Classification method for cigarette mainstream smoke based on oversampling self-encoding network spectral analysis

Abstract

The application discloses a cigarette mainstream smoke classification method based on oversampling self-encoding network spectrum analysis, which comprises the following steps: building an air-core waveguide Raman spectrum system; using an ISO suction mode, collecting the air-core waveguide Raman spectrum of the mainstream smoke of a single sample cigarette of a single kind of sample cigarette to be classified by using the air-core waveguide Raman spectrum system; processing the spectrum data by using a high-density discrete wavelet transform method to obtain high-density wavelet transform coefficients; obtaining a cigarette quality score by using an oversampling self-encoding network spectrum analysis model and the high-density wavelet transform coefficients; and obtaining the classification result of the mainstream smoke of the sample cigarette to be classified by using a quality control sample database and the cigarette quality score. The cigarette mainstream smoke classification method based on the oversampling self-encoding network spectrum analysis can organically combine the air-core waveguide Raman spectrum system and the oversampling self-encoding network, and can accurately evaluate the stability of the quality of tobacco products.

Description

Technical Field

[0001] This invention relates to the field of tobacco product quality evaluation technology, and in particular to a method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis. Background Technology

[0002] Cigarettes are a special consumer product whose effects are achieved through inhalation. Their main smoke composition is complex and can vary due to factors such as production processes and raw materials. These variations directly impact the stability of cigarette quality. Therefore, effectively differentiating cigarette products of varying quality and batches through the detection of their main smoke is a pressing issue for the tobacco industry.

[0003] Currently, the classification of cigarette quality mainly relies on expert evaluation. This method depends on the expert's experience and judgment, which has problems such as cumbersome operation and low efficiency, and makes it difficult to accurately judge the stability of cigarette quality.

[0004] Therefore, there is an urgent need for a method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis. Summary of the Invention

[0005] The purpose of this invention is to provide a method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis, in order to solve the problems in the prior art and accurately assess the stability of tobacco product quality.

[0006] This invention provides a method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis, comprising:

[0007] Construct a hollow waveguide Raman spectroscopy system;

[0008] The hollow waveguide Raman spectra of the mainstream smoke from a single cigarette sample of a single type of cigarette to be classified were collected using the ISO suction method and a hollow waveguide Raman spectroscopy system.

[0009] The high-density discrete wavelet transform method is used to process the spectral data to obtain the high-density wavelet transform coefficients;

[0010] Cigarette quality scores are obtained by using a pre-constructed oversampling autoencoder network spectral analysis model and high-density wavelet transform coefficients. The oversampling autoencoder network spectral analysis model is used to characterize the relationship between the high-density wavelet transform coefficients and the cigarette quality scores.

[0011] Using a pre-built quality control sample database and the cigarette quality score obtained based on the oversampling autoencoder network spectral analysis model, the mainstream smoke classification results of the cigarette samples to be classified are obtained.

[0012] The method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis, as described above, preferably includes the following: The construction of the hollow waveguide Raman spectroscopy system specifically includes:

[0013] The hollow waveguide Raman spectroscopy detection equipment is coupled and connected to the hollow waveguide gas cell;

[0014] A rotary smoke extractor is installed at the inlet of the hollow waveguide gas cell, and a peristaltic pump is installed at the outlet of the hollow waveguide gas cell.

[0015] A filter device and a pressure gauge are installed sequentially between the rotary smoke machine and the hollow waveguide gas pool;

[0016] A mass flow meter and a limit valve are installed sequentially between the hollow waveguide gas pool and the peristaltic pump.

[0017] In the above-described method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis, preferably, the hollow waveguide Raman spectroscopy detection device uses a laser with a wavelength of 532nm as the excitation source, and the output excitation light is transmitted to the Raman probe through an excitation fiber, and then the excitation light is focused into the hollow waveguide gas cell through a quartz plate.

[0018] In the above-described method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis, preferably, the inner diameter of the hollow waveguide gas cell is 1.2 mm, the excitation light emitted by the laser is focused into the hollow waveguide through an optical coupling system, the laser spot excites the gas in the hollow waveguide gas cell, and the generated Raman scattered light signal is collected and transmitted to the spectrometer through a collection fiber.

[0019] The cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis described above, preferably, involves using an ISO aspiration method and a hollow waveguide Raman spectroscopy system to collect the hollow waveguide Raman spectra of the mainstream smoke of a single sample cigarette of a single type to be classified, specifically including:

[0020] Using the ISO aspiration method, 10-20 cigarettes of a single type of cigarette were selected as test samples. The mainstream smoke of each cigarette sample was collected using a hollow waveguide Raman spectroscopy system. For each cigarette sample, the hollow waveguide Raman spectrum of the mainstream smoke was collected 20-30 times.

[0021] In the above-described method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis, preferably, during the acquisition of hollow waveguide Raman spectra, the laser output power is set to 400mW, the relative pressure in the hollow waveguide gas cell and gas path is maintained at atmospheric pressure, the gas temperature in the hollow waveguide gas cell is 20℃, the Raman spectrum measurement integration time is 5s, each sample is acquired 3 times, and the 3 spectral data are averaged into one spectral data.

[0022] The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis, as described above, preferably includes the following step: processing the spectral data using high-density discrete wavelet transform to obtain high-density wavelet transform coefficients.

[0023] The collected spectral data is filtered to remove individual abnormal spectral data;

[0024] The filtered spectral data to be evaluated are subjected to noise reduction and filtering to extract obvious characteristic peaks and characteristic signals in the spectrum;

[0025] The high-density discrete wavelet transform method was used to process the spectral data, and the time and frequency were oversampled by a factor of two to obtain the high-density wavelet transform coefficients.

[0026] The cigarette mainstream smoke classification method based on oversampled autoencoder network spectral analysis, as described above, preferably includes the following step: obtaining cigarette quality scores using a pre-constructed oversampled autoencoder network spectral analysis model and high-density wavelet transform coefficients, specifically including:

[0027] The high-density wavelet transform coefficients are input into the pre-constructed oversampled autoencoder network spectral analysis model to output the cigarette quality score.

[0028] The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis as described above, preferably, involves obtaining the mainstream smoke classification result of the cigarette sample to be classified using a pre-constructed quality control sample database and the cigarette quality score obtained based on the oversampled autoencoder network spectral analysis model, specifically including:

[0029] Construct a quality control sample database and select cigarette products with recognized quality as a reference for quality scoring;

[0030] If the cigarette quality score is within ±20% of the cigarette quality score range in the quality control sample database, then the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is reasonable.

[0031] If the cigarette quality score exceeds ±20% of the cigarette quality score range in the quality control sample database, the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is unreasonable and further improvements to the formula or production process are needed.

[0032] This invention provides a method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis. It organically combines a hollow waveguide Raman spectroscopy system with an oversampling autoencoder network, enabling accurate extraction of cigarette quality information from complex mainstream smoke Raman spectra. This allows for rapid classification and stability assessment of cigarette quality, effectively improving the efficiency and accuracy of mainstream cigarette smoke quality evaluation. Based on oversampling autoencoder network spectral analysis technology, it achieves the analysis and characterization of cigarette quality features without requiring the individual detection of various specific components in the volatile gases of tobacco. The method accurately determines the quality stability of different batches and brands of cigarettes using oversampling autoencoder network calculations. It is not dependent on a specific cigarette platform and can be widely applied to classifying different samples, identifying genuine and counterfeit cigarettes, and homogenizing evaluations. The entire evaluation process is fast, accurate, and highly sensitive, providing effective technical support for improving the quality and efficiency of cigarette products and developing new products. Attached Figure Description

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described below with reference to the accompanying drawings, wherein:

[0034] Figure 1 A flowchart illustrating an embodiment of the cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis provided by the present invention;

[0035] Figure 2 This is a schematic diagram of a hollow waveguide Raman spectroscopy system. Detailed Implementation

[0036] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. The descriptions of the exemplary embodiments are merely illustrative and are in no way intended to limit the present disclosure or its application or use. The present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided so that the present disclosure will be thorough and complete, and will fully express the scope of the disclosure to those skilled in the art. It should be noted that, unless specifically stated otherwise, the relative arrangement of components and steps, the composition of materials, numerical expressions, and values ​​set forth in these embodiments should be interpreted as exemplary only and not as limiting.

[0037] The terms “first,” “second,” and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different parts. Terms such as “including” or “contains” mean that the element preceding the term encompasses the element listed after it, and do not exclude the possibility of encompassing other elements as well. Terms such as “above” and “below” are used only to indicate relative positional relationships; when the absolute position of the described object changes, this relative positional relationship may also change accordingly.

[0038] In this disclosure, when a specific component is described as being located between a first component and a second component, an intermediary component may or may not be present between the specific component and the first or second component. When a specific component is described as connecting to other components, the specific component may be directly connected to the other components without having an intermediary component, or it may not be directly connected to the other components but may have an intermediary component.

[0039] All terms used in this disclosure (including technical or scientific terms) have the same meaning as understood by one of ordinary skill in the art to which this disclosure pertains, unless otherwise specifically defined. It should also be understood that terms defined in a general dictionary, such as a dictionary, should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.

[0040] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0041] like Figure 1 As shown, the cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis provided in this embodiment specifically includes the following in actual implementation:

[0042] Step S1: Construct a hollow waveguide Raman spectroscopy system.

[0043] This invention combines online Raman spectroscopy technology to accurately extract cigarette quality information from the complex and variable Raman spectra of mainstream cigarette smoke, and clarifies the impact of different formulation processes on mainstream cigarette smoke, enabling rapid judgment and classification of cigarette quality stability. Compared with other analytical methods, Raman spectroscopy can detect not only conventional volatile organic compounds but also diatomic gases such as oxygen and nitrogen. Therefore, it can fully assess the combustion status and quality information of mainstream cigarette smoke. In addition, compared with ordinary Raman spectroscopy, hollow waveguide Raman spectroscopy makes full use of the dual function of its hollow optical fiber as both a gas cell and an amplifier, greatly improving the detection sensitivity of Raman spectroscopy analysis, so as to capture more aroma and latent aroma substances in mainstream smoke.

[0044] like Figure 2 As shown, in one embodiment of the cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis of the present invention, step S1 may specifically include:

[0045] Step S11: Couple the hollow waveguide Raman spectroscopy detection device to the hollow waveguide gas cell.

[0046] The hollow waveguide Raman spectroscopy detection device uses a laser with a wavelength of 532nm as the excitation source, and the output excitation light is transmitted to the Raman probe through the excitation fiber. Then, the excitation light is focused into the hollow waveguide gas cell through a quartz plate.

[0047] Step S12: Install a rotary smoke extractor at the inlet end of the hollow waveguide gas cell and a peristaltic pump at the outlet end of the hollow waveguide gas cell.

[0048] The hollow waveguide gas cell has an inner diameter of 1.2 mm. The excitation light emitted by the laser is focused into the hollow waveguide through an optical coupling system. The laser spot excites the gas in the hollow waveguide gas cell, and the generated Raman scattered light signal is collected and transmitted to the spectrometer through a collecting optical fiber. Furthermore, the gas path of the hollow waveguide Raman spectroscopy system uses a stainless steel tube with a diameter of 3 mm as the gas channel.

[0049] Step S13: Install a filter device and a pressure gauge between the rotary smoke extractor and the hollow waveguide gas pool in sequence.

[0050] Step S14: Install a mass flow meter and a limit valve sequentially between the hollow waveguide gas pool and the peristaltic pump.

[0051] Pressure gauges, mass flow meters, and limit valves can be used to control the gas pressure in the hollow waveguide gas pool and gas path, ensuring seamless integration between the relevant instrument system and the rotary smoke extractor.

[0052] By constructing a hollow waveguide Raman spectroscopy system, seamless coupling between the hollow waveguide Raman spectroscopy detection equipment and the rotary smoke extraction machine can be achieved.

[0053] Step S2: Using the ISO suction method, the hollow waveguide Raman spectrum of the mainstream smoke of a single sample cigarette of a single type of cigarette to be classified is collected using a hollow waveguide Raman spectroscopy system.

[0054] Specifically, using the ISO aspiration method, 10-20 cigarettes of a single type of cigarette were selected as test samples. The mainstream smoke of each cigarette sample was collected using a hollow waveguide Raman spectroscopy system. The hollow waveguide Raman spectra of the mainstream smoke of each cigarette sample were collected 20-30 times.

[0055] Furthermore, during the acquisition of Raman spectra of the hollow waveguide, the output power of the laser was set to 400mW, the relative pressure in the gas cell and gas path of the hollow waveguide was maintained at atmospheric pressure, the gas temperature in the gas cell of the hollow waveguide was 20℃, the measurement integration time of the Raman spectrum was 5s, and the spectrum of each sample was acquired 3 times. The 3 spectral data were averaged into one spectral data to reduce the influence of random noise.

[0056] Step S3: Process the spectral data using the high-density discrete wavelet transform method to obtain the high-density wavelet transform coefficients.

[0057] In one embodiment of the cigarette mainstream smoke classification method based on oversampled autoencoder network spectral analysis of the present invention, step S3 may specifically include:

[0058] Step S31: Filter the collected spectral data to remove individual abnormal spectral data.

[0059] Data filtering ensures more accurate results in subsequent data processing.

[0060] Step S32: Perform noise reduction and filtering on the screened spectral data to be evaluated in order to extract obvious characteristic peaks and characteristic signals in the spectrum.

[0061] Step S33: The spectral data is processed using the high-density discrete wavelet transform method, which performs double oversampling on both time and frequency to obtain the high-density wavelet transform coefficients.

[0062] Among them, the spectral data after high-density wavelet transform is an oversampling process of the original spectral data, which ensures the oversampling effect of the spectral data.

[0063] Step S4: Using the pre-constructed oversampling autoencoder network spectral analysis model and high-density wavelet transform coefficients, the cigarette quality score is obtained. The oversampling autoencoder network spectral analysis model is used to characterize the relationship between the high-density wavelet transform coefficients and the cigarette quality score.

[0064] Specifically, the high-density wavelet transform coefficients are input into the pre-constructed oversampled autoencoder network spectral analysis model, which outputs cigarette quality scores. Using the oversampled autoencoder network spectral analysis model, scores for various cigarette quality aspects can be obtained based on the spectral data of mainstream cigarette smoke collected by the hollow waveguide Raman spectroscopy system. Through the oversampled autoencoder network spectral analysis model, Raman spectral feature compression and dimensionality reduction can be achieved, significantly reducing the data dependence on cigarette samples and obtaining good feature modeling results.

[0065] This invention organically combines a hollow waveguide Raman spectroscopy system with an oversampling autoencoder network, enabling accurate extraction of cigarette quality information from complex mainstream smoke Raman spectra, rapid classification and stability assessment of cigarette quality, and effectively improving the efficiency and accuracy of evaluating the quality of mainstream cigarette smoke.

[0066] Step S5: Using the pre-built quality control sample database and the cigarette quality score obtained based on the oversampling autoencoder network spectral analysis model, the mainstream smoke classification results of the cigarette samples to be classified are obtained.

[0067] In one embodiment of the cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis of the present invention, step S5 may specifically include:

[0068] Step S51: Construct a quality control sample database and select cigarette products with recognized quality as a reference for quality scoring.

[0069] Choosing cigarette products with generally recognized quality as a reference for quality scoring facilitates subsequent analysis of unknown samples.

[0070] Step S52: If the cigarette quality score is within ±20% of the cigarette quality score range in the quality control sample database, then the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is reasonable.

[0071] Step S53: If the cigarette quality score exceeds ±20% of the cigarette quality score range in the quality control sample database, the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is unreasonable and further improvements to the formula or production process are needed.

[0072] The cigarette mainstream smoke classification method based on oversampling autoencoder network spectral analysis provided in this invention organically combines a hollow waveguide Raman spectroscopy system and an oversampling autoencoder network. This allows for accurate extraction of cigarette quality information from complex mainstream smoke Raman spectra, enabling rapid classification and stability assessment of cigarette quality, effectively improving the efficiency and accuracy of mainstream smoke quality evaluation. Based on oversampling autoencoder network spectral analysis technology, it achieves the analysis and characterization of cigarette quality characteristics without requiring the individual detection of various specific components in the volatile gases of tobacco. The oversampling autoencoder network calculation method can accurately determine the quality stability of different batches and brands of cigarettes. It is not dependent on a specific cigarette platform and can be widely applied to classifying different samples, identifying genuine and counterfeit cigarettes, and homogenizing evaluations. The entire evaluation process is fast, accurate, and highly sensitive, providing effective technical support for improving the quality and efficiency of cigarette products and developing new products.

[0073] The embodiments of this disclosure have now been described in detail. To avoid obscuring the concept of this disclosure, some details known in the art have not been described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein based on the above description.

[0074] While specific embodiments of this disclosure have been described in detail by way of examples, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of this disclosure. Those skilled in the art should understand that modifications can be made to the above embodiments or equivalent substitutions can be made to some technical features without departing from the scope and spirit of this disclosure. The scope of this disclosure is defined by the appended claims.

Claims

1. A method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis, characterized in that, include: Construct a hollow waveguide Raman spectroscopy system; The hollow waveguide Raman spectra of the mainstream smoke from a single cigarette sample of a single type of cigarette to be classified were collected using the ISO suction method and a hollow waveguide Raman spectroscopy system. The high-density discrete wavelet transform method is used to process the spectral data to obtain the high-density wavelet transform coefficients; Cigarette quality scores are obtained by using a pre-constructed oversampling autoencoder network spectral analysis model and high-density wavelet transform coefficients. The oversampling autoencoder network spectral analysis model is used to characterize the relationship between the high-density wavelet transform coefficients and the cigarette quality scores. Using a pre-built quality control sample database and the cigarette quality score obtained based on the oversampling autoencoder network spectral analysis model, the mainstream smoke classification results of the cigarette samples to be classified are obtained. The construction of the hollow waveguide Raman spectroscopy system specifically includes: A hollow waveguide Raman spectroscopy detection device is coupled to a hollow waveguide gas cell. The hollow waveguide Raman spectroscopy detection device uses a laser with a wavelength of 532 nm as the excitation source, and the output excitation light is transmitted to the Raman probe through an excitation fiber. Then, the excitation light is focused into the hollow waveguide gas cell through a quartz plate. The inner diameter of the hollow waveguide gas cell is 1.2 mm. The excitation light emitted by the laser is focused into the hollow waveguide through an optical coupling system. The laser spot excites the gas in the hollow waveguide gas cell, and the generated Raman scattered light signal is collected and transmitted to the spectrometer through a collection fiber. A rotary smoke extractor is installed at the inlet of the hollow waveguide gas cell, and a peristaltic pump is installed at the outlet of the hollow waveguide gas cell. A filter device and a pressure gauge are installed sequentially between the rotary smoke machine and the hollow waveguide gas pool; A mass flow meter and a limit valve are installed sequentially between the hollow waveguide gas pool and the peristaltic pump.

2. The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis according to claim 1, characterized in that, The method employs ISO suction and utilizes a hollow waveguide Raman spectroscopy system to collect the hollow waveguide Raman spectra of the mainstream smoke from a single sample cigarette of a particular type to be classified. Specifically, this includes: Using the ISO aspiration method, 10-20 cigarettes of a single type of cigarette were selected as test samples. The mainstream smoke of each cigarette sample was collected using a hollow waveguide Raman spectroscopy system. For each cigarette sample, the hollow waveguide Raman spectrum of the mainstream smoke was collected 20-30 times.

3. The method for classifying mainstream cigarette smoke based on oversampling autoencoder network spectral analysis according to claim 2, characterized in that, During the acquisition of Raman spectra of hollow waveguides, the output power of the laser was set to 400 mW, the relative pressure in the gas cell and gas path of the hollow waveguide was maintained at atmospheric pressure, the gas temperature in the gas cell of the hollow waveguide was 20℃, the measurement integration time of the Raman spectrum was 5s, and the spectrum of each sample was acquired 3 times. The 3 spectral data were averaged into one spectral data.

4. The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis according to claim 1, characterized in that, The process of processing spectral data using the high-density discrete wavelet transform method to obtain high-density wavelet transform coefficients specifically includes: The collected spectral data is filtered to remove individual abnormal spectral data; The filtered spectral data to be evaluated are subjected to noise reduction and filtering to extract obvious characteristic peaks and characteristic signals in the spectrum; The high-density discrete wavelet transform method was used to process the spectral data, and the time and frequency were oversampled by a factor of two to obtain the high-density wavelet transform coefficients.

5. The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis according to claim 1, characterized in that, The method of obtaining cigarette quality scores using a pre-constructed oversampling autoencoder network spectral analysis model and high-density wavelet transform coefficients specifically includes: The high-density wavelet transform coefficients are input into the pre-constructed oversampled autoencoder network spectral analysis model to output the cigarette quality score.

6. The method for classifying mainstream cigarette smoke based on oversampled autoencoder network spectral analysis according to claim 1, characterized in that, The process of using a pre-built quality control sample database and the cigarette quality score obtained based on the oversampling autoencoder network spectral analysis model to obtain the mainstream smoke classification results of the cigarette samples to be classified specifically includes: Construct a quality control sample database and select cigarette products with recognized quality as a reference for quality scoring; If the cigarette quality score is within ±20% of the cigarette quality score range in the quality control sample database, then the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is reasonable. If the cigarette quality score exceeds ±20% of the cigarette quality score range in the quality control sample database, the mainstream smoke classification result of the cigarette sample to be classified is that the cigarette quality control is unreasonable and further improvements to the formula or production process are needed.

Images