Method for extracting ochratoxin A in coffee by using acetonitrile

Abstract

The invention relates to the technical field of toxin extraction, and discloses a method for extracting ochratoxin A in coffee by using acetonitrile, and the method comprises the following steps: collecting basic physicochemical characteristic data of a coffee sample, and dividing the sample into different matrix complexity levels according to the data; the method comprises the following steps: preparing different grades of MIP extraction columns by taking ochratoxin A as a template molecule; constructing a toxin extraction dynamic prediction model, predicting the extraction rate according to the multi-dimensional spectrum time sequence data, and judging whether the extraction rate reaches a preset threshold value or not; after extraction is completed, the concentration of the ochratoxin A is measured, and if the target concentration is not reached, closed-loop feedback control is started, extraction parameters are adjusted or an extraction column is replaced; the invention provides an efficient and intelligent ochratoxin A extraction method, the extraction process can be accurately regulated and controlled according to coffee samples with different matrix complexity degrees, the stability and the high efficiency of the extraction effect are ensured, and the accuracy and the reliability of the extraction process are improved through closed-loop feedback control.

Description

Technical Field

[0001] The present invention relates to the technical field of toxin extraction, and more particularly to a method for extracting ochratoxin A toxin in coffee by using acetonitrile. Background Art

[0002] Ochratoxin A (Aflatoxin A) is a common fungal toxin that is widely found in foods such as coffee, cereals, and dried fruits. Long-term intake may cause serious harm to human health, especially damage to the liver and immune system. Therefore, the development of an efficient and accurate detection method is crucial for food safety. Most of the existing detection methods focus on the extraction and analysis of samples using chromatographic technology combined with immunoaffinity columns, but these methods generally have problems such as high cost, complex operation, and limited scope of application. With the continuous improvement of food quality control and environmental protection requirements, how to optimize toxin extraction and detection methods to achieve efficient, low-cost, and easy-to-operate detection methods has become a difficult problem that needs to be solved urgently in current technology.

[0003] The Chinese patent with publication number CN106370747A discloses a method for detecting ochratoxin A in a pepper extract, which comprises the steps of ultrasonic extraction with an alcohol solution, refrigerated centrifugal separation, and dilution and filtration with a PBS solution, and finally uses liquid chromatography-fluorescence method to detect the content of ochratoxin A. This method removes impurities by refrigerated centrifugation and does not use an immunoaffinity column. It has good batch processing capabilities, can save costs, and improve detection efficiency. Although this method improves the simplicity and cost-effectiveness of operation, its scope of application is mainly limited to pepper extracts, and its applicability to other complex matrices such as coffee is relatively low, and the effects when treating different matrices cannot be uniformly evaluated. In addition, this method does not involve dynamic regulation of the complexity of the matrix, nor does it have a real-time monitoring and adjustment mechanism for the extraction process, which may lead to the problem of low extraction efficiency under different samples or extraction conditions.

[0004] The Chinese patent with the authorization announcement number CN112461973B provides a new method for detecting ochratoxin A in red wine based on hexafluorobutanol as an extractant. Hexafluorobutanol is used as an extractant, and the sample is extracted by shaking and centrifugation, and HPLC is used for quantitative and qualitative analysis. The method is simple to operate, time-saving, does not use volatile toxic solvents, and meets green environmental protection requirements. However, although hexafluorobutanol has a good effect as an extractant, this method is also mainly used for liquid samples such as red wine, and does not involve customized processing of samples with different matrix complexities. For solid matrices such as coffee, especially samples with strong matrix effects, the extraction efficiency of hexafluorobutanol may be limited, and dynamic monitoring and real-time adjustment strategies are not combined. Therefore, this method may not be able to guarantee the stability and consistency of extraction efficiency when dealing with samples with different matrices.

[0005] In summary, the existing methods fail to consider the complexity of the sample matrix, lack the classification of matrix complexity levels and targeted extraction schemes, which leads to unsatisfactory extraction effects when facing different types of samples. The failure to perform real-time prediction and dynamic adjustment during the extraction process leads to the inability to quickly respond and adjust the extraction process under different extraction conditions, affecting the stability of the extraction effect. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides an extraction method of ochratoxin A toxin in coffee using acetonitrile, which combines the matrix complexity classification based on the basic physicochemical characteristics of the sample, the customized preparation of the MIP extraction column, and the toxin extraction dynamic prediction model and closed-loop feedback control mechanism, aiming to solve these defects in the existing methods and provide an accurate and dynamically adjustable ochratoxin A extraction method for different matrix complexities.

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

[0008] A method for extracting ochratoxin A toxin from coffee using acetonitrile extraction comprises:

[0009] The basic physicochemical characteristic data of the target coffee samples are collected, and the target coffee samples are divided into L matrix complexity levels according to the basic physicochemical characteristic data; Ochratoxin A is used as a template molecule, the type of polymerization precursor is selected, the amount of polymerization precursor is optimized, and MIP extraction columns for target coffee samples with different matrix complexity levels are prepared;

[0010] Constructing a dynamic prediction model for toxin extraction, collecting a second multidimensional spectral time series of the coffee sample to be tested, and obtaining a predicted extraction rate value of the coffee sample to be tested according to the second multidimensional spectral time series and the dynamic prediction model for toxin extraction; judging whether the predicted extraction rate value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stopping the extraction process, otherwise continuing the extraction until the predicted extraction rate value reaches the extraction rate threshold;

[0011] After the extraction of the coffee sample is completed, the concentration of ochratoxin A in the extract is determined. A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

[0012] Furthermore, the dividing the target coffee sample into L matrix complexity levels includes:

[0013] Construct a sample feature vector based on the basic physical and chemical characteristic data of the target coffee sample;

[0014] Determine the optimal number of clusters L and divide the sample feature vectors into L clusters;

[0015] Calculate the feature mean of samples within each cluster and generate L cluster center vectors;

[0016] The L clusters are sorted from low to high according to their feature means to generate a matrix complexity level sequence {1, 2, ..., L}.

[0017] Further, the polymer precursor includes a functional monomer, a cross-linking agent and an eluent;

[0018] The selection of the type of polymer precursor includes: acetonitrile is selected as the eluent, methacrylic acid is selected as the functional monomer, and ethylene glycol dimethacrylate is selected as the cross-linking agent;

[0019] The MIP extraction column for preparing target coffee samples with different matrix complexity levels comprises:

[0020] The effect of the amount of polymer precursor on the performance of the MIP extraction column was comprehensively investigated to obtain the optimal MIP extraction column preparation parameter combination for the target coffee sample at different matrix complexity levels; the optimal MIP extraction column preparation parameter combination includes the optimal functional monomer amount, the optimal cross-linking agent amount and the optimal eluent amount;

[0021] According to the optimal MIP extraction column preparation parameter combination, a series of MIP extraction columns were prepared for target coffee samples with different matrix complexity levels.

[0022] Furthermore, the optimal MIP extraction column preparation parameter combination for obtaining target coffee samples at different matrix complexity levels includes:

[0023] The amount of polymer precursor is used as a parameter for preparing a MIP extraction column, and N experimental points are selected within the value range of the parameter for preparing a MIP extraction column to prepare N MIP extraction columns; the MIP extraction column preparation parameters include the amount of functional monomer, the amount of cross-linking agent, and the amount of eluent;

[0024] Using the N MIP extraction columns to extract target coffee samples with different matrix complexity levels, respectively, and measuring the performance indicators of the N MIP extraction columns; the performance indicators include second extraction rate, selectivity and anti-interference ability;

[0025] With the performance index as the response value and the preparation parameters of MIP extraction column as the independent variable, the response surface methodology was used to establish the nonlinear regression model between the preparation parameters of MIP extraction column and the performance index, and the MIP response surface model was obtained.

[0026] The MIP response surface model was solved to obtain the global optimal point of the second extraction rate, selectivity and anti-interference ability at each matrix complexity level, and the optimal functional monomer dosage, optimal cross-linking agent dosage and optimal eluent dosage of the target coffee sample at each matrix complexity level were obtained.

[0027] Furthermore, the construction of a dynamic prediction model for toxin extraction includes:

[0028] The target coffee samples with different matrix complexity levels were extracted using the corresponding MIP extraction columns, and the extract samples were collected at different time points to determine the extraction rate;

[0029] During the extraction process of the target coffee sample, multi-dimensional spectral time series data of the extract is continuously collected, which is marked as a first multi-dimensional spectral time series; the multi-dimensional spectral time series data includes a transmission spectrum, a reflection spectrum, a three-dimensional fluorescence spectrum and a SERS spectrum;

[0030] Extracting time series features from the first multidimensional spectral time series to obtain first spectral time series features, and matching them with extraction rates at corresponding time points to form a sample data set;

[0031] A dynamic prediction model for toxin extraction was constructed with the first spectral time series feature in the sample data set as the independent variable and the extraction rate at the corresponding time point in the sample data set as the dependent variable.

[0032] Further, obtaining the predicted value of the extraction rate of the coffee sample to be tested includes:

[0033] Extract the coffee sample to be tested;

[0034] Continuously collect multi-dimensional spectral time series data of the extract during the extraction process of the coffee sample to be tested, marked as the second multi-dimensional spectral time series; collect conductivity time series data of the extract;

[0035] Extracting time series features from the second multi-dimensional spectrum time series to obtain second spectrum time series features;

[0036] The second spectral time series feature is input into the toxin extraction dynamic prediction model to obtain the extraction rate prediction value of the coffee sample to be tested.

[0037] Furthermore, the extraction of the coffee sample to be tested includes:

[0038] Obtaining basic physicochemical characteristic data of the coffee sample to be tested, and determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade according to the basic physicochemical characteristic data of the coffee sample to be tested; if the coffee sample to be tested is a sample of known grade, determining the matrix complexity level of the coffee sample to be tested;

[0039] If the coffee sample to be tested is a sample of known grade, then according to the matrix complexity level of the coffee sample to be tested, a corresponding MIP extraction column is used to extract the coffee sample to be tested;

[0040] If the coffee sample to be tested is of unknown grade, a MIP extraction column corresponding to the matrix complexity level L is used for extraction.

[0041] Further, the step of determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade comprises:

[0042] According to the basic physical and chemical characteristic data of the coffee sample to be tested, a feature vector of the sample to be tested is constructed;

[0043] Calculate the similarity between the feature vector of the sample to be tested and the L cluster center vectors, and obtain the similarity set {S1, S2, ..., S L};

[0044] From the similarity set {S1,S2,...,S L}Select the maximum similarity S g , 1≤g≤L; if S g ≥S M , then the coffee sample to be tested is determined to be a sample of known grade; otherwise, the coffee sample to be tested is determined to be a sample of unknown grade; S M is the preset similarity threshold;

[0045] Determining the matrix complexity level of the coffee sample to be tested includes:

[0046] If the coffee sample to be tested is a sample of known grade, the matrix complexity grade corresponding to the cluster center vector having the highest similarity to the feature vector of the sample to be tested is taken as the matrix complexity grade of the coffee sample to be tested.

[0047] Further, the determination of whether the extraction process achieves the expected goal includes:

[0048] If C A ≥C θ , it is determined that the extraction process has achieved the expected goal;

[0049] If C A <C θ , it is determined that the extraction process has not reached the expected goal and closed-loop feedback control is required;

[0050] The closed-loop feedback control comprises:

[0051] According to the conductivity time series data of the extract during the extraction process of the coffee sample to be tested, the dynamic conductivity curve of the MIP extraction column is established; the statistical process control method is used to determine whether the performance of the MIP extraction column is abnormal based on the dynamic conductivity curve. If there is no abnormality, it is determined whether the adsorption capacity of the MIP extraction column is saturated. If not, the extraction rate threshold is increased and the amount of eluent is increased; if saturated, the MIP extraction column is replaced, and the extraction rate threshold is increased and the amount of eluent is increased; if the performance of the MIP extraction column is abnormal, the MIP extraction column is replaced;

[0052] The method for judging whether the adsorption capacity of the MIP extraction column is saturated comprises:

[0053] The curve slope of the dynamic conductivity curve is calculated. When the curve slope is lower than a preset slope threshold, it is determined that the adsorption capacity of the MIP extraction column has reached saturation; otherwise, it is determined that the adsorption capacity of the MIP extraction column is not saturated.

[0054] The extraction system of ochratoxin A toxin in coffee using acetonitrile is used to implement the above-mentioned extraction method of ochratoxin A toxin in coffee using acetonitrile, and the system comprises:

[0055] Grade system building module: used to collect basic physicochemical characteristic data of target coffee samples, and divide the target coffee samples into L matrix complexity levels according to the basic physicochemical characteristic data;

[0056] Extraction column preparation module: using ochratoxin A as the template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing MIP extraction columns for target coffee samples with different matrix complexity levels;

[0057] Prediction module: used to construct a dynamic prediction model for toxin extraction, collect the second multidimensional spectrum time series of the coffee sample to be tested, and obtain the extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectrum time series and the toxin extraction dynamic prediction model; determine whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stop the extraction process, otherwise continue the extraction until the extraction rate prediction value reaches the extraction rate threshold;

[0058] Evaluation module: After the extraction of the coffee sample to be tested is completed, the concentration of ochratoxin A in the extract is measured C A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

[0059] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit, wherein when the central processing unit executes the computer program, the method for extracting ochratoxin A from coffee using acetonitrile is implemented.

[0060] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed, the above-mentioned method for extracting ochratoxin A toxin in coffee using acetonitrile is implemented.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] The present invention divides coffee samples into different matrix complexity levels according to their basic physical and chemical characteristics, and tailors MIP (molecular imprinting polymer) extraction columns for each level. This customized method can accurately select suitable extraction conditions according to different characteristics of the sample (such as variety, roasting degree, etc.) to optimize the extraction effect of ochratoxin A. Compared with the traditional single extraction column, the adjustment for different sample characteristics greatly improves the extraction efficiency and accuracy.

[0063] The present invention constructs a dynamic prediction model for toxin extraction, which predicts the extraction rate in real time through multi-dimensional spectral time series data (such as transmission spectrum, fluorescence spectrum, etc.). This dynamic prediction model can not only provide timely feedback on changes in the extraction process, but also adjust operating parameters according to real-time data during the extraction process, such as extending the extraction time or adjusting the amount of extractant used, thereby ensuring accurate control of the extraction process. This real-time prediction and adjustment mechanism makes the extraction process more intelligent, avoids over-extraction or under-extraction, and improves efficiency and product quality.

[0064] When the extraction fails to reach the expected goal, the present invention optimizes through a closed-loop feedback control mechanism, thereby ensuring the stability and reliability of the extraction process and avoiding incomplete extraction or quality fluctuations caused by operating errors or equipment aging.

[0065] The present invention optimizes the MIP extraction column under different matrix complexity levels and comprehensively examines the effect of the amount of polymer precursor on the performance of the extraction column. A nonlinear regression model between the preparation parameters and performance indicators of the MIP extraction column is established by the response surface method to determine the optimal extraction column preparation parameters under each matrix complexity level. This precise optimization strategy not only improves the extraction efficiency of the MIP extraction column, but also enhances its selectivity and anti-interference ability for ochratoxin A, ensuring that the extraction effect of the target substance in the extraction process is not affected by the complexity of the matrix.

[0066] The control strategy of the present invention using multidimensional spectral time series data and dynamic prediction model makes the entire ochratoxin A extraction process highly automated and intelligent, reduces manual intervention and judgment errors, and improves production efficiency. Combined with closed-loop feedback control, the system can autonomously adjust and optimize to adapt to different samples and environmental conditions, ensuring the sustainability and efficiency of the production process.

[0067] The use of acetonitrile as an extraction solvent in the present invention can not only efficiently extract ochratoxin A, but also has good recovery and reuse performance, reduces solvent consumption and waste generation, and meets environmental protection requirements. At the same time, by optimizing the extraction conditions and the performance of the MIP extraction column, the present invention can reduce the use of solvents and improve the extraction efficiency while reducing production costs, and has high economic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 The principle flow chart of the method for extracting ochratoxin A toxin in coffee using acetonitrile in the present invention;

[0070] Figure 2 A flow chart of a method for dividing a target coffee sample into L matrix complexity levels in a method for extracting ochratoxin A toxin in coffee using acetonitrile according to the present invention;

[0071] Figure 3 A flow chart of a method for obtaining an optimal MIP extraction column preparation parameter combination for a target coffee sample at different matrix complexity levels in the method for extracting ochratoxin A toxin in coffee using acetonitrile according to the present invention;

[0072] Figure 4 A flow chart of a method for constructing a dynamic prediction model for toxin extraction in a method for extracting ochratoxin A toxin in coffee using acetonitrile according to the present invention;

[0073] Figure 5 A flow chart of a method for obtaining a predicted value of the extraction rate of a coffee sample to be tested in the method for extracting ochratoxin A toxin in coffee using acetonitrile according to the present invention;

[0074] Figure 6 A flow chart of a method for extracting ochratoxin A toxin from coffee using acetonitrile according to the present invention;

[0075] Figure 7 This is a functional module diagram of the extraction system for extracting ochratoxin A toxin in coffee using acetonitrile in the present invention. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] Example 1

[0078] See also Figure 1 As shown, this embodiment provides a method for extracting ochratoxin A toxin in coffee using acetonitrile, comprising:

[0079] Step S1000, collecting basic physicochemical characteristic data of target coffee samples, and dividing the target coffee samples into L matrix complexity levels according to the basic physicochemical characteristic data; using ochratoxin A as a template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing MIP extraction columns for target coffee samples with different matrix complexity levels;

[0080] Furthermore, step S1000 includes:

[0081] Step S1100, collecting basic physicochemical characteristic data of the target coffee sample, and dividing the target coffee sample into L matrix complexity levels according to the basic physicochemical characteristic data;

[0082] Furthermore, step S1100 includes:

[0083] Step S1110, constructing a sample feature vector according to the basic physical and chemical characteristic data of the target coffee sample; the basic physical and chemical characteristic data includes the variety, origin, roasting degree, crushed particle size and moisture content of the coffee sample;

[0084] Specifically, the basic physical and chemical characteristics data include the following:

[0085] Variety: The variety of coffee directly affects its flavor characteristics. Different varieties of coffee beans may have different chemical compositions, tastes, and aromas. Common coffee varieties are Arabica and Robusta.

[0086] Origin: The origin of coffee has a significant impact on its flavor. For example, coffee from different regions such as Colombia, Brazil and Ethiopia will have different components and flavors due to different geographical conditions, climate, soil and other factors.

[0087] Roasting degree: The roasting process has an important influence on the physical and chemical properties of coffee beans. Common roasting degrees include light roasting, medium roasting and deep roasting. The roasting degree will affect the taste, acidity and bitterness of coffee.

[0088] Particle size: Particle size will affect the extraction efficiency of coffee. Coffee powder with larger particle size requires longer extraction time to obtain sufficient flavor, while coffee powder with smaller particle size is extracted faster and prone to over-extraction, affecting the flavor.

[0089] Moisture content: The moisture content of coffee beans affects their shelf life and extraction effect. Coffee beans with higher moisture content may be more prone to mold, affecting the quality of the coffee.

[0090] The feature vector of each sample can be represented as a multidimensional vector, with specific dimensions corresponding to the five features mentioned above (variety, origin, roasting degree, crushed particle size and moisture content). At least 10 parallel samples are collected for each sample to ensure the reliability and representativeness of the data. This can avoid the interference of accidental factors and increase the accuracy of the sample feature vector. In this way, a numerical feature vector can be generated for each coffee sample, which is convenient for subsequent cluster analysis.

[0091] Step S1120, performing cluster analysis on the sample feature vectors, and dividing the target coffee samples into L matrix complexity levels.

[0092] Furthermore, if Figure 2 As shown, step S1120 includes:

[0093] Step S1121, determining the optimal number of clusters L, and dividing the sample feature vector into L clusters;

[0094] Step S1122, calculating the characteristic mean of samples within each cluster, generating L cluster center vectors, each representing a typical feature of the L matrix complexity levels;

[0095] Step S1123, sort the L clusters from low to high according to their feature means, and generate a matrix complexity level sequence {1, 2, ..., L}.

[0096] Specifically, in step S1120, cluster analysis is performed on the sample feature vectors, with the goal of classifying different coffee samples into L matrix complexity levels according to their features. Cluster analysis can help discover the intrinsic similarities and differences between samples, thereby grouping the samples according to the similarity of features.

[0097] Determining the number of clusters L is a crucial step in cluster analysis. If L is not chosen properly, the clustering results may be distorted. The commonly used method to determine the optimal number of clusters is the Elbow Method. The L value can be determined by the following steps:

[0098] Increase the number of clusters from 2 to 10.

[0099] After each clustering, the sum of squared errors (SSE) of the clustering results is calculated. SSE reflects the distance between the sample and the cluster center. The smaller the distance, the better the clustering effect.

[0100] Draw a curve between the number of clusters K and SSE, and select the "elbow" of the SSE curve as the optimal number of clusters L. The elbow refers to the point where the slope of the curve changes most significantly, which means that after this point, the decrease in SSE brought about by increasing the number of clusters is significantly reduced.

[0101] For example, if the decrease in SSE suddenly slows down at K=3, then L=3 may be the optimal number of clusters.

[0102] After determining the optimal number of clusters L, the feature mean can be calculated for the samples in each cluster, that is, the average value of all samples in each feature dimension. This mean vector represents the cluster center of the cluster, which is a typical feature of the cluster. For example, assuming L=3, 3 cluster center vectors will be obtained, each of which contains the mean of five features (variety, origin, roasting degree, crushed particle size and moisture content).

[0103] Each cluster center vector represents a level of matrix complexity. For example, the first cluster may correspond to low matrix complexity, the second cluster to medium complexity, and the third cluster to high complexity.

[0104] By sorting the characteristic means of the cluster center vectors from low to high, a matrix complexity level sequence {1, 2, ..., L} can be obtained. The higher the characteristic mean, the higher the matrix complexity. The cluster center corresponding to the matrix complexity level L has the highest characteristic mean, which means that the matrix complexity of this level is the greatest.

[0105] For example, if L=3, the sorting results may be: the first cluster (lowest matrix complexity) = level 1, the second cluster = level 2, and the third cluster (highest matrix complexity) = level 3. In this way, coffee samples can be clearly divided into different complexity levels, providing a basis for subsequent MIP extraction column customization and sample grading.

[0106] In step S1120, the target coffee samples are clustered according to their basic physical and chemical characteristics, and a coffee matrix complexity level system based on sample feature vectors is constructed. This system can effectively describe the complexity differences of different coffee samples, and thus lay the foundation for subsequent research and application (such as MIP extraction column customization). Through cluster analysis, the matrix complexity of coffee samples can be quantified, avoiding the deviation caused by subjective judgment. Based on the matrix complexity level system, MIP extraction columns can be customized for coffee samples of different complexities, optimizing the extraction effect and improving the overall process efficiency. By collecting multiple parallel sample data and performing cluster analysis, the reliability and scientificity of the research results are improved.

[0107] Step S1200, using ochratoxin A as a template molecule, selecting the type of polymer precursor, optimizing the amount of the polymer precursor, and preparing a MIP extraction column for target coffee samples with different matrix complexity levels; the polymer precursor includes a functional monomer, a cross-linking agent, and an eluent;

[0108] Further, step S1200 includes:

[0109] Step S1210, acetonitrile is selected as the eluent, methacrylic acid is selected as the functional monomer, and ethylene glycol dimethacrylate is selected as the cross-linking agent;

[0110] Specifically, the polymer precursor refers to a mixture of all participating substances before the polymerization reaction begins in the preparation process of molecular imprinting polymer (MIP), including functional monomers, cross-linking agents and eluents. In step S1210, acetonitrile is selected as the eluent, methacrylic acid (MAA) as the functional monomer, and ethylene glycol dimethacrylate (EGDMA) as the cross-linking agent. These chemical substances play a vital role in the preparation of molecular imprinting polymer (MIP). Acetonitrile is a common polar solvent and is widely used in the elution process of molecular imprinting polymers. Its selective solubility enables it to effectively remove template molecules (such as ochratoxin A) from MIP materials, thereby restoring the selectivity and affinity of MIP. Acetonitrile has a low polarity, which enables it to interact with polar molecules, dissolve and remove template molecules to a certain extent, and will not destroy the structure of MIP. Therefore, selecting acetonitrile as the eluent can ensure a high recovery rate and low interference during the extraction process.

[0111] Methacrylic acid (MAA) is a widely used functional monomer with strong hydrogen bond donor properties, which can form stable hydrogen bonds with template molecules (such as ochratoxin A). The affinity and selectivity of MAA make it the preferred monomer for constructing MIPs. Methacrylic acid has a small molecular structure, good polymerizability and compatibility, and can form highly specific binding sites during the polymerization process, thereby improving the selectivity and capture ability of MIPs. The binding force between it and the template molecule enables the template molecule to be effectively removed during elution, while the selectivity of the MIP itself will not be destroyed.

[0112] Ethylene glycol dimethacrylate (EGDMA) is a common cross-linking agent with low molecular weight, good solubility and small steric hindrance, which can evenly disperse functional monomers. EGDMA can effectively co-polymerize with functional monomers to form a three-dimensional network structure, thereby enhancing the mechanical strength, stability and selectivity of MIP. Due to its small molecular weight and small steric hindrance, EGDMA can form a uniform network structure during the polymerization process, increase the specific surface area of ​​the MIP material, and enhance its adsorption capacity for target molecules. Therefore, the selection of EGDMA helps to improve the performance of MIP, especially to enhance the adsorption capacity and selectivity of MIP during the extraction process.

[0113] Step S1220, comprehensively investigating the effect of the amount of polymer precursor on the performance of the MIP extraction column, and obtaining the optimal MIP extraction column preparation parameter combination for the target coffee sample at different matrix complexity levels; the optimal MIP extraction column preparation parameter combination includes the optimal functional monomer amount, the optimal cross-linking agent amount, and the optimal eluent amount;

[0114] Furthermore, if Figure 3 As shown, step S1220 includes:

[0115] Step S1221, using the amount of polymer precursor as a MIP extraction column preparation parameter, selecting N experimental points within the value range of the MIP extraction column preparation parameter, and preparing N MIP extraction columns; the MIP extraction column preparation parameters include the amount of functional monomer, the amount of cross-linking agent, and the amount of eluent;

[0116] The step of selecting N experimental points within the value range of the MIP extraction column preparation parameters to prepare N MIP extraction columns comprises:

[0117] The experimental parameter space is set as a three-dimensional space, where the parameters include the amount of functional monomers , cross-linking agent dosage and eluent dosage . Its mathematical expression is:

[0118]

[0119] in, Indicates the minimum amount of functional monomers used. Indicates the maximum amount of functional monomers used, and The value range of the functional monomer dosage in the experimental design is defined to ensure that the experimental points are selected within this range for optimization; Indicates the minimum amount of crosslinker used. Indicates the maximum amount of crosslinking agent used, and The range of the amount of cross-linking agent used in the experimental design is defined; Indicates the minimum amount of eluent used. Indicates the maximum amount of eluent used. and The range of eluent dosage in experimental design is defined.

[0120] Each experimental point is Indicates that is the number of the experimental point, .

[0121] In order to select experimental points more accurately in multidimensional space, weighted distance is used to optimize the selection of experimental points, considering the different value ranges of each dimension. The weighted distance formula can be expressed as:

[0122]

[0123] in, is the weight coefficient of the kth dimension, , used to control the impact of different dimensions on distance calculation. is the number of the experimental point, , i≠j; weighting coefficient The calculation can be based on the parameter range of this dimension:

[0124]

[0125] In this way, the parameter value ranges of all dimensions are standardized so that different dimensions contribute to the distance calculation to the same extent.

[0126] Next, a nonlinear objective function is constructed to maximize the weighted distance sum between all experimental points to ensure the uniformity of the distribution of experimental points. The objective function can be expressed as:

[0127]

[0128] in:

[0129] It is a positive adjustment factor used to control the influence of distance on the objective function.

[0130] is the weighted distance between experimental points i and j.

[0131] It is an exponential function with the base e of the natural logarithm as its base;

[0132] Constraints:

[0133] All experimental points need to be within the given parameter space, that is:

[0134]

[0135] In addition, the selection of experimental points also needs to meet the following constraints: each experimental point is not locally overlapped in the multidimensional space, that is, the distance between any two experimental points cannot be less than the preset distance threshold. :

[0136]

[0137] This constraint ensures that the experimental points are not too dense.

[0138] In summary, the entire optimization problem can be expressed as:

[0139]

[0140] Subject to:

[0141]

[0142] Maximizing the objective function F can ensure that the experimental points are evenly distributed in the multidimensional space and maintain a certain distance.

[0143] The influence of different dimensions is standardized by calculating the value range of each dimension to ensure that each dimension contributes equally to the selection of experimental points. The choice of depends on the accuracy requirements of the experiment and the requirements of the target distribution. In practical applications, The value of is used to balance the uniformity and the spacing of the experimental points. It is used to control the minimum distance of experimental points to prevent the experimental points from being overly concentrated in certain local areas, which would reduce the representativeness of the experimental results.

[0144] This formula can evenly select experimental points in the multidimensional parameter space and avoid excessive concentration of experimental points by introducing weighted distance and nonlinear optimization objective function. This can not only ensure the diversity and extensiveness of experimental data, but also reduce the waste of redundant experimental points and improve experimental efficiency. With the optimization of experimental point selection, the results of dependent variables (such as extraction rate, selectivity, etc.) will be closer to the global optimal solution, and the combination of independent variables (such as functional monomers, cross-linking agents, eluent dosage) will be able to explore the parameter space more comprehensively and improve the prediction ability of the model. This formula combines weighted distance, nonlinear objective function and local non-overlapping constraints to ensure the uniformity and diversity of experimental points in the parameter space. By optimizing the selection of experimental points, the number of experiments can be reduced, the experimental efficiency can be improved, and high-quality data support can be provided for subsequent modeling and optimization. This method is highly innovative and complex and is suitable for application scenarios that require high experimental accuracy.

[0145] Step S1222, using the N MIP extraction columns to respectively extract target coffee samples with different matrix complexity levels, and measuring performance indicators of the N MIP extraction columns; the performance indicators include a second extraction rate, selectivity, and anti-interference ability;

[0146] Step S1223, using the performance index as the response value and the MIP extraction column preparation parameters as the independent variable, a nonlinear regression model between the MIP extraction column preparation parameters and the performance index is established by using the response surface methodology to obtain a MIP response surface model;

[0147] Step S1224, solving the MIP response surface model to obtain the global optimal point of the second extraction rate, selectivity and anti-interference ability at each matrix complexity level, and obtaining the optimal functional monomer dosage, optimal cross-linking agent dosage and optimal eluent dosage of the target coffee sample at each matrix complexity level.

[0148] Specifically, in step S1220, the effects of three key preparation parameters, namely the amount of functional monomer, the amount of cross-linking agent and the amount of eluent, on the performance of the MIP extraction column are comprehensively investigated. The purpose is to determine the best combination of MIP extraction column preparation parameters based on target coffee samples of different matrix complexity levels. The selection of these parameters will directly affect the performance of the MIP, including extraction efficiency, selectivity and anti-interference ability. Ultimately, the goal is to provide a tailor-made MIP extraction column for each matrix complexity level.

[0149] Uniform design is an experimental design method that combines mathematical theory and statistical principles. It can evenly distribute experimental points in a multi-factor space with the least number of experiments, thereby significantly improving experimental efficiency. The core idea of ​​uniform design is: by scientifically selecting experimental points in a multidimensional space, the effects of different parameter combinations can be effectively evaluated and the interactions between parameters can be minimized. In step S1221, the amount of polymer precursor (the amount of functional monomer, cross-linking agent, and eluent) is used as the preparation parameter of the MIP extraction column, the value range of these parameters is set, and then N experimental points are selected by the uniform design method, that is, N different functional monomers, cross-linking agents, and eluent combinations are selected, so that N different MIP extraction columns can be prepared. These experimental points can cover the entire parameter space, ensuring that the effects of different parameter combinations on MIP performance can be evaluated. Through uniform design, the efficiency of the experiment is greatly improved without having to fully traverse all possible parameter combinations. For example, if three parameters need to be optimized (functional monomer, cross-linker, eluent), and each parameter has five choices, the number of combinations for a full experiment will be 125. A uniform design can reduce the number of experimental points to 20 or less, while still effectively evaluating the impact of different parameters.

[0150] In step S1222, the N MIP extraction columns are respectively tested for performance, and these MIP extraction columns are used to extract target coffee samples with different matrix complexity levels. The main performance indicators tested include:

[0151] Second extraction rate: This is an indicator to measure the efficiency of MIP extraction column in extracting target substances. A high second extraction rate means that the MIP extraction column can effectively extract the target substance, such as ochratoxin A, from the sample. An increase in the second extraction rate usually means that the MIP has a higher affinity for the target molecule and can effectively separate the target substance from a complex sample matrix.

[0152] Selectivity: Selectivity measures the specificity of the MIP column for the target molecule. Higher selectivity means that the MIP can preferentially bind to the target molecule and adsorb less other interfering components in the sample. MIP columns with high selectivity can help reduce matrix interference and improve the accuracy of analytical results.

[0153] Anti-interference ability: Anti-interference ability reflects the stability of MIP in complex matrices and its resistance to non-target substances. Complex matrices may contain a variety of interfering substances, such as other organic and inorganic substances. Strong anti-interference ability means that MIP can maintain a high extraction effect in a complex environment and is not easily affected by interfering substances.

[0154] These performance indicators will help evaluate the adaptability of different MIP extraction columns and provide a basis for subsequent optimization.

[0155] In step S1223, the response surface methodology (RSM) is used to establish a nonlinear regression model between the preparation parameters and performance indicators of the MIP extraction column. The response surface methodology is a multi-factor optimization technology. Through regression analysis, the nonlinear relationship between multiple independent variables (such as the amount of functional monomer, the amount of cross-linking agent and the amount of eluent) and dependent variables (such as the second extraction rate, selectivity, and anti-interference ability) can be found. The key to the response surface methodology is to explore the complex relationship between dependent and independent variables through mathematical modeling, and optimize the parameter combination to achieve the best performance.

[0156] By establishing a MIP response surface model, the performance prediction results of the MIP extraction column under different parameters can be obtained, thereby providing theoretical support for further parameter optimization. The model can help quickly determine which parameter combination can most effectively improve the performance of the MIP under the action of multiple factors. For example, if a response surface model is obtained, it can be known that at a specific cross-linking agent and functional monomer dosage, the selectivity and extraction rate will reach the optimal level.

[0157] In step S1224, the global optimal points of the second extraction rate, selectivity and anti-interference ability at each matrix complexity level are obtained by solving the MIP response surface model. The preparation parameter combinations corresponding to these optimal points are the optimal MIP extraction column preparation formulas for each matrix complexity level.

[0158] By solving the response surface model, the optimal combination of functional monomers, cross-linkers and eluents can be obtained for each matrix complexity level. These optimal parameters can ensure that the MIP extraction column has the best performance under different matrix complexities, minimize matrix interference, and improve the extraction efficiency of target molecules.

[0159] Step S1220 optimizes the selection of experimental points through uniform design, and sufficient information can be obtained with a smaller number of experiments, thereby significantly improving the efficiency of the experiment. Compared with traditional comprehensive experiments, uniform design greatly reduces the number of experiments, saves time and resources, and ensures the scientificity and accuracy of the experimental results. The regression model established by the response surface method can accurately find the optimal parameter combination of the MIP extraction column under different matrix complexities. This means that the most suitable MIP extraction column can be designed for each matrix complexity, thereby maximizing the selectivity, efficiency and anti-interference ability of the extraction. Traditional general-purpose MIP extraction columns may be interfered with in complex matrices, resulting in poor extraction effects. The graded customization scheme based on response surface analysis and uniform design can tailor MIP extraction columns for coffee samples of different complexities, reduce interference, and optimize the extraction efficiency of target substances, especially in the detection of trace toxins, which can improve sensitivity and accuracy.

[0160] For example, when processing coffee samples containing high concentrations of other organic matter, conventional MIP extraction columns may be affected by these interfering substances, resulting in poor extraction of ochratoxin A. By optimizing the preparation parameters of MIP using response surface methodology, a MIP extraction column tailored for this complex matrix can be obtained, which can improve the selectivity for ochratoxin A while reducing the interference of other components, thereby significantly improving the extraction effect and detection accuracy.

[0161] Step S1230, preparing a series of MIP extraction columns for target coffee samples with different matrix complexity levels according to the optimal MIP extraction column preparation parameter combination.

[0162] Specifically, in step S1230, a series of different MIP extraction columns are prepared according to the previously optimized MIP extraction column preparation parameter combination, aiming to meet the extraction requirements of target coffee samples with different matrix complexity levels. Each MIP extraction column will be customized to adapt to a coffee matrix of a specific complexity to ensure the best toxin extraction effect. Coffee samples with different matrix complexities have different requirements for MIP extraction columns. For example, simpler matrices (such as single-component coffee samples) have lower selectivity requirements for MIP extraction columns, while for complex coffee matrices (such as samples containing multiple interfering components), MIP extraction columns with higher selectivity and anti-interference ability are required. For each type of matrix complexity, the optimal parameters obtained by response surface analysis and uniform design will be applied in the preparation process. For example, in a high-complexity matrix, more cross-linking agents or different functional monomer ratios may be required to improve the selectivity and anti-interference ability of the extraction column. According to the optimized formula, MIP extraction columns are synthesized by traditional polymerization reactions or other suitable methods (such as solvent evaporation or sol-gel method). These extraction columns will show good selectivity for ochratoxin A with their specific functional monomer and cross-linking agent combinations.

[0163] Step S2000, constructing a toxin extraction dynamic prediction model, collecting a second multidimensional spectral time series of the coffee sample to be tested, and obtaining an extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectral time series and the toxin extraction dynamic prediction model; judging whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stopping the extraction process, otherwise continuing the extraction until the extraction rate prediction value reaches the extraction rate threshold;

[0164] Furthermore, step S2000 includes:

[0165] Step S2100, constructing a dynamic prediction model for toxin extraction;

[0166] Furthermore, if Figure 4 As shown, step S2100 includes:

[0167] Step S2110, extracting target coffee samples with different matrix complexity levels using corresponding MIP extraction columns, collecting extract samples at different time points, and determining the extraction rate;

[0168] Step S2120, during the extraction process of the target coffee sample, continuously collecting multidimensional spectral time series data of the extract, marked as a first multidimensional spectral time series; the multidimensional spectral time series data includes a transmission spectrum, a reflection spectrum, a three-dimensional fluorescence spectrum and a SERS spectrum;

[0169] Step S2130, extracting time series features from the first multidimensional spectral time series to obtain first spectral time series features, and matching them with the extraction rates at corresponding time points to form a sample data set;

[0170] Step S2140, constructing a toxin extraction dynamic prediction model using the first spectral time series feature in the sample data set as an independent variable and the extraction rate at the corresponding time point in the sample data set as a dependent variable.

[0171] Specifically, the goal of step S2100 is to establish a dynamic prediction model for toxin extraction to predict the toxin extraction rate of the coffee sample during the extraction process in real time. The model will be able to help more accurately grasp the progress of the extraction process, optimize operating parameters, and ensure the best extraction effect.

[0172] In step S2110, the target coffee samples with different matrix complexities are extracted using the corresponding MIP extraction columns, and the extract samples are collected at different time points to determine the extraction rate. The extraction rate is an indicator to measure the efficiency of the MIP extraction column in extracting the target substance (such as ochratoxin A) during the extraction process. Specifically, the extraction rate is defined as the ratio of the number of target molecules successfully extracted by the MIP extraction column to the number of target molecules in the initial sample within a certain period of time, usually expressed as a percentage.

[0173] Methods for determining extraction yield include:

[0174] 1. Extraction liquid sample collection: During the extraction process, samples are taken from the extraction liquid regularly.

[0175] 2. Measurement method: Determine the ochratoxin A concentration at each time point by high performance liquid chromatography (HPLC) or other appropriate analytical methods.

[0176] 3. Calculate the extraction rate: Calculate the extraction rate using the following formula:

[0177]

[0178] Through multiple sampling, the concentration changes of target molecules during the extraction process are monitored, and a curve of the extraction rate over time is drawn to help evaluate the efficiency of the extraction.

[0179] In step S2120, multi-dimensional spectral time series data is collected for the extraction process of the target coffee sample, which is marked as the first multi-dimensional spectral time series. The spectral data can be used to fully understand the changes in chemical components during the extraction process.

[0180] Multidimensional spectra include the following types:

[0181] Transmission and reflection spectra: Use a UV-Vis-NIR fiber optic spectrometer with a multi-channel fiber optic probe to collect the transmission and reflection spectra of the extract in real time. Transmission and reflection spectra mainly reflect the concentration changes of the target toxins and other components in the sample. For example, as the extraction process progresses, the concentration of ochratoxin A will increase, and changes in reflectivity and transmittance can reflect this process.

[0182] Three-dimensional fluorescence spectroscopy: Using a fluorescence spectrometer, the three-dimensional fluorescence spectrum of the extract is collected at multiple excitation wavelengths. The three-dimensional fluorescence spectrum provides a unique fingerprint for the toxin, which can effectively distinguish different compounds and has higher specificity.

[0183] SERS spectroscopy: Surface enhanced Raman scattering (SERS) technology is used to collect the Raman spectrum of the extract by configuring a SERS substrate. SERS technology can significantly enhance the Raman scattering signal of toxin molecules, thereby improving detection sensitivity, and is particularly suitable for monitoring trace toxins.

[0184] By collecting these multi-dimensional spectral time series data in real time, the changes in chemical composition during the extraction process can be fully characterized, thereby obtaining richer extraction information than a single spectrum.

[0185] In step S2130, time series feature extraction is performed on the first multi-dimensional spectral time series data. The purpose of this step is to extract representative features from the continuous spectral time series data so as to match them with the extraction rate and form a sample data set.

[0186] Common methods for time series feature extraction include:

[0187] Statistical features based on time windows: For example, the mean, standard deviation, maximum, minimum, and other statistics of the spectrum are calculated in each time period.

[0188] Frequency domain analysis: Convert the time domain signal to the frequency domain through Fourier transform to extract spectrum features such as peak frequency, bandwidth, etc.

[0189] Waveform features: Extract the morphological features of the spectral curve, such as rise time, fall time, peak position, etc.

[0190] These extracted features can comprehensively describe the trend of spectral data over time and correlate it with the extraction yield.

[0191] Step S2140 constructs a dynamic prediction model for toxin extraction based on the time series characteristics and the extraction rate data. The goal of the model is to predict the completion of the extraction process, especially the extraction rate of the target toxin, in real time.

[0192] Model structure and training:

[0193] Choose a suitable model: Common models include linear regression, support vector machine (SVM), artificial neural network (ANN), and random forest (RF). According to the characteristics of the data, choose a regression model that can handle time series data and multi-dimensional input.

[0194] Training model: The model is trained by collecting time series feature data and corresponding extraction rates at different time points. The training process includes data preprocessing, feature selection, model training, cross-validation and other steps.

[0195] Output prediction: The trained model can predict the extraction rate in the current extraction process in real time. Based on the prediction results, the extraction conditions are optimized to ensure the best extraction effect.

[0196] For example, assuming that an artificial neural network is selected as the model, by inputting the spectral features at different time points, the network can output the predicted value of the current extraction rate and adjust the extraction time or other parameters according to the predicted extraction rate.

[0197] Step S2100 can monitor the toxin extraction rate in real time during the extraction process by establishing a dynamic prediction model for toxin extraction, and adjust the operating parameters in time according to the prediction results of the model to improve the extraction efficiency. Multidimensional spectral data provides rich chemical information. Combined with chemometric methods, it can predict the extraction process in real time without waiting for complete experimental results, significantly improving detection efficiency. By establishing a prediction model, repeated tests in the experiment can be reduced, experimental costs can be reduced, and the reliability and stability of the extraction process can be improved. By monitoring the spectral data of the extract in real time and predicting the extraction rate, experimental failures caused by over-extraction or under-extraction can be effectively avoided, thereby improving the controllability and accuracy of the entire extraction process.

[0198] Step S2200, collecting a second multi-dimensional spectrum time series of the coffee sample to be tested, and obtaining an extraction rate prediction value of the coffee sample to be tested according to the second multi-dimensional spectrum time series and a toxin extraction dynamic prediction model;

[0199] Furthermore, if Figure 5 As shown, step S2200 includes:

[0200] Step S2210, obtaining basic physicochemical characteristic data of the coffee sample to be tested, and determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade according to the basic physicochemical characteristic data of the coffee sample to be tested; if the coffee sample to be tested is a sample of known grade, determining the matrix complexity level of the coffee sample to be tested;

[0201] Further, step S2210 includes:

[0202] Step S2211, constructing a feature vector of the sample to be tested according to the basic physical and chemical characteristic data of the coffee sample to be tested;

[0203] Step S2212, calculate the similarity between the feature vector of the sample to be tested and the L cluster center vectors, and obtain a similarity set {S1, S2, ..., S L};

[0204] The calculating of the similarity between the feature vector of the sample to be tested and the L cluster center vectors includes:

[0205]

[0206] :The characteristic vector of the sample to be tested and the The similarity of the cluster center vectors, .

[0207] : Dimension of the feature vector.

[0208] : The first feature vector of the sample to be tested feature components.

[0209] : No. The cluster center vector feature components.

[0210] : No. The importance weight of each feature component reflects the contribution of different features to the similarity calculation, which is usually determined by expert knowledge or data analysis.

[0211] : No. The distance measurement parameters of the feature components control the interaction and nonlinear relationship between different feature components, which are usually optimized through methods such as cross-validation.

[0212] : Ambient temperature (unit: Celsius).

[0213] : Standard ambient temperature (unit: degrees Celsius).

[0214] : The index of the cluster center vector.

[0215] : The influence coefficient of ambient temperature, reflecting the influence of temperature change on similarity calculation.

[0216] Introducing feature importance weights , the contribution of different feature components to the similarity calculation can be adjusted according to their importance. This helps to improve the rationality of the similarity calculation, especially when some feature components have a greater impact on the similarity, the weights of these features can be enhanced.

[0217] The normalized absolute difference metric is used in the similarity calculation:

[0218]

[0219] This measure ensures that the difference measure of each feature component is limited to , which facilitates the comparison and fusion of different features.

[0220] Introduced distance metric parameter , which can regulate the interactions and nonlinear relationships between different characteristic components. When , the feature components are independent; when When , it indicates that there is nonlinear correlation between features, which can better fit complex relationships.

[0221] The influence coefficient of ambient temperature is added to the formula:

[0222]

[0223] This coefficient reflects the impact of ambient temperature on similarity calculation. Close to standard temperature When the temperature deviates greatly from the standard temperature, the temperature influence coefficient decreases, thereby reducing the influence of ambient temperature on the similarity calculation and enhancing the robustness of the similarity calculation.

[0224] The sum of feature weights is introduced in the denominator of the formula:

[0225]

[0226] Through this normalization process, the similarities between different samples to be tested are comparable, which helps to unify the similarity calculation of the scale.

[0227] Distance metric parameters The value range is . It can be optimized by cross-validation and other methods. The specific method is to divide the data set into a training set and a validation set, and The cluster center vector is calculated by taking values, and then the similarity calculation effect is evaluated on the validation set, and the clustering performance is optimized. .

[0228] Ambient temperature influence coefficient The value range is ,in This means that temperature changes have little effect on similarity calculations. It indicates that temperature change has a great influence on similarity calculation. The best value can be obtained through historical data or expert experience, or by data fitting using regression analysis and other methods.

[0229] The similarity calculation formula comprehensively considers the influence of feature importance, nonlinear relationship of distance measurement and environmental factors. By introducing mechanisms such as importance weight, distance measurement parameter and temperature influence coefficient, the similarity calculation is made more accurate, flexible and adaptable to different environmental conditions. In practical applications, the reliability and accuracy of the model can be further improved by reasonably optimizing parameters and conducting sufficient verification.

[0230] Step S2213, from the similarity set {S1, S2, ..., S L}Select the maximum similarity S g , 1≤g≤L; if S g ≥S M , then the coffee sample to be tested is determined to be a sample of known grade; otherwise, the coffee sample to be tested is determined to be a sample of unknown grade; S M is the preset similarity threshold;

[0231] Step S2214: If the coffee sample to be tested is a sample of known grade, the matrix complexity grade corresponding to the cluster center vector having the highest similarity to the feature vector of the sample to be tested is used as the matrix complexity grade of the coffee sample to be tested.

[0232] Specifically, in step S2210, firstly, according to the basic physical and chemical characteristic data of the coffee sample to be tested, it is determined whether the sample to be tested belongs to a known grade sample or an unknown grade sample. The purpose of this process is to determine whether it belongs to the existing matrix complexity grade system by analyzing the characteristic data of the sample and determine its matrix complexity grade.

[0233] Known grade samples refer to samples that have been classified into the existing matrix complexity grading system through previous data clustering analysis. The physical and chemical characteristics of these samples have been fully collected and analyzed, and their matrix complexity has a clear classification in the existing system. For example, in coffee samples, if a batch of samples has been classified into matrix complexity grades and belongs to a specific grade (such as grade 1 or grade 2), then these samples are considered to be known grade samples. Unknown grade samples refer to samples whose matrix complexity level is not included in the existing matrix complexity grading system. These samples may be new varieties, new origins or new roasting degrees, etc., or they have not appeared in previous analysis for special reasons, so they do not belong to any level in the existing grading system. These samples require special processing and additional extraction analysis.

[0234] Step S2211 constructs a feature vector of the sample to be tested based on the basic physical and chemical characteristic data of the coffee sample to be tested. The feature vector of the sample to be tested is a set of numerical representations, which includes the key physical and chemical characteristics of the sample, which can help distinguish the matrix complexity of different samples. Specific basic physical and chemical characteristic data include variety, origin, roasting degree, crushed particle size, moisture content, etc.; these basic physical and chemical characteristic data will be converted into a multidimensional feature vector, in which each dimension represents a feature, and through numerical or standardized processing, ensure that they are compared on the same scale.

[0235] In step S2212, the similarity between the feature vector of the sample to be tested and the L cluster center vectors is calculated. Each cluster center vector represents a typical feature of a matrix complexity level, and calculating the similarity can help determine which level the sample to be tested is closest to.

[0236] Commonly used similarity calculation methods include:

[0237] Euclidean distance: used to measure the geometric distance between sample feature vectors. The smaller the distance, the more similar the samples are.

[0238] Cosine similarity: used to measure the angular similarity between sample feature vectors, ignoring the dimension size and only considering the similarity in direction. Cosine similarity is often used to deal with similarity problems in high-dimensional data.

[0239] In step S2213, from the similarity set {S1, S2, ..., S L} Select the maximum similarity value S g , and then compare it with the preset similarity threshold S M The comparison is performed to determine whether the sample to be tested belongs to a sample of known grade or a sample of unknown grade.

[0240] Known level samples: If the maximum similarity S g ≥SM , then the sample to be tested is determined to be a sample of known grade. At this time, the matrix complexity grade to which it belongs can be determined based on the similarity between the sample characteristics and the cluster center.

[0241] Unknown level sample: If the maximum similarity S g M , then the sample to be tested is determined to be a sample of unknown grade. At this time, the matrix complexity of the sample to be tested is not in the existing grade system.

[0242] Similarity threshold S M The choice of can be set through experimental data and experience, and is usually a value adjusted based on historical samples and expected error range.

[0243] In step S2214, if the sample to be tested is a sample of known grade, its matrix complexity grade is determined based on its feature vector and the cluster center vector with the highest similarity. Specifically, the coffee sample to be tested will be classified into the grade corresponding to the cluster center with the highest similarity.

[0244] Exemplarily, assuming that the maximum similarity between the sample to be tested and the cluster center vector is S1, and the cluster center vector corresponding to S1 belongs to the matrix complexity level 1, then the matrix complexity level of the sample to be tested is 1.

[0245] Step S2210 constructs a feature vector of the sample to be tested by acquiring basic physical and chemical characteristic data such as the variety, origin, roasting degree, crushed particle size and moisture content of the coffee sample, which can effectively classify and characterize the coffee sample preliminarily and ensure the accuracy of the sample in subsequent processing. By calculating the similarity between the feature vector of the sample to be tested and multiple cluster center vectors, the sample can be effectively classified. This can not only improve the accuracy of sample grade classification, but also reflect the characteristics of new samples by dynamically updating the cluster center. Similarity calculation is an efficient discrimination criterion that can automatically identify the similarity between the sample to be tested and the existing samples, thereby determining their grade. If the similarity is greater than the preset threshold, it can be confirmed that it belongs to a sample of known grade, reducing human errors and cumbersome manual operations. Similarity threshold S M The setting of can help to quickly determine whether a new sample belongs to the category of existing samples when it appears. If the similarity is low, it is marked as an unknown grade sample, which provides necessary space for further analysis or development of a new grade system. Set the similarity threshold (S M ) as the basis for judgment, ensuring that only when the similarity between the sample to be tested and the known sample is high enough, it is judged as a known grade sample. This provides a clear process for the processing of samples of unknown grade to avoid misjudgment.

[0246] ​Matching the sample to be tested with the cluster center vector with the highest similarity can assign the closest matrix complexity level to the sample, ensuring the rationality of the grade division. This helps to optimize subsequent extraction, processing and other treatment processes, and improve overall processing efficiency and effectiveness. Selecting the maximum similarity based on the similarity set and further determining the matrix complexity level of the sample to be tested can significantly improve the accuracy of sample classification and ensure that the subsequent processing of the sample is more in line with its actual characteristics. By combining new samples with cluster analysis of existing data sets, the matrix complexity judgment of the new samples can be adjusted and updated in real time, and the evaluation criteria can be continuously optimized to adapt to the increasingly diverse characteristics of coffee samples. When judging samples of unknown grades, the cluster centers can be adjusted and optimized by continuously accumulating and updating sample data, so as to dynamically adapt to new samples. This ensures the flexibility of the system and enables continuous improvement and updating of the grade system.

[0247] Step S2220, extracting the coffee sample to be tested;

[0248] Furthermore, if Figure 6 As shown, step S2220 includes:

[0249] Step S2221, if the coffee sample to be tested is a sample of known grade, then according to the matrix complexity level of the coffee sample to be tested, a corresponding MIP extraction column is used to extract the coffee sample to be tested;

[0250] Step S2222: If the coffee sample to be tested is a sample of unknown grade, extraction is performed using a MIP extraction column corresponding to the matrix complexity level L.

[0251] Specifically, in step S2220, according to the determination of the complexity level of the matrix of the coffee sample to be tested in step S2210, the corresponding MIP extraction column is used to extract the sample. The core purpose of this process is to selectively remove the target toxin (such as ochratoxin A) from the coffee sample through the MIP (molecular imprinting polymer) extraction column and ensure the optimization of the extraction effect. In order to cope with the matrix complexity of different samples, the MIP extraction column and operation method adopted will also be different.

[0252] If the coffee sample to be tested is a known grade sample, the appropriate MIP extraction column will be selected for extraction according to the matrix complexity level of the sample to be tested. This means that it is already known that the sample belongs to a certain matrix complexity level, so the extraction column most suitable for this level can be selected to obtain the best extraction effect.

[0253] Specifically, the characteristics and matrix complexity of known grade samples have been pre-classified and clarified. For example, coffee samples may have been classified as grade 1, grade 2, or grade 3, etc. based on their characteristics such as variety, origin, and roasting degree. Each grade of matrix complexity has different requirements. Simpler matrices (such as single-component coffee samples) may only require basic MIP extraction columns, while more complex matrices (such as samples containing multiple impurities or interfering components) require more selective MIP extraction columns. According to the matrix complexity of the sample, select a MIP extraction column designed specifically for that complexity. For example, samples with lower matrix complexity (such as lightly roasted, single-origin coffee) can use standard MIP extraction columns, while samples with higher complexity (such as dark roasted, coffee containing multiple impurities) may require more selective MIP extraction columns.

[0254] Step S2221 ensures that the matrix complexity of each sample is fully considered by selecting the MIP extraction column in a targeted manner, thereby achieving efficient and accurate extraction. Selecting a suitable extraction column can improve the affinity of the MIP for the target toxin, reduce interference from other components, and enhance the selectivity of the extraction column.

[0255] For example, suppose the matrix complexity of a coffee sample is level 2, which means that the sample contains a certain degree of impurities and interfering components. In this case, a highly selective MIP extraction column is selected, which is specially designed to remove toxins from complex matrices and has a high resistance to interference from other impurities.

[0256] If the coffee sample to be tested is judged as an unknown grade sample, that is, its matrix complexity cannot be found in the existing system. In this case, the MIP extraction column corresponding to the matrix complexity level L will be selected for extraction. Level L represents the highest complexity level in the existing grading system. The purpose of selecting this level of MIP extraction column is to ensure that even if the complexity of the sample exceeds the existing classification, it can cope with potential complexity and risks to the greatest extent.

[0257] Specifically, the matrix complexity of samples of unknown grade cannot be matched with the existing grade system, perhaps because their characteristics are relatively special or they are new sample types. For example, the matrix complexity of some new coffee varieties, origins or roasting degrees may not have appeared in previous sample analysis. In order to deal with this unknown risk and complexity, a MIP extraction column corresponding to the matrix complexity level L is selected. Level L usually represents the highest complexity level in the existing system. The use of this extraction column can minimize matrix interference and ensure efficient extraction of target toxins.

[0258] Step S2222 uses a level L MIP extraction column, which can effectively cope with the complexity of unknown samples and avoid the reduction of extraction effect due to the sample characteristics not matching the existing level. By selecting the MIP extraction column with the highest complexity, it can be ensured that even if the sample contains multiple interfering components, the target toxin can still be extracted efficiently.

[0259] For example, suppose a new coffee sample is encountered, and its variety and origin have no corresponding records in the existing database. Through the similarity calculation of the feature vector, it is found that its similarity with any existing level of matrix complexity is low, so it is judged as an unknown level sample. In order to ensure the extraction effect, a MIP extraction column with a matrix complexity level of L is selected, so that even if the matrix of the sample is more complex, the extraction effect can still be guaranteed.

[0260] Step S2230, continuously collecting multi-dimensional spectral time series data of the extract during the extraction process of the coffee sample to be tested, marked as a second multi-dimensional spectral time series; collecting conductivity time series data of the extract;

[0261] Step S2240, extracting time series features from the second multi-dimensional spectrum time series to obtain second spectrum time series features;

[0262] Step S2250: input the second spectral time series feature into the toxin extraction dynamic prediction model to obtain the extraction rate prediction value of the coffee sample to be tested.

[0263] Specifically, multidimensional spectral time series data refers to the data that combines multiple spectral data (such as transmission spectra, reflectance spectra, fluorescence spectra, and Raman spectra) collected over time during the extraction process with time series. As the extraction proceeds, the spectral data will change, reflecting the changes in the chemical composition of the sample, especially the concentration changes of target toxins (such as ochratoxin A). Multidimensional spectral time series data includes transmission spectra, reflectance spectra, three-dimensional fluorescence spectra, and Raman spectra; multidimensional spectral time series data provides an efficient way to monitor the changes in chemical composition during the extraction process in real time, especially the monitoring of target toxins. By combining different types of spectral data, sample changes can be analyzed from multiple angles, which is more comprehensive and accurate than a single spectrum.

[0264] Conductivity is an important indicator of ion concentration in the sample solution and is usually closely related to the concentration of dissolved substances (such as dissolved toxins or impurities) in the sample. During the extraction process, changes in conductivity can reflect changes in the concentration of dissolved substances in the solution and indirectly reflect the extraction progress of the target toxin.

[0265] Step S2240 extracts time series features from the second multidimensional spectral time series data. The purpose of time series feature extraction is to extract representative information from continuous spectral data. These features can help capture key changes in the extraction process, thereby more accurately predicting the extraction rate.

[0266] The content of time series feature extraction includes:

[0267] Statistical features: such as mean, standard deviation, maximum, minimum, peak, etc. These features can describe the trend and volatility of spectral signals over time. For example, during the extraction process, the concentration change of the target toxin is usually manifested as a peak change or an increase in absorption in certain bands.

[0268] Frequency domain features: By performing Fourier transform on time series data, we can obtain features in the frequency domain, such as frequency components, fluctuation amplitude, etc. These features can reveal periodic or repetitive change patterns in spectral signals.

[0269] Dynamic change characteristics: including the signal's rise time, fall time, slope, peak position change, etc., reflecting the changing dynamics of the target substance concentration during the extraction process.

[0270] For example, in the collected fluorescence spectrum time series data, the concentration change of toxins may appear as an increase or decrease in fluorescence intensity at a specific wavelength. Extracting these features (such as the rate of change of fluorescence intensity, peak position, etc.) can help identify the changing trend of toxins.

[0271] Time series feature extraction can help extract useful information from complex spectral data, especially in the process of dynamic changes, and can accurately capture the changing trend of toxin concentration. Through feature extraction, a small number of representative and predictive features can be extracted from a large amount of spectral data, reducing the computational complexity of subsequent analysis.

[0272] In step S2250, the second spectral time series feature is input into the toxin extraction dynamic prediction model to obtain the extraction rate prediction value of the coffee sample to be tested. The toxin extraction dynamic prediction model is a model established based on machine learning or statistical regression methods, which aims to predict the change of target toxin concentration during the extraction process through time series features, so as to obtain a real-time prediction of the extraction rate. The predicted extraction rate can help dynamically adjust the operating conditions during the extraction process, such as extraction time, solvent amount, etc., to improve the extraction efficiency and accuracy.

[0273] Step S2300, determining whether the predicted extraction rate value of the coffee sample to be tested reaches a preset extraction rate threshold; if it reaches the extraction rate threshold, the extraction process is stopped, otherwise the extraction is continued until the predicted extraction rate value reaches the extraction rate threshold.

[0274] Specifically, in step S2300, it is determined whether the predicted extraction rate of the coffee sample to be tested has reached a preset extraction rate threshold. The purpose of this determination is to determine whether the extraction needs to be continued or whether the extraction process can be stopped.

[0275] The preset extraction rate threshold is a value set according to experimental data and target toxin concentration requirements, indicating the minimum extraction rate of the target toxin. Generally speaking, this value is set according to actual needs, usually a critical value of toxin removal rate. For example, setting the threshold to 95% means that the extraction process can be stopped when the extraction rate reaches 95%. If the predicted extraction rate has reached or exceeded the extraction rate threshold, the extraction is stopped. If the predicted extraction rate does not reach the extraction rate threshold, the extraction continues until the preset extraction rate is reached. By predicting the extraction rate in real time and comparing it with the preset threshold, the extraction progress can be accurately controlled to avoid over-extraction or under-extraction, thereby improving the overall extraction efficiency. Avoid ineffective over-extraction and save time and resources. Only continue extraction when the extraction effect does not meet the standard to ensure the optimal use of resources.

[0276] Step S3000: After the extraction of the coffee sample to be tested is completed, the concentration of ochratoxin A in the extract is determined. A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

[0277] Furthermore, step S3000 includes:

[0278] Step S3100, if C A ≥C θ , it is determined that the extraction process has achieved the expected goal;

[0279] Step S3200, if C A <C θ , it is determined that the extraction process has not achieved the expected goal and closed-loop feedback control is required.

[0280] The closed-loop feedback control comprises:

[0281] According to the conductivity time series data of the extract during the extraction process of the coffee sample to be tested, the dynamic conductivity curve of the MIP extraction column is established; the statistical process control method is used to determine whether the performance of the MIP extraction column is abnormal based on the dynamic conductivity curve. If there is no abnormality, it is determined whether the adsorption capacity of the MIP extraction column is saturated. If not, the extraction rate threshold is increased and the amount of eluent is increased; if saturated, the MIP extraction column is replaced, and the extraction rate threshold is increased and the amount of eluent is increased; if the performance of the MIP extraction column is abnormal, the MIP extraction column is replaced.

[0282] The method for judging whether the adsorption capacity of the MIP extraction column is saturated comprises:

[0283] The curve slope of the dynamic conductivity curve is calculated. When the curve slope is lower than a preset slope threshold, it is determined that the adsorption capacity of the MIP extraction column has reached saturation; otherwise, it is determined that the adsorption capacity of the MIP extraction column is not saturated.

[0284] Specifically, in step S3000, after the extraction of the coffee sample to be tested is completed, the concentration of ochratoxin A (OTA) in the extract needs to be measured, and the concentration C A With the preset concentration threshold C θ Compare and judge whether the extraction process has achieved the expected goal.

[0285] Ochratoxin A is a common mycotoxin, which is usually quantitatively determined by high performance liquid chromatography (HPLC), enzyme-linked immunosorbent assay (ELISA) and other methods. Determination of the concentration of ochratoxin A in the extract C A , which can reflect the removal efficiency of the target toxin and provide a basis for the next step of judgment. By accurately measuring the concentration of ochratoxin A, it can be determined whether the extraction process has achieved the expected effect of removing the target toxin. If the concentration of ochratoxin A in the extract fails to reach the expected concentration level, it will enter the closed-loop feedback control link and adjust the extraction process to ensure the removal effect of the target toxin.

[0286] In step S3100, the concentration of ochratoxin A in the coffee sample extract to be tested is C A With the preset concentration threshold C θ For comparison:

[0287] If C A ≥C θ , it is judged that the extraction process has achieved the expected goal, which means that the removal efficiency of ochratoxin A is sufficient and the extraction process can be terminated.

[0288] If C A <C θ, it is determined that the extraction process has not reached the expected goal, the extraction rate is insufficient, and closed-loop feedback control is required.

[0289] By monitoring the concentration of ochratoxin A in real time, it is possible to determine whether the extraction has reached the expected target and avoid excessive or insufficient extraction. If the expected target is reached, the extraction process can be stopped to avoid waste of resources. If the target is not reached, the operation is adjusted through feedback control to ensure the best results.

[0290] In step S3200, if the ochratoxin A concentration C A If the expected target is not achieved, the closed-loop feedback control is entered to adjust the extraction process. This process includes the following key steps:

[0291] 1. Establish the dynamic conductivity curve of MIP extraction column

[0292] Conductivity is a sign of ion concentration in a solution. During the extraction process, conductivity changes with the concentration of the target toxin (such as ochratoxin A). By collecting conductivity time series data in real time, a dynamic conductivity curve can be established, which reflects the changing trend of the concentration of substances in the solution during the extraction process.

[0293] Dynamic conductivity curve: By continuously measuring the conductivity and recording the data, a conductivity curve can be constructed over time. As the concentration of ochratoxin A changes, the changes in conductivity will help understand the progress of the extraction process.

[0294] 2. Statistical Process Control (SPC)

[0295] After the conductivity curve is established, the curve is analyzed using statistical process control (SPC) methods. Statistical process control is a statistically based method used to monitor and control the quality of the process to ensure that the process is stable and meets the expected goals.

[0296] SPC method: mainly monitors process changes by setting control charts and evaluates whether there are any abnormalities in the process. Compare the dynamic conductivity curve with the control chart to determine whether the extraction process is normal.

[0297] Through the conductivity curve and SPC method, it is possible to evaluate in real time whether there are any abnormalities in the extraction process and ensure process stability. Statistical process control helps to detect and correct fluctuations in the process in a timely manner and prevent errors caused by improper operation.

[0298] 3. Determine whether the adsorption capacity of the MIP extraction column is saturated

[0299] During the extraction process, the adsorption capacity of the MIP extraction column will gradually decrease with the amount of toxin adsorbed. The slope of the conductivity curve can be used to determine whether the MIP extraction column has reached saturation.

[0300] Slope judgment method: Calculate the slope change of the conductivity curve. When the slope of the curve decreases significantly, it means that the adsorption capacity is close to saturation, and further addition of eluent or toxin will no longer significantly change the conductivity.

[0301] By monitoring the slope change, it is possible to promptly determine whether the extraction column has reached adsorption saturation, prevent continued extraction from being ineffective, and improve extraction efficiency.

[0302] 4. Adjust operating conditions according to adsorption capacity status

[0303] If the adsorption capacity of the MIP extraction column has not yet been saturated, the amount of toxin extracted can be increased by raising the extraction rate threshold and increasing the amount of eluent, thereby accelerating the extraction process and ensuring that the desired goal is achieved.

[0304] If the MIP extraction column is saturated, it means that its adsorption capacity for the target toxin has reached its limit and the extraction rate cannot be further improved. In this case, the MIP extraction column must be replaced, and the extraction rate threshold and eluent dosage must be increased to improve the extraction effect.

[0305] 5. Treatment of abnormal performance of MIP extraction column

[0306] Abnormal performance: If the SPC method finds that the performance of the MIP extraction column is abnormal (for example, the conductivity curve fluctuates too much or does not meet expectations), the MIP extraction column needs to be replaced immediately and the extraction process needs to be restarted.

[0307] By monitoring the adsorption capacity of the MIP extraction column, poor extraction results due to degradation of the extraction column performance can be prevented, ensuring the effectiveness of each step of the operation. Through closed-loop feedback control, operating conditions can be dynamically adjusted based on real-time data to maximize extraction efficiency and ensure that the removal of target toxins reaches the expected level.

[0308] In step S3000, by combining conductivity monitoring, statistical process control (SPC) and performance analysis of the MIP extraction column, the progress of the extraction process can be effectively judged and dynamically adjusted to ensure that the final extraction reaches the expected goal. Closed-loop feedback control ensures that each round of extraction can be carried out under optimal conditions by adjusting operating conditions in real time, thereby maximizing the removal effect of the target toxins and improving the overall extraction efficiency.

[0309] Example 2

[0310] This embodiment provides an extraction system for extracting ochratoxin A toxin in coffee using acetonitrile based on embodiment 1, such as Figure 7 As shown, including:

[0311] Grade system building module: used to collect basic physicochemical characteristic data of target coffee samples, and divide the target coffee samples into L matrix complexity levels according to the basic physicochemical characteristic data;

[0312] Extraction column preparation module: using ochratoxin A as the template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing MIP extraction columns for target coffee samples with different matrix complexity levels;

[0313] Prediction module: used to construct a dynamic prediction model for toxin extraction, collect the second multidimensional spectrum time series of the coffee sample to be tested, and obtain the extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectrum time series and the toxin extraction dynamic prediction model; determine whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stop the extraction process, otherwise continue the extraction until the extraction rate prediction value reaches the extraction rate threshold;

[0314] Evaluation module: After the extraction of the coffee sample to be tested is completed, the concentration of ochratoxin A in the extract is measured C A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

[0315] In the grade system building module, the classification of the target coffee sample into L matrix complexity levels includes:

[0316] Step S1110, constructing a sample feature vector according to the basic physical and chemical characteristic data of the target coffee sample; the basic physical and chemical characteristic data includes the variety, origin, roasting degree, crushed particle size and moisture content of the coffee sample;

[0317] Step S1120, performing cluster analysis on the sample feature vectors, and dividing the target coffee samples into L matrix complexity levels.

[0318] The step S1120 includes:

[0319] Step S1121, determining the optimal number of clusters L, and dividing the sample feature vector into L clusters;

[0320] Step S1122, calculating the characteristic mean of samples within each cluster, generating L cluster center vectors, each representing a typical feature of the L matrix complexity levels;

[0321] Step S1123, sort the L clusters from low to high according to their feature means, and generate a matrix complexity level sequence {1, 2, ..., L}.

[0322] In the extraction column preparation module, the polymer precursor includes a functional monomer, a cross-linking agent and an eluent; the preparation of the MIP extraction column for target coffee samples with different matrix complexity levels includes:

[0323] Step S1210, acetonitrile is selected as the eluent, methacrylic acid is selected as the functional monomer, and ethylene glycol dimethacrylate is selected as the cross-linking agent;

[0324] Step S1220, comprehensively investigating the effect of the amount of polymer precursor on the performance of the MIP extraction column, and obtaining the optimal MIP extraction column preparation parameter combination for the target coffee sample at different matrix complexity levels; the optimal MIP extraction column preparation parameter combination includes the optimal functional monomer amount, the optimal cross-linking agent amount, and the optimal eluent amount;

[0325] Step S1230, preparing a series of MIP extraction columns for target coffee samples with different matrix complexity levels according to the optimal MIP extraction column preparation parameter combination.

[0326] The step S1220 includes:

[0327] Step S1221, using the amount of polymer precursor as a MIP extraction column preparation parameter, selecting N experimental points within the value range of the MIP extraction column preparation parameter, and preparing N MIP extraction columns; the MIP extraction column preparation parameters include the amount of functional monomer, the amount of cross-linking agent, and the amount of eluent;

[0328] Step S1222, using the N MIP extraction columns to respectively extract target coffee samples with different matrix complexity levels, and measuring performance indicators of the N MIP extraction columns; the performance indicators include a second extraction rate, selectivity, and anti-interference ability;

[0329] Step S1223, using the performance index as the response value and the MIP extraction column preparation parameters as the independent variable, a nonlinear regression model between the MIP extraction column preparation parameters and the performance index is established by using the response surface methodology to obtain a MIP response surface model;

[0330] Step S1224, solving the MIP response surface model to obtain the global optimal point of the second extraction rate, selectivity and anti-interference ability at each matrix complexity level, and obtaining the optimal functional monomer dosage, optimal cross-linking agent dosage and optimal eluent dosage of the target coffee sample at each matrix complexity level.

[0331] In the prediction module, the construction of a dynamic prediction model for toxin extraction includes:

[0332] Step S2110, extracting target coffee samples with different matrix complexity levels using corresponding MIP extraction columns, collecting extract samples at different time points, and determining the extraction rate;

[0333] Step S2120, during the extraction process of the target coffee sample, continuously collecting multidimensional spectral time series data of the extract, marked as a first multidimensional spectral time series; the multidimensional spectral time series data includes a transmission spectrum, a reflection spectrum, a three-dimensional fluorescence spectrum and a SERS spectrum;

[0334] Step S2130, extracting time series features from the first multidimensional spectral time series to obtain first spectral time series features, and matching them with the extraction rates at corresponding time points to form a sample data set;

[0335] Step S2140, constructing a toxin extraction dynamic prediction model using the first spectral time series feature in the sample data set as an independent variable and the extraction rate at the corresponding time point in the sample data set as a dependent variable.

[0336] In the prediction module, obtaining the predicted value of the extraction rate of the coffee sample to be tested includes:

[0337] Step S2210, obtaining basic physicochemical characteristic data of the coffee sample to be tested, and determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade according to the basic physicochemical characteristic data of the coffee sample to be tested; if the coffee sample to be tested is a sample of known grade, determining the matrix complexity level of the coffee sample to be tested;

[0338] Step S2220, extracting the coffee sample to be tested;

[0339] Step S2230, continuously collecting multi-dimensional spectral time series data of the extract during the extraction process of the coffee sample to be tested, marked as a second multi-dimensional spectral time series; collecting conductivity time series data of the extract;

[0340] Step S2240, extracting time series features from the second multi-dimensional spectrum time series to obtain second spectrum time series features;

[0341] Step S2250: input the second spectral time series feature into the toxin extraction dynamic prediction model to obtain the extraction rate prediction value of the coffee sample to be tested.

[0342] The step S2210 includes:

[0343] Step S2211, constructing a feature vector of the sample to be tested according to the basic physical and chemical characteristic data of the coffee sample to be tested;

[0344] Step S2212, calculate the similarity between the feature vector of the sample to be tested and the L cluster center vectors, and obtain a similarity set {S1, S2, ..., SL};

[0345] Step S2213, from the similarity set {S1, S2, ..., S L}Select the maximum similarity S g , 1≤g≤L; if S g ≥S M , then the coffee sample to be tested is determined to be a sample of known grade; otherwise, the coffee sample to be tested is determined to be a sample of unknown grade; S M is the preset similarity threshold;

[0346] Step S2214: If the coffee sample to be tested is a sample of known grade, the matrix complexity grade corresponding to the cluster center vector having the highest similarity to the feature vector of the sample to be tested is used as the matrix complexity grade of the coffee sample to be tested.

[0347] The step S2220 includes:

[0348] Step S2221, if the coffee sample to be tested is a sample of known grade, then according to the matrix complexity level of the coffee sample to be tested, a corresponding MIP extraction column is used to extract the coffee sample to be tested;

[0349] Step S2222: If the coffee sample to be tested is a sample of unknown grade, extraction is performed using a MIP extraction column corresponding to the matrix complexity level L.

[0350] In the evaluation module, the concentration of ochratoxin A C A and the preset concentration threshold C θ , to determine whether the extraction process has achieved the expected goals include:

[0351] Step S3100, if C A ≥C θ , it is determined that the extraction process has achieved the expected goal;

[0352] Step S3200, if C A <C θ , it is determined that the extraction process has not reached the expected goal and closed-loop feedback control is required;

[0353] The closed-loop feedback control comprises:

[0354] According to the conductivity time series data of the extract during the extraction process of the coffee sample to be tested, the dynamic conductivity curve of the MIP extraction column is established; the statistical process control method is used to determine whether the performance of the MIP extraction column is abnormal based on the dynamic conductivity curve. If there is no abnormality, it is determined whether the adsorption capacity of the MIP extraction column is saturated. If not, the extraction rate threshold is increased and the amount of eluent is increased; if saturated, the MIP extraction column is replaced, and the extraction rate threshold is increased and the amount of eluent is increased; if the performance of the MIP extraction column is abnormal, the MIP extraction column is replaced.

[0355] The method for judging whether the adsorption capacity of the MIP extraction column is saturated comprises:

[0356] The curve slope of the dynamic conductivity curve is calculated. When the curve slope is lower than a preset slope threshold, it is determined that the adsorption capacity of the MIP extraction column has reached saturation; otherwise, it is determined that the adsorption capacity of the MIP extraction column is not saturated.

[0357] Example 3

[0358] This embodiment discloses an electronic device, which may include one or more processors and one or more memories. The memories store computer-readable codes, which, when executed by the one or more processors, may execute the above-described method for extracting ochratoxin A toxin from coffee using acetonitrile.

[0359] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store the extraction method of ochratoxin A toxin in coffee using acetonitrile provided in the present application. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction may, for example, include: collecting basic physicochemical characteristic data of the target coffee sample, and dividing the target coffee sample into L matrix complexity levels according to the basic physicochemical characteristic data; using ochratoxin A as a template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing a MIP extraction column for target coffee samples with different matrix complexity levels; constructing a toxin extraction dynamic prediction model, collecting a second multidimensional spectral time series of the coffee sample to be tested, and obtaining an extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectral time series and the toxin extraction dynamic prediction model; judging whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stopping the extraction process, otherwise continuing the extraction until the extraction rate prediction value reaches the extraction rate threshold; after the extraction of the coffee sample to be tested is completed, determining the concentration C of ochratoxin A in the extract A ; According to the concentration of ochratoxin A CA and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

[0360] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0361] Example 4

[0362] This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the method for extracting ochratoxin A toxin in coffee using acetonitrile extraction according to the embodiment of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0363] In addition, according to an embodiment of the present application, the process described above with reference to the flowchart may be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: collecting basic physicochemical characteristic data of the target coffee sample, and dividing the target coffee sample into L matrix complexity levels according to the basic physicochemical characteristic data; using ochratoxin A as a template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing MIP extraction columns for target coffee samples with different matrix complexity levels; constructing a toxin extraction dynamic prediction model, collecting a second multidimensional spectral time series of the coffee sample to be tested, and obtaining an extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectral time series and the toxin extraction dynamic prediction model; judging whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stopping the extraction process, otherwise continuing the extraction until the extraction rate prediction value reaches the extraction rate threshold; after the extraction of the coffee sample to be tested is completed, determining the concentration C of ochratoxin A in the extract A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ, determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required. When the computer program is executed by the central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0364] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0365] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0366] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for extracting ochratoxin A toxin from coffee using acetonitrile, characterized in that: The method comprises: The basic physicochemical characteristic data of the target coffee samples are collected, and the target coffee samples are divided into L matrix complexity levels according to the basic physicochemical characteristic data; Ochratoxin A is used as a template molecule, the type of polymerization precursor is selected, the amount of polymerization precursor is optimized, and MIP extraction columns for target coffee samples with different matrix complexity levels are prepared; Constructing a toxin extraction dynamic prediction model, collecting a second multidimensional spectral time series of the coffee sample to be tested, and obtaining an extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectral time series and the toxin extraction dynamic prediction model; judging whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stopping the extraction process, otherwise continuing the extraction until the extraction rate prediction value reaches the extraction rate threshold; After the extraction of the coffee sample is completed, the concentration of ochratoxin A in the extract is determined. A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

2. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 1, characterized in that: The method of dividing the target coffee sample into L matrix complexity levels includes: Construct a sample feature vector based on the basic physical and chemical characteristic data of the target coffee sample; Determine the optimal number of clusters L and divide the sample feature vectors into L clusters; Calculate the feature mean of samples within each cluster and generate L cluster center vectors; The L clusters are sorted from low to high according to their feature means to generate a matrix complexity level sequence {1, 2, ..., L}.

3. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 1, characterized in that: The polymer precursor includes a functional monomer, a cross-linking agent and an eluent; The selection of the type of polymer precursor includes: acetonitrile is selected as the eluent, methacrylic acid is selected as the functional monomer, and ethylene glycol dimethacrylate is selected as the cross-linking agent; The MIP extraction column for preparing target coffee samples with different matrix complexity levels comprises: The effect of the amount of polymer precursor on the performance of the MIP extraction column was comprehensively investigated to obtain the optimal MIP extraction column preparation parameter combination for the target coffee sample at different matrix complexity levels; the optimal MIP extraction column preparation parameter combination includes the optimal functional monomer amount, the optimal cross-linking agent amount and the optimal eluent amount; According to the optimal MIP extraction column preparation parameter combination, a series of MIP extraction columns were prepared for target coffee samples with different matrix complexity levels.

4. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 3, characterized in that: The optimal MIP extraction column preparation parameter combination for obtaining target coffee samples at different matrix complexity levels includes: The amount of polymer precursor is used as a parameter for preparing a MIP extraction column, and N experimental points are selected within the value range of the parameter for preparing a MIP extraction column to prepare N MIP extraction columns; the MIP extraction column preparation parameters include the amount of functional monomer, the amount of cross-linking agent, and the amount of eluent; Using the N MIP extraction columns to extract target coffee samples with different matrix complexity levels, respectively, and measuring the performance indicators of the N MIP extraction columns; the performance indicators include second extraction rate, selectivity and anti-interference ability; With the performance index as the response value and the preparation parameters of MIP extraction column as the independent variable, the response surface methodology was used to establish the nonlinear regression model between the preparation parameters of MIP extraction column and the performance index, and the MIP response surface model was obtained. The MIP response surface model was solved to obtain the global optimal point of the second extraction rate, selectivity and anti-interference ability at each matrix complexity level, and the optimal functional monomer dosage, optimal cross-linking agent dosage and optimal eluent dosage of the target coffee sample at each matrix complexity level were obtained.

5. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 3, characterized in that: The construction of a dynamic prediction model for toxin extraction comprises: The target coffee samples with different matrix complexity levels were extracted using the corresponding MIP extraction columns, and the extract samples were collected at different time points to determine the extraction rate; During the extraction process of the target coffee sample, multi-dimensional spectral time series data of the extract is continuously collected, which is marked as a first multi-dimensional spectral time series; the multi-dimensional spectral time series data includes a transmission spectrum, a reflection spectrum, a three-dimensional fluorescence spectrum and a SERS spectrum; Extracting time series features from the first multidimensional spectral time series to obtain first spectral time series features, and matching them with extraction rates at corresponding time points to form a sample data set; A dynamic prediction model for toxin extraction was constructed with the first spectral time series feature in the sample data set as the independent variable and the extraction rate at the corresponding time point in the sample data set as the dependent variable.

6. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 5, characterized in that: The method of obtaining the predicted value of the extraction rate of the coffee sample to be tested includes: Extract the coffee sample to be tested; Continuously collect multi-dimensional spectral time series data of the extract during the extraction process of the coffee sample to be tested, marked as the second multi-dimensional spectral time series; collect conductivity time series data of the extract; Extracting time series features from the second multi-dimensional spectrum time series to obtain second spectrum time series features; The second spectral time series feature is input into the toxin extraction dynamic prediction model to obtain the extraction rate prediction value of the coffee sample to be tested.

7. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 6, characterized in that: The extracting of the coffee sample to be tested comprises: Obtaining basic physicochemical characteristic data of the coffee sample to be tested, and determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade according to the basic physicochemical characteristic data of the coffee sample to be tested; if the coffee sample to be tested is a sample of known grade, determining the matrix complexity level of the coffee sample to be tested; If the coffee sample to be tested is a sample of known grade, then according to the matrix complexity level of the coffee sample to be tested, a corresponding MIP extraction column is used to extract the coffee sample to be tested; If the coffee sample to be tested is of unknown grade, a MIP extraction column corresponding to the matrix complexity level L is used for extraction.

8. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 7, characterized in that: The step of determining whether the coffee sample to be tested is a sample of known grade or a sample of unknown grade comprises: According to the basic physical and chemical characteristic data of the coffee sample to be tested, a feature vector of the sample to be tested is constructed; Calculate the similarity between the feature vector of the sample to be tested and the L cluster center vectors, and obtain the similarity set {S1, S2, ..., S L }; From the similarity set {S1,S2,...,S L }Select the maximum similarity S g , 1≤g≤L; if S g ≥S M , then the coffee sample to be tested is determined to be a sample of known grade; otherwise, the coffee sample to be tested is determined to be a sample of unknown grade; S M is the preset similarity threshold; Determining the matrix complexity level of the coffee sample to be tested includes: If the coffee sample to be tested is a sample of known grade, the matrix complexity grade corresponding to the cluster center vector having the highest similarity to the feature vector of the sample to be tested is taken as the matrix complexity grade of the coffee sample to be tested.

9. The method for extracting ochratoxin A toxin from coffee using acetonitrile extraction according to claim 6, characterized in that: Determining whether the extraction process achieves the expected goal includes: If C A ≥C θ , it is judged that the extraction process has achieved the expected goal; If C A <C θ , it is determined that the extraction process has not achieved the expected goal and closed-loop feedback control is required; The closed-loop feedback control comprises: According to the conductivity time series data of the extract during the extraction process of the coffee sample to be tested, the dynamic conductivity curve of the MIP extraction column is established; the statistical process control method is used to determine whether the performance of the MIP extraction column is abnormal based on the dynamic conductivity curve. If there is no abnormality, it is determined whether the adsorption capacity of the MIP extraction column is saturated. If not, the extraction rate threshold is increased and the amount of eluent is increased; if saturated, the MIP extraction column is replaced, and the extraction rate threshold is increased and the amount of eluent is increased; if the performance of the MIP extraction column is abnormal, the MIP extraction column is replaced; The method for judging whether the adsorption capacity of the MIP extraction column is saturated comprises: The curve slope of the dynamic conductivity curve is calculated. When the curve slope is lower than a preset slope threshold, it is determined that the adsorption capacity of the MIP extraction column has reached saturation; otherwise, it is determined that the adsorption capacity of the MIP extraction column is not saturated.

10. An extraction system for extracting ochratoxin A toxin from coffee using acetonitrile, which is used to implement the extraction method for extracting ochratoxin A toxin from coffee using acetonitrile according to any one of claims 1 to 9, characterized in that: The system comprises: Grade system building module: used to collect basic physicochemical characteristic data of target coffee samples, and divide the target coffee samples into L matrix complexity levels according to the basic physicochemical characteristic data; Extraction column preparation module: using ochratoxin A as the template molecule, selecting the type of polymer precursor, optimizing the amount of polymer precursor, and preparing MIP extraction columns for target coffee samples with different matrix complexity levels; Prediction module: used to construct a dynamic prediction model for toxin extraction, collect the second multidimensional spectrum time series of the coffee sample to be tested, and obtain the extraction rate prediction value of the coffee sample to be tested according to the second multidimensional spectrum time series and the toxin extraction dynamic prediction model; determine whether the extraction rate prediction value of the coffee sample to be tested reaches a preset extraction rate threshold; if the extraction rate threshold is reached, stop the extraction process, otherwise continue the extraction until the extraction rate prediction value reaches the extraction rate threshold; Evaluation module: After the extraction of the coffee sample to be tested is completed, the concentration of ochratoxin A in the extract is measured C A ; According to the concentration of ochratoxin A C A and the preset concentration threshold C θ , determine whether the extraction process reaches the expected goal; if the expected goal is reached, the extraction of ochratoxin A toxin is completed, otherwise closed-loop feedback control is required.

Images