Measuring method of specimen

Abstract

To provide a measuring method of a specimen whose physical properties are varied according to the lapse of time in a spectral analysis in which a peak correlated with the physical property can be easily extracted.SOLUTION: A measuring method of a specimen comprises: a preparation step S10 which prepares a plurality of specimens with different lapsed time; a measurement step S11 which measures spectrum measurement and physical property measurement of respective specimens; a first analysis step S12 which extracts a plurality of representative peaks from peaks contained in the obtained spectrum and executes a first multivariable analysis to obtain a plurality of main components; a second analysis step S15 which selects analysis peaks from the representative peaks contained in the main components and executes a second multivariable analysis to obtain similarities between the specimens; a grouping step S16 which groups the specimens into one or more groups based on the similarities; and a third analysis step S17 which executes a third multivariable analysis per group using the analysis peaks and the physical properties obtained by the physical measurement.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a method for measuring a sample. [Background technology]

[0002] Conventionally, in the field of material development, various methods have been proposed for analyzing material data using mathematical and statistical methods to improve the efficiency of development and analysis. For example, in the identification method described in Patent Document 1, the absorption spectrum of a known fiber material is first measured. The obtained spectral data is subjected to multivariate analysis to obtain analysis data. Then, a database is created from the analysis data group. Then, data obtained from the absorption spectrum of an unknown test fiber is compared with the database to identify similar fibers. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-34565 A Summary of the Invention [Problem to be solved by the invention]

[0004] When measuring various spectra of a sample containing many components, the information of those components appears as many peaks. If the physical properties of the sample change over time, those many peaks increase and decrease in a complex manner as the physical properties change.

[0005] Resins are an example of samples that contain many components. For example, durability is an issue for fiber-reinforced plastics (FRP) used in automobiles. For this reason, long-term durability tests are conducted using such fiber-reinforced resins, and changes in components and physical properties are evaluated over time by spectral and physical property measurements.

[0006] Conventionally, when analyzing such changes in the physical properties of resins, peaks with large change rates are visually selected from among the many peaks contained in the measured spectrum, and principal component analysis is performed using the peaks and physical property values. However, in order to clarify what changes are occurring in the sample at the atomic or molecular level, it is necessary to extract and analyze peaks that are correlated with changes in physical properties from among the many peaks contained in the measured spectrum. It is very difficult to select peaks that are correlated with changes in physical properties from among the many peaks that increase and decrease in a complex manner.

[0007] In addition, the mechanism of resin deterioration involves multiple reactions, and it is believed that the type of reaction changes over time. However, conventional analytical methods do not take into account the change in the type of reaction over time, and therefore cannot provide a detailed explanation of the mechanism of change in physical properties.

[0008] The present disclosure has been made in consideration of the above points, and has as its purpose to make it possible to easily extract peaks that are correlated with changes in physical properties in spectral analysis of a sample whose physical properties change over time. [Means for solving the problem]

[0009] In order to achieve the above object, in the present disclosure, spectral data of multiple samples with different elapsed times is measured, an analysis peak is selected from a large number of peaks using a first multivariate analysis, and the multiple samples are further grouped by a second multivariate analysis using the analysis peak.

[0010] Specifically, the method for analyzing a sample disclosed herein is a method for analyzing a sample whose physical properties change over time, and includes a preparation step of preparing a plurality of samples with different elapsed times; a measurement step of performing a spectrum measurement to analyze components of each of the samples and a physical property measurement to analyze the physical properties of each of the samples; a first analysis step of extracting a plurality of representative peaks from peaks included in the spectrum obtained by the spectrum measurement, and performing a first multivariate analysis on all of the samples using the representative peaks to obtain a plurality of principal components; a second analysis step of selecting a plurality of representative peaks from the representative peaks included in the principal components in order from the representative peak having the largest correlation coefficient indicating a correlation with the elapsed time as analysis peaks, and performing a second multivariate analysis on all of the samples using the analysis peaks to obtain a similarity between the samples; a grouping step of dividing the samples into one or more groups based on the similarity; and a third analysis step of performing a third multivariate analysis for each of the groups using the analysis peaks and the physical property values ​​obtained by the physical property measurement, and extracting a peak having the highest correlation with a rate of change in the physical property value from the obtained principal components.

[0011] Here, "similarity" is an index that indicates how similar the behavior of peaks that increase and decrease over time is between multiple samples. By using this similarity as an index, it becomes possible to group multiple samples according to the type of reaction. The similarity can be calculated as the Euclidean distance or cosine similarity.

[0012] In a sample whose physical properties change over time, many peaks appear in the spectrum measurement, and the many peaks increase and decrease in a complex manner over time. According to the analysis method of the present disclosure, the samples can be grouped according to the type of reaction that changes over time. Then, from the many peaks that increase and decrease in a complex manner, peaks that are highly correlated with the change in physical properties can be easily extracted for each type of reaction that changes over time. By using the peaks extracted in this way, it is possible to perform a more detailed and highly accurate analysis of the mechanism of the change in physical properties in a sample whose physical properties change over time.

[0013] In one embodiment, the method further includes a graphing step of obtaining a plurality of two-dimensional graphs using a plurality of principal components obtained in the first analysis step, the graphs having axes of two of the plurality of principal components, and an optimal graph selection step of selecting an optimal graph from the plurality of graphs obtained by the graphing step, the optimal graph being a graph that can connect each of the plotted samples in order of elapsed time without crossing each other, and the second analysis step selects the analysis peak from the representative peaks included in the two principal components constituting the optimal graph.

[0014] According to the analysis method of the present disclosure, the accuracy of the second multivariate analysis can be improved by selecting an analytical peak from the principal components that constitute the optimal graph.

[0015] In one embodiment, the first multivariate analysis and the third multivariate analysis are principal component analyses, and the second multivariate analysis is a cluster analysis.

[0016] In one embodiment, the two-dimensional graph is a scatter plot, the second multivariate analysis shows the relationship between the samples and the similarity as a tree diagram, and the grouping step divides the samples into one or more groups based on the tree diagram and the elapsed time of each of the samples.

[0017] In one embodiment, the sample is a resin, and the preparation step includes exposing a plurality of the samples to high temperature and high humidity conditions for a period of time to prepare a plurality of the samples having different elapsed times. Effect of the Invention

[0018] As described above, according to the present disclosure, it is possible to easily extract peaks that are highly correlated with changes in physical properties from among a large number of spectrum peaks that increase and decrease in a complex manner, for each type of reaction that changes over time. By using the peaks extracted in this way, it is possible to perform more detailed and highly accurate analysis of the mechanism of changes in physical properties in a sample whose physical properties change over time. [Brief description of the drawings]

[0019] [Figure 1] FIG. 1 is a flow chart showing the steps of the analysis method of the present disclosure. [Diagram 2] FT-IR spectrum data of Sample 8 in Example A. [Diagram 3] 1 is a graph created using principal components PC1 and PC2 of Example A. [Figure 4] 1 is a graph created using principal components PC1 and PC3 of Example A. [Diagram 5] 1 is a graph created using principal components PC2 and PC3 of Example A. [Figure 6] 1 is a table showing peaks and correlation coefficients contained in principal components PC1 and PC3 of Example A. [Figure 7] FIG. 13 is a diagram showing the results of analyzing the similarity using the analysis peaks of Example A. [Figure 8] 1 shows the analysis results using the physical property values ​​and analysis peaks for the group of samples 1 to 11 of Example A. [Figure 9] 1 shows the analysis results using the physical property values ​​and analysis peaks for the groups of Samples 1 to 5 of Example A. [Figure 10] 1 shows the analysis results using the physical property values ​​and analysis peaks for the group of samples 6 to 11 of Example A. [Figure 11] 1 shows the analysis results using the physical property values ​​and analysis peaks for the group of Samples 10 and 11 of Example A. [Figure 12] 1 is a graph showing the relationship between the rate of change in physical property value and the change in peak intensity in Example A. [Figure 13] 1 is a graph created using principal components PC1 and PC2 of Example B. [Figure 14] 1 is a graph created using principal components PC1 and PC3 of Example B. [Figure 15] 13 is a graph produced using principal components PC2 and PC3 of Example B. [Figure 16]1 is a table showing peaks and correlation coefficients contained in principal components PC1 and PC3 of Example B. [Figure 17] FIG. 13 is a diagram showing the results of analyzing the similarity using the analysis peaks of Example B. [Figure 18] 1 shows the analysis results using the physical property values ​​and analysis peaks for the groups of Samples 1 to 7 in Example B. [Figure 19] 1 shows the analysis results using the physical property values ​​and analysis peaks for the groups of Samples 1 to 5 of Example B. [Figure 20] 1 is a graph created using principal components PC1 and PC2 of Example C. [Figure 21] 1 is a graph produced using principal components PC1 and PC3 of Example C. [Figure 22] 13 is a graph produced using principal components PC2 and PC3 of Example C. [Figure 23] 1 is a table showing peaks and correlation coefficients contained in principal components PC1 and PC3 of Example C. [Figure 24] FIG. 13 is a diagram showing the results of analyzing the similarity using the analysis peaks of Example C. [Diagram 25] 1 shows the analysis results using the physical property values ​​and analysis peaks for the groups of Samples 1 to 5 of Example C. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0021] The analytical method disclosed herein is a method that employs various types of spectral analyses and is intended to analyze a sample whose physical properties change over time.

[0022] Fig. 1 is a flow diagram showing the procedure of the disclosed analysis method. As shown in Fig. 1, the analysis method of the present disclosure includes a preparation step S10, a measurement step S11, a first analysis step S12, a second analysis step S15, a grouping step S16, and a third analysis step S17. It is preferable to include a graphing step S13 and an optimal graph selection step S14 between the first analysis step S12 and the second analysis step S15.

[0023] (preparation process) In the preparation step S10, first, a plurality of identical samples are prepared. Then, the samples are exposed to a certain environment for a certain period of time, and a plurality of samples with different elapsed times are prepared. The sample is not particularly limited, but may be a material capable of spectrum measurement, and whose physical properties change when exposed to a certain environment, such as resin or metal. The environment to which the sample is exposed may be any environment as long as it brings about a change in the sample over time, and may be, for example, various stress environments that promote deterioration of the sample, such as heat, water, light, voltage, and salt.

[0024] (Measurement process) In the measurement step S11, a spectrum measurement for analyzing the components of each sample and a physical property measurement for analyzing the physical properties are performed on each of the multiple samples prepared in the preparation step S10 with different elapsed times. Any device for performing the spectrum measurement can be applied depending on the type of sample, and is not particularly limited as long as it can detect a spectral peak, and examples of such devices include nuclear magnetic resonance analysis (NMR), infrared spectroscopy (IR), Raman spectroscopy, gas chromatography (GC), liquid chromatography (LC), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), X-ray absorption fine structure analysis (XAFS), electrochemical impedance spectroscopy (EIS), differential scanning calorimetry (DSC), and energy dispersive spectroscopy (EDS). In addition, the physical properties to be measured may be any physical property depending on the type of sample and the change over time of the sample, and examples of such physical properties include mechanical properties such as bending strength, tensile strength, compressive strength, shear strength, and torsional strength.

[0025] (1st analysis step) In the first analysis step S12, first, a plurality of representative peaks are extracted from the peaks contained in the spectrum obtained by the spectrum measurement. The representative peaks may be all peaks detected from the spectrum. Then, the first multivariate analysis is performed for all samples using the representative peaks. Generally, the multivariate analysis method includes methods such as principal component analysis, independent component analysis, cluster analysis, multiple regression analysis, factor analysis, and discriminant analysis, but it is preferable to perform principal component analysis (PCA) as the first multivariate analysis to obtain a plurality of principal components. The second analysis step S15 can be performed using the plurality of principal components obtained in the first analysis step S12, but the accuracy of the analysis in the second analysis step S15 can be improved by going through a graphing step S13 and an optimal graph selection step S14 after the first analysis step S12.

[0026] The analysis software used for the first to third multivariate analyses is not particularly limited, but Mintab18 (manufactured by Mintab) was used in the following embodiment.

[0027] (Graphing process) In the graphing step S13, a plurality of two-dimensional graphs are obtained using the plurality of principal components obtained in the first analysis step S12. The graphs are graphs with two of the plurality of principal components as axes, for example, scatter plots plotted with the first principal component (PC1) as the horizontal axis and the second principal component (PC2) as the vertical axis. For example, when three principal components, PC1, PC2, and PC3, are obtained in the first analysis step S12, three scatter plots are obtained: a scatter plot of PC1 and PC2, a scatter plot of PC1 and PC3, and a scatter plot of PC2 and PC3. The optimum graph selection step S14 is to select the optimum graph from the plurality of graphs (scatter plots) obtained in this way.

[0028] (Optimal graph selection process) In the optimum graph selection step S14, a graph that can connect each plotted sample in chronological order without crossing is selected as the optimum graph. In other words, a graph in which the samples are arranged in chronological order and well dispersed is selected in a scatter plot. Specifically, a graph that can be judged to have the best dispersion is selected by connecting each sample with a line in chronological order, the lines do not cross, and each sample is far apart.

[0029] (Second analysis process) In the second analysis step S15, an analysis peak is selected from among the representative peaks contained in the principal components. Preferably, the analysis peak is selected from among the peaks contained in the two principal components constituting the optimum graph. Specifically, among the representative peaks contained in the principal components, a plurality of representative peaks are selected in order from the representative peak having the largest correlation coefficient indicating the correlation with the elapsed time, and are set as the analysis peaks. The correlation coefficient indicating the correlation with the elapsed time is obtained as information of the principal components obtained by the first multivariate analysis, and serves as an index for knowing the magnitude of the behavior of the representative peak over time. A representative peak with a large correlation coefficient is a peak whose behavior over time is more prominent than other peaks. The behavior is not limited to the behavior over the entire period from the elapsed time 0 hours to the end time, and may be the behavior during a certain period of time during the elapsed time. By selecting a plurality of representative peaks in order from the representative peak having the largest correlation coefficient as the analysis peaks and excluding the representative peaks having a small correlation coefficient, the analysis accuracy can be improved. For example, it is preferable to select 1 / 4 to 1 / 2 of all the representative peaks as the analysis peaks. Then, the second multivariate analysis is performed on all the samples using the analysis peaks to obtain the similarity between the samples.

[0030] Similarity is an index showing how similar the behavior of peaks that increase or decrease over time is between multiple samples. Using this similarity as an index, it is possible to group multiple samples according to the type of reaction. Similarity can be calculated as Euclidean distance or cosine similarity. As the second multivariate analysis, it is preferable to perform cluster analysis. As a result of cluster analysis, the relationship between samples and similarity can be shown as a dendrogram.

[0031] (Grouping process) In the grouping step S16, the samples are divided into one or more groups based on the similarity obtained in the second analysis step S15. This grouping is preferably performed in consideration of the tree diagram and the elapsed time of each sample. This is because it is believed that the reaction of multiple samples changes over time. Therefore, it is preferable to perform grouping based on the tree diagram after considering that the same or similar reaction occurs in samples with similar elapsed times. It is preferable that samples in the same group have consecutive elapsed times. A sample may be included in multiple groups in duplicate. For example, when samples 1 to 10 arranged in order of elapsed time are grouped, they may be grouped as follows: Group I including all samples 1 to 10, Group II including samples 1 to 5, Group III including samples 6 to 10, and Group IV including samples 9 and 10.

[0032] (Third analysis step) In the third analysis step S17, a third multivariate analysis is performed for each group divided in the grouping step S16 using the analysis peaks and the physical property values ​​obtained by the physical property measurement. As the third multivariate analysis, it is preferable to perform a principal component analysis (PCA) as in the first multivariate analysis to obtain a plurality of principal components. Then, a peak having the highest correlation with the rate of change of the physical property value is extracted from the obtained principal components. This makes it possible to identify a peak related to the change of the physical property. Then, by attributing to which structure or functional group in the sample the peak corresponds, it is possible to obtain a clue to clarify the mechanism of what change or reaction occurs in the sample to cause the change in the physical property.

[0033] (Example) The following description of the preferred embodiments is merely exemplary in nature and is in no way intended to limit the invention, its application, or its uses.

[0034] Table 1 shows the samples and test conditions used in the analytical methods of the present disclosure in Examples A, B, and C.

[0035] [Table 1]

[0036] Example A As shown in Table 1, in Example A, analysis was performed using resin X. Resin X is an amine-based high-toughness epoxy resin, and is a fiber-reinforced resin containing carbon fiber as a reinforcing fiber. When resin X was subjected to TMAH derivatization Py-GC / MS analysis (Shimadzu Corporation, TQ8030), bisphenol A, bisphenol F, and xanthene, which are epoxy resins, dicyandiamide and dicarboxylic acid, which are curing agents, and 3-(3,4-dichlorophenyl)-1,1-dimethylurea, which is a curing accelerator, were detected.

[0037] In Example A, as a preparation step, a moist heat degradation test was carried out on multiple samples (Resin X) by exposing them to high temperature and humidity conditions of 85°C and 95% RH for a certain period of time. Each sample was removed from the high temperature and humidity chamber at different times, and 11 samples were prepared with test times ranging from 0 hours to 2106 hours. Then, as a measurement step, FT-IR measurements and bending tests were carried out on each of the 11 samples, totaling Sample 1 to Sample 11, to analyze the components and physical properties of each sample. The measurement conditions for FT-IR and bending tests are as follows. FT-IR Equipment: Thermo Scientific Nicolet / iN10MX Measurement method: Microscope-ATR method (using germanium prism) Measurement area: 4000cm -1 ~600cm -1 Resolution: 4cm -1 Number of times accumulated: 128 times Aperture size: 100μm×100μm 3-point bending test Equipment: Electronic universal testing machine (Yonekura Manufacturing CATY2010S) Load cell capacity: 100kN The test room environment was adjusted to a temperature of 23°C ± 2°C and humidity of 50% RH ± 5%. The three-point bending test was performed in accordance with JIS K7074 Method A, with a crosshead displacement speed of 5 mm / min.

[0038] FIG. 2 shows FT-IR spectrum data of sample 8 in Example A. In FIG. 2, the portion marked with a thick line above the peak is the representative peak. In Example A, in the first analysis step, 47 representative peaks common to the FT-IR spectra of each sample were used, and a first multivariate analysis was performed using the representative peaks for all samples. As the first multivariate analysis, a principal component analysis was performed. As a result, three principal components, principal component PC1, principal component PC2, and principal component PC3, were obtained. In the graphing step, three graphs (scatter plots) were created using two of these three principal components.

[0039] Figure 3 is a graph created using principal components PC1 and PC2, Figure 4 is a graph created using principal components PC1 and PC3, and Figure 5 is a graph created using principal components PC2 and PC3. Of these three graphs, Figure 4 was selected as the optimal graph because when each sample is connected by a line in the order of elapsed time, the lines do not cross and the samples are far apart, resulting in good dispersion (optimal graph selection process). It was confirmed that when the results of Figures 3 and 5 were used in the subsequent analysis, the analytical accuracy was not as good as when the results of Figure 4 were used.

[0040] Next, in the second analysis step, analysis peaks are selected from the representative peaks contained in the principal components PC1 and PC3 constituting FIG. 4. FIG. 6 is a table showing correlation coefficients indicating the correlation between the representative peaks of the principal components PC1 and PC3 and the elapsed time. Among the representative peaks contained in the principal components PC1 and PC3, multiple representative peaks were selected in order from the representative peak with the largest correlation coefficient indicating the correlation with the elapsed time to be used as analysis peaks. Specifically, for the principal component PC1, since the correlation coefficients are close, 10 representative peaks were selected in order from the peak with the largest absolute value of the correlation coefficient as analysis peaks. For the principal component PC3, since the variation in the correlation coefficients was large, representative peaks with absolute values ​​of the correlation coefficients of 0.1 or more were selected as analysis peaks. In FIG. 6, the peaks marked with check marks are analysis peaks. Approximately 1 / 2 of the peaks in PC1 and PC3 were used as analysis peaks.

[0041] A second multivariate analysis was performed using the analysis peaks selected from the principal components PC1 and PC3 to obtain the similarity between the samples. As the second multivariate analysis, a cluster analysis-observation value was performed. The relationship between the similarity obtained as a result and the samples is shown in FIG. 7. FIG. 7 is a tree diagram showing the result of analyzing the similarity using the analysis peaks of Example A. This tree diagram was calculated using the square of the Euclidean distance by the Ward joining method. As shown by the double arrow in FIG. 7, the shorter the length (height of the vertical axis) of the line segments in the tree diagram to join, the more similar the samples are. In consideration of this tree diagram and the fact that samples 1 to 11 have passed through time in this order, in the grouping process, the samples were grouped as shown in Table 2.

[0042] [Table 2]

[0043] As shown in Table 2, samples 1 to 11, samples 1 to 5, samples 6 to 11, and samples 10 to 11 were each classified into one group.

[0044] Next, as a third analysis step, a third multivariate analysis was performed for each group classified in the grouping step, using the analysis peaks and the bending strength values ​​obtained by bending measurements. As the third multivariate analysis, a principal component analysis was performed. Figure 8 shows the principal components obtained as a result of performing a principal component analysis on groups of samples 1 to 11. Figure 9 shows the results for groups of samples 1 to 5, Figure 10 shows the results for groups of samples 6 to 11, and Figure 11 shows the results for groups of samples 10 to 11.

[0045] For each group, the principal component with the largest absolute value of the bending strength reduction rate among the calculated principal components was focused on, and the correlation coefficient showing the correlation with bending strength in that principal component was extracted with the largest absolute value. In Figures 8 to 11, the principal component of interest and the principal component with the largest correlation coefficient are checked. For example, in the group of samples 1 to 11 (Figure 8), the absolute value of the bending strength reduction rate in principal component PC6 is larger than the other principal components. Therefore, the peak with the largest absolute value of the correlation coefficient (1737.622 cm) in principal component PC6 was extracted. -1 In the groups of samples 1 to 5 (Figure 9), the peaks with the largest and second largest absolute values ​​of the correlation coefficient (1737.622 cm) were extracted. -1 ,3359.531cm -1 ) was extracted. In this way, if the absolute values ​​of the correlation coefficients are close, two or more peaks may be extracted. In the group of samples 10 to 11 (Figure 11), principal component analysis was performed on two samples, so it was not possible to narrow down the principal components, and so attention was focused on four principal components. In this way, attention may be focused on two or more principal components. Using the above method, the main peaks that are thought to have a high correlation with the bending strength reduction rate were extracted.

[0046] [Table 3]

[0047] Table 3 shows a summary of the results of identifying the peaks related to the change in bending strength. The identified peak numbers 1 to 5 correspond to the numbers given to the peaks in FIG. 2. By assigning these peaks, it is possible to know which functional groups in the sample are related to the change in bending strength. FIG. 12 shows a graph showing the relationship between the rate of change in physical properties and the change in peak intensity in Example A. FIG. 12 shows the main peak (approximately 1738 cm) obtained by the above method. -1 , approx. 3360cm -1 , approx. 2172cm -1 12 shows the change in peak strength and the change in bending strength reduction rate in a single graph with time on the horizontal axis. As shown in Fig. 12, it was confirmed that the change in peak strength obtained by the above method correlates with the change in bending strength reduction rate in a certain time period over the elapsed time.

[0048] From the above results, among the extracted peaks, the peak at approximately 1738 cm , which shows the absorption of the CO bond, -1 The peak at approximately 3360 cm corresponds to the bending strength during the elapsed time from 0 to 288 h and indicates the absorption of OH bonds. -1 The peak at approximately 2172 cm corresponds to the bending strength between 288 and 576 h and indicates the absorption of C–N double bonds. -1 It was found that the peak was related to bending strength between 576 and 2106 hours after the elapsed time. In the conventional method, in the analysis of similar resins, only the peak showing the absorption of the CO bond was used as an indicator, so it was not possible to clarify the structure that affects the change in mechanical strength other than the CO bond. By using the method disclosed herein, it was possible to clarify for the first time that the CN double bond constituting the crosslink is involved after 500 hours of elapsed time, when the rate of decrease in bending strength is large. This result is consistent with the components of resin A by GC / MS analysis, and it was predicted that the CN double bond is a bond derived from the curing agent dicyandiamide.

[0049] Example B Next, analysis was performed on Example B, which used the same resin X as in Example A but changed the high temperature and humidity conditions. In Example B, as a preparation step, a moist heat deterioration test was performed on multiple samples (resin X) by exposing them to high temperature and humidity conditions of 70°C and 95% RH for a certain period of time. Each sample was removed from the high temperature and humidity chamber at different times, and seven samples were prepared with test times ranging from 0 hours to 4320 hours. Then, as a measurement step, FT-IR measurement and bending test were performed on a total of seven samples from sample 1 to sample 7 to analyze the components and physical properties of each sample. The FT-IR measurement and bending test methods and analysis methods of Example B are the same as those of Example A, so detailed explanations are omitted.

[0050] Figures 13 to 15 are scatter plots created in the graphing process. Figure 13 is a graph created using principal components PC1 and PC2, Figure 14 is a graph created using principal components PC1 and PC3, and Figure 15 is a graph created using principal components PC2 and PC3. Of these three graphs, Figure 14 was selected as the optimal graph. Figure 16 is a table showing the correlation coefficients indicating the correlation between the representative peaks of the principal components PC1 and PC3 that make up the optimal graph and the elapsed time. For principal component PC1, the correlation coefficients were close, so seven representative peaks with the largest absolute value of the correlation coefficient were selected as analysis peaks. For principal component PC3, the variation in correlation coefficients was large, so representative peaks with absolute values ​​of the correlation coefficient of 0.1 or more were selected as analysis peaks. In Figure 16, the peaks marked with a check mark are analysis peaks. Approximately 2 / 3 of the peaks of PC1 and PC3 were selected as analysis peaks.

[0051] Using the analysis peaks selected from the principal components PC1 and PC3, a second multivariate analysis was performed to obtain the similarity between the samples. The relationship between the resulting similarity and the samples is shown in FIG. 17. FIG. 17 is a tree diagram showing the results of analyzing the similarity using the analysis peaks of Example B. In consideration of this tree diagram and the fact that samples 1 to 7 were grown in this order over time, the grouping step was performed as shown in Table 4.

[0052] [Table 4]

[0053] As shown in Table 4, samples 1 to 7 and samples 1 to 5 were classified into one group each. Next, in the third analysis step, a third multivariate analysis was performed for each group classified in the grouping step using the analysis peak and the bending strength value obtained by bending measurement. The results of the principal component analysis performed for the groups of samples 1 to 7 are shown in FIG. 18, and the results for the groups of samples 1 to 5 are shown in FIG. 19. In each group, the principal component with the largest absolute value of the bending strength reduction rate among the calculated principal components was focused on, and the correlation coefficient value showing the correlation with bending strength in that principal component was extracted with the largest absolute value. In FIG. 18 and FIG. 19, the principal component of interest and the largest correlation coefficient in that principal component are checked. For example, in the group of samples 1 to 7 (FIG. 18), the absolute value of the bending strength reduction rate in principal component PC2 is larger than that of the other principal components. Therefore, the peak with the largest absolute value of the correlation coefficient (3359.531 cm) in principal component PC2 was selected. -1 In the groups of samples 1 to 5 (Figure 9), we focused on the principal component PC8 and selected the peak with the largest absolute value of the correlation coefficient (1737.622 cm -1 ) was extracted.

[0054] [Table 5]

[0055] The results of identifying peaks related to changes in bending strength are summarized in Table 5. From the above results, in Example B, among the extracted peaks, the peak at about 1738 cm 2 , which indicates the absorption of a CO bond, was identified. -1 The peak at approximately 3360 cm indicates the absorption of OH bonds and is related to the bending strength during the elapsed time from 0 to 3456 h. -1It was found that the peak at about 2172 cm, which indicates the absorption of C-N double bonds, was related to the bending strength during the elapsed time period of 3456 to 4320 h. Since the high temperature condition was 70 °C, which is lower than that of Example A, the bending strength did not decrease again as in Example A. -1 No peaks were extracted.

[0056] Example C In Example C, as shown in Table 1, resin Y, which is different from those in Examples A and B, was analyzed. Resin X is an acid anhydride-based epoxy resin, and is a fiber-reinforced resin containing carbon fiber as a reinforcing fiber. When resin Y was subjected to TMAH derivatization Py-GC / MS analysis (Shimadzu Corporation, TQ8030), bisphenol A, which is an epoxy resin, 1,3-isobenzofurandione, which is a curing agent, and phosphine, which is a curing accelerator, were detected.

[0057] In Example C, as a preparation step, a moist heat degradation test was carried out on multiple samples (resin Y) by exposing them to high temperature and humidity conditions of 70°C and 95% RH for a certain period of time. Each sample was removed from the high temperature and humidity chamber at different times, and five samples were prepared with test times ranging from 0 hours to 3456 hours. Then, as a measurement step, FT-IR measurements and bending tests were carried out on a total of five samples, samples 1 to 5, to analyze the components and physical properties of each sample. The methods of FT-IR measurements and bending tests and the analysis methods in Example B are the same as those in Example A, so detailed explanations are omitted.

[0058] Figures 20 to 22 are scatter plots created in the graphing process. Figure 20 is a graph created using principal components PC1 and PC2, Figure 21 is a graph created using principal components PC1 and PC3, and Figure 22 is a graph created using principal components PC2 and PC3. Of these three graphs, Figure 21 was selected as the optimal graph. Figure 23 is a table showing the correlation coefficients indicating the correlation between the representative peaks of the principal components PC1 and PC3 that make up the optimal graph and the elapsed time. For principal component PC1, the correlation coefficients were close, so seven representative peaks were selected as analysis peaks in order of the largest absolute value of the correlation coefficient. For principal component PC3, the variation in correlation coefficients was large, so representative peaks with absolute values ​​of correlation coefficients of 0.2 or more were selected as analysis peaks. In Figure 23, the peaks marked with check marks are analysis peaks. Approximately 2 / 3 of the peaks of PC1 and PC3 were selected as analysis peaks.

[0059] A second multivariate analysis was performed using the analysis peaks selected from the principal components PC1 and PC3 to obtain the similarity between the samples. The relationship between the resulting similarity and the samples is shown in a tree diagram in Figure 24. In Example C, samples 1 to 5 were grouped into one group because the similarity between them was low.

[0060] Next, in the third analysis step, a third multivariate analysis was performed on samples 1 to 5 using the analysis peaks and the bending strength values ​​obtained by bending measurements. The results of the principal component analysis are shown in FIG. 25. The absolute value of the bending strength reduction rate is larger in principal component PC9 than in the other principal components. Therefore, the two peaks from the peak with the largest absolute value of the correlation coefficient (2925.607 cm -1 ,1797.407cm -1 ) was extracted. In addition, the peak with the largest absolute value of the correlation coefficient (1797.407 cm) was extracted from the principal component PC3, which has the second largest absolute value of the bending strength reduction rate. -1 ) was extracted.

[0061] [Table 6]

[0062] The results of identifying peaks related to changes in bending strength are summarized in Table 6. From the results, in Example C, among the extracted peaks, the peak at about 1798 cm 2 , which indicates the absorption of a C-O bond, was identified. -1 The peak of CH is related to bending strength between 0 and 1500 hours. 2 The bond absorption is approximately 2926 cm -1 It was found that the peak was related to the bending strength between about 1500 and 3500 h. [Industrial Applicability]

[0063] INDUSTRIAL APPLICABILITY The present disclosure makes it possible to easily extract peaks that are correlated with changes in physical properties, and is therefore extremely useful in the field of material development.

Claims

1. A method for analyzing a sample whose physical properties change over time, comprising the steps of: A preparation step of preparing a plurality of samples having different elapsed times; a measuring step of performing a spectrum measurement for analyzing components of each of the samples and a physical property measurement for analyzing physical properties of each of the samples; a first analysis step of extracting a plurality of representative peaks from peaks included in the spectrum obtained by the spectrum measurement, and performing a first multivariate analysis on all of the samples using the representative peaks to obtain a plurality of principal components; a second analysis step of selecting a plurality of representative peaks from among the representative peaks contained in the principal components in order from the representative peak having the largest correlation coefficient indicating a correlation with elapsed time as analysis peaks, and performing a second multivariate analysis on all of the samples using the analysis peaks to obtain a similarity between the samples; a grouping step of classifying the samples into one or more groups based on the similarity; and a third analysis step of performing a third multivariate analysis for each of the groups using the analysis peaks and the physical property values ​​obtained by the physical property measurements, and extracting from the obtained principal components a peak having the highest correlation with a rate of change in the physical property value.

2. a graphing step of obtaining a plurality of two-dimensional graphs using the plurality of principal components obtained in the first analysis step, the graphs being plotted on axes of two of the plurality of principal components; an optimal graph selection step of selecting an optimal graph from among the multiple graphs obtained by the graphing step, the optimal graph being a graph that can connect each of the plotted samples in order of elapsed time without crossing each other; Further comprising: the second analysis step includes selecting the analysis peak from the representative peaks included in two principal components constituting the optimal graph; A method for analyzing a sample according to claim 1.

3. The first multivariate analysis and the third multivariate analysis are principal component analysis, and the second multivariate analysis is cluster analysis. The method for analyzing a sample according to claim 2.

4. the two-dimensional graph is a scatter plot, The second multivariate analysis shows the relationship between the sample and the similarity as a tree diagram; The grouping step divides the samples into one or more groups based on the tree diagram and the elapsed time of each sample. The method for analyzing a sample according to claim 3.

5. The sample is a resin, The preparation step includes exposing a plurality of the samples to a high-temperature and high-humidity condition for a certain period of time to prepare a plurality of the samples having different elapsed times. A method for analyzing a sample according to any one of claims 1 to 4.

Images