A method for predicting rosin properties based on infrared spectroscopy

Abstract

The present invention discloses a method for predicting rosin properties based on infrared spectroscopy, the method comprising: S1, collecting rosin samples from wetland pine wood, dividing the rosin samples into a first rosin sample and a second rosin sample; S2, analyzing the sample components based on the first rosin sample to obtain the rosin components; S3, collecting near-infrared spectral data of the sample based on the second rosin sample to obtain near-infrared spectral data of the rosin; S4, correlating and fitting the rosin components with the near-infrared spectral data to establish a near-infrared correction model for rosin properties; S5, using an external validation set of rosin samples that do not participate in modeling to verify the near-infrared correction model for rosin properties; S6, predicting rosin properties based on the near-infrared correction model for rosin properties to obtain a rosin property prediction result. The present invention improves the efficiency of genetic improvement of resin-producing tree species and promotes the rapid development of forest genetic improvement research.

Description

Technical Field

[0001] The invention relates to the technical field of rosin property prediction, and in particular to a rosin property prediction method based on infrared spectroscopy. Background Art

[0002] Longleaf pine (Pinus elliottii Engelm) is a tree of the genus Pinus in the Pinaceae family. It is native to the southeastern United States and is one of the most important coniferous timber species in the world. Longleaf pine is an important ecological, timber and resin-collecting tree species in southern my country, with an artificial forest planting area of ​​2 million hm2. 2 In addition, it grows rapidly, has strong adaptability, and can produce high-quality forest by-products such as rosin and rosin, providing important support for the development of my country's rosin industry.

[0003] Rosin is a resin secreted by pine trees. Rosin properties include rosin yield and rosin chemical components; rosin chemical components mainly include resin acids, fatty acids, monoterpene hydrocarbons, monoterpene alcohols, sesquiterpene hydrocarbons, sesquiterpene alcohols, etc. At present, the functions of many rosin components have been identified, such as α-pinene is considered a natural insecticide, β-pinene can be developed as a potential biofuel to replace petroleum fuels; and pimaric acid-type resin acids with antibacterial properties play an important role in medicine and can be used to prepare anti-inflammatory and anti-cancer drugs. Therefore, determining the content of each chemical component in rosin is of great significance to the quality and effective use of rosin.

[0004] Near Infrared Reflectance Spectroscopy (NIRS) is a "green" non-destructive testing technology that can quickly, easily and accurately perform non-destructive testing on organic samples in various states, such as powder, solid, liquid and other organic samples. This method has been widely used in papermaking, agriculture, food, tobacco and other fields. In recent years, near infrared spectroscopy has also received attention from forestry workers and has been gradually applied to the detection of forest by-products. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method for predicting rosin properties based on infrared spectroscopy, characterized in that:

[0006] S1. collecting resin samples from the slash pine wood, and dividing the resin samples into a first resin sample and a second resin sample;

[0007] S2. Analyze the sample components based on the first rosin sample to obtain rosin components;

[0008] S3, collecting near infrared spectrum data of the sample based on the second rosin sample to obtain near infrared spectrum data of the rosin;

[0009] S4, performing correlation fitting between the rosin components and the near-infrared spectrum data to establish a near-infrared correction model for rosin properties;

[0010] S5. Using an external validation set of rosin samples that did not participate in modeling to verify the near-infrared calibration model of rosin properties;

[0011] S6. Predicting the properties of rosin based on the near-infrared correction model for rosin properties to obtain a prediction result of the properties of rosin.

[0012] Optionally, in S1, the process of collecting the rosin sample from the slash pine wood specifically includes:

[0013] A hole was drilled at a height of 1.3 m at the breast height of the active wood of the longleaf pine using a drill bit with a diameter of 16 mm and a depth of 0.5 cm. A standardized resin test collector was placed to collect the resin. After a preset time, it was sealed with a Parafilm sealing film, sealed with a screw-on plastic cap, and numbered to complete the collection of the longleaf pine wood resin sample.

[0014] Optionally, in S2, the process of analyzing the sample components of the first rosin sample respectively to obtain the rosin components includes:

[0015] Stirring the first rosin sample evenly in a test tube;

[0016] Take some rosin and dissolve it in anhydrous ethanol. Use phenolphthalein as an indicator and drip tetramethylammonium hydroxide ethanol solution until it turns red.

[0017] The composition of the first rosin sample was analyzed by gas chromatography-mass spectrometry.

[0018] Optionally, when the gas chromatography-mass spectrometry is used to perform component analysis on the first rosin sample, the chromatographic conditions are:

[0019] P-5MS quartz capillary column; interface temperature 250℃; maintain at 60℃ for 2min, increase to 80℃ at 5℃ / min, then increase to 230℃ at 30℃ / min, and finally increase to 240℃ at 5℃ / min, maintain for 10min, the temperature of the vaporizer and detector are both maintained at 260℃, and the injection volume is 1μL.

[0020] Optionally, in S3, the process of collecting near infrared spectrum data of the second rosin sample to obtain near infrared spectrum data of the rosin specifically includes:

[0021] Add anhydrous ethanol to the second rosin sample, shake to dissolve the sample, and pour the solution into a container equipped with a near-infrared spectrometer until the container is 80% full;

[0022] The sample solution is scanned using a Fourier near-infrared spectrometer to obtain the near-infrared original spectrum of the slash pine resin sample and the near-infrared spectrum data of the resin.

[0023] Optionally, in S4, the process of correlating and fitting the rosin components with the near-infrared spectrum data to establish a near-infrared correction model for rosin properties specifically includes:

[0024] Randomly selecting a preset number of rosin samples from the rosin samples as a modeling calibration set;

[0025] Based on the modeling correction set, the partial least squares method combined with the cross-validation method was used to establish a near-infrared correction model for rosin properties.

[0026] Optionally, in S5, the process of verifying the near-infrared calibration model of rosin properties using rosin samples from an external validation set that are not involved in modeling specifically includes:

[0027] The rosin samples other than the modeling calibration set are used as external validation data to verify the predictive ability of the model.

[0028] Optionally, in S6, the process of predicting the rosin properties based on the rosin properties near infrared calibration model to obtain the rosin properties prediction result includes:

[0029] Using a near infrared spectroscopy scanning instrument to measure the spectral data of the rosin sample;

[0030] The characteristic spectrum data is input into the near-infrared correction model for rosin properties to obtain the prediction results of rosin properties.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] Although the traditional determination of rosin components can be completed by GC-MS analysis, it is expensive, takes a long time to test, and cannot be applied on a large scale. The purpose of the present invention is to provide a method for predicting rosin properties using near-infrared spectroscopy technology, which is to provide a fast, accurate and simple method for large-scale detection of rosin by establishing a prediction model for rosin properties, greatly saving time, cost, manpower and material resources, and ensuring the prediction level of the model, thereby providing a fast and reliable detection method and scientific basis for the rational and efficient use of rosin properties and the cultivation of excellent forest trees. Improve the efficiency of genetic improvement work of resin-producing tree species and promote the rapid development of forest genetic improvement research. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0034] Figure 1 The near infrared original spectrum of the slash pine resin sample of the embodiment of the present invention;

[0035] Figure 2 The spectrum diagram here is different from the scans of the slash pine resin sample of the embodiment of the present invention;

[0036] Figure 3 This is an abnormal data diagram of excessively high residual values ​​in the deviation quadrant of an embodiment of the present invention;

[0037] Figure 4 A residual diagram showing the residual values ​​between the predicted values ​​and the measured values ​​in an embodiment of the present invention;

[0038] Figure 5 This is a near-infrared calibration model diagram of α-pinene according to an embodiment of the present invention;

[0039] Figure 6 This is a near-infrared calibration model diagram of β-pinene according to an embodiment of the present invention;

[0040] Figure 7 This is the near-infrared calibration model diagram of myrcene of the present invention;

[0041] Figure 8 is the abietic acid near infrared calibration model of the present invention;

[0042] Fig. 9 It is the near-infrared calibration model of levorotatory pimaric acid of the present invention;

[0043] Fig.10 It is the near infrared calibration model of dehydroabietic acid of the present invention;

[0044] Fig.11 It is the near infrared calibration model of the new abietic acid of the present invention;

[0045] Fig.12 The pimaric acid near infrared calibration model of the present invention;

[0046] Fig.13 This is the near-infrared calibration model of isopimaric acid of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. 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.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Embodiment 1

[0050] In this embodiment, a method for predicting resin properties based on infrared spectroscopy includes:

[0051] S1. collecting resin samples from the slash pine wood, and dividing the resin samples into a first resin sample and a second resin sample;

[0052] In S1, the process of collecting rosin samples from the slash pine wood specifically includes:

[0053] At the breast height of the active wood of the longleaf pine, 1.3m, it was collected after 24 hours. Use a drill bit with a diameter of 16mm to drill a hole with a depth of 0.5cm, place a standardized test collector for rosin to collect rosin, seal it with Parafilm sealing film after a preset time, seal it with a screw-on plastic cap, label it, and complete the collection of the rosin sample of the longleaf pine. During the collection of rosin, the rosin is isolated from the outside world. After removing the rosin collection centrifuge tube, it is sealed with Parafilm sealing film and placed in the freezer layer of the refrigerator. Each sample is divided into two parts, one for rosin composition data collection and the other for near-infrared spectroscopy data collection.

[0054] In the present embodiment, the slash pine rosin sample used when setting up the model is in the clear weather of July-August 2022, in the northern, central and southern parts of Jiangxi Province, and 120 (360 in total) slash pine with good growth and substantially the same age are randomly selected. At about 1.3m at breast height of live slash pine standing wood, a drill bit with a diameter of 16mm is used for drilling, and a hole depth of about 0.5cm is used to place a rosin standardization test collector to collect rosin, and after 24h, Parafilm sealing film is sealed, and a screw plastic cover is sealed, numbered, taken back to the laboratory, and placed in a refrigerator freezing layer, and chemical composition is analyzed within 5d (days). Each sample is divided into two parts, one for rosin composition data collection, and the other for near infrared spectrum data collection.

[0055] S2. Analyze the sample components based on the first rosin sample to obtain rosin components;

[0056] In S2, the process of analyzing the sample components of the first rosin sample to obtain the rosin components includes:

[0057] Stirring the first rosin sample evenly in a test tube;

[0058] Take part of the rosin and dissolve it in anhydrous ethanol. Use phenolphthalein as the indicator and tetramethylammonium hydroxide [C4H 13 NO] ethanol solution is dropped until it turns red;

[0059] The composition of the first rosin sample was analyzed using a gas chromatography-mass spectrometer (GC-MS) produced by Agilent.

[0060] When the gas chromatography-mass spectrometry is used to analyze the composition of the first rosin sample, the chromatographic conditions are:

[0061] P-5MS quartz capillary column (30M×0.25M×0.25M); interface temperature 250℃; 60℃ for 2min, 5℃ / min to 80℃, then 30℃ / min to 230℃, and finally 5℃ / min to 240℃, keep for 10min, the temperature of the vaporizer and detector are both kept at 260℃, and the injection volume is 1μL. The relative percentage of each main component was calculated by peak area normalization method, and 16 main components were listed to establish a model. The results are shown in Table 1.

[0062] Table 1

[0063]

[0064] As can be seen from Table 1, there are 16 main components of slash pine resin. Three categories with high content and strong actual economic benefits, totaling 9 components, were selected to establish the model. The components are monoterpenes: α-pinene, β-pinene, myrcene; diterpenes abietic acid: abietic acid, levopimaric acid, dehydroabietic acid, neoabietic acid; diterpenes pimaric acid: pimaric acid and isopimaric acid.

[0065] S3, collecting near infrared spectrum data of the sample based on the second rosin sample to obtain near infrared spectrum data of the rosin;

[0066] In S3, the process of collecting the near infrared spectrum data of the second rosin sample to obtain the near infrared spectrum data of the rosin specifically includes:

[0067] Add anhydrous ethanol to the second rosin sample, shake it thoroughly to fully dissolve the sample, and pour the solution into the container of the near-infrared spectrometer until the container is 80% full;

[0068] The sample solution was scanned and collected using a Fourier near-infrared spectrometer from PerkinELmer. The samples were scanned in the range of 667-2526 nm. Each sample was scanned 5 times, and the sample was repeated 3 times to take the average value and collect the original spectral data. The near-infrared original spectrum of the slash pine resin sample was obtained, and the near-infrared spectral data of the resin was obtained; the samples were randomly divided into two groups according to a ratio of 3:1, namely the modeling calibration set and the external validation set.

[0069] The near-infrared spectrometer needs to be turned on and preheated for 30 minutes, and then scanned after it stabilizes. When collecting the spectrum, first use a white plate for blank calibration, then place the dissolved sample in a sampling glass bottle, the scanning range is 667-2526nm, the indoor temperature is 24-26℃, and the air humidity is 45%-65%. Repeat the sample loading 3 times, scan 5 times each time, and then take the average value to reduce the error. The original near-infrared spectrum of the slash pine resin sample is as follows: Figure 1 shown.

[0070] S4, performing correlation fitting between the rosin components and the near-infrared spectrum data to establish a near-infrared correction model for rosin properties;

[0071] In S4, the process of correlating and fitting the rosin components with the near-infrared spectrum data to establish a near-infrared correction model for rosin properties specifically includes:

[0072] Randomly selecting a preset number of rosin samples from the rosin samples as a modeling calibration set;

[0073] The collected spectral data were imported into the chemical computing software PerkinElmer Spectrum Quant10 for data processing;

[0074] Based on the modeling calibration set, the partial least squares method combined with the cross-validation method was used to correlate the rosin components of the calibration set samples with their near-infrared spectral data for fitting, and a near-infrared calibration model for rosin properties was preliminarily established. The model was optimized by the following processing:

[0075] ① During scanning, the optimal conditions are selected by selecting the number of scans and resolution to establish a model; ② During modeling, the model is analyzed through the initially established regression model and outliers are eliminated to improve the accuracy of the model; ③ The model is established by selecting different pretreatments to screen out the optimal pretreatment combination.

[0076] In this example, α-pinene is taken as an example. 90 samples are randomly selected from 360 slash pine resin samples as an external validation set, and the remaining 270 samples are used as a modeling correction set. The partial least squares method combined with the interactive validation method is used to establish an α-pinene near-infrared prediction model. The screening process of several factors that affect the model prediction level, such as the number of scans, resolution, and preprocessing methods, is described, and the best result is selected.

[0077] ①Scan times filter

[0078] When scanning the spectrum, the near-infrared spectrometer is easily affected by changes in electrical signals caused by other external factors. Among them, the noise signal caused by the intensity of the light source, the electronic circuit itself, environmental interference, etc. is the most objective factor affecting the spectrum. Since the cause of the noise signal makes it randomly change, it will be superimposed on the absorption spectrum signal during scanning. Therefore, improving the signal-to-noise ratio (SNR) is an important means to improve the accuracy of the model. Increasing the number of spectral scans is a relatively effective way to improve the SNR. As the number of scans increases, the noise caused by random factors gradually decreases, the amount of signal absorbed by the spectrum gradually increases, the signal-to-noise ratio improves, and the model prediction ability is relatively improved. Figure 2 Shown are spectra at different scanning times.

[0079] Depend on Figure 2 It can be seen that when the number of scans is 2, the spectrum is rough and the noise has a great influence on the spectrum. As the number of scans increases, the spectrum gradually becomes smoother. This shows that increasing the number of scans may improve the signal-to-noise ratio and the accuracy of the model prediction ability within a certain range. The specific improvement effect is shown in Table 2.

[0080] Table 2

[0081]

[0082]

[0083] As shown in Table 2, when the number of scans is 32, the coefficient of determination in the calibration set is the highest compared with the others, reaching 94.73%, and the root mean square error of the validation set is the minimum, which is 0.493. 32 is determined as the optimal number of scans for the resin scanning of slash pine.

[0084] ② Resolution screening

[0085] The resolution of a near-infrared spectrometer refers to the minimum wavelength interval that the instrument can distinguish when facing closely adjacent peaks. It indicates the ability of the instrument to actually separate adjacent peaks. It is one of the important indicators of the instrument and one of the indicators that affect the accuracy of model prediction. Therefore, resolution will affect the quality of spectral scanning. Prediction models were established under 6 different scanning resolutions, and the differences in the prediction levels of the models established for samples at different resolutions are shown in Table 3.

[0086] Table 3

[0087]

[0088] It can be found from Table 3 that when the resolution is 16cm -1 With 32cm -1 When the resolution is 32cm, the coefficient of determination (RC2) is 90.15% and 91.93% respectively, and the data are relatively close; the root mean square error of the interactive validation (RMSECV) and the root mean square error of the validation set (RMSEP) are almost the same, but the resolution is 32cm -1 This set of data has a resolution higher than 16cm -1 Model data in, resolution 32cm -1 The model prediction ability is better. Therefore, based on the comprehensive inference of 5 data, 32cm -1 The model is built by scanning the sample at a resolution of 1000 nm and has the most accurate prediction ability.

[0089] ③ Screening of abnormal samples

[0090] The rosin composition data measured by GC-MS and the near-infrared spectral data were fitted in Spectrum Quant10 software. After the model was established using the partial least squares (PLS) method combined with the cross-validation method, the outliers with excessively high residual values ​​( Figure 3 ), residual plot of the residual values ​​between the predicted and measured values ​​( Figure 4 ) Deviations are combined to determine outliers.

[0091] The outliers were screened out by the above method and removed. The changes in the prediction level before and after removal are shown in Table 4. After removing the outliers, the determination coefficient of the calibration set of α-pinene increased from 69.75% to 82.26%, and the root mean square error of the calibration set and the root mean square error of the validation set also decreased.

[0092] Table 4

[0093]

[0094] ④ Pretreatment method screening

[0095] Due to the noise caused by the external environment and the interference of the instrument itself, when the near-infrared spectrometer scans the original spectrum of the sample, it will not only absorb the target information related to the composition, but also absorb interference signals such as noise, which will cause the spectrum to shift and affect the model prediction ability. Different preprocessing methods have different abilities to correct various interferences. Therefore, this experiment explores the effects of single preprocessing methods and combined preprocessing methods on the model prediction ability. Now, 11 preprocessing methods are selected, including smoothing (SG), standard normal variable transformation (SNV), first derivative (FD), second derivative (SD) and a combination of different preprocessing methods. 11 preprocessing methods are selected to establish a partial least squares model and the best treatment is selected.

[0096] It can be seen from Table 5 that the model has the highest prediction level when smoothing (SG) + second-order derivative (SD) is selected as the preprocessing method.

[0097] Table 5

[0098]

[0099] After comparing the pretreatment methods of the nine rosin components, the near-infrared calibration models of α-pinene, β-pinene, myrcene, abietic acid, levorotatory pimaric acid, dehydroabietic acid, neoabietic acid, pimaric acid, and isopimaric acid were established. The results are as follows:

[0100] α-pinene: select the full spectrum band, scan 32 times, and the resolution is 32cm -1 , smoothing (SG) + second-order derivative (SD) was used as the preprocessing method, 7 outliers were removed, the number of principal components was 12, and the near-infrared prediction model of α-pinene was established as follows Figure 5 As shown. The calibration set determination coefficient R C 2 It is 83.10, and the root mean square error of interactive validation RMSECV is 0.914.

[0101] β-pinene: select the full spectrum band, scan 32 times, and the resolution is 32cm -1 , using the preprocessing method of smoothing (SG) + standard normal variable transformation (SNV), 4 abnormal samples were removed, the number of principal components was 13, and the near-infrared prediction model of β-pinene was established as follows Figure 6 As shown. The calibration set determination coefficient R C 2 It is 83.39, and the root mean square error of interactive validation RMSECV is 0.751.

[0102] Myrcene: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 , the standard normal variable transformation (SNV) + second-order derivative (SD) preprocessing method was used to remove 14 abnormal samples, the number of principal components was 12, and the near-infrared prediction model of myrcene was established as follows Figure 7 As shown. The calibration set determination coefficient R C 2 It is 74.84, and the root mean square error of interactive validation RMSECV is 0.894.

[0103] Abietic acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 The standard normal variable transformation (SNV) was used as the preprocessing method, 4 outliers were removed, the number of principal components was 10, and the near-infrared prediction model of arginine was established. Figure 8 As shown. The calibration set determination coefficient R C 2 It is 79.61, and the root mean square error of interactive validation RMSECV is 0.490.

[0104] Levorotatory pimaric acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 The standard normal variable transformation (SNV) preprocessing method was used to remove three abnormal samples, the number of principal components was 11, and the near-infrared prediction model of levorotatory pimaric acid was established. Fig. 9 As shown. The calibration set determination coefficient R C 2 It is 82.45, and the root mean square error of interactive validation RMSECV is 1.396.

[0105] Dehydroabietic acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 , the standard normal variable transformation (SNV) preprocessing method was used to remove 7 abnormal samples, the number of principal components was 10, and the near-infrared correction model of dehydroabietic acid was established as follows Fig.10 As shown. The calibration set determination coefficient R C 2 It is 81.75, and the root mean square error of interactive validation RMSECV is 0.474.

[0106] Neoabietic acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 The standard normal variable transformation (SNV) preprocessing method was used to remove three abnormal samples, the number of principal components was 11, and the near-infrared calibration model of neoabietic acid was established as follows: Fig.11 As shown. The calibration set determination coefficient R C 2 It is 88.13, and the root mean square error of interactive validation RMSECV is 0.520.

[0107] Pimaric acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 , the standard normal variable transformation (SNV) preprocessing method was used to remove 5 abnormal samples, the number of principal components was 12, and the pimaric acid near-infrared correction model was established as follows Fig.12 As shown. The calibration set determination coefficient R C 2 It is 84.03, and the root mean square error of interactive validation RMSECV is 0.566.

[0108] Isopimaric acid: Select the full spectrum band, scan 32 times, and the resolution is 32cm -1 The standard SNV preprocessing method was used to remove 5 abnormal samples, the number of principal components was 11, and the near-infrared correction model of isopimaric acid was established as follows: Fig.13 As shown. The calibration set determination coefficient R C 2 It is 80.09, and the root mean square error of interactive validation RMSECV is 1.842.

[0109] S5. Using an external validation set of rosin samples that did not participate in modeling to verify the near-infrared calibration model of rosin properties;

[0110] In S5, the process of verifying the near-infrared calibration model of rosin properties using the external validation set rosin samples that are not involved in the modeling specifically includes:

[0111] The resin samples other than the modeling calibration set are used as external verification data, the constructed calibration model is predicted using the external verification set samples that are not involved in the modeling, the near-infrared spectrum of the verification set is input into the calibration model to obtain the predicted value, and the external verification method is used to verify the actual measurement ability of the model.

[0112] In this embodiment, after the calibration model was established, 90 samples were used to perform external validation on 9 models to test the prediction ability of the models. The near-infrared spectroscopy prediction model of α-pinene, β-pinene, myrcene, abietic acid, levopimaric acid, dehydroabietic acid, neoabietic acid, pimaric acid, and isopimaric acid was used to predict the validation set samples that did not participate in the modeling, and the determination coefficient Rv was 0.04. 2 All of them are above 70%. This shows that the prediction ability of the model is relatively high and can be used for the detection of unknown samples. The validation results of the near infrared prediction model for 9 rosin components of the present invention are shown in Table 6.

[0113] Table 6

[0114]

[0115] S6. Predicting the properties of rosin based on the near-infrared correction model for rosin properties to obtain a prediction result of the properties of rosin.

[0116] In S6, the process of predicting the rosin properties based on the near infrared calibration model for rosin properties to obtain the rosin properties prediction result includes:

[0117] During prediction, a near infrared spectroscopy scanning instrument is used to measure the spectral data of the rosin sample;

[0118] The characteristic spectrum data is input into the near-infrared correction model for rosin properties to obtain the prediction results of rosin properties.

[0119] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting resin properties based on infrared spectroscopy, characterized in that: The method comprises: S1. collecting resin samples from the slash pine wood, and dividing the resin samples into a first resin sample and a second resin sample; The process of collecting rosin samples from slash pine wood specifically includes: At 1.3m above the breast height of the active wood of the longleaf pine, a drill with a diameter of 16mm was used to drill a hole with a depth of 0.5cm, and a standardized test collector for rosin was placed to collect rosin. After a preset time, the hole was sealed with a Parafilm sealing film, sealed with a screw-on plastic cap, and numbered to complete the collection of rosin samples of the longleaf pine. S2. Analyzing sample components based on the first rosin sample to obtain rosin components, specifically comprising: stirring the first rosin sample evenly in a test tube; Take some rosin and dissolve it in anhydrous ethanol. Use phenolphthalein as an indicator and drip tetramethylammonium hydroxide ethanol solution until it turns red. The first rosin sample was subjected to component analysis using a gas chromatography-mass spectrometer. When conducting the component analysis, the chromatographic conditions were as follows: P-5MS quartz capillary column; interface temperature 250℃; 60℃ for 2 min, 5℃ / min to 80℃, then 30℃ / min to 230℃, and finally 5℃ / min to 240℃, hold for 10 min, the vaporizer and detector temperature are both kept at 260℃, injection volume is 1μL; S3. Collecting near infrared spectrum data of the sample based on the second rosin sample to obtain near infrared spectrum data of the rosin, specifically including: Add anhydrous ethanol to the second rosin sample, shake to dissolve the sample, and pour the solution into a container equipped with a near-infrared spectrometer until the container is 80% full; The sample solution is scanned by using a Fourier near-infrared spectrometer to obtain the near-infrared original spectrum of the slash pine resin sample and the near-infrared spectrum data of the resin; S4, correlating and fitting the rosin components with the near-infrared spectrum data respectively, and establishing a near-infrared correction model for rosin properties, specifically comprising: Randomly selecting a preset number of rosin samples from the rosin samples as a modeling calibration set; Based on the modeling correction set, a near-infrared correction model for rosin properties is established by using partial least squares method combined with cross-validation method; S5. Using an external validation set of rosin samples that are not involved in modeling to verify the near-infrared calibration model for rosin properties, specifically including: Using the rosin samples other than the modeling calibration set as external validation data to verify the prediction ability of the model; S6. Predicting the properties of rosin based on the near-infrared correction model for rosin properties to obtain a prediction result of rosin properties, specifically including: Using a near infrared spectroscopy scanning instrument to measure the spectral data of the rosin sample; The characteristic spectrum data is input into the near-infrared correction model for rosin properties to obtain the prediction results of rosin properties.

Images