Method and system for detecting organophosphorus pesticide residues and storage medium

Abstract

The invention provides a method and system for detecting organophosphorus pesticide residues and a storage medium, and relates to the technical field of organophosphorus pesticide residue detection.The method comprises the steps that a micro-fluidic chip is used for removing impurities from surface flushing liquid and homogeneous sample liquid, pesticide molecules are enriched through magnetic microspheres, the target molecule concentration is increased, and the detection sensitivity is high; the method comprises the following steps: firstly, preparing a surface-enhanced Raman scattering substrate, dropwise adding the enriched liquid onto the surface-enhanced Raman scattering substrate, enhancing Raman signals through a silver nanoparticle array, acquiring spectral data by using a portable Raman spectrometer, removing fluorescence background interference through polynomial fitting, extracting the characteristic peak intensity of target molecules, and calculating a comprehensive characteristic index by combining spectral characteristic parameters; finally, an SVM (support vector machine) model is trained based on historical sample data, the concentration of the sample to be detected is accurately predicted, and whether the concentration meets the food safety standard or not is judged.

Description

Technical Field

[0001] The present invention relates to the technical field of detecting organophosphorus pesticide residues, and specifically to a method, a system and a storage medium for detecting organophosphorus pesticide residues. Background Art

[0002] Organophosphorus pesticides are chemical pesticides widely used in agricultural production. They are characterized by high efficiency, broad spectrum and economy, so they are widely used to control pests and diseases of agricultural products. However, due to improper use of organophosphorus pesticides or insufficient residue control, the pesticide residues in agricultural products may exceed the standard, which may cause harm to human health, such as poisoning, nervous system damage and other problems. Therefore, rapid and accurate detection of organophosphorus pesticide residues in agricultural products has important practical significance, which is not only the basic requirement for ensuring food safety, but also a necessary means to promote the sustainable development of agriculture.

[0003] Currently, the detection methods for organophosphorus pesticide residues mainly include two categories: laboratory analysis and on-site rapid detection. In laboratory analysis, methods such as gas chromatography (GC), liquid chromatography (LC), gas chromatography-mass spectrometry (GC-MS) and liquid chromatography-mass spectrometry (LC-MS) are widely used in the detection of organophosphorus pesticide residues due to their high sensitivity and high accuracy. However, these methods usually require expensive instrument equipment, professional technical personnel and complex sample pretreatment processes, which are difficult to meet the needs of on-site rapid detection. In addition, the laboratory detection cycle is long, and it is difficult to achieve real-time monitoring and dynamic analysis. In terms of on-site rapid detection, enzyme inhibition method and immunoassay are relatively common technologies. These methods have the advantages of portability and rapidity, but there are certain deficiencies in terms of sensitivity, selectivity and anti-interference ability. Especially when multiple organophosphorus pesticides coexist, accurate detection cannot be achieved.

[0004] The deficiencies of the prior art are mainly reflected in the following aspects: First, although the laboratory detection means are accurate, the detection process is complex and time-consuming, and it is not suitable for on-site detection scenarios that require rapid response; Second, the existing rapid detection methods are limited by sensitivity and anti-interference ability, and it is difficult to meet the precise identification and quantification requirements for low-concentration organophosphorus pesticide residues; In addition, traditional methods may be interfered by background signals when dealing with complex matrix samples, such as surface residues of fruits and vegetables, resulting in inaccurate detection results.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The object of the present invention is to provide a method, a system and a storage medium for detecting organophosphorus pesticide residues, so as to solve the problems raised in the above-mentioned background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for detecting organophosphorus pesticide residues, the specific steps include:

[0009] Step 1: Pretreat the surface rinse liquid and the homogenized sample liquid of the agricultural product to be tested respectively, mix the two and add them to the sample inlet of the microfluidic chip. Remove impurities through the microfluidic chip, use magnetic microspheres to capture and enrich pesticide molecules, and then output a high-purity enriched liquid;

[0010] Step 2: Drop the high-purity enriched liquid collected in the outlet channel of the microfluidic chip onto the substrate of surface-enhanced Raman scattering. The substrate adopts a design with a periodic silver nanoparticle array. Use a portable Raman spectrometer to excite the surface-enhanced Raman scattering substrate, and calculate the spectral characteristic parameters;

[0011] Step 3: Perform polynomial fitting on the collected Raman spectrum data to remove the fluorescence background interference, extract the characteristic peak intensity of the target organophosphorus molecule after removing fluorescence, combine the characteristic peak intensity with the spectral characteristic parameters, calculate the comprehensive characteristic index of the target organophosphorus molecule, collect the comprehensive characteristic index of historical agricultural products as the training set, and the corresponding historical agricultural product concentration as the label to train the SVM machine learning model;

[0012] Step 4: Use the trained SVM model to predict the agricultural product sample to be tested, obtain the concentration of the target organophosphorus molecule in the agricultural product sample to be tested, and judge whether the concentration of the target organophosphorus molecule meets the national food safety standard.

[0013] Further, after the surface rinse liquid and the homogenized sample liquid of the agricultural product to be tested are pretreated respectively, the logic of mixing the two and adding them to the sample inlet of the microfluidic chip is as follows:

[0014] For samples with target pesticide residues on the surface of agricultural products, place the sample to be tested in a clean container, add 100 mL of deionized water, and use an ultrasonic instrument with a frequency of 40 kHz and a power of 200 W to oscillate for 3 minutes, so that the target pesticide residue molecules on the surface are eluted into the solution from the sample surface. Filter the rinse liquid through a filter membrane with a pore size of 0.45 μm to remove large-particle impurities, and collect the surface sample liquid for subsequent enrichment treatment and detection;

[0015] For samples with target pesticide residues inside agricultural products, the agricultural products were chopped into particles less than or equal to 2 cm in size, 10 g of the sample was weighed and placed in a homogenizer, an equal volume of 10 ml of acetonitrile extraction solvent was added, the homogenizer was started, and high-speed stirring was performed for 3 minutes to accelerate the molecular release to obtain uniform homogenization, the homogenate was transferred to a centrifuge tube, centrifuged at 8000 rpm for 10 minutes, the supernatant was taken and filtered again through a filter membrane with a pore size of 10 μm to remove particulate impurities, and the internal sample liquid was obtained for subsequent enrichment treatment;

[0016] The surface sample liquid and the internal sample liquid were mixed at a volume ratio of 1:1, 1 mL of acetonitrile was added as a mixed solvent before mixing, and a vortex mixer was used to fully oscillate to form a uniform liquid phase. The mixed liquid was then filtered through a filter membrane with a pore size of 10 μm to remove particulate impurities and obtain a high-purity mixed sample liquid;

[0017] The microfluidic chip contains a micron-scale liquid flow channel, a functional area and an outlet channel. The liquid channel injects the mixed sample liquid into the microfluidic chip at a flow rate of 10 μL / min through a syringe pump and enters the functional area; the functional area pre-stores acetonitrile extraction solvent and contains a filter membrane structure with a pore size of 10-100 μm for removing large particle impurities;

[0018] In the functional area, the high-speed rotating shear force of the spiral micro-mixing channel, i.e., the Reynolds number is 1000, is used to achieve efficient transfer from the aqueous phase to the organic phase, i.e., the target liquid entering the mixing area is selectively transferred to the acetonitrile phase according to the hydrophilicity or hydrophobicity of the target organophosphorus molecules;

[0019] The liquid enters the magnetic microsphere area, which includes magnetic microspheres with a particle size of 1-5μm. The surface of the magnetic microspheres is modified with MIP, which can selectively capture the target organophosphorus pesticide molecules through molecular imprinting. The captured magnetic microspheres control the flow trajectory through an external magnetic field and enrich them in the outlet area, and output high-purity enriched liquid in the outlet channel for subsequent experiments.

[0020] Furthermore, the logic of dropping the target liquid onto the periodic silver nanoparticle array substrate and exciting it using a portable Raman spectrometer is as follows:

[0021] 10 μL of high-purity enriched liquid is taken from the outlet channel of the microfluidic chip, and evenly dripped onto the periodic silver nanoparticle array substrate, wherein the periodic silver nanoparticle array is composed of silver particles with a particle size of 40 nm, a particle spacing of 10-15 nm, and an array period of 100 nm;

[0022] After dropping the sample, let it stand for 10 minutes. Through electrostatic adsorption and short-range chemical interactions, the target organophosphorus molecules are fixed in the hot spots between silver particles. After fixing the sample, place the substrate in a constant temperature condition of 37°C and dry it for 5 minutes to ensure that the liquid completely evaporates, leaving only the target organophosphorus molecules adsorbed on the surface of silver particles;

[0023] Place the dried substrate under a portable Raman spectrometer with a wavelength of 785 nm and a power of 10 mW for detection. The diameter of the laser spot is 10 μm, which is focused on the surface of the silver nanoparticle substrate. The Raman spectrum detection range is 400 - 2000 cm-1, covering the Raman characteristic peaks of organophosphorus pesticides;

[0024] The logic for calculating the spectral characteristic parameters is as follows: Use the LSPR effect of the silver nanoparticle array to introduce a signal enhancement factor, and the formula is:

[0025]

[0026] where, I raw is the original Raman spectrum intensity of the silver substrate, I ref is the Raman spectrum intensity of the substrate without enhancement, and η LSPR is the plasmon resonance enhancement factor;

[0027] For the SERS detection of target organophosphorus molecules, in addition to using the silver nanoparticle substrate, a gold nanoparticle array substrate is superimposed in the experiment. The target organophosphorus molecules are fixed on the silver nanoparticle array substrate and the gold nanoparticle array substrate respectively, and the Raman signals of the two substrates are calculated for their ratio. The formula is:

[0028]

[0029] where, I silver is the signal intensity of the silver substrate, I gold is the signal intensity of the gold substrate, and R is the signal ratio parameter of the double substrate;

[0030] If R > R threshold , then the target organophosphorus molecules exist and the signal is significantly enhanced. If R ≤ R threshold , then the interference of the target organophosphorus molecule signal is excluded, where R threshold is the threshold value under the double substrate combination;

[0031] Introduce an external electric field on the Raman substrate. By dynamically adjusting the enhancement effect of the local surface plasmon, further enhance the Raman signal of the target organophosphorus molecules. The formula for defining the dynamic electric field enhancement factor is:

[0032]

[0033] where, Iraw is the original Raman signal intensity, E est is the externally applied electric field intensity, k is the experimentally determined electric field enhancement coefficient, D enh is the dynamic electric field enhancement factor;

[0034] For each sample, record the above spectral characteristic parameters η LSPR (v), R, D enh and the original spectral signal I raw (ν).

[0035] Furthermore, the logic for performing polynomial fitting on the collected Raman spectra to remove fluorescence interference and extracting the Raman characteristic peak intensity of the target organophosphorus molecule is as follows:

[0036] Perform background fitting in the spectral region far from the Raman characteristic peak of the target organophosphorus molecule. The fluorescence background is usually a broadband continuous signal, and its change frequency is lower than that of the Raman characteristic signal, which can be regarded as a low-frequency variation function of the spectral shift ν. Use the polynomial P(v) to fit the background signal, and the formula is:

[0037] P(ν) = a 0 + a 1 v + a 2 v 2 + … + a n ν n

[0038] where a 0 , a 1 , …, a n represent the polynomial coefficients, n represents the polynomial order, and n ∈ [3, 5];

[0039] Subtract the background fitting P(v) from the original spectrum to obtain the fluorescence-removed signal, and the formula is:

[0040] I c (v) = I raw (v) - P(v)

[0041] where I c (v) is the Raman signal after removing fluorescence, and I raw (v) represents the original spectral signal intensity, including the Raman signal of the target molecule and background fluorescence interference;

[0042] Perform smoothing on I c (ν) using Savitzky-Golay filtering to reduce noise interference. The Savitzky-Golay filtering window size and polynomial order should be calculated and selected according to the noise level and Raman signal characteristics, and the window size is 5 or 7;

[0043] According to the characteristic peak position ν in the databased , find the corresponding characteristic peak position v in the spectrum I after fluorescence removal c (ν), and the formula for calculating the peak position difference based on the peak position offset criterion is: Δv = ν m - v m where v d is the measured peak position, and ν m is the peak position in the database. If the peak position offset range |Δv| ≤ 5 cm d , then it is determined that the characteristic peaks match; -1

[0044] Find the maximum intensity of the characteristic peak in the spectrum after fluorescence removal, and the formula is:

[0045]

[0046] where I peak represents the peak intensity of the characteristic peak of the target organophosphorus molecule, [ν 1 , ν 2 represents the peak position range set by the database;

[0047] Set a comprehensive characteristic index EI to characterize the target organophosphorus molecule, integrating the spectral signal intensity, enhancement effect, reliability, and dynamic characteristics of the organophosphorus. The formula is:

[0048]

[0049] where I peak is the characteristic peak intensity representing the target organophosphorus molecule, η LSPR is the surface plasmon resonance enhancement factor, R is the dual-substrate signal ratio parameter, D enh is the dynamic electric field enhancement factor, k 1 , k 2 is a constant parameter, and α is a parameter controlling the amplitude of the index change.

[0050] Furthermore, the logic of collecting the comprehensive characteristic index of historical samples as the training set and the corresponding known target organophosphorus concentration as the label to train the SVM machine learning model is as follows:

[0051] Obtain experimental data samples of multiple groups of historical organophosphorus solutions with different concentrations. For each sample, calculate the historical comprehensive characteristic index EI h , corresponding to the historical comprehensive characteristic index of each sample, record its true historical organophosphorus concentration as the label; normalize the historical comprehensive characteristic index EI h ;

[0052] Using the normalized comprehensive characteristic index EI of the historical samples hAs the input features of the training set and the corresponding known historical organophosphorus concentration c h As the label of the training set, use the SVR (Regression SVM) model to establish the mapping relationship between the historical comprehensive feature index and the concentration, and the comprehensive feature index EI h Contains non-linear features, calculated using the RBF kernel function formula, and the formula is:

[0053] K(x i ,x j )=exp(-γ·‖x i -x j ‖ 2 )

[0054] Among them, x i ,x j represents the input feature value of the training sample, that is, the normalized comprehensive feature index EI h ,‖x i -x j ‖ 2 represents the Euclidean distance, which measures the similarity between samples. γ is the kernel function width parameter, which is used to control the complexity of mapping to the high-dimensional space, and γ>0;

[0055] Use the objective function of the support vector machine to minimize the following equation:

[0056]

[0057] Among them, w is the weight vector of the regression hyperplane, C is the regularization parameter, which is used to balance the model fitting error and complexity, and ξ i , is the slack variable, which is used to tolerate the prediction error within a certain range;

[0058] To test the accuracy of the model prediction and adjust the model parameters, use the gradient descent method to solve the objective function, so as to determine the mapping function f(x):

[0059]

[0060] Among them, x is the comprehensive feature index of the sample to be predicted, K(x i ,x) is the kernel function value, δ i , is the weight of the support vector, and b is the bias term; use the mean square error or the coefficient of determination to evaluate the prediction performance of the SVR model.

[0061] Furthermore, use the trained SVM model to predict the sample of the agricultural product to be detected, and the logic for obtaining the concentration of the target organophosphorus molecule in the sample of the agricultural product to be detected is:

[0062] The comprehensive characteristic index of the target organophosphorus molecules extracted from the agricultural product samples to be detected is input into the trained support vector regression model f(x). The formula for calculating the predicted concentration is as follows:

[0063] c = f(x)

[0064] where c is the concentration of the agricultural product samples to be detected obtained by prediction, and f(x) is a mapping function determined in the training process, which is used to map the comprehensive characteristic index EI to the concentration;

[0065] According to the pesticide residue limits stipulated in the national food safety standards, the residue limit c' of the target organophosphorus molecules is obtained. The predicted concentration value c is compared with the limit c' in the pesticide residue limits stipulated in the national food safety standards to determine whether the agricultural product samples to be detected are qualified:

[0066]

[0067] where if c ≤ c', it is determined to be qualified; if c > c', it is determined to be unqualified. The predicted concentration value c and the determination result are output, indicating whether the agricultural product samples to be detected meet the national food safety standards.

[0068] The present invention further provides a system for detecting organophosphorus pesticide residues. The system is used to execute the above method for detecting organophosphorus pesticide residues, and includes:

[0069] A sample extraction module, which is used to preprocess the surface rinse liquid and the homogenized sample liquid of the agricultural product to be detected respectively, mix the two and add them to the sample inlet of the microfluidic chip, remove impurities through the microfluidic chip, and output a high-purity enriched liquid after capturing and enriching the pesticide molecules using magnetic microspheres;

[0070] A characteristic parameter module, which is used to drop the high-purity enriched liquid collected from the outlet channel of the microfluidic chip onto the substrate for surface-enhanced Raman scattering. The substrate adopts a design with a periodic silver nanoparticle array, use a portable Raman spectrometer to excite the surface-enhanced Raman scattering substrate, and calculate the spectral characteristic parameters;

[0071] A model establishment module, which is used to perform polynomial fitting on the collected Raman spectrum data to remove the fluorescence background interference, extract the characteristic peak intensity of the target organophosphorus molecules after removing fluorescence, calculate the comprehensive characteristic index of the target organophosphorus molecules based on the combination of the characteristic peak intensity and the spectral characteristic parameters, collect the comprehensive characteristic index of historical agricultural products as the training set, and the corresponding historical agricultural product concentration as the label to train the SVM machine learning model;

[0072] The result judgment module is used to predict the agricultural product sample to be detected by using the trained SVM model, obtain the concentration of the target organophosphorus molecule in the agricultural product sample to be detected, and judge whether the concentration of the target organophosphorus molecule meets the national food safety standard.

[0073] The present invention further provides a storage medium storing a computer program, which when executed by a processor, implements each step in the above method for detecting organophosphorus pesticide residues.

[0074] Compared with the prior art, the beneficial effects of the present invention are:

[0075] In the sample processing stage of the present invention, the microfluidic chip technology is introduced. Through the micron-level liquid flow channels and functional regions, the functions of automatic impurity removal of the surface rinse liquid and homogenized sample liquid of agricultural products and enrichment of pesticide molecules by magnetic microspheres are realized. The precise design of the microfluidic chip enables efficient removal of impurities such as particulate matter and other interfering substances in complex matrices during the pretreatment process, and the target organophosphorus molecules are efficiently captured and enriched by magnetic microspheres, greatly increasing the concentration of target molecules in the sample to be detected.

[0076] By adopting the SERS (Surface Enhanced Raman Scattering) technology and combining with the design of an enhanced substrate with a periodic silver nanoparticle array, the target molecule is excited by a portable Raman spectrometer, and high-sensitivity Raman spectral data is obtained. Compared with traditional spectral technologies, the present invention utilizes the local surface plasmon resonance effect of the silver nanoparticle array to significantly enhance the Raman signal intensity of the target molecule, so that the characteristic signal of the target molecule can be detected at low concentration conditions, such as ppb level. In addition, the enhanced substrate design in the invention has excellent uniformity and stability, effectively reducing the random volatility of the Raman signal, and solving the problems of unstable signal and limited sensitivity in existing rapid detection means for low-concentration samples.

[0077] The concentration prediction method of the present invention based on the SVM machine learning model converts the complex non-linear relationship between spectral characteristic parameters and target concentration into accurate concentration determination results, avoiding the problems of low accuracy and poor adaptability of traditional fitting methods, meeting the requirements of on-site rapid detection. Through the organic combination of the microfluidic chip, SERS enhanced substrate, polynomial fitting de-fluorescence algorithm, and SVM concentration prediction model, the present invention innovatively solves the problems of complex sample processing, low sensitivity, poor anti-interference ability, and inaccurate rapid detection in the prior art, realizing the rapid and accurate detection of organophosphorus pesticide residues in agricultural products, and having significant technical advantages and application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0079] Figure 2 This is a schematic diagram of the overall system module of the present invention. Detailed implementation manners

[0080] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0081] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this word cover the elements or objects listed after this word and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left" and "right" are only used to represent relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0082] Embodiment:

[0083] Please refer to Figure 1 , the present invention provides a technical solution:

[0084] A method for detecting organophosphorus pesticide residues, the specific steps include:

[0085] Step 1: Pretreat the surface rinse liquid and the homogenized sample liquid of the agricultural product to be tested respectively, mix the two and add them to the sample inlet of the microfluidic chip, remove impurities through the microfluidic chip, and use magnetic microspheres to capture and enrich pesticide molecules and then output a high-purity enriched liquid;

[0086] The logic of pretreating the surface rinse liquid and the homogenized sample liquid of the agricultural product to be tested respectively and then mixing the two and adding them to the sample inlet of the microfluidic chip is as follows:

[0087] The pesticide residues on the surface of the agricultural product are mainly attached to the surface layer in a physically adsorbed state. These residues may be trace components that have not been washed off after pesticide spraying, and direct contact with the human body may pose a health threat. Therefore, extracting the surface residues is the key first step in detection;

[0088] For samples with target pesticide residues on the surface of agricultural products, place the sample to be tested in a clean container, add 100 mL of deionized water, and use a 40 kHz, 200 W ultrasonic device to oscillate for 3 minutes to elute the surface target pesticide residue molecules from the sample surface into the solution. Filter the rinse liquid through a filter membrane with a pore size of 0.45 μm to remove large particle impurities, and collect the surface sample liquid for subsequent enrichment processing and detection;

[0089] The purpose of eluting the surface sample liquid is to extract the target pesticide residues on the surface of agricultural products to ensure the accuracy and representativeness of subsequent detection. The ultrasonic oscillation and filtration steps optimize the efficiency of sample pretreatment and reduce the interference of impurities. Ultrasonic oscillation applies shear force to the sample surface through sound wave energy, which can efficiently elute the attached pesticide molecules. The filtration step removes particulate impurities in the surface sample liquid to ensure the purity of the extract, thereby improving the accuracy of subsequent enrichment and detection;

[0090] Pesticide residue molecules in agricultural products may be deeply buried in the tissue structure and are not easy to be directly eluted. Physical and chemical methods such as homogenization and centrifugation can effectively extract these deep residues, laying the foundation for comprehensive detection:

[0091] For samples with target pesticide residues inside agricultural products, the agricultural products were chopped into particles less than or equal to 2 cm in size, 10 g of the sample was weighed and placed in a homogenizer, an equal volume of 10 ml of acetonitrile extraction solvent was added, the homogenizer was started, and high-speed stirring was performed for 3 minutes to accelerate the molecular release to obtain uniform homogenization, the homogenate was transferred to a centrifuge tube, centrifuged at 8000 rpm for 10 minutes, the supernatant was taken and filtered again through a filter membrane with a pore size of 10 μm to remove particulate impurities, and the internal sample liquid was obtained for subsequent enrichment treatment;

[0092] This process mainly targets pesticide residues inside agricultural product tissues. Through physical and chemical methods such as homogenization and centrifugation, the homogenization and centrifugation processes can effectively destroy the sample tissue structure and release deep target molecules. Acetonitrile is used as an extraction solvent, which has good polarity and organic solubility, and can efficiently extract target pesticide molecules.

[0093] The surface sample liquid and the internal sample liquid were mixed at a volume ratio of 1:1, 1 mL of acetonitrile was added as a mixed solvent before mixing, and a vortex mixer was used to fully oscillate to form a uniform liquid phase. The mixed liquid was then filtered through a filter membrane with a pore size of 10 μm to remove particulate impurities and obtain a high-purity mixed sample liquid;

[0094] The purpose of this step is to prepare a mixed sample liquid with high uniformity and purity. The vortex mixer achieves complete homogenization of the liquid phase through rotational shear force. Acetonitrile is used as an auxiliary solvent to further improve the solubility of the target molecule in the mixed liquid, ensuring the reliability of subsequent microfluidic chip operations and the accuracy of test results.

[0095] A microfluidic chip is a fluid operation platform at the micron scale, made of PDMS (polydimethylsiloxane), which can achieve efficient separation and enrichment of samples. It contains micron-scale liquid flow channels, functional regions, and outlet channels inside; the liquid channels are used to inject the mixed sample liquid into the microfluidic chip at a flow rate of 10 μL / min through a syringe pump and enter the functional region; the functional region pre-stores an acetonitrile extraction solvent and contains a filter membrane structure with a pore size of 10 - 100 μm to remove large particle impurities;

[0096] The application of the microfluidic chip can significantly improve the separation and enrichment efficiency of target molecules, while reducing reagent consumption and sample processing time. The magnetic microspheres are modified by MIP, which can achieve highly selective capture of target molecules, ensuring the high purity of the output liquid, and providing a rapid and sensitive solution for modern pesticide residue detection;

[0097] In the functional region, through the high-speed rotational shear force of the spiral micro-mixing channel, that is, the Reynolds number is 1000, efficient transfer from the aqueous phase to the organic phase is achieved, that is, the target liquid entering the mixing region is selectively transferred to the acetonitrile phase according to the hydrophilicity or hydrophobicity of the target organophosphorus molecule;

[0098] The liquid enters the magnetic microsphere region, which includes magnetic microspheres with a particle size of 1 - 5 μm, and its surface is modified with MIP (molecularly imprinted polymer). The target organophosphorus pesticide molecules are highly selectively captured through molecular imprinting. After capture, the magnetic microspheres control the flow trajectory through an external magnetic field and are enriched to the outlet region, and high-purity enriched liquid is output through the outlet channel for subsequent experiments.

[0099] Step 2: Drop the high-purity enriched liquid collected from the outlet channel of the microfluidic chip onto the substrate of surface-enhanced Raman scattering. The substrate adopts a design with a periodic silver nanoparticle array. Use a portable Raman spectrometer to excite the surface-enhanced Raman scattering substrate and calculate the spectral characteristic parameters;

[0100] The logic of dropping the target liquid onto the periodic silver nanoparticle array substrate and using a portable Raman spectrometer to excite it is as follows:

[0101] The silver nanoparticle array is a key substrate in the SERS, Surface-Enhanced Raman Scattering experiment. Through its strong LSPR, Localized Surface Plasmon Resonance effect, the Raman signal of target molecules can be significantly enhanced;

[0102] Take 10 μL of high-purity enriched liquid from the outlet channel of the microfluidic chip and evenly drop it onto the periodic silver nanoparticle array substrate. The periodic silver nanoparticle array is composed of silver particles with a particle size of 40 nm, a particle spacing of 10 - 15 nm, and an array period of 100 nm. The periodic silver nanoparticle array provides a super-strong electromagnetic field environment for target molecules through the LSPR effect, thereby improving the sensitivity of Raman signals.

[0103] After dropping the sample, let it stand for 10 minutes. Through electrostatic adsorption and short-range chemical interactions, fix the target organophosphorus molecules in the hot spot regions between the silver particles. After fixing the sample, place the substrate in a constant temperature condition of 37 °C and dry it for 5 minutes to ensure that the liquid completely evaporates, leaving only the target organophosphorus molecules adsorbed on the surface of the silver particles.

[0104] The purpose of standing for 10 minutes is to fix the target molecules on the surface of the silver particles, especially in the hot spot regions, through electrostatic adsorption and the chemical interaction between the molecules and the silver particles, such as the formation of coordination bonds. The hot spot region is the position with the smallest particle spacing and the strongest LSPR effect. When the target molecules are adsorbed here, the maximum signal enhancement can be obtained.

[0105] Through drying at a constant temperature of 37 °C, remove the excess solution in the sample to ensure that only the solid-state system of the interaction between the target molecules and the silver particles remains, thereby improving the accuracy of signal detection.

[0106] Place the dried substrate under a portable Raman spectrometer with a wavelength of 785 nm and a power of 10 mW for detection. The laser spot diameter is 10 μm, focused on the surface of the silver nanoparticle substrate. The Raman spectrum detection range is 400 - 2000 cm-1, covering the Raman characteristic peaks of organophosphorus pesticides. This range covers the main Raman characteristic peaks of organophosphorus pesticide molecules, such as C-H stretching vibration, P=O bond vibration, etc., which is convenient for characteristic identification.

[0107] The excitation light source of the Raman spectrum is selected as the near-infrared wavelength of 785 nm to avoid possible fluorescence interference in agricultural product samples and at the same time match the LSPR effect of the silver nanoparticles to ensure signal enhancement. The selection of the spot diameter matches the surface area of the silver nanoparticle array to ensure that the laser can fully excite the substrate surface and avoid uneven signals.

[0108] The logic for calculating the spectral characteristic parameters is as follows: Use the LSPR (Localized Surface Plasmon Resonance) effect of the silver nanoparticle array to introduce a signal enhancement factor, and the formula is:

[0109]

[0110] where I raw is the original Raman spectrum intensity of the silver substrate, I ref is the Raman spectrum intensity of the substrate without enhancement, ηLSPR is the surface plasmon resonance enhancement factor;

[0111] I raw is the original Raman spectral intensity of the silver substrate, representing the signal intensity of the target molecule on the enhanced substrate (silver nanoparticle array), I ref is the Raman spectral intensity without an enhanced substrate, such as on a common glass slide, representing the signal intensity when there is no LSPR effect, η LSPR is the local surface plasmon resonance enhancement factor, reflecting the enhancement ability of the silver nanoparticle array substrate for the Raman signal of the target molecule; η LSPR The larger it is, the stronger the enhancement ability of the silver substrate and the higher the signal sensitivity;

[0112] For the SERS (surface-enhanced Raman scattering) detection of target organophosphorus molecules, in addition to using a silver nanoparticle substrate, a gold nanoparticle array substrate is superimposed in the experiment. The target organophosphorus molecules are fixed on the silver nanoparticle array substrate and the gold nanoparticle array substrate respectively, and the Raman signals of the two substrates are calculated for their ratio. The formula is as follows:

[0113]

[0114] where, I silver is the signal intensity of the silver substrate, I gold is the signal intensity of the gold substrate, and R is the dual-substrate signal ratio parameter, reflecting the signal difference of the target molecule on the silver nanoparticle array substrate and the gold nanoparticle array substrate;

[0115] If R > R threshold , it indicates the presence of the target organophosphorus molecule, and its signal is significantly enhanced on the silver substrate. If R ≤ R threshold , it indicates that the signal of the target molecule is weak, there may be interference factors or the target molecule does not exist, where R threshold is the threshold under the dual-substrate combination;

[0116] An external electric field is introduced on the Raman substrate to further enhance the Raman signal of the target organophosphorus molecule by dynamically adjusting the enhancement effect of the local surface plasmon. The formula for defining the dynamic electric field enhancement factor is as follows:

[0117]

[0118] where, I raw is the original Raman signal intensity, E ext is the external electric field intensity, k is the experimentally determined electric field enhancement coefficient, D enh is the dynamic electric field enhancement factor, reflecting the enhancement effect of the external electric field on the original signal; D enh The larger it is, the more significant the contribution of the external electric field to Raman enhancement, E extThe larger, D enh The increase shows a square relationship, indicating that the electric field strength has a non-linear effect on signal enhancement;

[0119] For each sample, record the above spectral characteristic parameters: the enhancement ability η reflecting the silver substrate LSPR , R for judging the significance of the target molecule signal and excluding interference, D for quantifying the enhancement effect of the applied electric field enh and I providing the actual Raman signal intensity of the target molecule raw , laying a data foundation for subsequent analysis.

[0120] Step 3: Perform polynomial fitting on the collected Raman spectral data to remove fluorescence background interference, extract the characteristic peak intensity of the target organophosphorus molecule after removing fluorescence, and calculate the comprehensive characteristic index of the target organophosphorus molecule based on the combination of the characteristic peak intensity and spectral characteristic parameters. The comprehensive characteristic index of historical agricultural products is used as the training set, and the corresponding known concentration is used as the label to train the SVM machine learning model;

[0121] The logic of performing polynomial fitting on the collected Raman spectrum to remove fluorescence interference and extracting the Raman characteristic peak intensity of the target organophosphorus molecule is as follows:

[0122] Perform background fitting in the spectral region far from the Raman characteristic peak of the target organophosphorus molecule. The fluorescence background is usually a broadband continuous signal, and its change frequency is lower than that of the Raman characteristic signal, which can be regarded as a low-frequency variation function of the spectral shift v. Use the polynomial P(v) to fit the background signal, and the formula is:

[0123] P(v) = a 0 + a 1 v + a 2 v 2 + … + a n v n

[0124] where a 0 , a 1 , …, a n represent the polynomial coefficients, and the specific values are obtained by fitting the experimental spectrum. n represents the polynomial order, n ∈ [3, 5]; P(v) is the background signal fitted by the polynomial, reflecting the low-frequency variation of the fluorescence background;

[0125] The larger n is, the stronger the fitting flexibility, but it may lead to overfitting and reduce the accuracy of background removal. By reasonably selecting n, the fluorescence background P(v) is fitted so that the Raman signal intensity of the target organophosphorus molecule can be accurately separated;

[0126] Subtract the background fitting P(v) from the original spectrum to obtain the fluorescence-removed signal, and the formula is:

[0127] I c I(v) = I raw (v) - P(v)

[0128] wherein, I c (v) is the Raman signal after fluorescence removal, and I raw (v) represents the intensity of the original spectral signal, including the Raman signal of the target molecule and background fluorescence interference;

[0129] The larger the I c (v), the stronger the Raman signal of the target molecule. This formula subtracts the background signal P(v) from I raw (v) to obtain a clear target signal;

[0130] Apply Savitzky - Golay filtering to smooth I c (v) to reduce noise interference. The window size and polynomial order of Savitzky - Golay filtering should be calculated and selected according to the noise level and Raman signal characteristics, and the window size is 5 or 7;

[0131] According to the characteristic peak position v d in the database, find the corresponding characteristic peak position ν c in the spectrum I m after fluorescence removal. The formula for calculating the peak position difference based on the peak position offset criterion is: Δν = v m - v d , wherein, v m is the measured peak position, and v d is the database peak position. If the peak position offset range |Δν| ≤ 5 cm -1 , it is determined that the measured peak position matches the database peak position of the target molecule, that is, the target molecule is accurately identified; verify the characteristic peak of the target molecule through the peak position offset criterion |Δv|;

[0132] Find the maximum intensity of the characteristic peak in the spectrum after fluorescence removal, and the basis formula is:

[0133]

[0134] wherein, I peak represents the peak intensity of the characteristic peak of the target organophosphorus molecule, [ν 1 , ν 2 represents the peak position range set by the database; the higher the I peak , the higher the concentration of the target organophosphorus molecule, which is related to the signal enhancement factor. This formula extracts the maximum peak intensity in I c (v) as the signal intensity index of the target molecule;

[0135] Set a comprehensive characteristic index EI to characterize the target organophosphorus molecule, that is, the spectral signal intensity, enhancement effect, reliability, and dynamic characteristics of organophosphorus. The formula is as follows:

[0136]

[0137] where I peak represents the characteristic peak intensity of the target organophosphorus molecule, η LSPR is the surface plasmon resonance enhancement factor, reflecting the enhancement effect of the substrate, R is the dual-substrate signal ratio parameter, reflecting signal reliability, D enh is the dynamic electric field enhancement factor, reflecting the enhancement effect of the applied electric field, k 1 , k 2 are constant parameters obtained by experimental fitting, used to adjust the influence of dynamic enhancement, α is the parameter controlling the amplitude of the exponential change, and is optimized and adjusted according to experiments;

[0138] k 1 , k 2 are the parameters controlling the influence of the dynamic enhancement factor D enh on the comprehensive characteristic index EI. The goal is to fit these parameters through experimental data so that the dynamic enhancement term in the formula accurately describes the experimental results, avoiding the excessive influence of over-strong signals on the model and reasonably reflecting the dynamic enhancement effect;

[0139] Keep other factors, such as η LSPR , R, α constant, change the characteristic peak intensity I peak by adjusting concentration, laser power, etc., change the dynamic electric field enhancement factor D enh by adjusting the applied electric field intensity. Under these experimental conditions, measure a set of corresponding experimental characteristic indices EI exp , ensure that the data cover the change ranges of I peak and D enh to comprehensively fit the dynamic enhancement term. The dynamic enhancement term of the fitting formula: Make it able to accurately describe the EI exp in the experiment. The optimization goal is to minimize the sum of the squared errors between the theoretical value EI theory and the experimental value EI exp ;

[0140] In the dynamic enhancement term, k 1 plays a role in baseline adjustment, used to control the stability in the case of low signal intensity or low dynamic enhancement, avoiding the denominator approaching zero. k 2 is a dynamic adjustment parameter, used to adjust the non-linear influence of the signal intensity I peak on the enhancement effect; The initial guess value can be set by observing the approximate range of experimental data. For example, if the dynamic enhancement factor Denh Typically within the range of 10 to 100, the intensity I of the characteristic peak peak Within the range of 100 to 1000, k can be estimated 1 is 10 - 50, k 2 is 0.01 - 0.1;

[0141] α is a parameter that controls the amplitude of the exponential change and is used to adjust the weights of the characteristic index EI on the signal enhancement factor η LSPR and the signal reliability R. The setting of α needs to balance the sensitivity and stability of the signal characteristics;

[0142] Keep I peak 、D enh and other parameters constant, and change the plasma resonance enhancement factor η LSPR and the signal reliability parameter R separately, measure a set of experimental characteristic indices, and ensure that the data cover the change ranges of η LSPR and R; optimize the exponential term in the formula to make its description of the experimental data the best, which can not only amplify the significant signals but also not overly expand the influence of weak signals;

[0143] When the value of α is larger, the amplitude of the exponential change is more significant, which is suitable for detection scenarios with higher requirements for signal differentiation. When the value of α is smaller, the amplitude of the exponential change is smoother, which is suitable for detection scenarios to reduce signal fluctuations; take an initial guess value, such as α = 1, calculate the theoretical value, and by adjusting the value of α, such as 0.5, 1.0, 1.5, etc., observe the fitting effect between the theoretical value and the experimental value, and select the optimal value;

[0144] The fractional term This part is the normalization processing of the signal intensity with the enhancement factor and the reliability factor, reflecting the standardized intensity of the basic signal. The larger I peak is the larger, indicating that the signal intensity is high and the characteristics are significant; the dynamic enhancement term This part incorporates the influence of the dynamic electric field enhancement factor D enh and reduces the excessive influence of over-strong signals on the index by introducing the non-linear term k 1 +k 2 ·I peak to enhance the robustness of the model. The higher D enh is, the greater the contribution of dynamic enhancement; the exponential term By taking α as the control parameter and combining the non-linear normalization of R and η LSPR , the stability of the signal characteristics is further enhanced; the logarithmic correction term The logarithmic function is used to smooth the change range of high-sensitivity signal values and correct the overall characteristic index value in combination with the enhancement factor and the reliability factor;

[0145] EI is the comprehensive characteristic index of the target molecule, which combines signal intensity, substrate enhancement, dynamic enhancement, and reliability factor to comprehensively characterize the spectral characteristics of the target molecule. The larger the EI, the more significant the signal of the target molecule, the more obvious the enhancement effect, and the higher the detection reliability.

[0146] The logic of using the comprehensive characteristic index of historical samples as the training set and the corresponding known target organophosphorus concentration as the label to train the SVM machine learning model is as follows:

[0147] Obtain experimental data samples of organophosphorus solutions with multiple groups of different historical concentrations. For each sample, calculate the historical comprehensive characteristic index EI h , for the historical comprehensive characteristic index corresponding to each sample, record its true historical organophosphorus concentration as the label; for the historical comprehensive characteristic index EI h Perform normalization.

[0148] Use the normalized comprehensive characteristic index EI of historical samples h as the input feature of the training set and the corresponding known historical organophosphorus concentration c h as the label of the training set, and use the SVR (regression SVM) model to establish the mapping relationship between the historical comprehensive characteristic index and the concentration. The comprehensive characteristic index EI h contains non-linear characteristics and is calculated using the RBF kernel function formula. The formula is as follows:

[0149] K(x i , x j ) = exp(-γ·‖x i - x j ‖ 2 )

[0150] where x i , x j represents the input feature value of the training sample, that is, the normalized comprehensive characteristic index EI h , ‖x i - x j ‖ 2 represents the Euclidean distance, which measures the similarity between samples. γ is the kernel function width parameter, which is used to control the complexity of mapping to the high-dimensional space, and γ > 0;

[0151] The kernel function maps the training samples to a high-dimensional space to capture the non-linear relationship between the comprehensive characteristic index EI h and the concentration c h . The larger γ is, the higher the complexity of the kernel function, which may lead to overfitting;

[0152] Use the objective function of the support vector machine to minimize the following equation:

[0153]

[0154] Among them, w is the weight vector of the regression hyperplane, C is the regularization parameter used to balance the model fitting error and complexity, and ξ i , is the slack variable used to tolerate prediction errors within a certain range, and ‖w‖ 2 represents the complexity of the model;

[0155] Minimize L, optimize the weight w and bias b of the SVM model to ensure high accuracy and good generalization ability;

[0156] To verify the accuracy of the model prediction, adjust the model parameters, and use the gradient descent method to solve the objective function, thereby determining the mapping function f(x):

[0157]

[0158] Among them, x is the comprehensive feature index of the sample to be predicted, K(x i , x) is the kernel function value, δ i , is the weight of the support vector, b is the bias term; use the mean square error or the coefficient of determination to evaluate the prediction performance of the SVR model;

[0159] Through f(x), realize the non-linear mapping between the comprehensive feature index EI h and the target concentration c h to complete the concentration prediction.

[0160] Step 4: Use the trained SVM model to predict the samples of agricultural products to be detected, obtain the concentration of the target organophosphorus molecule in the samples of agricultural products to be detected, and determine whether the concentration of the target organophosphorus molecule meets the national food safety standards;

[0161] The logic of using the trained SVM model to predict the samples of agricultural products to be detected and obtain the concentration of the target organophosphorus molecule in the samples of agricultural products to be detected is as follows:

[0162] Input the comprehensive feature index of the target organophosphorus molecule extracted from the samples of agricultural products to be detected into the trained support vector regression model f(x), and the formula for calculating the predicted concentration is:

[0163] c = f(x)

[0164] Among them, c is the predicted concentration of the agricultural product sample to be detected, representing the actual concentration of the target organophosphorus pesticide in the agricultural product sample to be detected. f(x) is a mapping function obtained through training with an SVR (Support Vector Regression) model, a mapping function determined by the training process, and is used to map the comprehensive feature index EI to the concentration; x represents the input comprehensive feature index EI, which characterizes comprehensive information such as the signal intensity, enhancement effect, and dynamic characteristics of the target molecule;

[0165] c represents the concentration of the target organophosphorus molecule in the agricultural product sample to be detected, and is a key parameter for ultimately determining whether it meets the national food safety standards. The larger c is, the higher the residue of the organophosphorus pesticide in the agricultural product sample; the smaller c is, the lower the pesticide residue in the agricultural product sample;

[0166] The comprehensive feature index EI is a high-dimensional feature extracted from the Raman spectrum, reflecting the spectral signal intensity and enhancement effect of the target molecule. The larger its value, the stronger the signal of the target molecule, and the higher the concentration usually is;

[0167] According to the pesticide residue limit specified in the national food safety standards, obtain the residue limit c′ of the target organophosphorus molecule, and compare the predicted concentration value c with the limit c′ in the pesticide residue limit specified in the national food safety standards to determine whether the agricultural product sample to be detected is qualified:

[0168]

[0169] Among them, if c ≤ c′, it indicates that the concentration of the target organophosphorus molecule does not exceed the national allowable safety limit, and the sample is qualified; if c > c′, it indicates that the concentration of the target organophosphorus molecule exceeds the standard, and the sample is unqualified; output the predicted concentration value c and the determination result, indicating whether the agricultural product sample to be detected meets the national food safety standards;

[0170] Results is the final pass / fail determination result, indicating whether the agricultural product sample to be detected meets the national food safety standards, with values of Pass (qualified) or Fail (unqualified); c is the concentration of the target organophosphorus molecule predicted by f(x), and c′ is the residue limit of the target organophosphorus pesticide specified in the national food safety standards;

[0171] Results is the determination result based on the comparison of the concentration value c with the limit c′ in the national food safety standards; when Results = Pass, it indicates that the agricultural product meets the national food safety standards; when Results = Fail, it indicates that the agricultural product does not meet the national food safety standards.

[0172] Please refer to Figure 2 , the present invention also provides a system for detecting organophosphorus pesticide residues. The system is used to implement the above method for detecting organophosphorus pesticide residues, and specifically includes:

[0173] A sample extraction module is used to preprocess the surface rinse liquid and the homogenized sample liquid of the agricultural product to be tested respectively, mix the two and add them to the sample inlet of the microfluidic chip, remove impurities through the microfluidic chip, capture and enrich pesticide molecules using magnetic microspheres, and then output a high-purity enriched liquid.

[0174] A characteristic parameter module is used to drop the high-purity enriched liquid collected in the outlet channel of the microfluidic chip onto a substrate for surface-enhanced Raman scattering. The substrate adopts a design with a periodic silver nanoparticle array, use a portable Raman spectrometer to excite the surface-enhanced Raman scattering substrate, and calculate spectral characteristic parameters.

[0175] A model establishment module is used to perform polynomial fitting on the collected Raman spectral data to remove fluorescence background interference, extract the characteristic peak intensity of the target organophosphorus molecule after removing fluorescence, calculate the comprehensive characteristic index of the target organophosphorus molecule based on the combination of the characteristic peak intensity and the spectral characteristic parameters, collect the comprehensive characteristic index of historical agricultural products as a training set, and the corresponding historical agricultural product concentration as a label to train an SVM machine learning model.

[0176] A result judgment module is used to use the trained SVM model to predict the agricultural product sample to be detected, obtain the concentration of the target organophosphorus molecule in the agricultural product sample to be detected, and judge whether the concentration of the target organophosphorus molecule meets the national food safety standard.

[0177] The present invention further provides a storage medium, which stores a computer program. When the computer program is executed by a processor, each step in the above method for detecting organophosphorus pesticide residues is implemented.

[0178] The above formulas are all calculated by taking the numerical value after dimensionless. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation. The preset parameters in the formula are set by those skilled in the art according to the actual situation.

[0179] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0180] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, and it may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment. As mentioned above, the above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. A method for detecting organophosphorus pesticide residues, characterized in that: The specific steps include: Step 1: Pre-treat the surface washing liquid of the agricultural product to be tested and the homogenized sample liquid separately, mix the two and add them to the sample inlet of the microfluidic chip, remove impurities through the microfluidic chip, use magnetic microspheres to capture and enrich the pesticide molecules, and then output high-purity enriched liquid; Step 2: The high-purity enriched liquid collected in the outlet channel of the microfluidic chip is dripped onto a surface-enhanced Raman scattering substrate, which is designed with a periodic silver nanoparticle array. A portable Raman spectrometer is used to excite the surface-enhanced Raman scattering substrate and calculate the spectral characteristic parameters; Step 3: Perform polynomial fitting on the collected Raman spectral data to remove fluorescence background interference, and extract the characteristic peak intensity of the target organophosphorus molecule after defluorescence. Based on the combination of characteristic peak intensity and spectral characteristic parameters, calculate the comprehensive characteristic index of the target organophosphorus molecule, collect the comprehensive characteristic index of historical agricultural products as a training set, and the corresponding historical agricultural product concentrations as labels to train the SVM machine learning model; Step 4: Use the trained SVM model to predict the agricultural product samples to be tested, obtain the concentration of the target organophosphorus molecules in the agricultural product samples to be tested, and determine whether the concentration of the target organophosphorus molecules meets the national food safety standards.

2. A method for detecting organophosphorus pesticide residues according to claim 1, characterized in that: The logic of pre-treating the surface washing liquid of the agricultural product to be tested and the homogenized sample liquid and mixing them and adding them to the sample inlet of the microfluidic chip is as follows: For samples with target pesticide residues on the surface of agricultural products, place the sample to be tested in a clean container, add 100 mL of deionized water, and use a 40 kHz, 200 W ultrasonic device to oscillate for 3 minutes to elute the surface target pesticide residue molecules from the sample surface into the solution. Filter the rinse liquid through a filter membrane with a pore size of 0.45 μm to remove large particle impurities, and collect the surface sample liquid for subsequent enrichment processing and detection; For samples with target pesticide residues inside agricultural products, the agricultural products were chopped into particles less than or equal to 2 cm in size, 10 g of the sample was weighed and placed in a homogenizer, an equal volume of 10 ml of acetonitrile extraction solvent was added, the homogenizer was started, and high-speed stirring was performed for 3 minutes to accelerate the molecular release to obtain uniform homogenization, the homogenate was transferred to a centrifuge tube, centrifuged at 8000 rpm for 10 minutes, the supernatant was taken and filtered again through a filter membrane with a pore size of 10 μm to remove particulate impurities, and the internal sample liquid was obtained for subsequent enrichment treatment; The surface sample liquid and the internal sample liquid were mixed at a volume ratio of 1:1, 1 mL of acetonitrile was added as a mixed solvent before mixing, and a vortex mixer was used to fully oscillate to form a uniform liquid phase. The mixed liquid was then filtered through a filter membrane with a pore size of 10 μm to remove particulate impurities and obtain a high-purity mixed sample liquid; The microfluidic chip contains a micron-scale liquid flow channel, a functional area and an outlet channel. The liquid channel injects the mixed sample liquid into the microfluidic chip at a flow rate of 10 μL / min through a syringe pump and enters the functional area; the functional area pre-stores acetonitrile extraction solvent and contains a filter membrane structure with a pore size of 10-100 μm for removing large particle impurities; In the functional area, the high-speed rotating shear force of the spiral micro-mixing channel, i.e., the Reynolds number is 1000, is used to achieve efficient transfer from the aqueous phase to the organic phase, i.e., the target liquid entering the mixing area is selectively transferred to the acetonitrile phase according to the hydrophilicity or hydrophobicity of the target organophosphorus molecules; The liquid enters the magnetic microsphere area, which includes magnetic microspheres with a particle size of 1-5μm. The surface of the magnetic microspheres is modified with MIP, which can selectively capture the target organophosphorus pesticide molecules through molecular imprinting. The captured magnetic microspheres control the flow trajectory through an external magnetic field and enrich them in the outlet area, and output high-purity enriched liquid in the outlet channel for subsequent experiments.

3. A method for detecting organophosphorus pesticide residues according to claim 2, characterized in that: The logic of dropping the target liquid onto the periodic silver nanoparticle array substrate and exciting it using a portable Raman spectrometer is: 10 μL of high-purity enriched liquid is taken from the outlet channel of the microfluidic chip, and evenly dripped onto the periodic silver nanoparticle array substrate, wherein the periodic silver nanoparticle array is composed of silver particles with a particle size of 40 nm, a particle spacing of 10-15 nm, and an array period of 100 nm; After adding the sample, let it stand for 10 minutes. The target organophosphorus molecules are fixed in the hot spots between the silver particles through electrostatic adsorption and short-range chemical reactions. After the sample is fixed, the substrate is placed at a constant temperature of 37°C and dried for 5 minutes to ensure that the liquid is completely evaporated, leaving only the target organophosphorus molecules adsorbed on the surface of the silver particles. The dried substrate was placed under a portable Raman spectrometer with a wavelength of 785 nm and a power of 10 mW for detection. The laser spot diameter was 10 μm and focused on the surface of the silver nanoparticle substrate. The Raman spectrum detection range was 400-2000 cm -1 , covering the Raman characteristic peaks of organophosphorus pesticides; The logic for calculating the spectral characteristic parameters is to introduce a signal enhancement factor using the LSPR effect of the silver nanoparticle array, based on the formula: Among them, I raw is the original Raman spectrum intensity of the silver substrate, I ref is the Raman spectrum intensity without enhanced substrate, η LSPR is the plasmon resonance enhancement factor; For the SERS detection of the target organophosphorus molecules, in addition to using the silver nanoparticle substrate, a gold nanoparticle array substrate was superimposed in the experiment, the target organophosphorus molecules were fixed on the silver nanoparticle array substrate and the gold nanoparticle array substrate respectively, and the ratio of the Raman signals of the two substrates was calculated based on the formula: Among them, I silver is the signal intensity of the silver substrate, I gold is the signal intensity of the gold substrate, R is the double substrate signal ratio parameter; If R>R threshold , the target organophosphorus molecule exists and the signal is significantly enhanced. If R≤R threshold , then the interference of the target organophosphorus molecule signal is eliminated, where R threshold is the threshold under the double basis combination; By introducing an external electric field on the Raman substrate and dynamically adjusting the enhancement effect of the localized surface plasma, the Raman signal of the target organophosphorus molecule is further enhanced. The formula for defining the dynamic electric field enhancement factor is: Among them, I raw is the original Raman signal intensity, E ext is the external electric field strength, k is the experimentally determined electric field enhancement coefficient, D enh is the dynamic electric field enhancement factor; For each sample, record the above spectral characteristic parameter η LSPR (v), R, D enh And the original spectral signal I raw (ν).

4. A method for detecting organophosphorus pesticide residues according to claim 3, characterized in that: The logic of performing polynomial fitting on the collected Raman spectra to remove fluorescence interference and extracting the Raman characteristic peak intensity of the target organophosphorus molecule is as follows: Background fitting is performed in the spectral region far away from the Raman characteristic peak of the target organophosphorus molecule. The fluorescence background is usually a broadband continuous signal, and its change frequency is lower than the Raman characteristic signal. It can be regarded as a low-frequency change function of the spectral shift ν. The polynomial P(v) is used to fit the background signal. The formula based on it is: P(ν)=a0+a1v+a2v 2 +…+a n n n Among them, a0, a1, …, a n represents the polynomial coefficient, n represents the polynomial order, n∈[3,5]; The background fit P(v) is subtracted from the original spectrum to obtain the defluorescence signal according to the formula: I c (ν)=I raw (v)-P(v) Among them, I c (v) is the Raman signal after removing fluorescence, I raw (v) represents the original spectral signal intensity, including the Raman signal of the target molecule and the background fluorescence interference; The Savitzky-Golay filter is used to filter I c (ν) Smoothing is performed to reduce noise interference. The Savitzky-Golay filter window size and polynomial order should be selected based on the noise level and Raman signal characteristics. The window size is 5 or 7. According to the characteristic peak position ν in the database d , after defluorescence, the spectrum I c Find the corresponding characteristic peak position v in (ν) m The formula for calculating the peak position difference by the peak position shift criterion is: Δν=ν m -ν d , where ν m is the measured peak position, ν d is the database peak position, if the peak position offset range |Δv|≤5cm -1 , then the characteristic peaks are determined to match; The maximum intensity of the characteristic peak in the spectrum after defluorescence is found according to the formula: Among them, I peak represents the peak intensity of the characteristic peak of the target organophosphorus molecule, and [ν1, ν2] represents the peak position range set by the database; A comprehensive characteristic index EI is set to characterize the target organophosphorus molecule, which integrates the spectral signal intensity, enhancement effect, reliability and dynamic characteristics of the organophosphorus. The formula is: Among them, I peak is the characteristic peak intensity of the target organophosphorus molecule, η LSPR is the plasma resonance enhancement factor, R is the dual-base signal ratio parameter, and D enh is the dynamic electric field enhancement factor, k1 and k2 are constant parameters, and α is the parameter that controls the exponential variation.

5. A method for detecting organophosphorus pesticide residues according to claim 4, characterized in that: The logic of collecting the comprehensive characteristic index of historical samples as the training set and the corresponding known target organic phosphorus concentration as the label to train the SVM machine learning model is: Obtain multiple groups of historical organophosphorus solution experimental data samples of different concentrations, and calculate the historical comprehensive characteristic index EI for each sample h , corresponding to the historical comprehensive characteristic index of each sample, recording its real historical organic phosphorus concentration as a label; for the historical comprehensive characteristic index EI h Perform normalization; The comprehensive characteristic index EI of the normalized historical samples h As the input features of the training set and the corresponding known historical organic phosphorus concentration c h As the label of the training set, the SVR (regression SVM) model is used to establish the mapping relationship between the historical comprehensive characteristic index and the concentration. The comprehensive characteristic index EI h Contains nonlinear features and is calculated using the RBF kernel function formula based on the following formula: K(x i ,x j )=exp(-γ·‖x i -x j ‖ 2 ) Among them, x i ,x j Represents the input feature value of the training sample, that is, the normalized comprehensive feature index EI h , ‖x i -x j ‖ 2 represents the Euclidean distance, which measures the similarity between samples. γ is the kernel function width parameter, which is used to control the complexity of mapping to high-dimensional space, and γ>0; Using the objective function of the support vector machine, minimize the following equation: Among them, w is the weight vector of the regression hyperplane, C is the regularization parameter used to balance the model fitting error and complexity, ξ i , It is a slack variable, which is used to tolerate prediction errors within a certain range; To test the accuracy of the model prediction, adjust the model parameters, use the gradient descent method to solve the objective function, and thus determine the mapping function f(x): Among them, x is the comprehensive feature index of the sample to be predicted, K(x i ,x) is the kernel function value, δ i , is the weight of the support vector, b is the bias term, and the mean square error or determination coefficient is used to evaluate the prediction performance of the SVR model.

6. A method for detecting organophosphorus pesticide residues according to claim 5, characterized in that: The logic of using the trained SVM model to predict the concentration of the target organophosphorus molecules in the agricultural product samples to be tested is as follows: The comprehensive characteristic index of the target organophosphorus molecules extracted from the agricultural product samples to be tested is input into the trained support vector regression model f(x), and the formula for calculating the predicted concentration is: c=f(x) Where c is the predicted concentration of the agricultural product sample to be tested, and f(x) is the mapping function determined by the training process, which is used to map the comprehensive characteristic index EI to concentration; According to the pesticide residue limit specified in the national food safety standard, the residue limit c' of the target organophosphorus molecule is obtained, and the predicted concentration value c is compared with the limit c' in the pesticide residue limit specified in the national food safety standard to determine whether the agricultural product sample to be tested is qualified: Among them, if c≤c′, it is judged as qualified, and if c>c′, it is judged as unqualified; the predicted concentration value c and the judgment result are output to indicate whether the agricultural product sample to be tested meets the national food safety standards.

7. A system for detecting organophosphorus pesticide residues, characterized in that: The system is used to perform a method for detecting organophosphorus pesticide residues according to any one of claims 1 to 6, comprising: The sample extraction module is used to pre-treat the surface washing liquid and homogenized sample liquid of the agricultural products to be tested, respectively, mix the two and add them to the sample inlet of the microfluidic chip, remove impurities through the microfluidic chip, use magnetic microspheres to capture and enrich the pesticide molecules, and then output high-purity enriched liquid; The characteristic parameter module is used to drop the high-purity enriched liquid collected in the outlet channel of the microfluidic chip onto the surface-enhanced Raman scattering substrate, which is designed with a periodic silver nanoparticle array. The surface-enhanced Raman scattering substrate is excited by a portable Raman spectrometer and the spectral characteristic parameters are calculated. Model building: A model is used to perform polynomial fitting on the collected Raman spectral data to remove fluorescence background interference, and extract the characteristic peak intensity of the target organophosphorus molecule after defluorescence. Based on the combination of characteristic peak intensity and spectral characteristic parameters, the comprehensive characteristic index of the target organophosphorus molecule is calculated. The comprehensive characteristic index of historical agricultural products is collected as a training set, and the corresponding historical agricultural product concentrations are used as labels to train the SVM machine learning model; The result judgment module is used to use the trained SVM model to predict the agricultural product samples to be tested, obtain the concentration of the target organophosphorus molecules in the agricultural product samples to be tested, and judge whether the concentration of the target organophosphorus molecules meets the national food safety standards.

8. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, a method for detecting organophosphorus pesticide residues as described in any one of claims 1 to 6 is implemented.

Images