Method for nano-plastic analysis through AIOF-induced confinement localization and machine learning-assisted SERS (Surface Enhanced Raman Scattering)

Abstract

The invention discloses a method for nano-plastic analysis through AIOF induced confinement localization and machine learning assisted SERS, and belongs to the technical field of nano-plastic trace detection, high-precision quantification and classification, and the method comprises the following steps: step 1, preparing In2O3 / FTO; 2, preparing an AIOF substrate through a silver mirror reaction; step 3, preparing a nano plastic dispersion liquid, and collecting an SERS spectrum; 4, AIOF substrate performance characterization is carried out; step 5, preprocessing the spectral data; and step 6, classifying the nano plastics dispersed in different environmental samples through a machine learning model based on the spectral data preprocessed in the step 5. According to the method for nano plastic analysis through AIOF-induced confinement positioning and machine learning assisted SERS, the AIOF nano array SERS substrate obtained through the unique preparation method has a multi-layer nano wall array structure with mutually communicated holes, has a larger surface area, can provide more adsorption sites and hot spots, and can be applied to nano plastic analysis. Therefore, a stronger SERS signal is generated.

Description

Technical Field

[0001] The present invention relates to the technical fields of trace detection, high-precision quantification, and classification of nanoplastics, and particularly to a method for nanoplastics analysis using AIOF-induced confinement localization and machine learning-assisted SERS. Background Art

[0002] With the large-scale production and wide use of plastics, their accumulation in the environment is increasing. Microplastics are currently widely defined as solid plastic particles with a size ≤ 5 mm composed of polymers, functional additives, and other chemical substances, and nanoplastics (< 1 μm) are considered an extension of microplastics. Micro-nanoplastics in the environment are ingested by animals and accumulate in various tissues, and then are widely transferred through the food chain, causing harm to human health. However, the lack of a method for quickly and non-destructively identifying and quantifying these plastics in complex biological matrices has hindered the study of the accumulation of micro-nanoplastics in animals.

[0003] There are still many problems to be solved in the high-sensitivity detection of nanoplastics in complex samples by the current SERS technology. Although a coffee ring with the function of enriching micro-nanoplastics can be easily obtained by simply allowing a liquid droplet to dry naturally on a planar SERS substrate, achieving high SERS detection sensitivity, the random deposition of nanoparticles during the natural drying process of the colloidal droplet results in an irregular shape and position of the coffee ring formed on the solid substrate, making the distribution of the enriched micro-nanoplastics quite uneven and difficult to find with the naked eye or under a low-power microscope at low concentrations. Secondly, there are Raman peaks from different types of plastics as well as other impurities in the actual sample, which greatly increases the complexity of the Raman spectrum and brings great difficulties to quickly distinguish the types of plastics in the sample and accurately quantify them. Summary of the Invention

[0004] The object of the present invention is to provide a method for nanoplastics analysis using AIOF-induced confinement localization and machine learning-assisted SERS. The AIOF nanoarray SERS substrate obtained by a unique preparation method has a multi-layer nano-wall array structure with interconnected pores, a larger surface area, and can provide more adsorption sites and hot spots, thus generating a stronger SERS signal. The confinement effect brought by the scratches on it induces the deposition of micro-nanoplastics at the specified position to a certain extent, providing a localization strategy for the SERS detection of low-concentration nanoplastics, reducing the interference of other impurities on the SERS detection in a complex environment, and avoiding the problem that it is difficult to locate the position of the coffee ring formed by low-concentration plastics after drying with a portable Raman to a certain extent.

[0005] To achieve the above object, the present invention provides a method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis, comprising the following steps:

[0006] Step 1, prepare In2O3 / FTO;

[0007] Step 2, prepare the AIOF substrate through a silver mirror reaction;

[0008] Step 3, prepare a nanoplastics dispersion solution and collect SERS spectra;

[0009] Step 4, characterize the performance of the AIOF substrate;

[0010] Step 5, preprocess the spectral data;

[0011] Step 6, classify the nanoplastics dispersed in different environmental samples based on the preprocessed spectral data in Step 5 through a machine learning model.

[0012] Preferably, the specific operation of Step 1 is: place the cleaned FTO substrate with the conductive surface facing down into a polytetrafluoroethylene reaction kettle containing InCl3·4H2O and C2H5NS ethanol solution for reaction. When a golden yellow film grows on the FTO surface, take it out and rinse it repeatedly with ethanol and distilled water, and then put it into an oven for drying;

[0013] After that, anneal the product to obtain In2O3 / FTO. Take out the In2O3 / FTO and cut it into small pieces, perform scribing treatment, and then set aside.

[0014] Preferably, in Step 1, the dosage of InCl3·4H2O is 5-10 mM, the dosage of C2H5NS is 10-30 mM, the dosage of ethanol solvent is 40 mL, the reaction time is 10-15 h, and the reaction temperature is 150-200 °C;

[0015] The annealing parameters are set as: anneal at 400-500 °C at a heating rate of 2-3 °C / min for 3-5 h.

[0016] Preferably, the specific operation of Step 2 is: put the In2O3 / FTO obtained in Step 1 into a clean beaker, first add silver ammonia solution, then add glucose solution. After the reaction, take out the product and rinse it clean to finally obtain an Ag / In2O3 / FTO substrate with high-density silver nanoparticles uniformly loaded on the surface, that is, the AIOF substrate.

[0017] Preferably, in Step 2, the concentration of the silver ammonia solution is 0.1-0.2 M, the dosage is 1-2 mL, the concentration of glucose is 0.1-0.5 M, the dosage is 0.1-1.0 mL, and the reaction is maintained for 10-15 min.

[0018] Preferably, the specific operation of step three is as follows: suspend PET solid particles in a sodium dodecyl sulfate solution, perform ultrasonic treatment on it to obtain a PET monodisperse mother liquor, and prepare PS monodisperse mother liquor and PMMA monodisperse mother liquor with different particle sizes; subsequently, add different volumes of PET monodisperse mother liquor, PS monodisperse mother liquor and PMMA monodisperse mother liquor into different environmental samples respectively to prepare nano-plastic dispersions with different concentrations.

[0019] Drop the above nano-plastic dispersion on the AIOF substrate, dry it, and collect the SERS signal.

[0020] Preferably, in step three, the dosage of PET solid particles is 20 - 30 mg, the solution is 10 - 15 mL of pure water containing 2% sodium dodecyl sulfate, and the ultrasonic time is 30 - 45 min.

[0021] Preferably, the specific operation of step four is as follows: use Gemini 500SEM to perform SEM image characterization on the AIOF substrate, use a rotating anode X-ray diffractometer equipped with Cu Kα radiation to record the crystal structure, and collect data in the range of 20° to 80°, use JEM 2100F for TEM, energy-dispersive X-ray spectroscopy and selected area electron diffraction characterization, measure the ultraviolet-visible absorption spectrum in a TU-1950 double-beam ultraviolet-visible spectrophotometer, and determine the Raman mapping by a confocal Raman spectrometer equipped with a 532 nm laser.

[0022] Preferably, the specific operation of step five is as follows: the initial range of the SERS spectrum is 600 - 1800 cm -1 , truncated to 700 - 1700 cm -1 , use BWSpec software to perform background subtraction on different nano-plastic spectra, the peak intensities that vary in different nano-plastic spectra are normalized to a standard scale between 0 - 1 using Min-Max normalization to eliminate the influence of intensity differences on the model, use principal component analysis and t-distributed stochastic neighbor embedding method to project the high-dimensional data onto a two-dimensional plane, and visualize the data of nano-plastic categories dispersed in different environmental samples in the principal component analysis and t-distributed stochastic neighbor embedding scatter plot.

[0023] Preferably, the specific operation of step six is as follows: use KNN, GDBT, CNN and Transformer algorithms to classify the nano-plastics dispersed in different environmental samples, and reveal the relationship between the SERS characteristic peak intensity and the nano-plastic concentration in the sample by performing linear fitting on the data. For each concentration, take three different measurement points to calculate the average value and standard deviation of its intensity, and the model fitting degree is determined by the coefficient of determination R 2Evaluate, analyze and predict exponential data, draw a fitting curve by taking the logarithm of the concentration and the corresponding characteristic peak intensity respectively;

[0024] Describe the performance of the machine learning model in distinguishing different types of nanoplastics in environmental samples graphically through a confusion matrix, and evaluate the prediction performance using a series of metrics, including accuracy, recall, precision and F1 score:

[0025]

[0026] Among them, tp, fp, tn and fn represent the number of true positive, false positive, true negative and false negative predictions respectively;

[0027] For external validation, use the established machine learning model to classify nanoplastics in different environmental samples to determine the classification accuracy.

[0028] Therefore, the present invention adopts the above-mentioned method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis, and has the following beneficial effects:

[0029] (1) Efficient substrate preparation and performance advantages: The AIOF nanoarray SERS substrate obtained by a unique preparation method has a multi-layer nano-wall array structure with interconnected pores, has a larger surface area, can provide more adsorption sites and hot spots, thus generating a stronger SERS signal. The confinement effect brought by the scratches on it induces the deposition of micro-nanoplastics at the specified position to a certain extent, providing a localization strategy for the SERS detection of low-concentration nanoplastics, reducing the interference of other impurities in the complex environment on the SERS detection and avoiding the problem that it is difficult to locate the coffee ring position formed by low-concentration nanoplastics after drying with a portable Raman to a certain extent;

[0030] (2) Accurate classification ability in multiple scenarios: The four developed classification models based on machine learning can accurately classify different types of nanoplastics in multiple environments after scientific preprocessing and dimensionality reduction of Raman spectral data.

[0031] The technical solution of the present invention will be further described in detail below with reference to the drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a flowchart of an embodiment of the method of the present invention for AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis;

[0033] Figure 2It is the topographic, structural, and property diagram of the AIOF substrate in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0034] Figure 3 It is the performance characterization diagram of the AIOF substrate in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0035] Figure 4 It is the data preprocessing, data dimensionality reduction, and ML classification model flowchart in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0036] Figure 5 It is the classification analysis diagram of nanoplastics in different environments by a machine learning model in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0037] Figure 6 It is the quantitative analysis diagram of nanoplastics dispersed in samples of different environments in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0038] Figure 7 It is the Raman spectrum of 4-MBA on the blank substrate and the SERS spectrum of 4-MBA adsorbed on the AIOF substrate in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0039] Figure 8 It is the relationship diagram between the amount of InCl3·4H2O used and the SERS performance with 4-MBA (10 -8 M) when the silver plating time is 11 minutes in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0040] Figure 9 It is the relationship between the silver plating time and the SERS intensity of 4-MBA (10 -8 M) when the amount of InCl3·4H2O used is 8 mM in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0041] Figure 10 It is the SERS spectra of nanoplastics with different particle sizes in pure water with AIOF as the substrate in an embodiment of the method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastics analysis of the present invention;

[0042] Figure 11 This is the evaluation of the classification performance of four models for nanoplastics in different environments in an embodiment of the method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis according to the present invention. The indicators include accuracy, recall, precision, and F1 score;

[0043] Figure 12 This is the decision boundary of different environments determined by KNN in an embodiment of the method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis according to the present invention;

[0044] Figure 13 This is the limit of detection (LOD) and actual detection sensitivity (True) of nanoplastics in different environments in an embodiment of the method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis according to the present invention. Detailed implementation mode

[0045] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0046] Unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meaning understood by those of ordinary skill in the field to which the present invention belongs.

[0047] The present invention provides a method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis, including the following steps:

[0048] Step 1. Preparation of In2O3 / FTO: Place the clean FTO substrate with the conductive side facing down into a polytetrafluoroethylene reaction kettle containing an ethanol solution of InCl3·4H2O and C2H5NS for reaction. A golden-yellow film grows on the FTO surface. Take it out and rinse it repeatedly with ethanol and distilled water, and then put it into an oven overnight;

[0049] After that, anneal the product to obtain In2O3 / FTO. Take out the In2O3 / FTO and cut it into small pieces, and then perform scribing treatment for later use.

[0050] The dosage of InCl3·4H2O is 5-10 mM, the dosage of C2H5NS is 10-30 mM, the dosage of ethanol solvent is 40 mL, the reaction time is 10-15 h, and the reaction temperature is 150-200 °C;

[0051] The annealing parameters are set as follows: anneal at 400-500 °C at a heating rate of 2-3 °C / min for 3-5 h.

[0052] Step 2. Preparation of the AIOF substrate through the silver mirror reaction: Put the In2O3 / FTO obtained in Step 1 into a clean beaker, first add the silver ammonia solution, then add the glucose solution. After the reaction, take out the product and rinse it clean. Finally, obtain the Ag / In2O3 / FTO substrate with high-density silver nanoparticles uniformly loaded on the surface, that is, the AIOF substrate.

[0053] The concentration of the silver ammonia solution is 0.1 - 0.2 M, and the dosage is 1 - 2 mL. The concentration of glucose is 0.1 - 0.5 M, and the dosage is 0.1 - 1.0 mL. The reaction lasts for 10 - 15 min.

[0054] Step 3. Preparation of the nanoplastics samples and collection of SERS spectra: Suspend the PET solid particles in the sodium dodecyl sulfate solution, perform ultrasonic treatment on it to obtain the PET monodisperse mother liquor, and prepare the PS monodisperse mother liquor and PMMA monodisperse mother liquor with different particle sizes; subsequently, add different volumes of the PET monodisperse mother liquor, PS monodisperse mother liquor, and PMMA monodisperse mother liquor into different environmental samples respectively to prepare nanoplastics dispersions with different concentrations;

[0055] Drop the above nanoplastics dispersions on the AIOF substrate, dry them, and collect the SERS signals.

[0056] The dosage of the PET solid particles is 20 - 30 mg, the dispersion is 10 - 15 mL of pure water containing 2% sodium dodecyl sulfate, and the ultrasonic time is 30 - 45 min.

[0057] Step 4. Performance characterization of the AIOF substrate: Perform SEM image characterization on the AIOF substrate using Gemini 500SEM, record the crystal structure using a rotating anode X-ray diffractometer equipped with Cu Kα radiation, and collect data in the range of 20° to 80°. Perform TEM, energy-dispersive X-ray spectroscopy, and selected area electron diffraction characterization using JEM 2100F. The ultraviolet-visible absorption spectrum is measured in a TU-1950 double-beam ultraviolet-visible spectrophotometer, and the Raman mapping is determined by a confocal Raman microscope equipped with a 532 nm laser. The above steps are as Figure 1 shown in (a) of the above.

[0058] Step 5. Preprocessing of the spectral data: The initial range of the SERS spectrum is 600 - 1800 cm -1 , truncated to 700 - 1700 cm -1, the background of different nanoplastics spectra was subtracted using the BWSpec software. The changing peak intensities shown in different nanoplastics spectra were normalized to a standard scale between 0 and 1 using Min - Max normalization to eliminate the influence of intensity differences on the model. The principal component analysis and t - distributed stochastic neighbor embedding methods were used to project the high - dimensional data onto a two - dimensional plane, and the data of different nanoplastics categories were visualized in the scatter plots of the principal component analysis and t - distributed stochastic embedding methods.

[0059] Step Six: Classify the nanoplastics samples based on the pre - processed spectral data in Step Five through a machine learning model: The KNN, GDBT, CNN, and Transformer algorithms were used to classify the nanoplastics dispersed in different environmental samples. By linearly fitting the data, the relationship between the SERS characteristic peak intensity and the concentration of the nanoplastics samples was revealed. For each concentration, three different measurement points were taken to calculate the average value and standard deviation of its intensity. The goodness of fit of the model was evaluated using the coefficient of determination R 2 for evaluation. The exponential - type data was analyzed and predicted. By taking the logarithm of the concentration and the corresponding characteristic peak intensity respectively, a fitting curve was plotted;

[0060] The performance of the machine learning model in distinguishing nanoplastics in different environmental samples was graphically described through a confusion matrix. A series of metrics were used to evaluate the prediction performance, including accuracy, recall, precision, and F1 - value:

[0061]

[0062] where tp, fp, tn, and fn represent the number of true positive, false positive, true negative, and false negative predictions respectively;

[0063] For external validation, the established machine learning model was used to classify all the nanoplastics added to pure water, Yangtze River water, and fish meat samples to determine the classification accuracy, as shown in (b) of Figure 1 .

[0064] Example One

[0065] The present invention provides a method for AIOF - induced confinement localization and machine - learning - assisted SERS for nanoplastics analysis, including the following steps:

[0066] Step 1. Preparation of In2O3 / FTO: Place the clean FTO substrate with the conductive side facing down into a polytetrafluoroethylene reaction kettle containing InCl3·4H2O and C2H5NS ethanol solution for reaction. After reacting at 160 °C for 14 hours, a golden-yellow thin film grows on the surface of the FTO. Take it out and rinse it repeatedly with ethanol and distilled water, and then place it in an oven at 160 °C overnight. Subsequently, anneal the product at 450 °C at a heating rate of 2 °C / min for 4 h to obtain In2O3 / FTO. Finally, take out the annealed In2O3 / FTO and cut it into small pieces of 5×5 mm, and perform scribing treatment for later use.

[0067] Step 2. Preparation of the AIOF substrate through the silver mirror reaction: Place the prepared glass slide flat in a clean beaker, first add 1 mL of 0.2 M silver ammonia solution, and then add 0.5 mL of 0.3 M glucose solution. After maintaining the reaction for 11 min, take it out and rinse it clean. Finally, an Ag / In2O3 / FTO substrate with a high density of silver nanoparticles uniformly loaded on the surface, that is, the AIOF substrate, can be obtained.

[0068] The four manufacturing steps of AIOF include the growth of In2S3 (ISF) on the FTO surface, the thermal annealing of In2S3 on the FTO surface to generate In2O3 (IOF), the scribing treatment of the IOF surface, and the in-situ growth of Ag nanoparticles on the IOF to generate Ag / In2O3 / FTO (AIOF). The In2S3 grown on the FTO in the first step is a three-dimensional structure composed of a multi-layered nanowall array structure with interconnected pores. The In2O3 obtained after thermal annealing effectively retains the three-dimensional structure of In2S3, as shown in Figure 2 (a) in. Compared with the unplated Ag IOF substrate, SEM analysis after in-situ silver plating shows that the entire surface of the three-dimensional (3D) network architecture is uniformly decorated with densely packed Ag nanoparticles (Ag NPs), and the average particle size of these particles is 97 nm, as shown in Figure 2 (b) in and Figure 2 (d) in. TEM images further prove that Ag NPs are uniformly attached to all surfaces of In2O3. Energy-dispersive X-ray (EDX) analysis is used to analyze the elemental composition and distribution of the post-silver-plated AIOF, and it is proved that it is composed of In, O, and Ag, as shown in Figure 2 (c) in. The XRD pattern directly confirms that In2S3 (47.99°) is completely transformed into In2O3 (30.58°) during thermal annealing, and the subsequently grown Ag NPs are elemental silver (38.12°, 44.28°, 64.43°), as shown in Figure 2As shown in (f). The STEM image shows a distinct boundary at the interface between Ag and In2O3, with lattice spacings of 0.146 and 0.115 nm corresponding to the (220) crystal plane of Ag and the (742) crystal plane of In2O3, respectively, as Figure 2 shown in (e). In2O3 has light absorption ability in the range of 200 - 350 nm. To improve the SERS performance of the AIOF substrate under 785 nm laser, the AIOF obtained by in-situ growth of AgNPs on the IOF has a strong LSPR effect in the range of 400 - 800 nm, as Figure 2 shown in (g), which is attributed to the local surface plasmon resonance of Ag nanoparticles.

[0069] To better understand the enhancement mechanism of the AIOF substrate, FDTD simulations were carried out on Ag NPs and Ag / In2O3. The FDTD results show that from the Z-axis direction, strong hotspots are generated in the nano-gaps between Ag NPs; weak enhancement also occurs at the interface between Ag NPs and In2O3, and by comparing the electric field intensities in the hotspots, it is found that the electric field intensity in the hotspots on Ag / In2O3 is higher than that in the hotspots of Ag NPs alone under the same conditions, which is attributed to the synergistic enhancement between Ag and In2O3, as Figure 2 shown in (h) and (i).

[0070] Since the structure of the IOF determines the arrangement state of the subsequently grown Ag NPs, the number of hotspots formed per unit volume, and the adsorption effect on substance molecules, the amount of InCl3·4H2O during the growth of In2S3 was explored under fixed reaction time, temperature, and raw material ratio. Figure 8 (a) shows the relationship between the amount of InCl3·4H2O and the SERS performance between 4-MBA (10 -8 M) when the silver plating time is 11 minutes. According to the histogram of the peak intensity at 1074 cm -1 , it can be seen that the SERS performance of the AIOF substrate is the best when the amount of InCl3·4H2O is 8 mM, as Figure 8 shown in (b). The SEM images show that when the amount of InCl3·4H2O is 8 mM, the In2O3 on the surface of the IOF exhibits the highest roughness and the most pore structures for connecting different layers, as Figure 8 shown in (c)-(f). This provides more growth sites for Ag NPs, thereby increasing the number of hotspots formed per unit volume, and also provides more adsorption sites for the probes. Figure 3 (a)-(b) and Figure 9Figures (a)-(b) show the relationship between the silver plating time and the SERS performance of 4-MBA (10-8M) at the optimal InCl3·4H2O dosage (8 mM). When the silver plating time is 11 minutes, the AIOF substrate has the best SERS performance.

[0071] The SEM images show that before the silver plating time reaches 11 minutes, as the time increases, the accumulation of Ag nanoparticles on the entire surface of the three-dimensional network structure gradually becomes denser and the nano-gap gradually decreases. When the silver plating time is 12 minutes, a lot of silver dendrites grow on the surface, which seriously weakens the LSPR effect of Ag NPs, as Figure 9 shown in (c)-(f). This optimized AIOF substrate is used for the detection of probe molecules and microplastics in different environmental samples subsequently.

[0072] Step 3: Prepare the nanoplastics dispersion and collect the SERS spectra. First, suspend 25 mg of PET solid particles in 10 mL of 2% sodium dodecyl sulfate (SDS) solution and perform ultrasonic treatment for 30 min to obtain 10 mL of PET monodisperse mother liquor (2.5 mg / mL). The monodisperse mother liquor of PS and PMMA with different particle sizes (2.5 mg / mL) are obtained by direct purchase. The nanoplastics dispersions in different environments are subsequently obtained by mixing the mother liquor and real samples. Drop the above nanoplastics dispersion on the AIOF substrate, dry it, and collect the SERS signal.

[0073] Use a portable Raman spectrometer (i-Raman Plus, B&W Tek, USA) to obtain the SERS spectra. The instrument uses a 50× microscope objective lens, operates with an excitation laser wavelength of 785 nm, an integration time of 3 s, and a spectral range of 600-1800 cm -1 . All spectral data are collected with the AIOF substrate as the SERS substrate.

[0074] Step 4: Characterize the performance of the AIOF substrate. Use Gemini 500 SEM for SEM image characterization, use a rotating anode X-ray diffractometer (Rigaku D / MAX2500VL / PC) equipped with Cu Kα radiation to record the crystal structure and collect data in the range of 20° to 80°. Use JEM 2100F for TEM, energy-dispersive X-ray spectroscopy (EDS), and selected area electron diffraction characterization. The ultraviolet-visible absorption spectrum is measured in a TU-1950 double-beam ultraviolet-visible spectrophotometer. Raman mapping is determined by a confocal Raman microscope (LabRam HR Evolution) equipped with a 532 nm laser.

[0075] Step 5. Preprocess the spectral data: The initial range of the SERS spectrum is 600 - 1800 cm -1 , truncated to 700 - 1700 cm -1 , as shown in (a) of Figure 4 . Background subtraction is performed on different nanoplastics spectra using BWSpec software. The peak intensities that vary in different nanoplastics spectra are normalized to a standard scale between 0 - 1 using Min - Max normalization to eliminate the influence of intensity differences on the model. For dimensionality reduction, principal component analysis (PCA) and t - distributed stochastic neighbor embedding (t - SNE) methods are used to project the high - dimensional data onto a two - dimensional (2D) plane. Data visualization of three different nanoplastics categories (PS, PMMA, PET) is performed in the PCA and t - SNE scatter plots, and these illustrations are drawn based on two principal components.

[0076] Figure 4 . (b) in

[0077] is the preprocessing process diagram, where Ⅰ is the baseline correction diagram, Ⅱ is the smoothing process diagram, and Ⅲ is the normalization process diagram. When using Raman spectral data for machine learning, data preprocessing steps such as baseline correction, smoothing, and normalization are key steps to ensure the performance of the model. These operations can effectively eliminate noise interference, unify the data scale, and highlight the true chemical information, thereby significantly improving the accuracy, robustness, and generalization ability of the model.

[0078] First is the baseline correction. The obtained Raman data is fitted and the background signal is subtracted using BWSpec software to restore the true peak height and peak shape.

[0078] The subsequent smoothing process is to reduce noise. Spectral smoothing is performed using the Savitzky - Golay filter because it can reduce noise while retaining the peak shape, which can reduce the interference of noise on the machine learning model and make it easier to learn the true features in the spectral data.

[0079] Finally is the normalization. Under different experimental conditions, such as different laser powers or sample concentrations, the intensity of the Raman signal may vary greatly. The Min - Max normalization is used to process the spectral data to scale the data to the range of 0 - 1, eliminating the influence of intensity differences on the model.

[0080] After the above preprocessing of the Raman spectral data, certain features can be effectively prevented from dominating the training process due to large scales. After the above preprocessing of the Raman spectral data, principal component analysis (PCA) and t - distributed stochastic neighbor embedding (t - SNE) methods are respectively used for dimensionality reduction. Both the PCA and t - SNE scatter plots show that different types of plastic samples have the potential to be accurately distinguished despite some overlaps. Figure 4Among them, (c) shows that using t-SNE dimensionality reduction produces better clustered data than PCA. This is because t-SNE, as a non-linear dimensionality reduction method, focuses on preserving the local structure between data points and is particularly good at visualizing cases like Raman spectra with a lot of high-dimensional data containing characteristic information.

[0081] Step 6: Classify the nanoplastics samples based on the preprocessed spectral data in Step 5 through machine learning models. Among them, the four machine learning models are: two traditional machine learning models (KNN and GDBT) and two deep learning models (CNN and Transformer). And after preprocessing the original spectra data, evaluate their classification capabilities in different scenarios, and use the confined localization SERS on the AIOF nanowire array and the customized machine learning-based Raman spectroscopy processing method to achieve accurate classification and trace measurement of nanoplastics under multi-scenario complex conditions.

[0082] The KNN, GDBT, CNN, and Transformer models are all trained using only the preprocessed data. Use the KNN, GDBT, CNN, and Transformer models to classify the nanoplastics samples. To ensure comparability and consistency, all models use the same set of training sets and test sets. Visualize the decision boundaries of the KNN dimensionality reduction components, highlighting the different regions corresponding to each nanoplastics type, emphasizing the model's ability to distinguish between PS, PMMA, and PET plastic samples, and demonstrating the stability of the AIOF substrate in complex environments in practical applications and the applicability of the model.

[0083] K-Nearest Neighbor Algorithm (KNN): Optimize the key parameters through the grid search method, evaluate the values of the number of neighborhood samples in the range of [3, 5], and at the same time test the performance differences between the two weight assignment strategies of "uniform" (uniform weighting) and "distance" (inverse distance weighting).

[0084] Gradient Boosting Decision Tree (GBDT): Adopt the Bayesian optimization strategy for hyperparameter tuning. The parameters to be adjusted mainly include: the number of base learners (100 decision trees), the maximum depth of a single tree ([3, 5, 10]), the minimum number of samples for node splitting ([1, 2, 5]), the minimum number of samples for leaf nodes ([1, 3, 5]), and the maximum depth of the tree structure ([3, 5]).

[0085] Convolutional Neural Network (CNN): Construct a one-dimensional convolutional architecture with deep feature extraction capabilities. The network consists of three feature extraction modules (each module is composed of Conv1D + MaxPooling1D + BatchNorm). The receptive field is gradually expanded by stacking convolutional layers to capture global features and effectively avoid local feature interference. The classifier adopts a three-layer fully connected structure, and the classification decision is implemented by a sigmoid activation function at the end.

[0086] Transformer Network: For the characteristics of high-dimensional feature data, construct a hybrid architecture: the first layer uses one-dimensional convolution for local feature extraction and dimension compression; then, three Transformer encoding layers (including multi-head self-attention mechanism and feed-forward network) are used to capture long-range dependencies; finally, feature integration is performed through one-dimensional convolutional layers and fully connected layers, and the classification task is also completed by a sigmoid activation function at the end.

[0087] Quantitative Analysis of Nanoplastics:

[0088] To reveal the relationship between the intensity of SERS characteristic peaks and the concentration of nanoplastics, linear fitting is performed on the data. For each concentration, three different measurement points are taken to calculate the average value and standard deviation of their intensities. The goodness of fit of the model is evaluated using the coefficient of determination (R 2 ). For exponential data, analysis and prediction are carried out. By taking the logarithm of the concentration and the corresponding characteristic peak intensity respectively, and then plotting the fitting curve.

[0089] The performance of the machine learning model in distinguishing nanoplastics is graphically described through a confusion matrix. A series of metrics are used to evaluate the prediction performance, including accuracy, recall, precision, and F1 value:

[0090]

[0091] Where tp, fp, tn, and fn represent the number of true positive, false positive, true negative, and false negative predictions respectively.

[0092] For external validation, the established machine learning model is used to classify nanoplastics dispersed in different environmental samples to determine the classification accuracy.

[0093] To quantify the contribution of the AIOF substrate to the enhancement effect, its enhancement factor EF is calculated according to the following formula:

[0094] EF = (I SERS / I BULK ) × (N BULK / N SERS );

[0095] Where, I SERSis the intensity of the SERS spectrum, I BULK is the intensity of the ordinary Raman spectrum, N SERS is the average number of molecules within the laser spot for SERS excitation, N BULK is the average number of molecules within the laser spot for ordinary Raman excitation.

[0096] Example Two

[0097] To prove the applicability of this method for the detection of nanoplastics under natural environmental water and in vivo conditions, taking the Yangtze River ecosystem as an example, Yangtze River water and bream widely distributed in the Yangtze River basin were selected as models to evaluate the precise quantification ability of the AIOF substrate. Three types of nanoplastics (PS, PMMA, PET) were introduced into pure water, Yangtze River water, and fish meat.

[0098] For pure water samples: Different concentrations of PS (20 - 400 nm), PMMA, and PET solutions (W-PS20, W-PS100, W-PS200, W-PS400, W-PMMA, and W-PET) were all obtained by directly diluting the respective stock solutions with ultrapure water.

[0099] For Yangtze River water samples: Different concentrations of PS200 and PMMA solutions (Y-PS, Y-PMMA) were respectively obtained by directly diluting the stock solutions with Yangtze River water.

[0100] For biological samples: Step 1: Approximately 20 g of fish meat tissue was obtained from bream caught in the Yangtze River basin and thoroughly ground into fish mince; Step 2: 2 g of fish mince was respectively added with different masses of PS2000 and PET plastics to prepare mixed samples with different mass gradients; Step 3: The above mixed samples were respectively added to 20 mL of 15% (w / w) tetramethylammonium hydroxide and ultrasonically treated for 60 min at room temperature to fully perform alkaline digestion on the mixed samples; Step 4: Centrifuged at 2000 rpm for 2 min to completely precipitate the protein-loaded nanoplastics with a small amount of protein; Step 5: After removing the supernatant, 100 μL of ultrapure water was added to redisperse the remaining trace protein and the extracted nanoplastics in the precipitate (B-PS, B-PET).

[0101] For the detection of probe molecules, the AIOF substrate was immersed in 200 μL of 4-MBA solution with different concentrations for 30 min and then taken out to collect SERS signals; for the detection of nanoplastics in water samples, 60 μL of the nanoplastic dispersion was dropped on the substrate surface and dried before collecting SERS signals; for the detection of fish meat samples, 60 μL of the redispersion with different mass concentrations was dropped on the AIOF substrate and dried before collecting SERS signals.

[0102] First, the prepared AIOF substrate was tested for SERS performance after being soaked in the classical Raman reporting dye 4-mercaptobenzoic acid (4-MBA). Raman spectra of 4-MBA at gradient concentrations were obtained using a portable Raman spectrometer. The measurement conditions were as follows: laser wavelength 785 nm, exposure time 3 s. The AIOF substrate was soaked in 4-MBA solutions with different concentrations (10 -8 M to 10 -12 M) for 30 min. The lowest detectable concentration was 10 -12 M, as shown in (d) of Figure 3 . Then, the Raman intensity at 1074 cm -1 was used to generate a standard curve, which showed a good linear correlation at various 4-MBA concentrations, as shown in (e) of Figure 3 . According to the 3N formula, the detection limit (LOD) of the AIOF substrate for 4-MBA was calculated to be 2.4×10 -13 .

[0103] As shown in Figure 7 , the enhancement factor (EF) of the substrate after dropping and drying the 4-MBA solution on the AIOF substrate was calculated, and the EF was obtained as 5.045×10 10 . In addition, the uniformity of the SERS signal is a key factor in determining the reliability of the SERS substrate. Using W-PS200 in the actual sample to be measured as a model, Raman mapping within a range of 20×20 μm -1 containing 400 measurement points at the main peak position of 1002 cm 2 showed the uniformity of the Raman signal within a small range, as shown in (g) of Figure 3 ; at the main peak position of 1002 cm -1 , the uniformity of the Raman signal within a large range on the scratch of the substrate with an area of 5×5 mm 2 containing 140 measurement points was shown, and the RSD was 7.03%, as shown in (f) of Figure 3 .

[0104] Since the specific intensity in the SERS spectrum is proportional to the analyte concentration, accurate and sensitive quantitative analysis can be achieved by detecting the intensity of the characteristic peaks. By reading their respective highest characteristic peaks (PS: 1002 cm -1 , PMMA: 812 cm -1 , PET: 1616 cm -1 ) as features, the concentration gradients of nanoplastics above the LOD and the intensities of the corresponding features collected were logarithmically processed and then linearly fitted to achieve high-precision quantification.

[0105] To verify the detection ability of the AIOF substrate for different types of common nanoplastics in the environment, W-PS20, W-PS100, W-PS200, W-PS400, W-PMMA, and W-PET dispersed in pure water were detected. Figure 3 The characteristic peaks identified for each nanoplastic in the SERS spectrum are shown in (c) of Figure 3 : PS at 1002 cm -1 (in-plane breathing vibration of the benzene ring), 1030 cm -1 and 1600 cm -1 : symmetric and asymmetric stretching vibrations of the benzene ring (C-C bond vibration); PMMA at 811 cm -1 (C-O-C bond stretching vibration) and 1452 cm -1 (C-H bond bending vibration), 1730 cm -1 (C=O bond stretching vibration in the ester group); PET at 1100 cm -1 (bending vibration of O-CH2 in the ethylene glycol unit), 1616 cm -1 (C-C bond stretching vibration), 1280 cm -1 (C-O bond stretching vibration in the ester group) and 1730 cm -1 (C=O bond stretching vibration in the ester group). The detection sensitivities of this substrate for PS of the same particle size and different types of plastics are 500 ppb (W-PS20), 250 ppb (W-PS100), 1 ppm (W-PS400), 1 ppm (W-PS200), 2.5 ppm (W-PMMA), and 5 ppm (W-PET) respectively and have a good linear relationship, as shown in (h)-(i) of Figure 3 Figure 3 and (a)-(f) of Figure 10 Figure 10 shown. The significantly lower detection limit of PS can be attributed to its inherently higher Raman activity, especially compared to PMMA and PET.

[0106] In the present invention, 800 Raman spectral data of W-PS200, W-PMMA, and W-PET nanoplastics with different concentrations dispersed in pure water were collected by the AIOF substrate respectively, and their concentrations are all higher than their respective LODs. After preprocessing them, they were used for the training of the four models as shown in (d) of Figure 4 Figure 4 . Then, Raman data with different concentrations dispersed in a pure water environment were collected by the AIOF substrate respectively (94 for W-PET, 94 for W-PMMA, and 103 for W-PS); Raman data with different concentrations dispersed in the Yangtze River water (110 for Y-PMMA and 116 for Y-PS); Raman data with different concentrations dispersed in fish meat samples (89 for B-PET and 110 for B-PS) and after data preprocessing, they were used as the test sets of the four models in three environments, and the concentrations of these data are all higher than the LOD.

[0107] Figure 6 showed that nanoplastics in all environments exhibited good linear relationships within the LOD range ( Figure 12 ), and the SERS spectra of microplastics in these actual environments were highly consistent with their intrinsic Raman spectra. No additional Raman characteristic peaks appeared in the Yangtze River water samples, but Raman characteristic peaks at 757 cm -1 and 950 cm -1 appeared in the fish muscle samples, which were attributed to the incompletely separated fish digestion products, and these substances did not interfere with the reading and identification of the intensities of microplastic characteristic peaks at all.

[0108] In common pure water samples, the Raman spectra of W-PMMA and W-PET on the AIOF substrate were highly consistent with their intrinsic Raman spectra ( Figure 6 in (a), (c)), and the detection sensitivities reached a very high level (W-PMMA: 2.5 ppm, W-PET: 5 ppm) ( Figure 6 in (b), (d)). The Raman spectra of Y-PS and Y-PMMA in the Yangtze River water samples with a similar environment showed no obvious changes ( Figure 6 in (e), (g)), and the detection sensitivities (Y-PS: 1 ppm, Y-PMMA: 10 ppm) did not show obvious reduction ( Figure 6 in (f), (h)), which was attributed to the fact that there were no large pollutants in the Yangtze River water that were likely to cover the substrate, so there was no strong interference with the signals of dried nanoplastics.

[0109] However, the detection sensitivities of the AIOF substrate for B-PS and B-PET in biological samples were significantly reduced (B-PS: 10 μg / g; B-PET: 50 μg / g) ( Figure 6 in (j), (l)), and new Raman characteristic peaks at 757 cm -1 and 950 cm -1 appeared in both B-PS and B-PET samples ( Figure 6 in (i), (k)), which was attributed to the fact that even after alkali digestion and centrifugation separation steps, a small amount of fish digestion products remained in the biological samples mixed with microplastics and adhered to the AIOF substrate during the drying process, thus hindering the entry of nanoplastics into the hot spot area of the substrate to a certain extent, resulting in a decrease in the SERS signal.

[0110] Further evaluated two traditional machine learning models (KNN and GDBT) and two deep learning models (CNN and Transformer) ( Figure 4(d) in it. KNN and GDBT are common supervised learning multi-classification models. Considering the complexity of Raman spectral data in actual samples, especially fish samples, two deep learning models, CNN and Transformer, were developed for plastic sample classification. For water environments such as pure water and Yangtze River water, there is not much difference between traditional machine learning models and deep learning. However, for the classification of nanoplastics in biological samples, the deep learning Transformer (96%) and CNN (99%) models are significantly superior to the traditional machine learning KNN (52%) and GDBT (62%) models in terms of accuracy, as shown in Figure 5 (a) in it and Figure 10 (a)-(b) in it. This may be due to the fact that when testing plastics in bream fish meat, the digestion products of fish meat cannot be completely separated from microplastics, thus introducing additional Raman characteristic peaks that are not in the training set, as shown in Figure 5 (b) in it.

[0111] Figure 5 (c) in it shows that the clusters of W-PS, W-PMMA, and W-PET dispersed in pure water are in three different regions. Only the clusters of W-PS and W-PMMA have slight overlap, which means that the confusion matrix diagram shows that the KNN model in pure water has a 1.94% probability of misidentifying PS as PMMA, but the recognition accuracy for PMMA is 100%, as shown in Figure 5 (e) in it, which is consistent with the decision boundary diagram of KNN showing that only a small number of points representing PS enter the region of PMMA, as shown in Figure 5 (d) in it; GDBT has a 0.97% probability of identifying W-PS as W-PMMA and a 1.06% probability of misidentifying W-PMMA as W-PS, as shown in Figure 5 (f) in it, and the recognition accuracy of both models for W-PET is 100%; the clusters of Y-PS and Y-PMMA dispersed in the Yangtze River water are in the same region or even completely overlap and are completely independent of the clusters of the same type of nanoplastics dispersed in pure water, which corresponds to the confusion matrix diagrams shown in Figure 4 (i) and (j) in it showing that the classification accuracy of KNN and GDBT for nanoplastics in the Yangtze River water is 100%, as shown in Figure 11as shown in (a) of [reference]; while B-PS and B-PET dispersed in biological samples are in completely different regions from the same type of nanoplastics clustered in pure water. The reason for this phenomenon may be that the Raman characteristic peaks of proteins with high intensity after digestion of fish meat are doped in the fish meat samples. Among them, some of the more dispersed B-PET scatter points are located within the clustering regions of W-PS, W-PMMA, and W-PET dispersed in pure water, which is manifested as the KNN and GDBT models misidentifying a large number of PETs as PS and PMMA in the confusion matrix diagram, as Figure 4 shown in (m) and (n) of [reference].

[0112] The recognition accuracies of the two deep learning models, Transformer and CNN, for nanoplastics dispersed in pure water and Yangtze River water are 100%. The recognition accuracies for PS and PET dispersed in fish meat samples are 96% and 99% respectively. The confusion matrix diagram of the classification accuracy of Transformer is as Figure 5 shown in (g), (k), and (o) of [reference], and the confusion matrix diagram of the classification accuracy of CNN is as Figure 5 shown in (h), (l), and (p) of [reference]. This is attributed to the fact that the convolutional layer of CNN has good local feature extraction ability. By stacking convolutional layers, the receptive field is increased, and a better extraction effect on the characteristic peaks is achieved. Therefore, through the training with only a small amount of pure water data, it is possible to achieve an accuracy of nearly 100% for nanoplastics in Yangtze River water and fish meat samples.

[0113] Transformer has strong global feature extraction ability and can also accurately identify characteristic peaks. However, the attention mechanism of Transformer calculates the similarity for each peak in the data, resulting in the interference of miscellaneous peaks in fish meat samples on the classification performance of the model for nanoplastics. This also reflects that the two developed deep learning models, CNN and Transformer, can identify nanoplastics in different real samples with extremely high accuracy and have wide applicability after only being trained with a small amount of Raman data of micro-nano plastics dispersed in pure water.

[0114] Therefore, the present invention adopts the above method of AIOF-induced confinement localization and machine learning-assisted SERS for nanoplastics analysis. The AIOF nanoarray SERS substrate obtained by a unique preparation method has a multi-layer nano-wall array structure with interconnected pores, a larger surface area, and can provide more adsorption sites and hot spots, thus generating a stronger SERS signal.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and such modifications or equivalent replacements do not cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis, characterized in that: The following steps are involved: Step 1: Preparation of In2O3 / FTO; Step 2: preparing the AIOF substrate through a silver mirror reaction; Step 3, preparing a nanoplastic dispersion and collecting a SERS spectrum; Step 4: Characterization of AIOF substrate performance; Step 5: preprocess the spectral data; Step 6: Based on the spectral data preprocessed in step 5, classify the nanoplastics dispersed in different environmental samples through a machine learning model.

2. The method of using AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operation of step 1 is: put the cleaned FTO substrate with the conductive surface facing down into a polytetrafluoroethylene reactor filled with InCl3·4H2O and C2H5NS ethanol solution for reaction. A golden film grows on the surface of FTO, take it out and rinse it repeatedly with ethanol and distilled water, and then put it in an oven to dry; The product is then annealed to obtain In2O3 / FTO, which is then taken out and cut into small pieces for scoring before use.

3. The method of using AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 2, characterized in that: In step 1, the amount of InCl3·4H2O is 5-10mM, the amount of C2H5NS is 10-30mM, the amount of ethanol solvent is 40mL, the reaction time is 10-15h, and the reaction temperature is 150-200℃; The annealing parameters were set as follows: annealing at 400-500°C for 3-5h at a heating rate of 2-3°C / min.

4. The method of AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operation of step 2 is: put the In2O3 / FTO obtained in step 1 into a clean beaker, first add silver ammonia solution, then add glucose solution, take out the product after the reaction and rinse it, and finally obtain the Ag / In2O3 / FTO substrate with high-density silver nanoparticles uniformly loaded on the surface, that is, the AIOF substrate.

5. The method of using AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 4, characterized in that: In step 2, the concentration of the silver ammonia solution is 0.1-0.2M, the dosage is 1-2mL, the concentration of glucose is 0.1-0.5M, the dosage is 0.1-1.0mL, and the reaction is maintained for 10-15min.

6. The method of AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operation of step three is: suspending PET solid particles in a sodium dodecyl sulfate solution, subjecting the suspension to ultrasonic treatment to obtain a PET monodisperse mother liquor, and preparing PS monodisperse mother liquor and PMMA monodisperse mother liquor with different particle sizes; subsequently, adding different volumes of PET monodisperse mother liquor, PS monodisperse mother liquor and PMMA monodisperse mother liquor to different environmental samples to prepare nanoplastic dispersions with different concentrations; The above nanoplastic dispersion was dropped onto the AIOF substrate and dried, and then the SERS signal was collected.

7. The method of using AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 6, characterized in that: In step 3, the amount of PET solid particles used is 20-30 mg, the solution is 10-15 mL of pure water containing 2% sodium dodecyl sulfate, and the ultrasonic time is 30-45 min.

8. The method of AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operations of step four are as follows: the AIOF substrate is characterized by SEM images using Gemini 500SEM, the crystal structure is recorded using a rotating target X-ray diffractometer equipped with Cu Kα radiation, and data are collected in the range of 20° to 80°, TEM, energy dispersive X-ray spectroscopy and selected area electron diffraction are performed using JEM 2100F, the UV-visible absorption spectrum is measured in a TU-1950 double-beam UV-visible spectrophotometer, and the Raman mapping is determined by a micro-confocal Raman spectrometer equipped with a 532nm laser.

9. The method of AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operation of step 5 is: the initial range of SERS spectrum is 600-1800cm -1 , cut off to 700-1700cm -1 , BWSpec software was used to perform background subtraction on different nanoplastic spectra. The peak intensities showing variations in different nanoplastic spectra were normalized to a standard scale between 0 and 1 using Min-Max to eliminate the impact of intensity differences on the model. Principal component analysis and t-distributed random neighbor embedding methods were used to project high-dimensional data onto a two-dimensional plane. Data visualization of nanoplastic categories scattered in different environmental samples was performed in principal component analysis and t-distributed random neighbor embedding scatter plots.

10. The method of using AIOF-induced confined localization and machine learning-assisted SERS for nanoplastic analysis according to claim 1, characterized in that: The specific operation of step six is: use KNN, GDBT, CNN and Transformer algorithms to classify nanoplastics dispersed in different environmental samples, and reveal the relationship between the SERS characteristic peak intensity and the concentration of nanoplastics in the sample by linear fitting of the data. For each concentration, three different measurement points are taken to calculate the average and standard deviation of its intensity. The model fitting degree is measured by the determination coefficient R 2 To evaluate, analyze and predict the exponential data, and draw a fitting curve by taking the logarithm of the concentration and the corresponding characteristic peak intensity; The performance of the machine learning model in distinguishing the types of nanoplastics in different environmental samples is graphically described through the confusion matrix. The prediction performance is evaluated using a series of metrics, including accuracy, recall, precision, and F1 value: Where tp, fp, tn, and fn represent the number of true positive, false positive, true negative, and false negative predictions, respectively; For external validation, the established machine learning model was used to classify nanoplastics in different environmental samples to determine the classification accuracy.

Images