Soil heavy metal ion detection method

Abstract

The invention discloses a soil heavy metal ion detection method which comprises the following steps: selecting a plurality of heavy metal ions, and preparing a detection sample; collecting a Raman spectrum of the detection sample, and preprocessing the Raman spectrum to obtain processed spectrum data; carrying out characteristic analysis on the processed spectrum data, and identifying a characteristic peak position; taking the data including the characteristic peak position as a data set, and performing network model training on the neural network model by using the data set to obtain a trained network model; and inputting a to-be-detected substance into the trained network model, and outputting a heavy metal ion detection result. The method can effectively distinguish heavy metal forms, is rapid and sensitive, can realize on-site detection, and provides technical support for soil heavy metal monitoring.

Description

Technical Field

[0001] The present invention relates to the field of environmental pollution detection, and particularly to a method for detecting heavy metal ions in soil. Background Art

[0002] Heavy metals widely exist in the environment and are easily enriched into the human body through the food chain, posing a potential threat to health, such as causing chronic damage to the nervous, liver and kidney systems. Therefore, carrying out heavy metal detection plays an important role in pollution prevention and control and public health protection.

[0003] Currently, the detection of heavy metals mainly relies on traditional technologies such as atomic absorption spectrometry and inductively coupled plasma mass spectrometry. Although these methods have high sensitivity, they can only measure the total content and cannot distinguish the types, valence states and bioavailability of heavy metals, which may lead to deviation in risk assessment. In addition, these technologies also have limitations such as complex pretreatment, large equipment consumption, professional operation and dependence on laboratory environment, and it is difficult to meet the needs of on-site rapid detection.

[0004] Therefore, to solve the above problems, a method for detecting heavy metal ions in soil is needed, which can effectively distinguish the forms of heavy metals, be rapid, sensitive and capable of on-site detection, providing technical support for soil heavy metal monitoring. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to overcome the defects in the prior art and provide a method for detecting heavy metal ions in soil, which can effectively distinguish the forms of heavy metals, be rapid, sensitive and capable of on-site detection, providing technical support for soil heavy metal monitoring.

[0006] The method for detecting heavy metal ions in soil of the present invention includes:

[0007] Select several heavy metal ions and prepare a detection sample;

[0008] Collect the Raman spectrum of the detection sample, preprocess the Raman spectrum to obtain processed spectral data;

[0009] Perform feature analysis on the processed spectral data to identify the characteristic peak positions;

[0010] Use the data including the characteristic peak positions as a data set, and use the data set to train a neural network model to obtain a trained neural network model;

[0011] Input the substance to be measured into the trained neural network model and output the heavy metal ion detection result.

[0012] Further, preparing the detection sample specifically includes:

[0013] Select healthy mung bean seeds, rinse the seeds several times with ultrapure water to remove surface impurities;

[0014] Transfer the water-absorbing and swelling seeds to flower pots filled with soils containing different types and concentrations of heavy metal ions respectively;

[0015] Irrigate the flower pots with ultrapure water in a fixed amount every day, keep the soil temperature appropriate, and culture in an incubator for several days.

[0016] Furthermore, before collecting the Raman spectrum of the test sample, the test sample is processed, specifically including:

[0017] Take several seedling samples respectively, rinse them with ultrapure water for multiple times, then grind them into a paste with a mortar, weigh x g of the ground sample, and mix it with x1 mL of ultrapure water;

[0018] Ultrasonicate the mixture for t1 min and vortex for t2 s to achieve uniform dispersion; then, centrifuge x2 mL of the mixture at n1 rpm for t3 min, and take x3 mL of the supernatant for Raman spectrum collection.

[0019] Furthermore, collect the Raman spectrum of the test sample, specifically including:

[0020] Prepare silver nanoparticles using silver nitrate and sodium citrate, centrifuge x4 mL of the silver nanoparticles at n2 rpm for t4 min, and discard x5 μL of the supernatant to concentrate the silver nanoparticles;

[0021] Then, mix x6 μL of the concentrated silver nanoparticles with x7 μL of the processed seedling sample; use a perforated mold to deposit the obtained mixture onto a Teflon tape to form a liquid film substrate;

[0022] Collect the Raman spectrum at an excitation wavelength of λ nm, with an integration time of t5 seconds, an integration number of m1 times, and a spectral range of l1 - l2 cm -1 , and collect m2 spectra for each sample.

[0023] Furthermore, preprocess the Raman spectrum, specifically including: performing baseline correction, smoothing, normalization, and averaging on the Raman spectrum.

[0024] Furthermore, perform feature analysis on the processed spectral data, specifically including:

[0025] Project the original high-dimensional spectral data into the principal component space with the largest variance through orthogonal transformation;

[0026] Map the high-dimensional spectral data into a 2D or 3D space while preserving the neighborhood structure.

[0027] Furthermore, the neural network model uses an artificial neural network.

[0028] The beneficial effects of the present invention are as follows: A method for detecting heavy metal ions in soil disclosed by the present invention constructs a collaborative detection platform for plant stress effects and surface-enhanced Raman spectroscopy (SERS) technology, and realizes the intelligent identification and quantitative analysis of heavy metal ions in soil by combining machine learning algorithms. Using mung bean (Vigna radiata) seedlings as biological sensing units, by capturing the SERS fingerprint spectra of metabolites under the stress of Cd 2+ 、Cr 3+ 、Cu 2+ 、Ni 2+ ,a neural network-driven intelligent heavy metal detection model is constructed, solving the technical bottleneck that traditional methods can only detect the total content. The detection of a single sample only takes 5 minutes, and a portable instrument can be used to support on-site rapid detection, making the detection convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below in conjunction with the drawings and embodiments:

[0030] Figure 1 It is a schematic flow diagram of the heavy metal ion detection method of the present invention;

[0031] Figure 2 It is a schematic diagram of the plant stress response induced by heavy metals of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0032] The following further describes the present invention in conjunction with the drawings of the specification, as shown in the figure:

[0033] This embodiment discloses a method for detecting heavy metal ions in soil, including the following steps:

[0034] S1. Select several heavy metal ions and prepare test samples;

[0035] S2. Collect the Raman spectra of the test samples, preprocess the Raman spectra, and obtain the processed spectral data;

[0036] S3. Conduct feature analysis on the processed spectral data to identify the characteristic peak positions;

[0037] S4. Use the data including the characteristic peak positions as a data set to train the neural network model with the data set to obtain a trained network model;

[0038] S5. Input the substance to be tested into the trained network model and output the heavy metal ion detection result.

[0039] The present invention can not only effectively distinguish the stress characteristic spectra of different heavy metals, but also quantitatively evaluate the heavy metal concentration in soil through a regression prediction model, providing a new biosensing solution for soil environmental monitoring.

[0040] In this embodiment, in step S1, a number of heavy metal ions are selected to prepare a test sample, which specifically includes:

[0041] Select healthy mung bean seeds, rinse the seeds three times with ultrapure water to remove surface impurities;

[0042] Transfer the water-absorbed and swollen seeds to flower pots filled with soil containing different types and concentrations of heavy metal ions; among them, the heavy metal ions include Cd 2+ , Cr 3+ , Cu 2+ , Ni 2+ ;

[0043] Irrigate the flower pots with a fixed amount of ultrapure water every day, keep the soil temperature appropriate, for example, set the temperature to 25 °C, and cultivate in an incubator for ten days.

[0044] By selecting healthy mung bean seeds as biological indicator materials, the absorption process of heavy metal ions in the actual soil environment is simulated. Using mung bean seeds, which are sensitive to environmental changes and have a short growth cycle, can reflect the pollution situation of heavy metal ions in the soil in a short time, improving the detection efficiency. By controlling the types and concentrations of heavy metal ions and unifying the cultivation conditions, the repeatability of the experiment and the comparability of the data are ensured, which helps to establish a standardized detection system. Moreover, by simulating the absorption behavior of plants to heavy metals in the actual environment, a stable and reliable sample basis is provided for subsequent Raman spectroscopic analysis based on plant tissues, which is beneficial to improving the accuracy of subsequent model training and recognition.

[0045] In this embodiment, before collecting the Raman spectrum of the test sample, the test sample is processed on a clean workbench, which specifically includes:

[0046] Take several mung bean seedling samples respectively, rinse them three times with ultrapure water, then grind them into a paste with a mortar, weigh 1 g of the ground sample, and mix it with 9 mL of ultrapure water;

[0047] Ultrasonicate the mixture for 10 min and vortex for 10 s to achieve uniform dispersion; then, centrifuge 0.5 mL of the mixture at 5000 rpm for 5 min, and take 0.2 mL of the supernatant for Raman spectrum collection.

[0048] Systematically processing the mung bean seedling samples before Raman spectrum collection can improve the quality of spectral data and detection accuracy. By rinsing with ultrapure water multiple times, surface impurities of the samples are effectively removed, ensuring that the detection results are not interfered by exogenous pollution. After grinding the samples into a paste and mixing them with ultrapure water, ultrasonic and vortex operations are used to enhance the release and uniform distribution of the target substances in the samples, which is beneficial to obtaining stable spectral signals in subsequent analysis.

[0049] The precipitate is further removed by centrifugation, and the clarified supernatant is taken for Raman spectrum acquisition, which can effectively reduce stray light interference and background noise, thereby improving the clarity and reproducibility of the spectral signal. The overall processing flow is simple and efficient, providing a high-quality spectral data basis for model training, thus enhancing the reliability and accuracy of the heavy metal ion recognition model.

[0050] In this embodiment, in step S2, the Raman spectrum of the detection sample is collected, specifically including:

[0051] Silver nanoparticles are prepared using silver nitrate and sodium citrate. 1 mL of the silver nanoparticles is centrifuged at 10,000 rpm for 10 min, and 975 μL of the supernatant is discarded to concentrate the silver nanoparticles;

[0052] Then, 15 μL of the concentrated silver nanoparticles is mixed with 15 μL of the treated seedling sample; the obtained mixture is deposited on a Teflon tape using a perforated mold to form a liquid film substrate;

[0053] The Raman spectrum is collected at an excitation wavelength of 532 nm (output power 21 mW), the integration time is 1 second, the integration times is 1 time, and the spectral range is 450 - 1700 cm -1 , and 400 spectra are collected for each sample.

[0054] By introducing silver nanoparticles as a surface enhancement material (SERS substrate), the sensitivity and signal-to-noise ratio of the Raman spectrum can be improved. Silver nanoparticles are prepared using silver nitrate and sodium citrate, and after being concentrated by centrifugation and mixed into the treated sample, "hot spots" can be formed in the sample to enhance the Raman scattering signal of the target molecule, thereby realizing highly sensitive detection of biochemical changes caused by heavy metal pollution.

[0055] The mixture is uniformly deposited on the Teflon tape through a perforated mold to ensure the consistency of the thickness and distribution of the liquid film, further improving the repeatability and stability of spectral acquisition. A large amount of spectral data is collected in batches under fixed excitation wavelength and parameter conditions, providing sufficient and high-quality data support for subsequent feature extraction and model training.

[0056] In this embodiment, in step S2, the Raman spectrum is preprocessed, specifically including: baseline correction, smoothing, normalization, and averaging of the Raman spectrum.

[0057] Among them, the ALS algorithm or the sliding fitting background method can be used for baseline correction to remove the fluorescence background or baseline drift in the SERS spectrum and make the signal clearer; the Savitzky-Golay filter can be used for smoothing to remove high-frequency noise and enhance signal continuity; maximum-minimum normalization can be used for normalization to eliminate the influence of intensity differences and make the spectra of different samples comparable; by averaging multiple collected repeated spectra, noise can be reduced and representativeness can be enhanced.

[0058] By preprocessing the Raman spectral data, the accuracy and reliability of subsequent analysis can be effectively improved. Baseline correction can remove background interference in the spectrum and highlight the true Raman signal; smoothing is conducive to reducing instrument noise and random fluctuations, making the spectral curve smoother and clearer; normalization operation unifies the spectral intensity range under different samples or experimental conditions and enhances the comparability between different samples; and taking the average further reduces the influence brought by individual differences and improves the stability and representativeness of the overall data. Through these preprocessing steps, the characteristic information related to heavy metal ions is retained to the greatest extent, while irrelevant interferences are excluded, providing data support for subsequent feature extraction.

[0059] In this embodiment, in step S3, feature analysis is performed on the processed spectral data, which specifically includes:

[0060] Project the original high-dimensional spectral data into the principal component space with the largest variance through orthogonal transformation;

[0061] Map the high-dimensional spectral data into a 2D or 3D space while retaining the neighborhood structure.

[0062] By performing feature analysis on the processed Raman spectral data, the visualization ability and modeling efficiency of the data can be improved. Projecting the original high-dimensional spectral data into the principal component space with the largest variance through orthogonal transformation effectively reduces the data dimension while retaining most of the original information, which is conducive to highlighting the important features related to heavy metal ions and reducing the interference of redundant information on the model.

[0063] Through a non-linear dimensionality reduction method that maintains the local neighborhood structure of the data, map the high-dimensional spectral data into a two-dimensional or three-dimensional space, making the internal structure and distribution relationship of the data more intuitive and facilitating pattern recognition and sample classification. This kind of feature extraction and dimensionality reduction method not only benefits the efficient learning of features by the subsequent neural network model, but also improves the interpretability and accuracy of the overall model, providing strong technical support for the rapid identification and classification of heavy metal ions.

[0064] As Figure 2 shown, by performing feature analysis on the processed spectral data, characteristic peaks related to plant metabolism can be identified.

[0065] In this embodiment, a comparative study is carried out by constructing a variety of machine learning models, including decision trees, linear regression, support vector machines, Gaussian kernel SVM, k-nearest neighbor algorithms, and artificial neural networks. Through model performance evaluation, the artificial neural network (ANN) model performs optimally in heavy metal identification and concentration prediction.

[0066] Artificial neural networks have good non-linear modeling capabilities, can automatically learn and extract potential features from complex Raman spectral data, and effectively capture the subtle differences in spectra of different heavy metal ions. Compared with traditional statistical analysis methods, ANN is more robust and has better generalization ability when dealing with high-dimensional and multi-variable data, and can adapt to the variability between different samples. Moreover, neural networks have self-learning and optimization capabilities, and with the continuous enrichment of training data, their recognition accuracy and stability will continue to improve. By inputting the processed spectral data into the ANN for training and prediction, rapid and accurate classification and quantitative analysis of heavy metal ions in samples can be achieved.

[0067] In step S4, the data including the characteristic peak positions is used as the data set, 70% of the data set is used as the training set, and the remaining 30% is used as the test set to train the network model of the artificial neural network, obtaining a trained network model.

[0068] To further verify the universality of the heavy metal ion detection method of the present invention, a transfer learning strategy is adopted to optimize the model for experimental data of different batches. The experimental results show that this method can not only effectively distinguish the characteristic spectra of different types of heavy metal stresses, but also quantitatively evaluate the heavy metal concentration in the soil through a regression prediction model. In addition, the simultaneous ICP-MS analysis not only verifies the accumulation pattern of heavy metal ions in mung bean sprouts, but also reveals the toxicity differences of different heavy metals to plants.

[0069] The present invention realizes the rapid and accurate detection of heavy metal ions in soil through the synergistic effect of plant metabolic response and SERS technology, combined with machine learning algorithms, providing a new technical means for environmental monitoring. The present invention has strong application potential in soil heavy metal detection based on SERS technology combined with machine learning algorithms. This method has good environmental adaptability and can be extended to the monitoring of other environmental media and key pollutants.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not 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 the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for detecting heavy metal ions in soil, characterized in that: Including: Select several heavy metal ions to prepare a detection sample; Collect the Raman spectrum of the detection sample, preprocess the Raman spectrum to obtain the processed spectral data; Conduct feature analysis on the processed spectral data to identify the characteristic peak positions; Use the data including the characteristic peak positions as a data set to train a neural network model with the data set to obtain a trained neural network model; Input the substance to be measured into the trained neural network model and output the heavy metal ion detection result.

2. The soil heavy metal ion detection method according to claim 1, characterized in that: Prepare the detection sample, specifically including: Select healthy mung bean seeds, rinse the seeds several times with ultrapure water to remove surface impurities; Transfer the water-absorbed and swollen seeds to flower pots filled with soils containing different types and concentrations of heavy metal ions respectively; Quantitatively irrigate the flower pots with ultrapure water every day, keep the soil temperature appropriate, and cultivate in an incubator for several days.

3. The soil heavy metal ion detection method according to claim 2, wherein: Before collecting the Raman spectrum of the detection sample, process the detection sample, specifically including: Take several seedling samples respectively, rinse them multiple times with ultrapure water, then grind them into a paste with a mortar, weigh x g of the ground sample, and mix it with x1 mL of ultrapure water; Ultrasonicate the mixture for t1 min and vortex for t2 s to achieve uniform dispersion; then, centrifuge x2 mL of the mixture at n1 rpm for t3 min, and take x3 mL of the supernatant for Raman spectrum collection.

4. The soil heavy metal ion detection method according to claim 3, characterized in that: Collect the Raman spectrum of the detection sample, specifically including: Prepare silver nanoparticles using silver nitrate and sodium citrate, centrifuge x4 mL of the silver nanoparticles at n2 rpm for t4 min, and discard x5 μL of the supernatant to concentrate the silver nanoparticles; Then, mix x6 μL of the concentrated silver nanoparticles with x7 μL of the processed seedling sample; use a perforated mold to deposit the obtained mixture on a Teflon tape to form a liquid film substrate; Collect Raman spectra at an excitation wavelength of λ nm, with an integration time of t5 seconds, an integration number of m1 times, and a spectral range of l1 to l2 cm -1 , and collect m2 spectra for each sample.

5. The soil heavy metal ion detection method according to claim 1, wherein: Preprocess the Raman spectrum, specifically including: perform baseline correction, smoothing, normalization, and averaging on the Raman spectrum.

6. The soil heavy metal ion detection method according to claim 1, wherein: Conduct feature analysis on the processed spectral data, specifically including: Project the original high-dimensional spectral data into the principal component space with the largest variance through orthogonal transformation; Map the high-dimensional spectral data into a 2D or 3D space while retaining the neighborhood structure.

7. The soil heavy metal ion detection method according to claim 1, characterized in that: The neural network model uses an artificial neural network.

Images