Harmful organic matter infrared spectrum detection method based on segmented multi-dimensional harmonic decomposition

Abstract

The invention provides a harmful organic matter infrared spectrum detection method based on segmented multi-dimensional harmonic decomposition, relates to the technical field of infrared spectrum analysis, and aims to solve the problems of noise interference, peak overlapping and inaccurate signal extraction during harmful organic matter detection in the prior art. Obtaining a sample solution to be detected and obtaining an original infrared spectrum of the sample solution to be detected; the method comprises the following steps of: segmenting a dynamic wave band, and performing local signal decomposition on each segment by adopting a harmonic primary function with a proper order; the absorption peak of the target harmful organic compound is accurately extracted by utilizing wave band coupling and characteristic peak identification; stable and accurate quantitative analysis is realized by combining partial least squares, so that the detection sensitivity and accuracy are improved, and the safety of food contact materials is ensured; according to the method, the accuracy and reliability of signal denoising, peak identification and quantitative analysis under a complex matrix are effectively improved by adopting a wave band segmentation strategy, segmented multi-dimensional harmonic decomposition and wave band coupling and characteristic peak identification.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared spectroscopy analysis, and more specifically, to a method for detecting harmful organic compounds in infrared spectra based on segmented multi-dimensional harmonic decomposition. Background Art

[0002] With the increasing prominence of food safety issues, the safety of food contact materials such as tableware and kitchen utensils has attracted much attention. Coatings are often used on the surface of tableware to improve appearance and durability, but harmful organic compounds that may be present in the coatings may dissolve and migrate into food during actual use, posing a risk to consumer health. Infrared spectroscopy analysis technology has been widely used in the detection of organic compounds due to its non-destructive, rapid, and highly sensitive characteristics. However, the complex coating matrix and weak target signal make traditional methods face problems such as noise interference, peak overlap, and inaccurate signal extraction, and there is an urgent need for a more advanced signal processing solution.

[0003] Currently, the following problems mainly exist in the detection of harmful organic compounds in tableware coatings using conventional infrared spectroscopy analysis methods:

[0004] 1) Since the concentration of harmful organic compounds in the coating samples is relatively low, there is still a lot of noise and baseline interference in the spectral data, resulting in blurred signals;

[0005] 2) The absorption peaks of multi-component organic compounds often overlap and superimpose, and it is difficult for traditional denoising and smoothing methods to achieve accurate peak separation and identification;

[0006] 3) Global signal processing methods cannot fully capture the local signal characteristics, resulting in large fitting errors, which in turn affect the stability and prediction accuracy of the quantitative analysis model;

[0007] 4) When the quantitative regression model processes unoptimized high-dimensional spectral variables, there is a problem of multicollinearity, resulting in low prediction accuracy when detecting low-concentration samples. Summary of the Invention

[0008] In order to overcome the problems of noise interference, peak overlap, and inaccurate signal extraction existing in the above-mentioned prior art when detecting harmful organic compounds in coatings, the present invention provides a method for detecting harmful organic compounds in infrared spectra based on segmented multi-dimensional harmonic decomposition. This method performs dynamic band segmentation, uses harmonic basis functions of appropriate orders for local signal decomposition for each segment, then accurately extracts the absorption peaks of target harmful organic compounds by means of band coupling and characteristic peak identification, and combines the partial least squares (PLS) method to achieve stable and accurate quantitative analysis, thereby improving the sensitivity and accuracy of detection and meeting the requirements of food safety monitoring.

[0009] To solve the above technical problems, the technical solution of the present invention is as follows:

[0010] An infrared spectroscopy detection method for harmful organic compounds based on segmented multi-dimensional harmonic decomposition, comprising the following steps:

[0011] S1: Sample acquisition and pretreatment: Obtain a sample to be measured and immerse it in a preset specific solution. After the harmful organic compounds in the sample to be measured are dissolved, obtain the dissolved solution; concentrate and correct the baseline of the dissolved solution to obtain a sample solution to be measured;

[0012] S2: Spectral signal segmented multi-dimensional harmonic decomposition and denoising: Use an infrared spectrometer to collect the original infrared spectrum of the sample solution to be measured; dynamically divide the original infrared spectrum into several segments, and perform multi-dimensional harmonic decomposition within each segment using sine and cosine basis functions of different orders, and solve the harmonic coefficients of each segment by the least squares method to reconstruct the infrared spectrum of each segment and obtain the denoised infrared spectrum;

[0013] S3: Spectral band coupling positioning and characteristic peak identification: Extract the absorption peak section of the denoised infrared spectrum and perform coupling measurement, and at the same time combine a preset threshold and a characteristic peak database to obtain the characteristic absorption peak data of harmful organic compounds;

[0014] S4: Quantitative analysis by partial least squares method: Input the characteristic absorption peak data of harmful organic compounds into a prediction model based on the partial least squares regression algorithm for concentration prediction, obtain the concentration detection result of harmful organic compounds, and complete the quantitative detection of harmful organic compounds.

[0015] Preferably, the step S1 includes:

[0016] S1.1: Preparation of dissolved solution: Obtain a sample to be measured and immerse it in a pre-prepared acidic or neutral aqueous solution, set the soaking temperature, and soak for several hours under the soaking temperature condition. The harmful organic compounds in the sample to be measured are dissolved, and the soaking ends; filter or centrifuge the solution obtained after the soaking ends to obtain the dissolved solution;

[0017] S1.2: Solution concentration: Set the concentration temperature and concentration multiple, and concentrate the dissolved solution by rotary evaporation or vacuum concentration to obtain a concentrated solution;

[0018] S1.3: Baseline correction: Dilute or redissolve the concentrated solution with an inert carrier liquid, and then drop-coat or cover it on an ATR crystal to achieve baseline correction and obtain a sample solution to be measured.

[0019] Preferably, in the step S2, the original infrared spectrum of the sample solution to be measured is collected by an infrared spectrometer in the mid-infrared band, and the original infrared spectrum has noise and background interference;

[0020] The wavelength range of the mid-infrared band is 2.5 μm to 25 μm.

[0021] Preferably, in step S2, obtaining the denoised infrared spectrum includes the following steps:

[0022] Define the collected original infrared spectrum as a vector X = [X(w 1 ), X(w 2 ),..., X(w n )], where X(w i ) is the spectral intensity data collected from the wavenumber position w i , i = 1, 2,..., n, and n is the maximum wavenumber;

[0023] According to the synchronous change characteristics of the spectral absorption peak signals, divide the original infrared spectrum into K segments with different characteristics, expressed as:

[0024] X = [X 1 , X 2 ,..., X K

[0025] where X k is the k-th segment, k = 1, 2,..., K;

[0026] Perform multi-dimensional harmonic decomposition within each segment using sine and cosine harmonic basis functions of different orders. The harmonic basis function φ k (w) is specifically:

[0027]

[0028] where w min and w max are respectively the minimum and maximum wavenumbers within the k-th segment; w is the current wavenumber point;

[0029] For each segment, use the least squares method to fit and calculate the harmonic coefficients The fitting error E k of each segment is expressed as:

[0030]

[0031] where n k is the maximum wavenumber within the k-th segment; X k (w i ) is the spectral intensity data collected at the wavenumber position w i within the k-th segment; m k is the harmonic order of the k-th segment; M k is the maximum harmonic order of the k-th segment; represents the k-th segment m​k Harmonic basis functions of order

[0032] By minimizing the fitting error E of each segment k To solve for the optimal harmonic coefficients Thus obtaining the fitting results of each segment and reconstructing the infrared spectra of each segment;

[0033] Add the fitting results of each segment to obtain the fitting result of the entire spectrum Expressed as:

[0034]

[0035] Take the fitting result of the entire spectrum As the denoised infrared spectrum.

[0036] Preferably, the step S3 includes the following steps:

[0037] Extract several segments Δw where harmful organic matter absorption peaks may exist in the denoised infrared spectrum i = 1, 2, …, L, where L is the maximum number of segments;

[0038] Set the following coupling metric function to measure whether the peak intensity within the segment Δw i Conforms to the typical characteristic absorption pattern:

[0039]

[0040] Where Is the spectral intensity at wavenumber w in the denoised infrared spectrum; w i (w) is the weighting function of the segment Δw i ;

[0041] Calculate the coupling metric function values C of each segment i , when C i Is greater than or equal to the preset characteristic peak threshold, then there is a harmful organic matter absorption peak within the corresponding segment Δw i ;

[0042] Within the segment Δw where there is a harmful organic matter absorption peak i , find The maximum point w * , and through comparison with the known characteristic peak database for characteristic peak identification, determine the category of harmful organic matter corresponding to the absorption peak at the maximum point w * And obtain the characteristic absorption peak data of this category of harmful organic matter.

[0043] Preferably, the weighting function w i(w) It is set according to the position and shape of the absorption peak of the known target compound. The closer the wave number is to the peak center, the higher the weight value;

[0044] The characteristic peak threshold is the basic spectral signal intensity statistically obtained based on the blank baseline or the prior noise level.

[0045] Preferably, in step S4, the prediction model is obtained through the following steps:

[0046] S4.1: Obtain the historical data of the characteristic absorption peaks of several harmful organic compounds with known concentrations and standardize them;

[0047] S4.2: Use the standardized historical data of the characteristic absorption peaks as the input data X in , and use the standardized known concentrations of the harmful organic compounds as the response variable Y out , calculate the covariance matrix between the input data X in and the response variable Y out ;

[0048] S4.3: According to the covariance matrix, decompose the input data X in into several principal components by singular value decomposition, extract the first t principal components, and obtain the principal component matrix T 0 ;

[0049] S4.4: Use the principal component matrix T 0 and the standardized response variable Y out , and calculate the regression coefficient β by the least squares method;

[0050] S4.5: Based on the regression coefficient β, construct the following prediction model:

[0051] Y predicted = X new ·β

[0052] where Y predicted is the predicted concentration of the harmful organic compound; X new is the characteristic absorption peak data of the harmful organic compound to be detected.

[0053] Preferably, in step S4.1, the standardization is performed according to the following formula:

[0054]

[0055]

[0056] where X 0 and Y 0 are respectively the historical data of the characteristic absorption peaks and the known concentrations of the harmful organic compounds; μ x and μ yare the mean values of the characteristic absorption peak historical data and the known concentrations, respectively; σ x and σ y are the variances of the characteristic absorption peak historical data and the known concentrations, respectively.

[0057] Preferably, in the step S4.4, the regression coefficient β is calculated according to the following formula:

[0058] β = (T 0 T T 0 ) -1 T 0 T Y out

[0059] where T represents matrix transpose; (·) -1 represents the inverse matrix.

[0060] Preferably, the method is used for qualitative and quantitative analysis of harmful organic compounds in tableware coatings.

[0061] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0062] The present invention provides an infrared spectrum detection method for harmful organic substances based on segmented multi-dimensional harmonic decomposition. First, sample acquisition and pretreatment are carried out to obtain a sample solution to be measured and its original infrared spectrum; then, through dynamic band segmentation, local signal decomposition is carried out for each segment using a harmonic basis function of an appropriate order, and then the absorption peaks of target harmful organic compounds are accurately extracted by using band coupling and characteristic peak identification. Finally, stable and accurate quantitative analysis is achieved by combining the partial least squares method, thereby improving the sensitivity and accuracy of detection and ensuring the safety of food contact materials;

[0063] The present invention has the following beneficial effects:

[0064] 1) The signal-to-noise ratio and detection sensitivity of infrared spectrum detection are significantly improved, and the local signal characteristics are optimized through the segmentation strategy;

[0065] 2) The overlapping absorption peaks in the complex spectrum are effectively separated and identified, and the characteristic signals of target harmful substances are accurately extracted, providing a clear basis for qualitative analysis;

[0066] 3) By inputting the optimized spectrum data into the PLS model, the stability and prediction accuracy of quantitative regression are significantly improved, which is particularly suitable for the detection of low-concentration samples;

[0067] 4) The overall detection process realizes efficient and automated processing, reduces manual operation and errors, and is convenient for large-scale popularization and application. Description of the Drawings

[0068] Figure 1 Flow chart of a method for detecting infrared spectra of harmful organic compounds based on segmented multi-dimensional harmonic decomposition provided in Example 1.

[0069] Figure 2 Sample spectrum of a single leachate provided in Example 2.

[0070] Figure 3 Sample spectra of 200 leachates provided in Example 2. Detailed implementation manners

[0071] The accompanying drawings are only for illustrative purposes and should not be construed as limitations on this patent;

[0072] To better illustrate this embodiment, some components in the accompanying drawings are omitted, enlarged or reduced, and do not represent the dimensions of actual products;

[0073] For those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0074] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0075] Example 1

[0076] As Figure 1 shown, this embodiment provides a method for detecting infrared spectra of harmful organic compounds based on segmented multi-dimensional harmonic decomposition, which is used for qualitative and quantitative analysis of harmful organic compounds in tableware coatings, and includes the following steps:

[0077] S1: Sample acquisition and pretreatment: Obtain a sample to be tested and soak it in a preset specific solution. After the harmful organic compounds in the sample to be tested are dissolved, obtain the dissolved solution; concentrate and correct the baseline of the dissolved solution to obtain a sample solution to be tested;

[0078] S2: Spectral signal segmented multi-dimensional harmonic decomposition and denoising: Use an infrared spectrometer to collect the original infrared spectrum of the sample solution to be tested; dynamically divide the original infrared spectrum into several segments, and perform multi-dimensional harmonic decomposition within each segment using sine and cosine basis functions of different orders, and solve the harmonic coefficients of each segment by the least squares method to reconstruct the infrared spectra of each segment and obtain the denoised infrared spectrum;

[0079] S3: Spectral band coupling positioning and characteristic peak identification: Extract the absorption peak sections of the denoised infrared spectrum and perform coupling measurement, and at the same time combine a preset threshold and a characteristic peak database to obtain the characteristic absorption peak data of harmful organic compounds;

[0080] S4: Quantitative analysis by partial least squares method: Input the characteristic absorption peak data of harmful organic substances into a prediction model based on the partial least squares regression algorithm for concentration prediction, obtain the concentration detection results of harmful organic substances, and complete the quantitative detection of harmful organic compounds.

[0081] In the specific implementation process, taking the detection of harmful organic substances in tableware coatings as an example, this embodiment first prepares and preprocesses the samples. Under simulated food contact conditions, immerse the tableware coating samples in a specific solution, and strictly control the environment, duration, temperature, and solvent conditions to ensure that harmful organic compounds are fully dissolved. Subsequently, concentrate and correct the baseline of the dissolved solution to eliminate solvent and matrix interference and provide a representative sample for subsequent spectral analysis.

[0082] Then, perform spectral signal segmented multi-dimensional harmonic decomposition and denoising. Use the segmented multi-dimensional harmonic decomposition algorithm to process the infrared spectral data, dynamically divide the spectrum into regions according to local signal characteristics, decompose within each segment using fixed sine and cosine basis functions, and solve the corresponding harmonic coefficients by the least squares method to achieve effective noise reduction and smoothing, and extract clear signal characteristics.

[0083] After that, perform spectral band coupling positioning and characteristic peak identification. Use the band coupling positioning technology to perform weighted summation processing on the denoised spectral data to accurately lock the characteristic absorption peaks of target harmful organic substances. At the same time, combine the characteristic peak identification method to match the positioning results with the preset standards, distinguish interference signals from target peaks, and ensure high-precision peak extraction and recognition under complex backgrounds.

[0084] Finally, perform quantitative analysis by partial least squares method. Input the characteristic peak data obtained through the above steps into a partial least squares (PLS) regression model to construct a concentration prediction model and achieve accurate quantification of the dissolved harmful organic compounds. With the support of optimized data, this conventional quantitative method significantly improves the stability and prediction accuracy of the model and meets the requirements of food safety monitoring.

[0085] This method constructs a set of infrared spectral detection methods for harmful organic substances in tableware coatings by adopting the band segmentation strategy, segmented multi-dimensional harmonic decomposition, and band coupling and characteristic peak identification, effectively improving the accuracy and reliability of signal denoising, peak recognition, and quantitative analysis under complex matrices.

[0086] Embodiment 2

[0087] This embodiment provides an infrared spectral detection method for harmful organic substances based on segmented multi-dimensional harmonic decomposition, which is used for qualitative and quantitative analysis of harmful organic compounds in tableware coatings, and includes the following steps:

[0088] S1: Sample acquisition and pretreatment: Obtain a sample to be measured and immerse it in a preset specific solution. After the harmful organic compounds in the sample to be measured are dissolved, obtain the dissolved solution; Concentrate and perform baseline correction on the dissolved solution to obtain a sample solution to be measured;

[0089] S2: Spectral signal segmented multi-dimensional harmonic decomposition and denoising: Use an infrared spectrometer to collect the original infrared spectrum of the sample solution to be measured; Dynamically divide the original infrared spectrum into several segments, and perform multi-dimensional harmonic decomposition within each segment using sine and cosine basis functions of different orders. Solve the harmonic coefficients of each segment by the least squares method, reconstruct the infrared spectrum of each segment, and obtain the denoised infrared spectrum;

[0090] S3: Spectral band coupling localization and characteristic peak identification: Extract the absorption peak section and perform coupling measurement on the denoised infrared spectrum. At the same time, combine a preset threshold and a characteristic peak database to obtain the characteristic absorption peak data of harmful organic substances;

[0091] S4: Quantitative analysis by partial least squares method: Input the characteristic absorption peak data of harmful organic substances into a prediction model based on the partial least squares regression algorithm for concentration prediction, obtain the concentration detection result of harmful organic substances, and complete the quantitative detection of harmful organic compounds;

[0092] The step S1 includes:

[0093] S1.1: Preparation of dissolved solution: Obtain a sample to be measured and immerse it in a pre-prepared acidic or neutral aqueous solution, set the soaking temperature, and soak for several hours under the soaking temperature condition. The harmful organic compounds in the sample to be measured are dissolved, and the soaking ends; Filter or centrifuge the solution obtained after the soaking ends to obtain the dissolved solution;

[0094] S1.2: Solution concentration: Set the concentration temperature and concentration multiple, and concentrate the dissolved solution by rotary evaporation or vacuum concentration to obtain a concentrated solution;

[0095] S1.3: Baseline correction: After diluting or redissolving the concentrated solution with an inert carrier liquid, drop-coat or cover it on the ATR crystal to achieve baseline correction and obtain a sample solution to be measured;

[0096] In the step S2, use an infrared spectrometer to collect the original infrared spectrum of the sample solution to be measured in the mid-infrared band, and there is noise and background interference in the original infrared spectrum;

[0097] The wavelength range of the mid-infrared band is 2.5μm - 25μm;

[0098] In the step S2, obtaining the denoised infrared spectrum includes the following steps:

[0099] Define the collected original infrared spectrum as a vector X = [X(w 1 ), X(w 2 ),..., X(w n )], where X(w i ) is the spectral intensity data collected from the wavenumber position w i , i = 1, 2,..., n, and n is the maximum wavenumber;

[0100] According to the synchronous change characteristics of the spectral absorption peak signal, divide the original infrared spectrum into K segments with different characteristics, expressed as:

[0101] X = [X 1 , X 2 ,..., X K

[0102] where X k is the k-th segment, k = 1, 2,..., K;

[0103] Within each segment, perform multi-dimensional harmonic decomposition using sine and cosine harmonic basis functions of different orders. The harmonic basis function φ k (w) is specifically:

[0104]

[0105] where w min and w max are respectively the minimum and maximum values of the wavenumber within the k-th segment; w is the current wavenumber point;

[0106] For each segment, use the least squares method to fit and calculate the harmonic coefficients The fitting error E k of each segment is expressed as:

[0107]

[0108] where n k is the maximum wavenumber within the k-th segment; X k (w i ) is the spectral intensity data collected at the wavenumber position w i within the k-th segment; m k is the harmonic order of the k-th segment; M k is the maximum harmonic order of the k-th segment; represents the m k -order harmonic basis function of the k-th segment;

[0109] Solve for the optimal harmonic coefficients k by minimizing the fitting error E ​Thus, the fitting results of each segment are obtained, and the infrared spectra of each segment are reconstructed;

[0110] Add the fitting results of each segment to obtain the fitting result of the entire spectrum Expressed as:

[0111]

[0112] Take the fitting result of the entire spectrum As the denoised infrared spectrum;

[0113] The step S3 includes the following steps:

[0114] Extract several sections Δw where there may be absorption peaks of harmful organic substances in the denoised infrared spectrum i = 1, 2, …, L, where L is the maximum number of sections;

[0115] Set the following coupling metric function to measure whether the peak intensity in the section Δw i Meets the typical characteristic absorption pattern:

[0116]

[0117] Among them, Is the spectral intensity at the wavenumber w in the denoised infrared spectrum; w i (w) is the weighting function of the section Δw i ;

[0118] Calculate the coupling metric function values C of each section i , when C i Is greater than or equal to the preset characteristic peak threshold, then there are absorption peaks of harmful organic substances in the corresponding section Δw i ;

[0119] In the section Δw where there are absorption peaks of harmful organic substances i , find The maximum point w * , identify the characteristic peaks by comparing with the known characteristic peak database, judge the type of harmful organic substances corresponding to the absorption peaks at the maximum point w * , and obtain the characteristic absorption peak data of this type of harmful organic substances;

[0120] The weighting function w i (w) is set according to the absorption peak position and shape of the known target compound, and the closer the wavenumber is to the peak center, the higher the weight value;

[0121] The characteristic peak threshold is the basic spectral signal intensity statistically obtained based on the blank baseline or the prior noise level;

[0122] In step S4, the prediction model is obtained through the following steps:

[0123] S4.1: Obtain the historical data of the characteristic absorption peaks of several harmful organic substances with known concentrations and standardize them. The standardization formula is as follows:

[0124]

[0125] where X 0 and Y 0 are the historical data of the characteristic absorption peaks of harmful organic substances and the known concentrations respectively; μ x and μ y are the means of the historical data of the characteristic absorption peaks and the known concentrations respectively; σ x and σ y are the variances of the historical data of the characteristic absorption peaks and the known concentrations respectively.

[0126] S4.2: Use the standardized historical data of the characteristic absorption peaks as the input data X in , and the standardized known concentrations of harmful organic substances as the response variable Y out , and calculate the covariance matrix between the input data X in and the response variable Y out .

[0127] S4.3: According to the covariance matrix, decompose the input data X in into several principal components through singular value decomposition, extract the first t principal components, and obtain the principal component matrix T 0 .

[0128] S4.4: Use the principal component matrix T 0 and the standardized response variable Y out , and calculate the regression coefficient β through the least squares method:

[0129] β = (T 0 T T 0 ) -1 T 0 T Y out

[0130] where T represents matrix transpose; (·) -1 represents the inverse matrix.

[0131] S4.5: Based on the regression coefficient β, construct the following prediction model:

[0132] Y predicted = X new ·β

[0133] where Y predictedis the predicted concentration of harmful organic substances; X new is the characteristic absorption peak data of the harmful organic substances to be detected.

[0134] In the specific implementation process, in this method, in order to accurately capture the characteristic absorption peaks of common organic compounds (such as phenols, ketones, alcohols, etc.) in the infrared region, the mid-infrared (Mid-IR) band is mainly used, usually in the range of 4000 cm -1 ~400 cm -1 ; this band covers the characteristic vibration intervals of most organic functional groups (such as hydroxyl groups, carbonyl groups, ether bonds, benzene rings, etc.), so it is suitable for qualitative and quantitative analysis of organic extracts in tableware coatings;

[0135] This method mainly includes the following steps:

[0136] Step 1: Preparation and pretreatment of samples. Under simulated food contact conditions, immerse the tableware coating sample in a specific solution, and strictly control the environment, duration, temperature, and solvent conditions to ensure that harmful organic compounds are fully dissolved; then concentrate and baseline correct the dissolved solution to eliminate solvent and matrix interference, and provide a representative sample for subsequent spectral analysis;

[0137] Specifically, first determine the actual use scenario of the tableware coating, including factors such as the common food acidity, temperature, and contact time, and use these as the simulation conditions for the dissolution environment, so as to ensure that the collected solution sample can truly reflect the dissolution of harmful organic compounds that may occur during the daily use of the tableware. The specific process is as follows:

[0138] 1) Solution preparation: Usually select a simulated solution with a pH range between 4 and 7 (such as a weakly acidic or neutral aqueous solution) to simulate the environment of daily food or cleaning agents; if the tableware is used in some special scenarios (such as high acidity or alkalinity environments), the pH range and salinity can be adjusted accordingly to be closer to the actual application;

[0139] 2) Determination of soaking conditions: Immerse the tableware to be tested completely in the above solution and place it under constant temperature conditions (such as 37°C to simulate body temperature or 50 - 80°C to simulate the high-temperature tableware washing environment), and the soaking time can be adjusted according to the product instructions and the expected contact time, such as 1 hour, 2 hours or longer;

[0140] 3) Solution collection: After the soaking is completed, perform a simple filtration or centrifugation operation on the solution to remove large particles or suspended impurities so as not to affect the subsequent infrared spectroscopy test; the filtration can use a filter membrane with a pore size of 0.45 μm or smaller; if necessary, the solution can also be temporarily stored in a container that does not react with the compound to be tested (such as an inert glass bottle or a polytetrafluoroethylene bottle) to prevent secondary pollution;

[0141] Example: To evaluate the dissolution risk of a certain coating under acidic conditions, prepare an acetic acid buffer solution with pH = 4, soak the tableware in an environment at 60 °C for 2 hours, and then collect the filtrate to obtain a sample of the main dissolved organic compounds;

[0142] Next, perform sample concentration and baseline correction. In this step, the key is to appropriately concentrate and pretreat the obtained solution sample, reduce the interference of solvents and other matrices on the infrared spectrum, and enhance the signal intensity of trace organic compounds in the spectrum;

[0143] Solution concentration: According to the properties of the dissolved organic compounds, select a suitable concentration method, such as rotary evaporation or vacuum concentration; it should be noted that during the concentration process, try to maintain a low or moderate temperature to avoid the destruction or volatilization of some thermosensitive organic components at high temperatures; if the concentration multiple is too large, it may cause the loss of some components, which needs to be verified in advance according to the characteristics of the analyte;

[0144] Baseline correction: In infrared spectroscopy analysis, water or other solvents often produce significant absorption bands, which can mask the key peaks of the organic compounds to be measured or interfere with their quantitative analysis; using attenuated total reflection (ATR), drop or cover a small amount of the concentrated solution on the ATR crystal, and utilize its detection ability for trace substances on the surface to reduce solvent interference;

[0145] Example: For example, when detecting phenol or phenolic derivatives in an experiment, the solution volume can be reduced to 1 / 5 of the original volume by rotary evaporation, and then diluted or redissolved using an inert carrier liquid (such as heavy water or a specific organic solvent), and finally measured by thin film coating on the ATR crystal to obtain relatively clear characteristic absorption peaks;

[0146] Step 2: Perform spectral signal segmented multi-dimensional harmonic decomposition and denoising. Use the segmented multi-dimensional harmonic decomposition algorithm to process the infrared spectrum data, dynamically divide the spectrum into regions according to local signal characteristics, decompose within each segment using fixed sine and cosine basis functions, and solve the corresponding harmonic coefficients by the least squares method to achieve effective noise reduction and smoothing, and extract clear signal characteristics;

[0147] After the sample preparation and pretreatment in Step 1, a concentrated and baseline-corrected sample solution is obtained; using an infrared spectrometer in the mid-infrared band (4000 cm -1 ~400 cm -1 ), with a sampling step of 1 cm -1 for discrete sampling, there are a total of n = w max - w min + 1 sampling points (3601 points), and a relatively clean but still possibly containing some noise and background interference original infrared spectrum X can be obtained, such as Figure 2 and 3As shown, they are respectively the spectral diagram of a single leachate sample and the spectral diagram of 200 leachate samples. The absorption peak positions of some toxic compounds are shown in Table 1;

[0148] Toxic compounds corresponding to the absorption peaks in the infrared absorption spectrum in Table 1

[0149] <![CDATA[Absorption peak (cm- 1 )]]> Corresponding vibration type Corresponding toxic compound 3400 -OH or -NH vibration Formaldehyde, bisphenol A, phenolic compounds, amines 2950 C-H stretching vibration Phthalates, polycyclic aromatic hydrocarbons, alkylphenols 1720 C=O stretching vibration Formaldehyde, acetaldehyde, organic esters, ketone compounds 1600 Benzene ring skeletal vibration Benzene compounds, phenolic compounds, azo dye decomposition products 1250 C-O stretching vibration Organophosphates, bisphenol A and its derivatives, ether compounds 700 C-H out-of-plane bending vibration Benzene series, halogenated hydrocarbons, PVC degradation products

[0150] After collecting the original infrared spectrum, the piecewise multi-dimensional harmonic decomposition algorithm is used to denoise and background subtract the spectrum;

[0151] Specifically, the collected original infrared spectrum is defined as the vector X = [X(w 1 ), X(w 2 ),..., X(w n )], where X(w i ) is the spectral intensity data collected from the wavenumber position w i , i = 1, 2,..., n, and n is the maximum wavenumber;

[0152] According to the synchronous change characteristics of the spectral absorption peak signals, the original infrared spectrum is divided into K segments with different characteristics, expressed as:

[0153] X = [X 1 , X 2 ,..., X K

[0154] where X k is the kth segment, k = 1, 2,..., K;

[0155] For example, Table 2 gives an example of a segment:

[0156] Table 2 Example of segment

[0157] Segment numbering <![CDATA[Band range (cm- 1 )]]> <![CDATA[X 1 > 400～1000 <![CDATA[X 2 > 1000～1450 <![CDATA[X 3 > 1450～1660 <![CDATA[X 4 > 1660～1800 <![CDATA[X 5 > 1800～3200 <![CDATA[X 6 > 3200～4000

[0158] Within each segment, multi-dimensional harmonic decomposition is performed using sine and cosine harmonic basis functions of different orders m k ; The purpose of harmonic decomposition is to decompose complex periodic or quasi-periodic signals into a series of simple sine or cosine waves (harmonics) in signal processing to reveal their frequency components and characteristics; In the data processing of the spectral detection of tableware leachates, harmonic decomposition can be used to remove noise, enhance characteristic absorption peaks, identify weak signals, and improve the resolution of spectral data; For example, for some signal regions, low-order harmonics (such as the 2nd order) can be selected, while for other signal regions, high-order harmonics (such as the 3rd order) can be selected to adapt to the changes of local peaks;

[0159] The harmonic basis function φ k (w) is specifically:​

[0160]

[0161] wherein, w min and w max are respectively the minimum and maximum values of the wave number within the k-th segment; w is the current wave number point; k is the harmonic order (1, 2, 3,...), that is, which order of sine / cosine;

[0162] For each segment, a group of specially designed functions is adopted. This function is used to determine the most suitable harmonic order and its corresponding sine and cosine basis functions within this segment in subsequent steps, and to adjust the weights of each fixed basis function in signal reconstruction by solving the optimal harmonic coefficients, so that the linear combination of each segment can approximate the original signal as accurately as possible;

[0163]

[0164] wherein, is the harmonic basis function (such as sine, cosine, etc.) used in the k-th segment, is the harmonic coefficient to be solved; the harmonic coefficients are used to determine the weights that each fixed harmonic basis function should account for in reconstructing the original signal, and they reflect the contributions of different frequency components to the overall signal; by linearly combining the fixed sine and cosine basis functions, the harmonic coefficients adjust the amplitude of each basis function, so that the combined signal can approximate the original signal as accurately as possible;

[0165] For each segment, the least squares method is adopted in this embodiment to fit and calculate the harmonic coefficients The fitting error E of each segment k is expressed as:

[0166]

[0167] wherein, n k is the maximum wave number within the k-th segment; X k (w i ) is the spectral intensity data collected at the wave number position w i within the k-th segment; m k is the harmonic order of the k-th segment; M k is the maximum harmonic order of the k-th segment; represents the harmonic basis function of the m k -th order of the k-th segment;

[0168] By minimizing the fitting error E of each segment k to solve the optimal harmonic coefficients thus obtaining the fitting results of each segment and reconstructing the infrared spectra of each segment;

[0169] Add the fitting results of each segment to obtain the fitting result of the entire spectrum Expressed as:

[0170]

[0171] Take the fitting result of the entire spectrum As the denoised infrared spectrum, this fitting result will be used as the input data for the subsequent PLS algorithm;

[0172] Reasons for using the above fitting result as the input data for the PLS algorithm:

[0173] 1) Denoising and signal clarification;

[0174] In this embodiment, harmonic basis functions (sine and cosine functions) are designed to capture periodic and trend changes in the signal; random noise and aperiodic interference generally do not conform to the characteristics of these harmonic basis functions, so they are effectively suppressed during the fitting process; the fitting signal mainly contains the main characteristic peaks and trends of the signal, and noise and random interference are greatly reduced;

[0175] 2) Enhancing the interpretability of signal features;

[0176] Each harmonic basis function corresponds to a specific frequency component in the signal; by adjusting the harmonic coefficients, the contribution of each frequency component can be precisely controlled, making the fitting signal more interpretable; this method makes the source of the characteristic peaks clearer and facilitates the subsequent quantitative analysis model (PLS) to accurately identify and quantify target organic compounds;

[0177] 3) Optimizing the performance of the quantitative analysis model;

[0178] The fitting signal provides a smoother and more stable data set, reducing the model instability caused by multicollinearity and high-dimensional features. High-quality input data helps the PLS regression model establish a more robust and accurate prediction relationship, especially when dealing with low-concentration samples, and the performance is particularly excellent;

[0179] Example:

[0180] The mid-infrared band is [400, 4000]. To simplify, two wavenumber ranges [1718, 1723] and [3398, 3403] are selected as examples;

[0181] Segment 1: w ∈ [1718, 1723], Segment 2: w ∈ [3398, 3403];

[0182] The spectral data is shown in Table 3:

[0183] Table 3 Spectral data

[0184] <![CDATA[Wave number ω (cm -1 )]]> Original spectral intensity X(ω) Belonging segment 1718 1.15 1 1719 1.1 1 1720 1.05 1 1721 0.95 1 1722 0.9 1 1723 0.85 1 3398 1 2 3399 0.95 2 3400 0.9 2 3401 0.85 2 3402 0.8 2 3403 0.75 2

[0185] For the spectral data segment 1 (w ∈ [1718, 1723]), the signal intensity gradually decreases, and it is suitable to use the second-order harmonic basis function for fitting; this smooth signal change is suitable for using low-order harmonic basis functions (such as the first or second order) for fitting because low-order harmonics can effectively capture the overall trend change without introducing too much fluctuation or complexity, expressed as:

[0186]

[0187]

[0188] Basis function matrix φ 1 As shown in Table 4:

[0189] Table 4 Basis function matrix φ 1

[0190]

[0191] Construct the matrix equation:

[0192] φ 1 ·α 1 =X 1

[0193] That is:

[0194]

[0195] Define the error as:

[0196]

[0197] Solve using the least squares method to obtain the harmonic coefficients:

[0198]

[0199] Then the fitting result is:

[0200]

[0201] For the spectral data segment 2 (w ∈ [3398, 3403]), the signal intensity gradually increases, and it is suitable to use the third-order harmonic basis function for fitting; high-order harmonic basis functions (such as the third order or higher orders) can capture these complex signal features more precisely, ensuring that both the overall trend can be accurately reflected and the subtle changes can be identified and retained during the fitting process;

[0202]

[0203] Basis function matrix φ2 As shown in Table 5:

[0204] Table 5 Basis function matrix φ 2

[0205]

[0206] Construct a matrix equation:

[0207] φ 2 ·α 2 =X 2

[0208] That is:

[0209]

[0210] Define the error:

[0211]

[0212] Solve using the least squares method to obtain the harmonic coefficients:

[0213]

[0214] Then the fitting result is:

[0215]

[0216] The goal of using the least squares method in this embodiment is to minimize the above error, and this error term represents the gap between the actual signal and the fitting signal; by calculating the error of each data point and then summing them up, the best harmonic coefficients that minimize this error or reach the error threshold are found;

[0217] In the above way, adjust the harmonic basis function as needed and calculate the harmonic coefficients to make the error reach the specified threshold, then the relevant harmonic basis function and harmonic coefficients can be determined, and the segmented fitting spectral data can be obtained. Finally, the spectral data of each segment are globally fitted to obtain the spectral data with clear signals

[0218] Step 3: Perform spectral band coupling positioning and characteristic peak identification. Using the band coupling positioning technology, perform weighted summation processing on the denoised spectral data to accurately lock the characteristic absorption peaks of the target harmful organic substances; at the same time, combine the characteristic peak identification method to match the positioning result with the preset standard to distinguish the interference signal from the target peak, ensuring high-precision peak extraction and identification under complex backgrounds;

[0219] After completing the segmented multi-dimensional harmonic decomposition and denoising, the obtained spectrum Smoother and with lower noise; in order to quickly and accurately lock the characteristic absorption peaks of different organic compounds, the present embodiment proposes a band coupling positioning algorithm; its core idea is to combine the absorption regions of common functional groups in the infrared spectrum (such as benzene ring - OH vibration, - NH vibration, etc.) with the actually detected spectrum, and identify the interval where the target peak is located by calculating the band coupling metric;

[0220] Because in the spectrum, different absorption peaks may contain information of both the target organic compound and other substances or background noise; the purpose of band coupling positioning and characteristic peak identification is to determine which peaks or bands mainly come from the target organic compound and which mainly belong to interfering components; for non - target interfering peaks, they will be marked, and their influence will be controlled or reduced in subsequent quantitative or qualitative analysis; this step helps to exclude background information irrelevant to the detection target and ensure that subsequent analysis focuses on the most critical signal region;

[0221] Specifically, several sections Δw where absorption peaks of harmful organic compounds may exist are extracted from the denoised infrared spectrum i = 1, 2, …, L, where L is the maximum number of sections;

[0222] The following coupling metric function is set to measure whether the peak intensity within the section Δw i complies with the typical characteristic absorption pattern:

[0223]

[0224] Among them, is the spectral intensity at wavenumber w in the denoised infrared spectrum; w i (w) is the weighting function of the section Δw i , and the weighting function w i (w) is set according to the absorption peak position and shape of the known target compound. The closer the wavenumber is to the peak center, the higher the weight value. For example, for common phenolic compounds, the vibration peaks in the vicinity of 3400 cm -1 are mainly concerned, and larger weights can be assigned in this interval;

[0225] Calculate the coupling metric function value C i for each section. When C i is greater than or equal to the preset characteristic peak threshold, the corresponding section Δw iThe memory has absorption peaks of harmful organic substances; in this embodiment, the characteristic peak threshold is a key parameter for judging whether there are characteristic peaks of harmful organic substances in the spectral signal, and this threshold is obtained through a statistical method based on the blank baseline or the prior noise level to ensure the accurate extraction of the target signal; among them, the blank baseline refers to the infrared spectral data measured by the instrument in the absence of the sample to be measured; these data usually reflect environmental noise, the response of the instrument itself, and any background interference during the experiment (such as the preset solution used for sample soaking); the blank baseline is the interference signal without the target substance and is collected before measuring the actual sample; for example, for the blank baseline without the sample to be measured, the spectral intensities of N preset solutions are collected, the baseline intensity is μ = 0.03, and the standard deviation is σ = 0.005, then the threshold can be set as the baseline intensity plus 3 times the standard deviation increment, and the result is μ + 3*σ = 0.045; if the characteristic absorption peak of harmful organic substance A appears at a wavelength of 1600 cm -1 , and its intensity is 0.06, exceeding the set threshold of 0.045, so this peak is confirmed to be valid, indicating that harmful organic substance A may exist in the tableware coating; if the characteristic absorption peak of harmful organic substance B appears at a wavelength of 1590 cm -1 , and its intensity is 0.038, lower than 0.045, so this peak is not considered a valid absorption peak and may be background noise or other non-target signals;

[0226] In the section Δw where there are absorption peaks of harmful organic substances i , find the maximum point w * , identify the characteristic peaks by comparing with the known characteristic peak database, judge the category of harmful organic substances corresponding to the absorption peak at the maximum point w * , and obtain the characteristic absorption peak data of this category of harmful organic substances; if there are multiple peaks or overlapping with other matrix peaks in the band, then combine the sample background information obtained in the previous step one and the possible compound database to exclude the peak shapes that do not match the expectations and improve the reliability of peak identification;

[0227] Example 1: Pay attention to the benzene ring skeletal vibration peaks (commonly found in phenols and aromatic compounds) in the range of 1600 - 1580 cm -1 ; in this range, the average intensity value of the spectrum is concentrated around 0.80 - 0.86. After calculation by the band coupling positioning function, C i ≈12.7 (assuming there are several wavenumber points in this range and this value is obtained after weighted summation); this result is significantly higher than the preset threshold (for example, 10.0), so it can be determined that in the range of 1600 - 1580 cm -1There are significant aromatic ring absorption peaks; if further combined with the sample information obtained in Step 1 and compared with the database, and it is confirmed that the peaks in this region may come from phenolic derivatives, quantitative analysis can be carried out in the next step (Step 4);

[0228] Example 2: If another interval (such as 1700~1680 cm -1 ) corresponds to the C of the carbonyl absorption peak i The calculated value is only 5.8, which is lower than the threshold of 10.0. Then it can be determined that there are no obvious strong absorption peaks of ketones or aldehydes in this interval, excluding the possibility of characteristic peaks; this can more effectively lock the truly suspicious peak segments and save subsequent data processing resources;

[0229] Example 3: This example gives a specific calculation process. The denoised spectral data is shown in Table 6:

[0230] Table 6 Denoised spectral data

[0231]

[0232] Design of the weighting function w i (w): At 1598 and 1597 cm -1 , assign a value of 1.1 (the highest), indicating that the center of the estimated aromatic ring absorption peak may be in this region; the further the wavenumber deviates from this center, the slightly lower the weight value, so as to reflect the idea that "the peak center is more important";

[0233] Calculation of the band coupling metric C i : Assume that Δw = {1600, 1599, 1598, 1597, 1596} is the test sub-interval, and calculate its coupling metric. According to the data in the "weighted product" column of Table 6, adding them up gives:

[0234] C i = 0.425 + 0.645 + 1.10 + 1.05 + 0.42 = 3.64

[0235] In this embodiment, it is assumed that the determination threshold corresponding to this interval is 2.00 (determined by combining the prior noise level or the blank baseline evaluation); since C = 3.64 > 2.00, it can be preliminarily determined that this band is very likely to contain the target aromatic ring peak;

[0236] In Step 3 of this method, the band coupling localization algorithm is used to identify the interval where the target peak is located, forming a preliminary qualitative judgment on the organic compounds that may be leached from the tableware coating; the "target peak segment" obtained in this step will be further processed in the quantitative analysis based on partial least squares in Step 4 to obtain the accurate content of the target compound, thus completing the detection target; through the combination of Steps 2 to 4, high-resolution qualitative and quantitative analysis of low-content multi-component leachates can be achieved, providing a reliable basis for evaluating the safety of tableware;

[0237] Step 4: Perform partial least squares (PLS) quantitative analysis. Input the characteristic peak data obtained through the above steps into the PLS regression model to construct a concentration prediction model, and achieve accurate quantification of the dissolved harmful organic compounds. With the support of the optimized data, this conventional quantification method significantly improves the stability and prediction accuracy of the model, meeting the requirements of food safety monitoring.

[0238] For the organic extracts from tableware coatings, PLS can "learn" the relationship between spectral variables and actual concentrations through a limited number of calibration samples (known concentrations), and quickly predict the concentration values after the spectral measurement of unknown samples. This not only meets the requirements of food safety detection for quantitative accuracy but also simplifies the algorithm implementation and calculation amount.

[0239] Specifically, first obtain the historical data of the characteristic absorption peaks of several harmful organic compounds with known concentrations and perform standardization. The purpose of standardization is to eliminate the influence of the dimension on the model results and ensure that each variable has the same influence. The standardization formula is as follows:

[0240]

[0241] where X 0 and Y 0 are the historical data of the characteristic absorption peaks and the known concentrations of the harmful organic compounds respectively; μ x and μ y are the means of the historical data of the characteristic absorption peaks and the known concentrations respectively; σ x and σ y are the variances of the historical data of the characteristic absorption peaks and the known concentrations respectively.

[0242] Take the standardized historical data of the characteristic absorption peaks as the input data X in , and the standardized known concentrations of the harmful organic compounds as the response variable Y out . Calculate the covariance matrix between the input data X in and the response variable Y out . The covariance matrix reflects the relationship between the input data X in and the response variable Y out , and determines the direction of subsequent principal component extraction.

[0243] According to the covariance matrix, decompose the input data X in into several principal components through singular value decomposition (SVD). Each principal component is a new variable that can effectively represent the variability of the original data. Extract the first t principal components to maximize the explanation of the variance in the data and obtain the principal component matrix T 0 .

[0244] Subsequent regression analysis is performed. By projecting the input data X in onto the principal component space, PLS establishes the regression relationship between the input data X in and the response variable Y out ; the regression coefficients are calculated by minimizing the sum of squared residuals and are used to represent the contribution of each principal component to the prediction result; the regression model formula is:

[0245] Y out = X in ·β + ∈

[0246] where ∈ is the error term;

[0247] In this embodiment, the regression coefficient β is calculated by the least squares method:

[0248] β = (T 0 T T 0 ) -1 T 0 T Y out

[0249] where T represents matrix transpose; (·) -1 represents the inverse matrix;

[0250] Based on the regression coefficient β, the following prediction model can be constructed to predict the response variable value of new samples:

[0251] Y predicted = X new ·β

[0252] where Y predicted is the predicted concentration of harmful organic substances; X new is the characteristic absorption peak data (after standardization) of the harmful organic substances to be detected;

[0253] The role of PLS is to establish an effective prediction model through spectral data, which can extract the most useful information from multivariate data, avoid noise interference, and improve the prediction accuracy;

[0254] Finally, the characteristic absorption peak data of the harmful organic substances obtained in step three are standardized and then input into the prediction model for prediction, and the accurate concentration detection result of the harmful organic substances is obtained, thus completing the quantitative detection of the harmful organic compounds;

[0255] This method constructs a set of infrared spectroscopy detection methods for harmful organic substances in tableware coatings by adopting the band segmentation strategy, segmented multi-dimensional harmonic decomposition, and band coupling and characteristic peak identification, effectively improving the accuracy and reliability of signal denoising, peak identification, and quantitative analysis under complex matrices.

[0256] Like or similar reference numerals correspond to like or similar components;

[0257] The terms used in the drawings to describe the positional relationships are for illustrative purposes only and should not be construed as limiting the present patent;

[0258] Obviously, the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention and are not intended to limit the embodiments of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the embodiments here. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition, characterized in that: The following steps are involved: S1: Sample acquisition and pretreatment: Obtain the sample to be tested and soak it in a preset specific solution, and obtain the dissolution solution after the harmful organic compounds in the sample to be tested are dissolved; Concentrating and baseline correcting the dissolution solution to obtain a sample solution to be tested; S2: Segmented multi-dimensional harmonic decomposition and denoising of spectral signals: using an infrared spectrometer to collect the original infrared spectrum of the sample solution to be tested; dynamically dividing the original infrared spectrum into several segments, using sine and cosine basis functions of different orders to perform multi-dimensional harmonic decomposition in each segment, and solving the harmonic coefficient of each segment by the least square method, reconstructing the infrared spectrum of each segment, and obtaining the denoised infrared spectrum; S3: Spectral band coupling positioning and characteristic peak identification: Extract the absorption peak segment and measure the coupling of the de-noised infrared spectrum, and combine the preset threshold and characteristic peak database to obtain the characteristic absorption peak data of harmful organic matter; S4: Partial least squares quantitative analysis: Input the characteristic absorption peak data of harmful organic matter into the prediction model based on the partial least squares regression algorithm to predict the concentration, obtain the concentration detection results of harmful organic matter, and complete the quantitative detection of harmful organic compounds.

2. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 1, characterized in that: The step S1 comprises: S1.1: Preparation of dissolution solution: obtain the sample to be tested and soak it in a pre-prepared acidic or neutral aqueous solution, set the soaking temperature, and soak it for several hours under the soaking temperature condition, so that the harmful organic compounds in the sample to be tested are dissolved, and the soaking is completed; filter or centrifuge the solution obtained after the soaking to obtain the dissolution solution; S1.2: Solution concentration: setting the concentration temperature and concentration multiple, and concentrating the solution by rotary evaporation or reduced pressure concentration to obtain a concentrated solution; S1.3: Baseline correction: dilute or dissolve the concentrated solution with an inert carrier liquid, and then drip or cover it on the ATR crystal to achieve baseline correction and obtain a sample solution to be tested.

3. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 1, characterized in that: In the step S2, an infrared spectrometer is used to collect the original infrared spectrum of the sample solution to be tested in the mid-infrared band, and the original infrared spectrum has noise and background interference; The wavelength range of the mid-infrared band is 2.5 μm to 25 μm.

4. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 1, characterized in that: In step S2, obtaining the denoised infrared spectrum includes the following steps: The collected original infrared spectrum is defined as a vector X = [X(w1), X(w2), ..., X(w n )], where X(w i ) is the wave number position w i The collected spectral intensity data, i = 1, 2, ..., n, n is the maximum wave number; According to the synchronous change characteristics of the spectral absorption peak signal, the original infrared spectrum is divided into K segments with different characteristics, which can be expressed as: X=[X1,X2,...,X K ] Among them, X k is the kth segment, k = 1, 2, ..., K; In each segment, multidimensional harmonic decomposition is performed using sine and cosine harmonic basis functions of different orders. The harmonic basis function φ k (w) specifically: Among them, w min and w max are the minimum and maximum wavenumbers in the kth segment respectively; w is the current wavenumber point; For each segment, the least squares method is used to fit and calculate the harmonic coefficients. The fitting error E of each segment k It is expressed as: Among them, n k is the maximum wave number in the kth segment; X k (w i ) is the wave number position w in the kth segment i The collected spectral intensity data; m k is the harmonic order of the kth segment; M k is the maximum harmonic order of the kth segment; represents the kth segment m k Harmonic basis functions of order; By minimizing the fitting error E of each segment k To find the optimal harmonic coefficients Thus, the fitting results of each segment are obtained and the infrared spectra of each segment are reconstructed; Add the fitting results of each segment to get the fitting result of the entire spectrum It is expressed as: The fitting result of the whole spectrum As the denoised infrared spectrum.

5. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 4, characterized in that: The step S3 comprises the following steps: Extract several sections Δw where there may be absorption peaks of harmful organic matter in the de-noised infrared spectrum i =1,2,…,L, L is the maximum number of segments; Set the following coupling metric function to measure the coupling between the segments Δw i Whether the peak intensity in the sample conforms to the typical characteristic absorption pattern: in, is the spectral intensity at wave number w in the denoised infrared spectrum; w i (w) is the segment Δw i The weighting function of Calculate the coupling metric function value C of each segment i , when C i When it is greater than or equal to the preset characteristic peak threshold, the corresponding segment Δw i There are absorption peaks of harmful organic matter inside; In the section where there is an absorption peak of harmful organic matter, Δw i Inside, look for The maximum point w * , by comparing the known characteristic peak database to identify the characteristic peak, determine the maximum point w * The harmful organic matter category corresponding to the absorption peak at is determined, and the characteristic absorption peak data of the harmful organic matter of this category is obtained.

6. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 5, characterized in that: The weighting function w i (w) is set according to the absorption peak position and shape of the known target compound, and the closer the wave number is to the peak center, the higher the weight value; The characteristic peak threshold is the basic spectral signal intensity obtained based on blank baseline or prior noise level statistics.

7. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 1, characterized in that: In step S4, the prediction model is obtained by the following steps: S4.1: Obtain historical data of characteristic absorption peaks of several hazardous organic substances of known concentrations and standardize them; S4.2: Take the standardized characteristic absorption peak historical data as input data X in , the known concentration of the standardized harmful organic matter is used as the response variable Y out , calculate the input data X in With the response variable Y out The covariance matrix between ; S4.3: According to the covariance matrix, the input data X is decomposed by singular value decomposition. in Decompose into several principal components, extract the first t principal components, and obtain the principal component matrix T0; S4.4: Using the principal component matrix T0 and the standardized response variable Y out , the regression coefficient β is calculated by the least squares method; S4.5: Construct the following prediction model based on the regression coefficient β: Y predicted =X new ·b Among them, Y predicted is the predicted concentration of harmful organic matter; X new It is the characteristic absorption peak data of the harmful organic matter to be detected.

8. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 7, characterized in that: In step S4.1, standardization is performed according to the following formula: Among them, X0 and Y0 are the historical data of characteristic absorption peaks and known concentrations of harmful organic matter respectively; μ x and μ y are the historical data of characteristic absorption peak and the mean of known concentration respectively; σ x and σ y are the variances of characteristic absorption peak historical data and known concentrations respectively.

9. The method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to claim 7, characterized in that: In step S4.4, the regression coefficient β is calculated according to the following formula: Where T represents the matrix transpose; (·) -1 Represents the inverse matrix.

10. A method for detecting harmful organic matter by infrared spectrum based on segmented multidimensional harmonic decomposition according to any one of claims 1 to 9, characterized in that: The method is used for qualitative and quantitative analysis of harmful organic compounds in tableware coatings.

Images