthermore, with regard to the quality, attributes, analytical sample, and instrumental analysis, the same ones as those adopted in creating the quality prediction model or attribute estimation model are applied.

[0029] <Processing unit> The multivariate analysis, quality prediction, and attribute estimation in this embodiment are performed using a calculation processing device or a device having the function thereof. A suitable calculation processing device includes a first storage unit that stores first instrumental analysis data of butters of known quality or attributes, a second storage unit that stores input quality data or attribute data, a first calculation unit that performs multivariate analysis of the first instrumental analysis data stored in the first storage unit and the quality data or attribute data stored in the second storage unit based on a preset program to create a quality prediction model or attribute estimation model, a third storage unit that stores the quality prediction model or attribute estimation model created by the first calculation unit, a fourth storage unit that stores second instrumental analysis data of butters of unknown quality or attributes, and a second calculation unit that compares the second instrumental analysis data stored in the fourth storage unit with the quality prediction model or attribute estimation model stored in the third storage unit based on a preset program to calculate the quality predicted or the attribute estimated from the second instrumental analysis data.

[0030] For example, when the instrumental analysis is GC / MS analysis, when the first instrumental analysis data obtained by GC / MS analysis of a plurality of butter samples with known sensory evaluation scores are input to the arithmetic processing device, they are stored in the first memory unit, and separately, when the sensory evaluation scores of the plurality of butter samples are input to the arithmetic processing device, they are stored in the second memory unit. Then, in the first calculation unit of the arithmetic processing device, a multivariate analysis is performed by PLS regression analysis using the sensory evaluation data of the butters and the instrumental analysis data as the objective variable and the explanatory variable, respectively, to create a sensory evaluation score prediction model. In addition, a multivariate analysis is performed by PCA with a previously known attribute, for example, the type of feed given to dairy cows, linked to the instrumental analysis data.

[0031] When predicting a sensory evaluation score, second instrumental analysis data obtained by a similar GC / MS analysis of butters with unknown sensory evaluation scores is input to the arithmetic processing device, which stores the second instrumental analysis data in the fourth memory unit. Next, in the second calculation unit of the arithmetic processing device, the input second instrumental analysis data is compared with the sensory evaluation score prediction model, and a predicted value of the sensory evaluation score is calculated.

[0032] The multivariate analysis in this embodiment makes it possible to select peaks according to the accuracy required for prediction, rather than all of the obtained data, for example, in the case of GC / MS, all of the peak information derived from chemical substances in chromatography, and it is possible to select some or all of the peak information required for prediction.

[0033] Accumulation of information linking the quality and attributes to be predicted or estimated other than those exemplified in this embodiment can further improve the application range and accuracy of the present disclosure. For example, it is self-evident that the aroma components generated change depending on the difference in storage temperature conditions of butter. It is also possible to prepare a plurality of storage temperature conditions as standards, accumulate the changes in the aroma components (fatty acids, lactones, alcohols, aldehydes) for each standard, evaluate the butter by sensory evaluation, etc., and newly link the quality and attributes. It becomes possible to make a highly accurate prediction according to the amount of accumulated information. In other words, if it is possible to link the aroma components generated after a storage temperature of a certain temperature or higher and / or after a certain time has passed with the quality and attributes, it becomes possible to estimate the quality and attributes. EXAMPLES

[0034] The present invention will be described in more detail with reference to the following examples, but the present invention is not limited to these examples.

[0035] <Production of butter products> Raw milk from different collection areas was ordered and salted butter was produced according to the established method. It was filled into approximately 200g containers measuring 12 x 6.5 x 3cm in length x width x height and individually wrapped in aluminum pouches. These were stored at 10°C. The milk collection areas were Region O, which has a mixture of grassland and field crops and tends to feed dairy cows corn for feed, and Region S, which is a grassland area where many dairy cows are pastured and tend to be fed mainly sugar-rich grass. Salted butter was also produced according to the established method using raw milk ordered in different seasons from each collection area.

[0036] <Sensory evaluation> Sensory evaluation of flavor quality was performed based on the evaluation of flavor, texture, and color indices shown in Table 1, in accordance with the items shown in the General Dictionary of Milk, Table II-3.11-13 (Asakura Publishing Co., Ltd., page 204), with reference to the Testing Methods for Dairy Products, Commentary [Revised 2nd Edition] (Kanbara Publishing Co., Ltd., p. 289). Each butter was evaluated by eight people, and the following overall evaluation score was recorded as the average of the sensory testers. [Determination of evaluation points] The flavor, texture, and color were considered comprehensively and the samples were scored as follows. The sensory evaluators were made to evaluate the samples of each score and memorize the scores to standardize the sensitivity.

[0037] [Overall score] 10 points: The sweetness and flavor derived from milk is fully felt, the texture and color are not confirmed to have any of the indicators shown in Table 1, and when squeezed with a butter knife, no indicators such as water separation are confirmed, and the spreadability is very good. In terms of flavor, the desired flavor can be felt perfectly. 9 points: Slightly inferior to 10 points, or 10 points with other indicators. 8 points: A slight sweetness and flavor derived from milk is felt, one of the indicators shown in Table 1 is confirmed for the texture and color tone, and no indicators such as water separation are confirmed when squeezed with a butter knife, and the spreadability is good. The flavor that is desired to be appealed can also be felt. 7 points: Slightly inferior to 8 points, or 8 points plus other indicators. 6 points: A slight sweetness and flavor derived from milk is felt, two or less of the indicators shown in Table 1 are confirmed for the texture and color tone, and two or less indicators such as water separation are confirmed when squeezed with a butter knife, and the spreadability is good. As for the flavor, it can be felt if you are conscious of the flavor you want to appeal. 5 points: Slightly inferior to 6 points, or 6 points plus other indicators. 4 points: No milk-derived sweetness or flavor, 3 or less indicators in Table 1 confirmed for texture and color, 3 or less indicators confirmed for water separation when squeezed with a butter knife, and not very good spreadability. Mouthfeel and melting in the mouth are not very good, and some melted oil remains when swallowed. As for flavor, it is difficult to feel the flavor that is being promoted. 3 points: Slightly inferior to 4 points, or 4 points plus other indicators. 2 points: No milk-derived sweetness or flavor was felt, and the indicators in Table 1 were found in the texture and color. When squeezed with a butter knife, indicators such as water separation were found, and the spreadability was poor and crunchy. The mouthfeel and melting in the mouth were poor, and molten oil remained when swallowed. As for the flavor, the desired flavor was not felt at all. 1 point: Slightly inferior to 2 points, or other indicators were confirmed in 2 points.

[0038] [Table 1]

[0039] <Aroma component extraction process> After storing the samples for a specified number of days in a refrigerator set at 10℃, the packaging was removed, the butter on the surface was scraped off, and the butter was sampled from the center of the 200g sample, and precisely weighed out into dedicated vials at 1.0±0.05g each. This sampling was done so that each sample was approximately the same. For the measurements, samples were always prepared before opening, and the 200g sample taken after opening was not used as the next measurement sample. 10μl of 1000ppm trans-2-methyl-2-butenal was added to each vial as an internal standard. A 50 / 30μm DVB / CAR / PDMS (MERCK) SPME fiber was used, and aroma components were collected from the headspace at 60℃ for 1 hour. A GESTEL MPS autosampler was used.

[0040] <Instrumental analysis process> For GC / MS, a GC System 7890A manufactured by Agilent Technologies was used, and for mass spectrometry, a 5975C was used. The GC oven was heated from 40°C to 250°C at a rate of 10°C / min. For the GC column, a DB-WAX UI (manufactured by Agilent Technologies) with a length of 30 m, an inner diameter of 0.25 mm, and a film thickness of 0.25 μm was used. A total of nine components were selected: four fatty acids (9-decenoic acid, n-decanoic acid, 1-octanoic acid, and butanoic acid), three lactones (δ-dodecalactone, δ-decalactone, and δ-octalactone), one alcohol (1-octanol), and one aldehyde (2-nonenal). These standards were mixed and the mixture was subjected to GC / MS. The retention times and mass spectra were compared with those in the library (NIST library) to identify the various aroma components contained in the measured sample. The peak area ratios to the peak area of ​​trans-2-methyl-2-butenal used as the internal standard were calculated and recorded.

[0041] Table 2 shows an example of a table summarizing the collection region of the raw milk used as the raw material for the evaluation samples, the collection season of the raw milk, the storage period of the evaluation samples, the peak area ratios of the above nine aroma components to the peak area of ​​the internal standard, the sensory evaluation scores, and the sensory evaluation comments.

[0042] [Table 2]

[0043] <Multivariate analysis process / quality prediction process> [Example 1-1_Prediction of sensory evaluation scores using a PLS regression analysis model] A PLS regression analysis was performed on 76 evaluation samples using the data on the peak area ratios of the above nine aroma components and the sensory evaluation score of salted butter using the statistical analysis software JMP (Ver.14.3.0) (SAS Institute) to create a prediction model. PLS regression analysis finds the relationship between two sets of variables, namely the measured values ​​of the sensory evaluation score and the response values ​​of the sensory evaluation score. The relationship is shown in Figure 1. The plots of ● in Figure 1 indicate the evaluation samples used in the analysis. As shown in Figure 1, the predicted values ​​of the sensory evaluation were calculated from the actual measured values ​​of the sensory evaluation of the evaluation samples, making it possible to predict the evaluation score without actually conducting a sensory evaluation. When the number of factors was set to 7, the PRESS mean square root was the minimum value, 0.59288, and the cumulative Q^2 at that time was 0.997585. The plots of △ and □ in Figure 1 indicate test samples for the prediction model. The scores predicted from the aroma component analysis results were 7.2 points and 8.7 points, respectively. The actual values ​​measured in the sensory evaluation were 7 and 9 points, respectively.

[0044] [Example 1-2_Prediction of sensory evaluation scores using a PLS regression analysis model] A prediction model was created by performing PLS regression analysis similar to that in Example 1 using the peak area ratio data of four aroma components consisting of 1-octanol, δ-octalactone, δ-decalactone, and 9-decenoic acid among the nine aroma components and the sensory evaluation score of salted butter. As a result, when the number of factors was 4, the PRESS mean square root was the minimum value of 0.86187 and the cumulative Q^2 was 0.658244. Compared to the case where nine aroma components were used as in Example 1, the PRESS mean square root increased, the cumulative Q^2 decreased, and the prediction accuracy decreased.

[0045] [Example 2-1_Predicting changes over time using a PLS regression analysis model] A PLS regression analysis was performed on 76 evaluation samples using the statistical analysis software JMP (Ver.14.3.0) (SAS Institute) for the peak area ratio data of the above nine aroma components and the shelf life of salted butter, and a prediction model was created. PLS regression analysis finds the relationship between two sets of variables, namely the measured value of shelf life and the response value of shelf life. The relationship is shown in Figure 2. The plots with ● in Figure 2 indicate the evaluation samples used in the analysis. As shown in Figure 2, the predicted value of the shelf life of the evaluation samples was calculated from the actual measured value of the shelf life of the evaluation samples, and the number of days from the date of manufacture could be predicted by the aroma component analysis. When the number of factors was set to 6, the PRESS mean square root was the minimum value of 0.42444, and the cumulative Q^2 at that time was 0.999874. The △ and □ plots in Figure 2 indicate test samples for the prediction model. The shelf life predicted from the aroma component analysis results was 318 days and 145 days, respectively. The actual storage days were 365 and 180 days, respectively.

[0046] [Example 2-2_Predicting changes over time using a PLS regression analysis model] A prediction model was created by performing PLS regression analysis similar to that in Example 2 using the peak area ratio data of four aroma components consisting of 1-octanol, δ-octalactone, δ-decalactone, and 9-decenoic acid among the nine aroma components and the storage period of salted butter. As a result, when the number of factors was 4, the PRESS mean square root was the minimum value of 0.75569 and the cumulative Q^2 was 0.862847. Compared to the case where nine aroma components were used as in Example 2, the PRESS mean square root increased, the cumulative Q^2 decreased, and the prediction accuracy decreased.

[0047] <Multivariate analysis process / attribute estimation process> [Example 3: Estimation of milk-collection areas using PCA] For the data on the peak area ratios of the above nine aroma components and the collection area of ​​the raw milk, which is the raw material for salted butter, a PCA was performed on 32 evaluation samples using JMP (Ver.14.3.0) (SAS Institute) as statistical analysis software, and each evaluation sample was plotted in a two-dimensional matrix. The results are shown in Figure 3. The plots marked with a black circle in Figure 3 indicate the evaluation samples used in the analysis. The plots marked with a black circle with an arrow "S" indicate evaluation samples using raw milk collected in the S region. The plots marked with a black circle with an arrow "O" indicate evaluation samples using raw milk collected in the O region. The plots marked with a black circle in Figure 3 are the average of each principal component of the evaluation samples with the attributes of the S region, or the average of each principal component of the evaluation samples with the attributes of the O region. From Figure 3, the plots marked with S indicate that the feed is close to that fed in the S region, that is, that a grassland-type feed was fed, and the plots marked with O indicate that the feed is that fed in the O region, that is, that a grassland-type feed was fed with a field crop-type feed.

[0048] [Example 4: Estimation of milk collection season using PCA] Using the statistical analysis software JMP (Ver.14.3.0) (SAS Institute), PCA was performed on the 26 evaluation samples for the peak area ratio data of the above 9 aroma components and the collection season of raw milk, the raw material for salted butter, and each evaluation sample was plotted in a two-dimensional matrix. The evaluation samples used were those in which the butter had been stored for 0 or 30 days and the raw milk was collected in region S or region O. The results are shown in Figure 4.

[0049] In Figure 4, the plots marked with a black circle indicate the evaluation samples used in the analysis. The plots marked with a triangle with an arrow pointing to "spring" indicate test samples using raw milk collected in spring (April to May). The plots marked with a triangle with an arrow pointing to "summer" indicate test samples using raw milk collected in summer (June to August). The plots marked with a triangle with an arrow pointing to "autumn" indicate test samples using raw milk collected in autumn (September to October). The plots marked with a triangle with an arrow pointing to "winter" indicate test samples using raw milk collected in winter (November to March). The butter used in this analysis was raw milk ordered from production areas in Hokkaido, so the seasons were defined as above. However, the definitions of attributes such as season and production area are not limited to this.

[0050] The plots marked with ■ in FIG. 4 are obtained by averaging the principal components of the evaluation samples with spring attributes, the principal components of the evaluation samples with summer attributes, the principal components of the evaluation samples with autumn attributes, or the principal components of the evaluation samples with winter attributes. From FIG. 4, it can be confirmed that the production seasons of butter produced from raw milk collected in different seasons can be separated and classified. Also, as shown in the figure, each of the △ plots of the test samples is plotted near the plots marked with ■, which are plotted by averaging the seasons. In this way, it is possible to estimate the collection season of the raw materials for butter when the collection season is unknown.

[0051] Although the present invention has been described in detail with reference to specific embodiments, it will be apparent to those skilled in the art that various changes and modifications can be made without departing from the spirit and scope of the invention.

Claims

1. A method for predicting the quality of butter, comprising the steps of: A step of subjecting an analytical sample prepared from butters of known quality to instrumental analysis to obtain first instrumental analysis data; and performing a multivariate analysis of the relationship between the first instrumental analysis data and the quality by PLS regression analysis to obtain a quality prediction model for butters, A method for predicting the quality of butter of unknown quality by subjecting an analytical sample prepared from the butter of unknown quality to an instrumental analysis and comparing the second instrumental analysis data obtained with the quality prediction model, comprising: the first instrumental analysis data and the second instrumental analysis data are data on aroma components of the butters, and the data on the aroma components are explanatory variables in a multivariate analysis; The method for predicting the quality of butters, wherein the data on aroma components includes data on at least 1-octanol, δ-octalactone, δ-decalactone, and 9-decenoic acid.

2. 2. The method for predicting the quality of butters according to claim 1, wherein the data on the aroma components of the butters includes data on fatty acids, lactones, alcohols and aldehydes contained in the butters, and the data on the fatty acids, lactones, alcohols and aldehydes are explanatory variables in a multivariate analysis.

3. The method for predicting the quality of butters according to claim 1 or 2, wherein the instrumental analysis is gas chromatography analysis.

4. The method for predicting the quality of butters according to any one of claims 1 to 3, wherein extraction of aroma components in an analysis sample prepared from butters is carried out by a method selected from the group consisting of a headspace-SPME method, a dynamic headspace method, a steam distillation extraction method, a SAFE-SPME method, and a solvent extraction method.

5. The method for predicting the quality of butters according to any one of claims 1 to 4, wherein the detection of aroma components in an analytical sample prepared from the butters is carried out using a method selected from the group consisting of mass spectrometry, sniffing mass spectrometry, and flame ionization detection.

6. The quality is an evaluation value based on a sensory evaluation or the number of days of storage from the date of manufacture. A method for predicting the quality of butters according to any one of claims 1 to 5.

7. A step of subjecting an analytical sample prepared from butters of known quality to instrumental analysis to obtain first instrumental analysis data; performing a multivariate analysis of the relationship between the first instrumental analysis data and the quality by PLS regression analysis; the first instrumental analysis data is data on aroma components of the butters, and the data on the aroma components is an explanatory variable in a multivariate analysis; The method for creating a quality prediction model for butters, wherein the data on aroma components includes data on at least 1-octanol, δ-octalactone, δ-decalactone, and 9-decenoic acid.

Images