Method for identifying fresh cocoon / dried cocoon raw silk based on ultraviolet absorption spectrum
By using ultraviolet absorption spectroscopy and linear discriminant analysis, the problems of simplicity and accuracy in identifying raw silk from fresh and dried cocoons have been solved, achieving low-cost and efficient raw silk quality control.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2026-04-13
- Publication Date
- 2026-07-10
AI Technical Summary
Existing technologies cannot easily and accurately distinguish between fresh and dried silk cocoons, resulting in unstable quality of silk products, and the testing equipment is expensive and complex.
Using a method based on ultraviolet absorption spectroscopy, raw silk samples were pretreated and extracted, and spectral data were collected using a micro spectrophotometer. A discriminant model was established for identification by combining standard normal transformation and linear discriminant analysis.
It has achieved a low-cost, simple and rapid identification method that can distinguish between fresh cocoons and dried cocoons with high accuracy, making it suitable for large-scale industrial applications.
Smart Images
Figure CN122361333A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for identifying raw silk from cocoons, and to the field of raw silk inspection technology, specifically to a method for identifying raw silk from fresh / dried cocoons based on ultraviolet absorption spectroscopy. Background Technology
[0002] Traditionally, the silk industry has widely adopted the dried cocoon reeling process to produce raw silk, using dried mulberry cocoons that have been dried with hot air. In recent years, fresh cocoon reeling technology has been gradually promoted due to its significant economic benefits. Fresh mulberry cocoons are not dried; instead, they are refrigerated or frozen. However, the quality and performance of the raw silk produced from fresh cocoons have not been fully verified. As a result, a mixture of dried and fresh cocoon raw silk has appeared in the market, seriously affecting the quality stability of silk products and hindering the standardized development of fresh cocoon reeling technology.
[0003] To date, there is no simple and accurate method for identifying raw silk from fresh / dried cocoons. Patent CN202410940006.0 discloses a method using acid hydrolysis and stable isotope ratio mass spectrometry to test the nitrogen content and its relative isotope content. This process is relatively complex and requires advanced instruments. Patent CN201610997303.4 discloses a GC-MS method that tests for the presence of 1-hexadecyl acetate. This method also requires expensive instruments and the test results are not reliable. Patents CN201710199278.X, CN201710280139.X, and CN201510037780.1 differentiate raw silk based on its macroscopic properties, such as hydrophilicity / hydrophobicity and sericin gelation characteristics, but their accuracy is insufficient. Summary of the Invention
[0004] To address the problems existing in the background technology, this invention provides a method for identifying fresh / dried raw silk cocoons based on ultraviolet absorption spectroscopy. This invention overcomes the shortcomings of existing technologies, such as expensive equipment, complex detection processes, and low accuracy, by providing a simple method with high detection accuracy.
[0005] The technical solution adopted in this invention is: The method of the present invention for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy includes: Step 1) Pre-treat and extract raw silk samples from fresh and dried cocoons to obtain their respective test solutions.
[0006] Step 2) Collect the UV absorption spectra of the test solution and ultrapure water for each raw silk sample. Based on the UV absorption spectrum of ultrapure water, perform background signal correction on the UV absorption spectrum of each test solution to obtain the corrected UV absorption spectrum.
[0007] Step 3) Preprocess the calibrated UV absorption spectrum of each raw silk sample to obtain the preprocessed spectrum and its derivative spectrum; perform difference analysis on the preprocessed spectrum and its derivative spectrum of each raw silk sample, screen out the difference scoring index with significant distinguishing ability, and then construct a two-dimensional feature space with xy axis to perform visual cluster analysis, and then establish a discrimination model between fresh cocoons and dry cocoons.
[0008] Step 4): Obtain the raw silk sample to be identified and perform the same operations as in Steps 1) to 3) to obtain the preprocessed spectrum and its derivative spectrum. Then, input the preprocessed spectrum into the discrimination model to output the category of the raw silk sample to be identified, thus completing the identification of fresh cocoon and dried cocoon raw silk.
[0009] In step 1), the raw silk samples include fresh cocoon and dried cocoon raw silk. For each raw silk sample, pretreatment is performed by dividing the raw silk sample into several raw silk parts, specifically into 3 to 5 different parts for sampling. A predetermined mass of raw silk fragments is cut out from each raw silk part, specifically into fragments with a length of 0.5cm to 2cm. Each raw silk fragment of the raw silk sample is placed in a centrifuge tube, and then an extraction solvent is added for ice bath extraction. Finally, the supernatant in the centrifuge tube is extracted as the test solution.
[0010] The extraction solvent is an aqueous solution of an alcohol or chlorinated hydrocarbon organic solvent. The ratio of raw silk fragments to the extraction solvent in the raw silk sample is 1:20 to 1:60 g / mL. In specific implementation, the extraction solvent is an 80% (v / v) aqueous solution of an alcohol, and the ratio of raw silk fragments to the extraction solvent is 1:40 g / mL. Multiple random sampling combined with a specific ratio range can maximize the representativeness of the extract and the stability of its characteristic concentration, avoiding the influence of excessively high or low concentrations on the spectral characteristics.
[0011] The ice bath extraction is ultrasonic extraction at 0°C for 20 to 60 minutes, specifically 30 minutes. Ice bath ultrasound can effectively prevent the degradation or structural changes of heat-sensitive identification markers in raw silk due to local overheating during the extraction process.
[0012] In step 2), a micro spectrophotometer is used to collect the ultraviolet absorption spectra of the test liquid and ultrapure water, and to correct the background signal of the ultraviolet absorption spectrum of the test liquid; the ultraviolet band range during collection is 190nm ~ 400nm.
[0013] In step 3), each corrected UV absorption spectrum is processed using standard normal variate transformation (SNV) to obtain a preprocessed spectrum. Then, the derivative spectra of the preprocessed spectrum are obtained, including the first derivative spectrum and the second derivative spectrum. The first and second derivative values of the absorbance of the preprocessed spectrum are calculated. The preprocessing using standard normal variate transformation (SNV) is used to eliminate the effects of baseline drift and light scattering. The first and second derivative spectra are used to enhance hidden subtle feature signals.
[0014] In step 3), the differences in the pretreatment spectra, first derivative spectra, and second derivative spectra of each raw silk sample are analyzed, and a discriminant model is established, as follows: The absorbance values of the pretreated, first-derivative, and second-derivative spectra of each raw silk sample were extracted and used as basic features. The ratio features formed by pairwise combinations of these basic features were then used as candidate discriminant indicators. Each candidate discriminant indicator was divided into fresh cocoon and dried cocoon groups according to the type of raw silk sample it corresponds to. The mean and standard deviation of each candidate discriminant indicator within each fresh and dried cocoon group were calculated. Then, the difference score for each candidate discriminant indicator was calculated. For each candidate discriminant indicator in the fresh and dried cocoon groups, the ratio of the absolute value of the difference between the means of two candidate discriminant indicators to the sum of their standard deviations was used as the difference score. The difference score index can effectively reduce the interference caused by the difference in substrate absorption of different batches of silkworm cocoons and accurately anchor the essential difference between dried and fresh cocoons and raw silk. All candidate discrimination indicators are sorted from high to low according to their difference scores. If the feature wavelength of the current candidate discrimination indicator is less than the feature wavelength of any candidate discrimination indicator that has been retained in the previous list, it is removed. Finally, the remaining candidate discrimination indicators are constructed into a set of independent and high-quality one-dimensional features. The discrimination indicators in the one-dimensional feature set are combined in pairs to construct a two-dimensional feature space. The two-dimensional feature spaces are cross-validated by the linear discriminant analysis (LDA) algorithm. The two-dimensional feature space with the highest classification accuracy is selected to establish a discrimination model between dried and fresh cocoons.
[0015] During discrimination, the following parameters are used: Feature values: the original absorbance values at wavelengths of 200-400 nm, and the first and second derivatives, such as A200 (absorbance value at 200 nm), D1_220 (first derivative value at 220 nm), and D2_230 (second derivative value at 230 nm); Discrimination indices: such as A200 (single feature value) and A200 / D1_230 (ratio of feature value combinations); Difference score: the difference score of the discrimination indices between the dry cocoon group and the fresh cocoon group, calculated using the mean and standard deviation of the discrimination indices. Discrimination indices with a spacing smaller than the wavelength tolerance are removed: for example, if the wavelength tolerance is set to 2 nm, then if A228 is selected as part of the discrimination indice combination, then A226, A227, A229, and A230 are not considered.
[0016] In each discrimination spectrum of the raw silk samples of fresh and dried cocoons, a difference scoring index is constructed between each pair of discrimination spectra for each spectral group. The difference scoring index is positively correlated with the absolute value of the difference between the mean values of the spectral data of the two discrimination spectra and negatively correlated with the sum of the standard deviations of the two discrimination spectra.
[0017] The beneficial effects of this invention are: 1. Low detection cost, simple and rapid operation, and easy to promote on a large scale: Existing identification methods mostly rely on extremely expensive and complex precision instruments such as stable isotope ratio mass spectrometers or gas chromatography-mass spectrometry (GC-MS); while the testing of macroscopic physical properties (such as hydrophilicity / hydrophobicity, gel properties) suffers from strong subjectivity and insufficient accuracy. This invention creatively introduces micro-ultraviolet absorption spectroscopy technology, which can complete the test with only a conventional micro-spectrophotometer combined with simple organic solvent ultrasonic extraction, greatly reducing the equipment threshold and detection time cost, and has high value for popularization and promotion.
[0018] 2. Minimal sampling volume, high representativeness, perfectly suited to industrial sampling inspection needs: This invention employs a multi-point random sampling method for sample pretreatment. Each test requires only a trace amount of raw silk fragments (e.g., 0.1g), barely damaging the macroscopic morphology and subsequent performance of the raw silk, making it an extremely low-destructive test. Its simple operation process and highly automated model discrimination mechanism perfectly meet the current industrial application needs of the silk industry for large-scale, rapid quality sampling inspection of raw silk from dried / fresh cocoons. Attached Figure Description
[0019] Figure 1 The images show the ultraviolet absorption spectra of fresh and dried raw silk from cocoons in Example 1. Figure 2 The ultraviolet absorption spectra of fresh and dried raw silk from cocoons in Example 2 are shown below. Figure 3 The ultraviolet absorption spectra of fresh and dried raw silk from cocoons in Example 3 are shown. Figure 4 The ultraviolet absorption spectra of fresh and dried raw silk from cocoons in Example 4 are shown. Figure 5 This is a diagram of the LDA two-dimensional partitioning model of the dataset in Example 4; Figure 6 Example 4 shows the accuracy representation of the validation set partitioning based on the LDA model of the dataset; Figure 7 The ultraviolet absorption spectra of fresh and dried raw silk cocoons are shown in Comparative Example 1. Figure 8 The image shows the ultraviolet absorption spectra of fresh and dried raw silk cocoons in Comparative Example 2. Detailed Implementation
[0020] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0021] Specific embodiments of the present invention are as follows: Example 1: (1) Three samples each of fresh and dried raw silk from known sources were selected and cut into fragments approximately 1 cm in length. 0.1 g of each raw silk fragment was accurately weighed and placed in a 5 mL centrifuge tube, and 4 mL of 80% methanol solution (solid-to-liquid ratio 1:40 g / mL) was added. Extraction was performed for 30 min using an ultrasonic cleaner under ice bath conditions at 0°C. The solid raw silk was then removed, and the supernatant was collected for analysis.
[0022] (2) The raw UV absorption spectrum of the supernatant in the 190–400 nm wavelength range was collected using a micro spectrophotometer. Ultrapure water was used as a blank control, and the spectrum of 80% methanol solution was measured simultaneously for background signal correction.
[0023] (3) Perform standard normal variable transformation (SNV) preprocessing on the collected raw spectral data and calculate its first and second derivative spectra.
[0024] (4) Extract feature indicators, construct a two-dimensional feature space to perform visual cluster analysis on raw silk data, and initially substitute them into the linear discriminant analysis (LDA) algorithm to verify the linear separability of features.
[0025] Experimental results: such as Figure 1As shown, 80% methanol combined with an ice bath can extract characteristic difference substances between fresh and dried cocoon raw silk, which are reflected in the wavelength-absorbance ultraviolet spectral curves, with the blue and red curves showing different distribution patterns. Quantitative evaluation shows that the DB (Davies-Bouldin) Score in its two-dimensional feature space is as low as 1.29~1.73, and the Pseudo-F value is as high as 2.94~4.99, as shown in Table 1. Table 1 shows the distribution difference index of the absorbance and derivative curves of fresh and dried cocoon raw silk in Example 1, reflecting the overall spectral differences, indicating high intra-class cohesion and low inter-class overlap. Based on the above-mentioned superior characteristics, the LDA classification model can successfully achieve accurate classification and identification of fresh and dried cocoon raw silk.
[0026] Table 1 Example 2: The only difference from Example 1 is that in step (1), the extraction solvent is replaced with an ethanol solution with a volume fraction of 80%, while the rest of the steps are completely consistent with Example 1.
[0027] Experimental results: such as Figure 2 As shown, ethanol, as an organic solvent, can also effectively extract characteristic difference substances. The extracted characteristic indicators can still form their own cluster centers in spatial projection, with a DB Score of 1.57~3.63 and a Pseudo-F value of 0.58~3.3 in the two-dimensional feature space. As shown in Table 2, the distribution difference index of absorbance and derivative of fresh and dried cocoon raw silk in Example 2 is shown. The cluster density is slightly more dispersed than that in Example 1, and the inter-cluster overlap is slightly increased.
[0028] Table 2 Example 3: The only difference from Example 1 is that in step (3), the original spectral data is not preprocessed by standard normal variable transformation (SNV), but the first and second derivative spectra are calculated directly. The remaining steps are completely consistent with Example 1.
[0029] Experimental results: such as Figure 3As shown, the blue and red curves are distinct. The derivative spectrum itself has the function of eliminating baseline shift, and it can still effectively highlight the latent characteristic peaks even without SNV processing to address light scattering errors. After extracting feature indices and constructing a two-dimensional feature space, the DB Score of the two-dimensional feature space is 2.56~3.99, and the Pseudo-F value is 0.46~0.98. As shown in Table 3, the absorbance and its derivative distribution difference index of fresh and dried cocoon raw silk in Example 3 are shown. The cluster density is slightly more dispersed than that in Example 1, and the inter-class overlap is slightly increased.
[0030] Table 3 Example 4: The difference from Example 1 is that in step (1), 18 fresh silk cocoons and 30 dried silk cocoons from known sources are selected. In step (4), the LDA algorithm is used to construct a classification model, and 12 fresh silk cocoons and 16 dried silk cocoons are randomly selected as independent validation sets as blind samples to be input into the model for accuracy prediction.
[0031] Experimental results: such as Figure 4 It can be observed that the blue and red curves have different distribution trends. Their DB scores in the two-dimensional feature space range from 2.92 to 15.35, and their Pseudo-F values range from 0.22 to 8.04, as shown in Table 4. Table 4 shows the distribution difference index of absorbance and its derivative of fresh and dried raw silk cocoons in Example 4. Relatively stable and accurate characteristic difference indicators can be found, with concentrated clustering and low inter-class overlap. Figure 5 As shown, the obtained feature indices are: the horizontal axis D1_213 / D1_229, which refers to the ratio of the first derivative of the ultraviolet spectrum at 213nm to the first derivative at 229nm; and the vertical axis D2_263 / D2_212, which refers to the ratio of the second derivative of the ultraviolet spectrum at 263nm to the second derivative at 212nm. Substituting these high-quality features into a supervised LDA classification model, the accuracy (Acc) for distinguishing between fresh and dried cocoon raw silk reached 89.6%, where the black dashed line represents the boundary between dried and fresh cocoon raw silk. Furthermore, blind testing on an independent validation set showed a stable accuracy of 85.71% in the confusion matrix. Figure 6 As shown, when faced with large-scale samples with certain basis differences between batches, the feature indicators extracted by this invention exhibit a certain degree of stability and anti-interference ability.
[0032] Table 4 Comparative Example 1: The only difference from Example 1 is that in step (1), the extraction solvent is replaced with ultrapure water, and the rest of the steps are completely consistent with Example 1.
[0033] Experimental results: such as Figure 7 As shown, pure water cannot extract lipid-soluble markers or denatured hydrophobic substances. In the feature space, the data points from dried and fresh cocoons exhibit severe random confounding. The DB Score in the two-dimensional feature space ranges from 1.87 to 2.34, and the Pseudo-F value drops sharply to 1.06 to 1.78. Table 5 shows the distribution difference index of absorbance and its derivative for raw silk from fresh and dried cocoons in Comparative Example 1. The clustering density is dispersed, and the inter-class overlap is high, causing the established LDA model to fail and making it impossible to define effective classification boundaries, thus demonstrating the necessity of using organic solvents.
[0034] Table 5 Comparative Example 2: The only difference from Example 1 is that in step (1), ice bath conditions were not used, and ultrasonic extraction was performed directly at room temperature. All other steps are completely consistent with Example 1.
[0035] Experimental results: Ultrasonic cavitation heat generation caused secondary thermal denaturation of fresh cocoon markers, leading to convergence of characteristics between the two types of samples and a sharp increase in spatial overlap, such as... Figure 8 As shown in Table 6, the DB Score of its two-dimensional feature space increased to 6.11~10.11, and the Pseudo-F value decreased to 0.21~0.55. The table shows the distribution difference index of absorbance and derivative of fresh and dried raw silk cocoons in Comparative Example 2. The cluster density is dispersed and the inter-class overlap is high. Under these poor features, the misclassification rate of the LDA model output is extremely high, which proves the necessity of ice bath conditions.
[0036] Table 6 Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, this invention is intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.
[0037] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations of this application fall within the scope of the equivalent technology of this invention, this application also intends to include these modifications and variations.
Claims
1. A method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy, characterized in that, include: Step 1) Pre-treat and extract raw silk samples from fresh and dried cocoons to obtain their respective test solutions; Step 2) Collect the UV absorption spectra of the test solution and ultrapure water for each raw silk sample. Based on the UV absorption spectrum of ultrapure water, perform background signal correction on the UV absorption spectrum of each test solution to obtain the corrected UV absorption spectrum. Step 3) Preprocess the calibrated UV absorption spectrum of each raw silk sample to obtain the preprocessed spectrum and its derivative spectrum; Difference analysis was conducted on the pretreatment spectra and derivative spectra of each raw silk sample to screen out difference scoring indicators, and then a discrimination model between fresh cocoons and dried cocoons was established. Step 4): Obtain the raw silk sample to be identified and perform the same operations as in Steps 1) to 3) to obtain the preprocessed spectrum and its derivative spectrum. Then, input the preprocessed spectrum into the discrimination model to output the category of the raw silk sample to be identified, thus completing the identification of fresh cocoon and dried cocoon raw silk.
2. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 1, characterized in that: In step 1), the raw silk samples include fresh cocoon and dried cocoon raw silk. For each raw silk sample, pretreatment is performed by dividing the raw silk sample into several raw silk parts evenly. A preset mass of raw silk fragments is cut out from each raw silk part. Each raw silk fragment of the raw silk sample is placed in a centrifuge tube, and then an extraction solvent is added for ice bath extraction. Finally, the supernatant in the centrifuge tube is extracted as the test solution.
3. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 2, characterized in that: The extraction solvent is an aqueous solution of an alcohol or chlorinated hydrocarbon organic solvent, and the ratio of raw silk fragments to extraction solvent in the raw silk sample is 1:20 ~ 1:60 g / mL.
4. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 2, characterized in that: The ice bath extraction is ultrasonic extraction at 0°C for 20 to 60 minutes.
5. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 1, characterized in that: In step 2), a micro spectrophotometer is used to collect the ultraviolet absorption spectra of the test liquid and ultrapure water, and to correct the background signal of the ultraviolet absorption spectrum of the test liquid; the ultraviolet band range during collection is 190nm ~ 400nm.
6. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 1, characterized in that: In step 3), each corrected UV absorption spectrum is processed using the Standard Normal Variable Transform (SNV) to obtain a preprocessed spectrum, and then the derivative spectrum of the preprocessed spectrum is obtained, including the first derivative spectrum and the second derivative spectrum.
7. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 6, characterized in that: In step 3), the differences in the pretreatment spectra, first derivative spectra, and second derivative spectra of each raw silk sample are analyzed, and a discriminant model is established, as follows: The absorbance values of the pretreated, first-derivative, and second-derivative spectra of each raw silk sample were extracted and used as basic features. The ratio features formed by pairwise combinations of these basic features were then used as candidate discriminant indicators. Each candidate discriminant indicator was divided into fresh cocoon and dried cocoon groups according to the type of raw silk sample it corresponded to. The mean and standard deviation of each candidate discriminant indicator within each fresh and dried cocoon group were calculated. Then, the difference score for each candidate discriminant indicator was calculated. For each candidate discriminant indicator in the fresh and dried cocoon groups, the absolute value and standard deviation of the difference between the means of two candidate discriminant indicators were calculated. The ratio of the sum of differences is used as the difference score. All candidate discriminant indicators are sorted from high to low according to their respective difference scores. If the feature wavelength of the current candidate discriminant indicator is less than the feature wavelength of any candidate discriminant indicator that has been retained in the previous list, it is removed. Finally, the remaining candidate discriminant indicators are used to construct a one-dimensional feature set. The discriminant indicators in the one-dimensional feature set are combined in pairs to construct a two-dimensional feature space. The two-dimensional feature spaces are cross-validated by the Linear Discriminant Analysis (LDA) algorithm. The two-dimensional feature space with the highest classification accuracy is selected to establish a discrimination model between dry cocoons and fresh cocoons.
8. The method for identifying fresh / dried raw silk from cocoons based on ultraviolet absorption spectroscopy according to claim 7, characterized in that: In the discrimination spectra of the raw silk samples of fresh and dried cocoons, a difference scoring index is constructed for each pair of discrimination spectra for each spectral group. The difference scoring index is positively correlated with the absolute value of the difference between the mean values of the spectral data of the two discrimination spectra and negatively correlated with the sum of the standard deviations of the two discrimination spectra.