Tobacco powder particle size detection model establishment method based on near infrared spectrum

Abstract

The invention discloses a tobacco powder particle size detection model establishment method based on near infrared spectroscopy, which comprises the following steps: preparing and obtaining tobacco powder samples with different mesh numbers, detecting the particle size distribution of the tobacco powder samples, then collecting near infrared spectroscopy data of the tobacco powder samples according to preset scanning parameters, and determining the particle size distribution of the tobacco powder samples according to the near infrared spectroscopy data. The near infrared spectrum data and the particle size distribution data are utilized to establish a particle size prediction model, the particle size prediction model is optimized, finally, unknown particle size samples used for non-modeling are adopted to construct an external verification set, and the optimized particle size prediction model is predicted to verify the model. The main conception of the method is that a spectrum mathematical model for rapid analysis is established by using tobacco powder with different morphological structures and particle sizes and is used for predicting unknown tobacco samples, so that a convenient, efficient and effective scheme is provided for particle size detection of powder with different mesh numbers.

Description

Technical Field

[0001] The invention relates to the technical field of tobacco quality inspection, and in particular to a method for establishing a tobacco powder particle size detection model based on near infrared spectroscopy. Background Art

[0002] Near-infrared is the earliest non-visible light region discovered. Near-infrared mainly absorbs the double and combined frequencies of the vibration of hydrogen-containing groups XH (X=C, N, O). Different types of organic matter have different degrees of absorbance or transmittance in different bands of the near-infrared spectrum, which means that the spectra of different compounds have different shapes. At the same time, the concentration will also affect the absorbance or transmittance, which conforms to the Lambert-Beer law.

[0003] Therefore, near-infrared spectroscopy analysis technology has been widely used in various fields such as food, pharmaceuticals, environmental monitoring, and agriculture. Near-infrared spectroscopy analysis technology has the advantages of high sensitivity to material feedback and fast information transmission. Based on this, the tobacco industry urgently needs a solution for rapidly detecting the particle size content of tobacco raw materials using near-infrared spectroscopy analysis technology. Summary of the invention

[0004] In view of the above, the present invention aims to provide a method for establishing a smoke powder particle size detection model based on near infrared spectroscopy to solve the above-mentioned technical problems.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The present invention provides a method for establishing a smoke powder particle size detection model based on near infrared spectroscopy, which includes:

[0007] Prepare and obtain tobacco powder samples of different mesh sizes;

[0008] Testing the particle size distribution of tobacco powder samples;

[0009] Collect near infrared spectrum data of the tobacco powder sample according to preset scanning parameters;

[0010] Establishing a particle size prediction model using the near infrared spectrum data and the particle size distribution data;

[0011] Optimizing the particle size prediction model;

[0012] An external validation set is constructed using samples of unknown particle size that are not used for modeling, and the optimized particle size prediction model is used to make predictions to validate the model.

[0013] In at least one possible implementation, establishing a particle size prediction model includes:

[0014] Presetting target evaluation parameters of the particle size prediction model;

[0015] Combining the near infrared spectrum data and the sample particle size distribution data, partial least squares regression analysis modeling is performed, and evaluation parameters of the model are calculated;

[0016] If the calculated evaluation parameters do not meet the target evaluation parameters, the tobacco powder samples involved in the modeling will be eliminated based on the origin;

[0017] Until the final prediction model architecture containing the most production areas that meets the target evaluation parameter requirements is obtained.

[0018] In at least one possible implementation, the establishment method also includes preprocessing the near-infrared spectral data before establishing the particle size prediction model: performing first-order derivative, or vector normalization preprocessing, or first-order derivative and vector normalization preprocessing on the original near-infrared spectral data.

[0019] In at least one possible implementation manner, the optimizing the particle size prediction model includes optimizing by cross-checking and eliminating outliers.

[0020] In at least one possible implementation, the optimization method for removing outliers specifically includes:

[0021] After removing outliers, the optimal component dimension is selected;

[0022] The near-infrared spectrum data after removing the outliers and the particle size distribution data are again matched one to one;

[0023] The partial least squares method is used to fit the re-correlated data.

[0024] In at least one possible implementation, the preparation and obtaining of tobacco powder samples of different mesh sizes includes: baking the tobacco raw material under set temperature conditions to a constant weight, and grinding it with a pulverizer according to a predetermined mesh size requirement to obtain tobacco powder samples of different mesh sizes.

[0025] In at least one possible implementation manner, the tobacco raw material includes a single tobacco component, or a combination of a tobacco component and a non-tobacco substance.

[0026] Compared with the prior art, the main design concept of the present invention is to establish a spectral mathematical model for rapid analysis using tobacco powders of different morphological structures and particle sizes. Specifically, after tobacco leaves are baked to a constant mass state under preset temperature conditions, tobacco powder samples with different gradients of tobacco particle size are prepared using an ultrafine pulverizer, and the particle size distribution instrument is used to detect the particle size; then, the raw powder sample is scanned by an infrared spectrometer, and its spectrum is preprocessed using the vector normalization method and the first-order derivative method, and finally a quantitative model is established in combination with the least squares method for predicting unknown tobacco samples. The present invention provides a convenient, efficient and effective solution for the detection of powder particle sizes of different mesh sizes. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be further described below with reference to the accompanying drawings, in which:

[0028] Figure 1 A schematic flow chart of a method for establishing a smoke powder particle size detection model based on near infrared spectroscopy provided in an embodiment of the present invention;

[0029] Figure 2 A fitting diagram of the model prediction value and the true value provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] Embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be interpreted as limiting the present invention.

[0031] The present invention proposes an embodiment of a method for establishing a smoke powder particle size detection model based on near infrared spectroscopy. Specifically, Figure 1 shown, including:

[0032] Step S1, preparing and obtaining tobacco powder samples of different mesh sizes;

[0033] Specifically, the tobacco raw material can be placed under set temperature conditions and baked until the mass is constant, and then ground using an ultrafine grinder according to the established mesh size requirements to obtain tobacco powder samples of different mesh sizes; wherein the tobacco raw material can be a pure tobacco component, and can also be combined with one or more other non-tobacco substances including but not limited to plums, orange peels, tree moss, star anise, fennel, etc.

[0034] Step S2, detecting the particle size distribution of the tobacco powder sample;

[0035] Specifically, the median particle sizes of tobacco powder of different mesh sizes can be tested by using special instruments based on industry standards. Of course, in order to obtain a test result that is closer to reality, the total number of sample samples is preferably not less than 400.

[0036] Step S3, collecting near infrared spectrum data of the tobacco powder sample according to preset scanning parameters;

[0037] Step S4, establishing a particle size prediction model using the near infrared spectrum data and the particle size distribution data;

[0038] Specifically, this link may also include the following processes:

[0039] (1) presetting target evaluation parameters of the particle size prediction model;

[0040] (2) performing partial least squares regression analysis modeling in combination with the near-infrared spectral data and the sample particle size distribution data, and calculating the evaluation parameters of the model; it can be supplemented here that the near-infrared spectral data can also be preferably preprocessed, for example, the spectral preprocessing can be a first-order derivative, vector normalization preprocessing, or first-order derivative + vector normalization preprocessing of the original spectral data, so as to reduce the influence of non-target factors;

[0041] (3) If the calculated evaluation parameters do not meet the target evaluation parameters, the tobacco powder samples involved in the modeling will be eliminated based on the origin;

[0042] (4) Thus, the final prediction model architecture containing the most production areas while meeting the target evaluation parameter requirements is obtained. That is, the final prediction model architecture can achieve the prediction of ultrafine smoke particle size that meets the target evaluation parameter requirements.

[0043] Continuing from the above, step S5: optimizing the particle size prediction model;

[0044] In actual operation, optimization can be achieved by cross-checking and eliminating outliers. For example, one sample can be selected from the modeling samples to test the model, and the remaining samples can be used for modeling. The above process is repeated until all samples have been used for testing at least once. In addition, the optimization process mentioned here can also include selecting the optimal component dimension after eliminating outliers; matching the near-infrared spectral data with the particle size distribution data after eliminating outliers one by one, and fitting the matched data again using the partial least squares method to optimize the particle size prediction model;

[0045] Step S6: construct an external validation set using samples of unknown particle size that are not used for modeling, and perform predictions on the optimized particle size prediction model to validate the model.

[0046] In combination with the above embodiments, in order to facilitate understanding of the concept of the present invention, specific practical application examples are provided below for implementation reference:

[0047] First, the tobacco raw material was placed at 45°C and baked until the sample had a constant weight. An ultrafine grinder was then used to prepare tobacco powders of 100, 200, 300, 400, 600, 800, 1000, 1200, 1400, and 1600 meshes. Twenty samples of each were prepared for use.

[0048] Next, the BT-9300ST laser particle size analyzer can be used to test the particle size distribution of tobacco ultrafine powder samples with different gradients of 100, 200, 300, 400, 600, 800, 1000, 1200, 1400, and 1600 mesh, with no less than 20 samples tested for each gradient.

[0049] Then, put the tobacco ultrafine powder samples of different particle sizes of 100, 200, 300, 400, 600, 800, 1000, 1200, 1400, and 1600 mesh into the sample cup in turn, and gently flatten the sample in the cup with a sample press, fix the thickness of each sample in the cup to 6 cm, and then place the sample on the rotating table of the spectrometer for scanning (preferably, the gold-plated inner wall of the integrating sphere can be used as the background during scanning), and the scanning range of the spectrum acquisition can be set: 3600~12500cm -1 ; Resolution 8 cm -1 ; Number of scans: 64 times, in order to collect the spectrum and obtain the spectrum of the sample.

[0050] After obtaining the near-infrared spectrum of the sample, the partial least squares (PLS) method is used in combination with the feature extraction algorithm to establish a quantitative analysis model between the near-infrared spectrum and the particle size. For the actual operation process, the near-infrared spectra of 400 samples are transferred into the OPUS software, and then the particle sizes are input one by one, and the cross-check mode is selected. The preprocessing of the spectral data can be achieved using the optimization function of the software. Then, the best spectral area, spectral preprocessing method and number of principal components are selected, and the particle size quantitative model of tobacco ultrafine powder is established using the optimal conditions. The coefficient of determination (R 2 ) and the root mean square error of cross validation (RMSECV), as well as the comparison of the sample spectrum with the calibration set sample spectrum for abnormal verification. When the model continuously cross-checks the results, the samples that deviate seriously from the fitting line or have a deviation greater than 3 times the average deviation are eliminated, and finally the best modeling scheme is determined.

[0051] For reference Figure 2Schematic diagram of the model prediction value and true value fitting, where the optimal component dimension after removing outliers is 5 and the model correlation coefficient R 2 It reached 98.77%, RMSECV was 3.28, and residual deviation RPD was 0.0326. It can be seen that there is a significant linear relationship between the spectral data and the quantitative index of the sample, which also verifies that the near-infrared spectrum of the sample contains information closely related to the particle size index.

[0052] Finally, the smoke powder particle size distribution model can be verified: use the established particle size model to predict it, test the linear relationship between the predicted value and the reference value, and examine the linear correlation coefficient R 2 , root mean square error RMSECV and residual deviation RPD, as shown in the following Appendix 1.

[0053] Table 1 True value and predicted value of smoke powder particle size at different mesh sizes

[0054]

[0055] It can be seen from the model data of the true value and predicted value of the particle size of tobacco powder of different mesh sizes in Table 1 that the deviation range is 0.00~0.21μm, and the average value of the difference is 0.06μm, which is within the acceptable range, that is, there is no significant difference between the true value of the median particle size of tobacco powder and the predicted value output by the model constructed by the present invention.

[0056] In summary, the present invention prepares and obtains tobacco powder samples of different mesh sizes, detects the particle size distribution of the tobacco powder samples, then collects near-infrared spectral data of the tobacco powder samples according to preset scanning parameters, then uses the near-infrared spectral data and particle size distribution data to establish a particle size prediction model, and optimizes the particle size prediction model, and finally uses unknown particle size samples not used for modeling to construct an external validation set, and predicts the optimized particle size prediction model to verify the model. The main idea of ​​the present invention is to use tobacco powders of different morphological structures and particle sizes to establish a spectral mathematical model for rapid analysis, which is used to predict unknown tobacco samples, thereby providing a convenient, efficient and effective solution for the detection of powder particle sizes of different mesh sizes.

[0057] If the expressions expressing orientation are mentioned in the embodiments of the present invention, they are relative concepts based on the embodiments. In addition, "at least one" means one or more, and "plurality" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, c can be single or multiple.

[0058] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A method for establishing a smoke powder particle size detection model based on near infrared spectroscopy, characterized in that: include: Prepare and obtain tobacco powder samples of different mesh sizes; Testing the particle size distribution of tobacco powder samples; Collect near infrared spectrum data of the tobacco powder sample according to preset scanning parameters; Establishing a particle size prediction model using the near infrared spectrum data and the particle size distribution data; Optimizing the particle size prediction model; An external validation set is constructed using samples of unknown particle size that are not used for modeling, and the optimized particle size prediction model is used to make predictions to validate the model.

2. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to claim 1, characterized in that: The particle size prediction model is established comprising: Presetting target evaluation parameters of the particle size prediction model; Combining the near infrared spectrum data and the sample particle size distribution data, partial least squares regression analysis modeling is performed, and evaluation parameters of the model are calculated; If the calculated evaluation parameters do not meet the target evaluation parameters, the tobacco powder samples involved in the modeling will be eliminated based on the origin; Until the final prediction model architecture containing the most production areas that meets the target evaluation parameter requirements is obtained.

3. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to claim 1, characterized in that: The establishment method also includes preprocessing the near-infrared spectrum data before establishing the particle size prediction model: performing first-order derivative, or vector normalization preprocessing, or first-order derivative and vector normalization preprocessing on the original near-infrared spectrum data.

4. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to claim 1, characterized in that: The optimizing the particle size prediction model includes optimizing by cross-checking and eliminating outliers.

5. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to claim 4, characterized in that: The optimization methods for removing outliers include: After removing outliers, the optimal component dimension is selected; The near-infrared spectrum data after removing the outliers and the particle size distribution data are again matched one to one; The partial least squares method is used to fit the re-correlated data.

6. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to any one of claims 1 to 5, characterized in that: The method of preparing and obtaining tobacco powder samples with different mesh sizes includes: baking tobacco raw materials under set temperature conditions to constant weight, and grinding them with a pulverizer according to predetermined mesh requirements to obtain tobacco powder samples with different mesh sizes.

7. The method for establishing a smoke powder particle size detection model based on near infrared spectroscopy according to claim 6, characterized in that: The tobacco raw material includes a single tobacco component, or a combination of tobacco components and non-tobacco substances.

Images