Rapid detection method for ammonia nitrogen content of sewage generated in seasoning production process

Abstract

The invention discloses a method for rapidly detecting the ammonia nitrogen content of sewage generated in a seasoning production process. The method comprises the following steps: (1) collecting a sewage sample, and filtering; (2) scanning the sample by using an intermediate infrared spectrometer to obtain an infrared wave number-absorbance spectrogram; (3) carrying out numerical processing; (4) determining the ammonia nitrogen content of the sample according to a national standard method; (5) dividing the samples into a modeling group and a verification group, and establishing an optimal mapping relation by using a PLS algorithm; (6) taking the prediction deviation of the verification group as a standard for evaluating the advantages and disadvantages of the model, and obtaining a stable and reliable prediction model through multiple times of training and verification; and (7) selecting a to-be-detected sewage sample, obtaining the mid-infrared spectrogram of the to-be-detected sewage sample, and substituting the mid-infrared spectrogram into the prediction model to obtain a predicted value of the ammonia nitrogen content of the to-be-detected sewage sample. Compared with a traditional method, the deviation of the method is within + / -10%, chemical reagents are not consumed, the method is environment-friendly, the detection time can be greatly shortened, and a large amount of labor investment can be reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of water quality detection, and in particular relates to a method for quickly detecting the ammonia nitrogen content of wastewater generated in a condiment production process. Background Art

[0002] Condiments such as soy sauce and paste usually require soybeans as the basic raw material. The production process is prone to produce a variety of nitrogen-containing wastewaters, among which ammonia nitrogen is a key indicator in wastewater discharge, and its content needs to be closely monitored to meet discharge requirements.

[0003] At present, for the detection of ammonia nitrogen in sewage, the national standard method (HJ 535-2009 Determination of ammonia nitrogen in water quality Nessler reagent spectrophotometry) is usually used in the laboratory. This method requires the consumption of mercury-containing chemical reagents, has a long detection cycle, and cannot meet the needs of rapid detection. Therefore, it is necessary to develop a method that can quickly, reagent-free, and accurately detect the ammonia nitrogen content in sewage. Summary of the invention

[0004] The purpose of the present invention is to provide a rapid detection method for the ammonia nitrogen content of wastewater generated in the condiment production process. Compared with the traditional method, the method has a deviation within ±10%, does not consume chemical reagents, is environmentally friendly, can greatly reduce the detection time, and can also reduce a large amount of labor input.

[0005] The above-mentioned object of the present invention can be achieved by the following technical scheme: A method for quickly detecting the ammonia nitrogen content of wastewater generated in the condiment production process, comprising the following steps:

[0006] (1) Collect wastewater samples of different types and concentrations generated at wastewater discharge points at different production stages in the entire condiment production process to ensure the comprehensiveness and representativeness of the data, and filter the samples;

[0007] (2) scanning each sample in step (1) using a mid-infrared spectrometer to obtain a mid-infrared wavenumber-absorbance spectrum of the sewage sample;

[0008] (3) performing numerical processing on the mid-infrared spectrum of the sewage sample obtained in step (2), that is, converting each wavenumber-absorbance information in the characteristic spectrum into a string of wavenumber value-absorbance;

[0009] (4) Determine the ammonia nitrogen content of each sample in step (1) according to the national standard method;

[0010] (5) Dividing the samples into a modeling group and a validation group, using the PLS algorithm to extract the principal components in the modeling group data, using the mid-infrared spectrum data after the digitization processing in step (3) as the independent variable, and the ammonia nitrogen content measured in step (4) as the dependent variable, and establishing the optimal mapping relationship between the infrared spectrum data after the digitization processing and the ammonia nitrogen content;

[0011] (6) Using the prediction deviation of the validation group in step (5) as a criterion for evaluating the quality of the model, a stable and reliable prediction model is obtained through multiple training and validation;

[0012] (7) Select a wastewater sample to be tested, scan the sample using a mid-infrared spectrometer to obtain a mid-infrared spectrum of the wastewater sample to be tested, perform digitization on the mid-infrared spectrum of the wastewater sample to be tested, substitute it into the prediction model of step (6), and obtain a predicted value of the ammonia nitrogen content of the wastewater sample to be tested.

[0013] In the rapid detection method of ammonia nitrogen content in wastewater generated in the above condiment production process:

[0014] Preferably, the number of sewage samples in step (1) is 150 to 250.

[0015] Preferably, a 600-mesh filter is used to filter the sample in step (1). The sample is filtered through a 600-mesh filter. The filter with this pore size can remove impurities that affect the transmittance of mid-infrared light while retaining the effective information of the sample to the greatest extent.

[0016] Preferably, when each sample in step (1) is scanned by a mid-infrared spectrometer in step (2), the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, the filtered sample is statically scanned 8 times, the absorbance of the 8 times is averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

[0017] Preferably, in step (3), each wavenumber-absorbance information in the characteristic spectrum is converted into a 7-digit string of wavenumber value-absorbance, such as the absorbance at wavenumber 1110 is 0.76, which is a string of 1110076.

[0018] Preferably, in step (4), the ammonia nitrogen content of each sample in step (1) is measured according to the national standard method HJ 535-2009.

[0019] Preferably, the ratio of the modeling group to the validation group in step (5) is 8:2.

[0020] Preferably, in step (5), the PLS algorithm is used to first standardize the data, that is, the independent variable and dependent variable data are respectively deducted by the mean (the respective means, that is, each independent variable is deducted from the mean value of the independent variable set, and the dependent variable is the same) to achieve data standardization, and the first principal components t1 and u1 are extracted by the NIPALS iterative algorithm, and the second, third and more principal components are extracted by repeated iterations. The first 6-10 principal components are selected to establish a regression model, and the standard deviation of the verification group results is used to evaluate the effect of the regression model.

[0021] Preferably, the prediction deviation in step (6) is within 10%, and the prediction deviation = (prediction value - national standard test value) × 100% / national standard test value.

[0022] Preferably, the prediction model in step (6) includes the selected spectral segment, the number of principal component factors, the weight of each factor and the bias correction.

[0023] The present invention has the following advantages: Through experiments, the present invention found that mid-infrared spectroscopy exhibits significant technical advantages in ammonia nitrogen detection, including wavelength selectivity and avoidance of water absorption, wide wavelength range and sensitivity, excellent penetration depth, high resolution and strong anti-interference ability. These advantages make mid-infrared spectroscopy an ideal choice in the field of ammonia nitrogen detection, especially in complex application scenarios such as condiment sewage treatment, where its advantages are more prominent, as shown below:

[0024] 1) Significant advantages over traditional chemical methods

[0025] Traditional ammonia nitrogen detection methods mainly rely on chemical reagents such as Nessler's reagent. These reagents not only pose safety hazards during the operation process, but long-term exposure may also cause health and ecological risks to operators and the environment. In contrast, the method of combining mid-infrared spectroscopy with PLS algorithm completely abandons the use of harmful chemical reagents and significantly improves the safety of the detection process. In addition, this method simplifies the cumbersome pre-treatment, experimental operation and data analysis processes, greatly shortens the detection cycle, reduces costs, and is more environmentally friendly and avoids the generation of harmful waste liquids.

[0026] 2) Advantages compared to general optical (visible light, near-infrared light) rapid detection methods

[0027] a) Wavelength selectivity and avoidance of moisture absorption

[0028] Mid-infrared spectroscopy is rich in molecular fundamental frequency vibration information. Compared with visible light and near-infrared light, it is more characteristic and more suitable for multi-component analysis. More importantly, in sewage samples containing a large amount of water, mid-infrared spectroscopy can effectively avoid the main peak area of ​​water absorption and reduce the interference of water on spectral information. It has more available information wavelength bands and can significantly reduce the interference of combined frequency vibration compared to near-infrared light and visible light, thereby significantly improving the efficiency and accuracy of ammonia nitrogen detection.

[0029] b) Wavelength range and sensitivity

[0030] The wavelength range of mid-infrared spectroscopy (2.5 to 25 microns) covers a wide range of vibration and rotational energy level transitions of organic molecules. This feature makes mid-infrared spectroscopy extremely sensitive and selective in ammonia nitrogen detection, and can accurately capture the spectral characteristics of ammonia nitrogen molecules and achieve quantitative detection of ammonia nitrogen indicators.

[0031] Mid-infrared spectroscopy originates from the fundamental vibration of molecules, that is, the simple harmonic vibration of molecules near the equilibrium position. This fundamental vibration spectrum contains the basic information of molecular vibration, and its absorption intensity is linearly related to the content of the corresponding molecule; at the same time, compared with the near-infrared band, the mid-infrared band has fewer harmonic vibrations, that is, less noise interference, and is more suitable for quantitative analysis of material content.

[0032] In contrast, near-infrared spectra are mainly produced by the doublet and sum frequency of molecules. These doublet and sum frequency information contain information about the composition and molecular structure of most types of organic compounds and are more suitable for qualitative analysis.

[0033] c) Penetration Depth

[0034] Compared with visible light or near-infrared spectroscopy, mid-infrared spectroscopy has deeper penetration ability. This excellent penetration allows mid-infrared spectroscopy to penetrate deep into sewage samples. Even when facing turbid and opaque samples, sufficient spectral information can be obtained, improving the versatility of the detection method.

[0035] d) Anti-interference ability

[0036] Mid-infrared spectroscopy is based on the fundamental frequency vibration principle of molecules and has strong resistance to interference from factors such as sample temperature and transparency. During the ammonia nitrogen detection process, even if the sample conditions change, mid-infrared spectroscopy can maintain stable detection results, ensuring the reliability and accuracy of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for rapidly detecting the ammonia nitrogen content of wastewater generated in the condiment production process in Examples 1-4 of the present invention;

[0038] Figure 2It is a collection of infrared spectra of 150 sewage samples in Example 1 of the present invention. The blue line is the average line of the sample set, and the yellow line is the difference between the samples (i.e., the standard deviation of the absorbance of the sample set at this wave number point);

[0039] Figure 3 is the sewage ammonia nitrogen prediction model for sewage discharge points at different production stages in Example 1 of the present invention, R 2 =0.964 (number of principal component factors = 8), RMSEP = 0.20;

[0040] Figure 4 The predicted value and chemical value bias distribution of the sewage ammonia nitrogen prediction model at different production stages of sewage discharge points in Example 1 of the present invention (referring to the distribution of deviations of multiple verification samples, and the deviation refers to the deviation when the modeling is completed and verified), the bias value is -0.03 (the bias value is the average value of the deviation values ​​of 30 groups of comparison data);

[0041] Figure 5 is the prediction model of ammonia nitrogen in nitrogen-containing wastewater generated in the soybean fermentation liquid production process in Example 2 of the present invention, R 2 =0.934 (number of principal component factors = 10), RMSEP = 0.47;

[0042] Figure 6 The biased distribution of the predicted value and chemical value of nitrogen-containing wastewater generated in the soybean fermentation liquid production process in Example 2 of the present invention, with an offset value of +0.08;

[0043] Figure 7 is the pipeline flushing water ammonia nitrogen prediction model in Example 3 of the present invention, R 2 =0.956 (number of principal component factors = 6), RMSEP = 0.13;

[0044] Figure 8 is the biased distribution of the predicted value and chemical value of nitrogen-containing wastewater generated in the pipeline flushing water in Example 3 of the present invention, and the offset value is +0.01;

[0045] Fig. 9 is the prediction model of ammonia nitrogen in wastewater from seasoning sauces in Example 4 of the present invention, R 2 =0.92 (number of principal component factors = 9), RMSEP = 0.23;

[0046] Fig.10 The deviation distribution of the predicted value and chemical value of the seasoning sauce wastewater in Example 4 of the present invention, the deviation value is -0.07;

[0047] Fig.11 This is a typical absorption spectrum of the mid-infrared liquid sample in Comparative Example 1 of the present invention;

[0048] Fig.12This is a typical near-infrared absorption spectrum of the liquid sample in Comparative Example 1 of the present invention;

[0049] Fig.13 This is the mid-infrared absorption spectrum of the high-concentration sewage at the discharge port in Comparative Example 2 of the present invention. The sample information can be fully obtained, and there is no situation where the absorption value is too high and the information is lost. DETAILED DESCRIPTION

[0050] In order to better explain the device of the present invention, the specific implementation methods under some actual conditions targeted by the present invention are further described below. It should be noted that the description of these implementation methods is used to help understand the present invention, but does not constitute a limitation of the present invention. In addition, the technical features involved in each implementation method of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0051] Example 1

[0052] like Figure 1 As shown, the method for rapid detection of ammonia nitrogen content in wastewater generated in the condiment production process provided in this embodiment includes the following steps:

[0053] (1) 150 wastewater samples of different types and concentrations generated at wastewater discharge points at different production stages in the entire condiment production process were collected to ensure the comprehensiveness and representativeness of the data. The samples were filtered through a 600-mesh filter. The filter with this pore size can remove impurities that affect the transmittance of mid-infrared light while retaining the effective information of the sample to the greatest extent.

[0054] (2) Scan each sample in step (1) using a mid-infrared spectrometer to obtain a mid-infrared wavenumber-absorbance spectrum of the sewage sample, such as Figure 2 As shown;

[0055] In step (2), when each sample in step (1) is scanned by a mid-infrared spectrometer, the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, and the sample filtered through a 600-mesh filter is statically scanned 8 times. The absorbances of the 8 times are averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

[0056] (3) performing numerical processing on the mid-infrared spectrum of the sewage sample obtained in step (2);

[0057] In step (3), the mid-infrared spectrum of the sewage sample obtained in step (2) is digitized, that is, each wave number-absorbance information in the characteristic spectrum is converted into a 7-digit string of "wave number value-absorbance", such as the absorbance at wave number 1110 is 0.76, then it is 1110076 string;

[0058] (4) Determine the ammonia nitrogen content of each sample in step (1) according to the national standard method;

[0059] (5) Dividing the samples into a modeling group and a validation group, using the PLS algorithm to extract the principal components in the modeling group data, using the mid-infrared spectrum data after the numerical processing in step (3) as the independent variable, and the ammonia nitrogen content measured in step (4) as the dependent variable, and establishing the optimal mapping relationship (regression model) between the infrared spectrum data after the numerical processing and the ammonia nitrogen content;

[0060] In step (5), the PLS algorithm is used to first standardize the data (i.e., the means of the independent variable and dependent variable data are deducted from their respective sets), and the first principal components t1 and u1 are extracted through the NIPALS iterative algorithm. The second, third and more principal components are extracted repeatedly, and the first 8 principal components are selected to establish a regression model. The standard deviation of the verification group results is used to evaluate the effect of the regression model.

[0061] (6) Using the prediction deviation of the validation group in step (5) as the criterion for evaluating the quality of the model, a stable and reliable prediction model is obtained through multiple training and validation, such as Figure 3 As shown;

[0062] The prediction model includes the selected spectral segment, the number of principal component factors, the weight of each factor and the bias correction.

[0063] from Figure 2 As can be seen in the figure, the spectral segments were selected based on their relevance and information content for the prediction of ammonia nitrogen content. The wavenumbers were 1000-1800 cm -1 , 2500-3000cm -1 There are relatively rich information differences (the yellow Std line is the absorption difference between samples. Sample difference information can be extracted from this area to facilitate model establishment). The number of principal component factors is usually selected from the first 6 to 10, and generally does not exceed 10, to ensure the prediction ability of the model and avoid overfitting. The weight of each factor is iteratively calculated by the PLS algorithm. The bias is calculated based on the data of the validation group to verify the overall deviation of the data. By introducing the intercept value (that is, the average deviation value of the validation group data, Figure 4 The final fine-tuning correction is done by using the offset value in Figure 4 shown.

[0064] The relative deviation value calculation formula is: (the detection value of the method of the present invention - the detection value of the national standard) × 100% / the detection value of the national standard.

[0065] Table 1 Example 1 Validation Group Data and Deviation

[0066]

[0067]

[0068] The relative deviation in Table 1 is distributed within ±3.9%, which meets the method comparison requirements (relative deviation < 10%).

[0069] (7) 12 wastewater samples to be tested were selected, and the samples were scanned using a mid-infrared spectrometer to obtain mid-infrared spectra of the wastewater samples to be tested. The mid-infrared spectra of the wastewater samples to be tested were digitized and substituted into the prediction model of step (6) to obtain the predicted values ​​of the ammonia nitrogen content of the wastewater samples to be tested, as shown in Table 2.

[0070] Table 2 Application effect of Example 1 (compared with national standard method)

[0071]

[0072] Example 2

[0073] like Figure 1 As shown, in this embodiment, the method of the present invention is used to quickly detect nitrogen-containing wastewater generated during the production of soybean fermentation liquid:

[0074] (1) Five types and concentrations of wastewater samples were collected from the wastewater treatment systems of discharge points at different production stages, with 30-50 samples of each type, for a total of 197 samples.

[0075] These samples cover the entire production process from raw material pretreatment, fermentation, extraction to final discharge, ensuring the comprehensiveness and representativeness of the data.

[0076] The sample was filtered through a 600-mesh filter. This pore size filter can remove impurities that affect the transmittance of mid-infrared light while retaining the effective information of the sample to the greatest extent.

[0077] (2) scanning each sample using a mid-infrared spectrometer to obtain a mid-infrared spectrum of the sample;

[0078] In step (2), when each sample in step (1) is scanned by a mid-infrared spectrometer, the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, and the sample filtered through a 600-mesh filter is statically scanned 8 times. The absorbances of the 8 times are averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

[0079] (3) numerically processing the spectrum;

[0080] The mid-infrared spectrum of the sewage sample obtained in step (2) is digitized, that is, each wave number-absorbance information in the characteristic spectrum is converted into a 7-digit string of wave number value-absorbance, such as 1110076 characters for the absorbance at wave number 1110;

[0081] (4) At the same time, the ammonia nitrogen content of each sewage sample was measured according to the national standard method (HJ 535-2009).

[0082] (5) After obtaining 197 mid-infrared spectra and ammonia nitrogen data, the samples were randomly divided into a modeling group (157 samples) and a validation group (40 samples) in a ratio of 8:2.

[0083] The PLS algorithm is used to extract the principal components in the modeling group data, analyze the relationship between the independent variables and the dependent variables, and establish a prediction model.

[0084] In this embodiment, the digitized mid-infrared spectrum is used as the independent variable, and the ammonia nitrogen content is used as the dependent variable, and the optimal mapping relationship between them is found through the PLS algorithm.

[0085] Using the PLS algorithm, the data was first standardized (i.e., the means of the independent and dependent variable data were deducted from their respective sets), and the first principal components t1 and u1 were extracted through the NIPALS iterative algorithm. The second, third and more principal components were extracted repeatedly, and the first 10 principal components were selected to establish a regression model. The standard deviation of the verification group results was used to evaluate the effectiveness of the regression model.

[0086] The main function of extracting principal components in the PLS algorithm is to reduce dimension, extract information, and deal with multicollinearity.

[0087] (6) Using the prediction deviation of the validation group as the criterion for evaluating the quality of the model, a stable and reliable prediction model was obtained through multiple training and validation, such as Figure 5 shown.

[0088] The prediction model includes the selected spectral segment, the number of principal component factors, the weight of each factor and the bias correction.

[0089] from Figure 2 It can be seen that the selection of spectral segments is based on their relevance and information content for the prediction of ammonia nitrogen content. The wavenumbers 1000-1800 and 2500-3000 have richer information differences (the yellow Std line is the absorption difference between samples. Sample difference information can be extracted from this area to facilitate model establishment). The number of principal component factors is usually selected from the first 6-10, and generally does not exceed 10, to ensure the prediction ability of the model and avoid overfitting. The weight of each factor is iteratively calculated by the PLS algorithm. The bias is calculated based on the data of the validation group to verify the overall deviation of the data. The final fine-tuning correction is made by introducing the intercept value (that is, the average deviation value of the validation group data, the offset distribution diagram below each model diagram), as shown in the figure. Figure 6 shown.

[0090] The spectral segment is usually the full mid-infrared spectrum, that is, the wavelength segment of 2.5 to 25 μm. The weak information wavenumber segment with absorbance greater than 2.0 is eliminated. The number of principal component factors is determined by the standard deviation value of the validation group, and the number of principal component factors with the minimum deviation value is selected. Generally, a few principal components that can explain most of the variance are selected.

[0091] In this embodiment, the prediction deviation of the method of the present invention is within ±6.2% compared with the traditional national standard method. The verification data is shown in Table 3 below.

[0092] The relative deviation value calculation formula is: (the detection value of the method of the present invention - the detection value of the national standard) × 100% / the detection value of the national standard.

[0093] The relative deviation of this embodiment is distributed within ±6.9%, which meets the method comparison requirement (the comparison requirement is met when the relative deviation is <10%).

[0094] Table 3: Example 2 Validation Group Data and Deviations

[0095]

[0096] (7) Detection application: After 16 unknown sewage samples were sent for inspection, the samples were scanned using a mid-infrared spectrometer to obtain a mid-infrared spectrum. The spectrum was digitized and substituted into the above-verified prediction model to get the predicted value of the ammonia nitrogen content in the sewage, as shown in Table 4 below.

[0097] Table 4 Application effect of Example 2 (compared with national standard method)

[0098]

[0099] Example 3

[0100] In food production, pipeline flushing water is a common nitrogen-containing wastewater. This embodiment uses the method of the present invention to quickly detect the ammonia nitrogen content of pipeline flushing water (pipes that convey soy sauce and fermented semi-finished liquids) to improve detection efficiency, reduce detection costs, and provide a basis for pipeline cleaning effects during the production process.

[0101] (1) 185 samples were collected from discharge outlets at different production stages. The samples covered the entire production process from pipeline flushing to final wastewater discharge in multiple production process stages, ensuring the comprehensiveness and representativeness of the data.

[0102] The sample was filtered through a 600-mesh filter. This pore size filter can remove impurities that affect the transmittance of mid-infrared light while retaining the effective information of the sample to the greatest extent.

[0103] (2) scanning each collected sample using a mid-infrared spectrometer to obtain a mid-infrared spectrum of each sample;

[0104] In step (2), when each sample in step (1) is scanned by a mid-infrared spectrometer, the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, the filtered sample is statically scanned 8 times, the absorbance of the 8 times is averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

[0105] (3) These spectra are digitized for subsequent analysis.

[0106] In step (3), each wavenumber-absorbance information in the characteristic spectrum is converted into a 7-digit string of wavenumber value-absorbance.

[0107] (4) At the same time, the ammonia nitrogen content of each sewage sample is accurately measured according to the national standard method (HJ 535-2009) as reference data.

[0108] (5) After obtaining 185 mid-infrared spectra and ammonia nitrogen data, the samples were randomly divided into a modeling group (148 samples) and a validation group (37 samples) in a ratio of 8:2.

[0109] The partial least squares (PLS) algorithm was used to extract the principal components in the modeling group data, and the relationship between the independent variable (the digitized mid-infrared spectrum) and the dependent variable (ammonia nitrogen content) was analyzed to establish a prediction model.

[0110] In step (5), the PLS algorithm is used to first standardize the data (i.e., the means of the independent variable and dependent variable data are deducted from their respective sets), and the first principal components t1 and u1 are extracted through the NIPALS iterative algorithm. The second, third and more principal components are extracted repeatedly, and the first 6 principal components are selected to establish a regression model. The standard deviation of the verification group results is used to evaluate the effect of the regression model.

[0111] (6) The prediction deviation of the validation group is used as the criterion for evaluating model performance. Through multiple model training and validation, a stable and reliable prediction model is obtained, such as Figure 7 shown.

[0112] The model information includes key parameters such as the selected spectral segment, the number of principal component factors, the weight of each factor, and bias correction.

[0113] from Figure 2It can be seen that the selection of spectral segments is based on their relevance and information content for the prediction of ammonia nitrogen content. The wavenumbers 1000-1800 and 2500-3000 have richer information differences (the yellow Std line is the absorption difference between samples. Sample difference information can be extracted from this area to facilitate model establishment). The number of principal component factors is usually selected from the first 6 to 10, and generally does not exceed 10, to ensure the prediction ability of the model and avoid overfitting. The weight of each factor is iteratively calculated by the PLS algorithm. The bias is calculated based on the data of the validation group to verify the overall deviation of the data. The final fine-tuning correction is made by introducing the intercept value (that is, the average deviation value of the validation group data), as shown in the figure. Figure 8 shown.

[0114] In this embodiment, the prediction deviation of the method of the present invention is within ±7.8% compared with the traditional national standard method, showing good prediction accuracy, as shown in Table 5.

[0115] The relative deviation value calculation formula is: (patent method test value - national standard test value) / national standard test value

[0116] Relative deviation distribution is within ±7.8%, meeting the method comparison requirements (relative deviation <10%)

[0117] Table 5: Validation group data and deviation

[0118]

[0119]

[0120] (7) Detection application: For the 10 unknown pipe flushing water samples received, the mid-infrared spectrometer was used to scan the samples, and the infrared spectra were obtained, and the spectra were processed into data. Then, the processed data was substituted into the above-mentioned established model for calculation, and the predicted value of the ammonia nitrogen content in the sewage was quickly obtained. The results are shown in Table 6. This method greatly shortens the detection time and improves the detection efficiency, providing an effective means for the rapid monitoring of nitrogen-containing wastewater in pipe flushing water.

[0121] Table 6 Application effect of Example 3 (compared with national standard method)

[0122]

[0123] Example 4

[0124] In the process of producing sauce, the wastewater produced usually contains a high concentration of ammonia nitrogen. In order to achieve rapid detection of ammonia nitrogen content in such wastewater, improve detection efficiency and reduce detection cost, this embodiment adopts the method of the present invention to perform rapid detection of ammonia nitrogen content in wastewater produced by sauce production, thereby providing an effective basis for wastewater control in the production process.

[0125] (1) Data Collection

[0126] A total of 168 wastewater samples were collected from the discharge port of the sauce production workshop. These samples covered the entire process from container cleaning, product packaging cleaning, to final wastewater discharge, ensuring the comprehensiveness and representativeness of the data.

[0127] The sample was filtered through a 600-mesh filter. This pore size filter can remove impurities that affect the transmittance of mid-infrared light while retaining the effective information of the sample to the greatest extent.

[0128] (2) Spectral scanning and reference data determination

[0129] (2.1) Use a mid-infrared spectrometer to perform a detailed scan on each collected sample to obtain a mid-infrared spectrum of each sample;

[0130] When each sample in step (1) is scanned by a mid-infrared spectrometer, the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, the filtered sample is statically scanned 8 times, the absorbance of the 8 times is averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

[0131] (2.2) These spectra are converted into numerical data to facilitate subsequent data processing and analysis.

[0132] Convert each wavenumber-absorbance information in the characteristic spectrum into a 7-digit string of wavenumber value-absorbance. For example, if the absorbance at wavenumber 1110 is 0.76, it will be 1110076 characters.

[0133] (2.3) At the same time, according to relevant national standards (such as HJ535-2009), the ammonia nitrogen content of each sewage sample was accurately measured as reference data for subsequent modeling.

[0134] 3. Modeling and Verification

[0135] (3.1) After obtaining 168 mid-infrared spectra and ammonia nitrogen content data, the samples were randomly divided into a modeling group (135 samples) and a validation group (33 samples) in a ratio of approximately 8:2.

[0136] The partial least squares (PLS) algorithm was used to extract the principal components of the data of the modeling group, and the relationship between the independent variable (the digitized mid-infrared spectrum) and the dependent variable (ammonia nitrogen content) was studied to establish a prediction model.

[0137] Using the PLS algorithm, the data was first standardized (i.e., the means of the independent and dependent variable data were deducted from their respective sets), and the first principal components t1 and u1 were extracted through the NIPALS iterative algorithm. The second, third and more principal components were extracted repeatedly, and the first 9 principal components were selected to establish a regression model. The standard deviation of the verification group results was used to evaluate the effectiveness of the regression model.

[0138] (3.2) The prediction deviation of the validation group is used as the criterion for evaluating the model performance. Through model training and validation, a stable and reliable prediction model is obtained, such as Fig. 9 shown.

[0139] The model information includes key parameters such as the selected spectral segment, the number of principal component factors, the weight of each factor, and bias correction.

[0140] from Figure 2 It can be seen that the selection of spectral segments is based on their relevance and information content for the prediction of ammonia nitrogen content. The wavenumbers 1000-1800 and 2500-3000 have richer information differences (the yellow Std line is the absorption difference between samples. Sample difference information can be extracted from this area to facilitate model establishment). The number of principal component factors is usually selected from the first 6 to 10, and generally does not exceed 10, to ensure the prediction ability of the model and avoid overfitting. The weight of each factor is iteratively calculated by the PLS algorithm. The bias is calculated based on the data of the validation group to verify the overall deviation of the data. The final fine-tuning correction is made by introducing the intercept value (that is, the average deviation value of the validation group data), as shown in the figure. Fig.10 shown.

[0141] Compared with the traditional national standard method, the prediction deviation of the method of this embodiment is within ±5.5%, showing good prediction accuracy, as shown in Table 7 below.

[0142] The relative deviation value calculation formula is: (the detection value of the method of this embodiment - the detection value of the national standard) × 100% / the detection value of the national standard.

[0143] The relative deviation of this embodiment is distributed within ±5.5%, which meets the method comparison requirement (relative deviation <10%).

[0144] Table 7: Validation group data and deviation

[0145]

[0146] 4. Detection application: 10 wastewater samples from the seasoning sauce production process with unknown results were scanned using a mid-infrared spectrometer to obtain the infrared spectra, which were then digitized.

[0147] Then, the processed data is substituted into the above-established model for calculation to obtain the predicted value of the ammonia nitrogen content in the sewage, and the results are shown in Table 8. This method greatly shortens the detection time, improves the detection efficiency, and provides an effective means for monitoring sewage discharge in production workshops.

[0148] Table 8 Application effect of Example 4 (compared with national standard method)

[0149]

[0150] Comparative Example 1

[0151] Different from Example 1, when using near infrared and mid infrared testing:

[0152] The mid-infrared band can bypass the peak area of ​​water absorption and effectively avoid the interference of water on spectral information. In sewage samples containing a large amount of water, the mid-infrared spectrum can more effectively utilize the information wavelength band. Its available information wavelength band is 4 times that of near-infrared, thereby significantly improving the efficiency and accuracy of ammonia nitrogen detection.

[0153] At the same time, compared with the near-infrared band, the mid-infrared band has fewer frequency harmonic vibrations, that is, less noise interference, and is more suitable for quantitative analysis of material content.

[0154] like Figure 11-12 As shown, compared with mid-infrared absorption, near-infrared wavenumber 5000-7500cm -1 ,Due to the strong absorption of water in the samples, there is no available information to extract, ,and the difference information between samples is difficult to identify, ,making modeling difficult.

[0155] Comparative Example 2

[0156] Different from Example 1, chromatogram comparisons of turbid, opaque and clean samples are provided.

[0157] Compared with visible light or near-infrared spectroscopy, mid-infrared spectroscopy has deeper penetration. This excellent penetration allows mid-infrared spectroscopy to penetrate deep into sewage samples. Even when facing turbid and opaque samples, it can obtain sufficient spectral information, improving the versatility of detection methods, such as Fig.13 shown.

[0158] The embodiments of the present invention are described in detail above, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments are made without departing from the principles and spirit of the present invention, and still fall within the protection scope of the present invention.

Claims

1. A rapid detection method for the ammonia nitrogen content of wastewater generated in the condiment production process, characterized in that: The following steps are involved: (1) Collect wastewater samples of different types and concentrations generated at wastewater discharge points at different production stages in the entire condiment production process to ensure the comprehensiveness and representativeness of the data, and filter the samples; (2) scanning each sample in step (1) using a mid-infrared spectrometer to obtain a mid-infrared wavenumber-absorbance spectrum of the sewage sample; (3) performing numerical processing on the mid-infrared spectrum of the sewage sample obtained in step (2), that is, converting each wavenumber-absorbance information in the characteristic spectrum into a string of wavenumber value-absorbance; (4) Determine the ammonia nitrogen content of each sample in step (1) according to the national standard method; (5) Dividing the samples into a modeling group and a validation group, using the PLS algorithm to extract the principal components in the modeling group data, using the mid-infrared spectrum data after the digitization processing in step (3) as the independent variable, and the ammonia nitrogen content measured in step (4) as the dependent variable, and establishing the optimal mapping relationship between the infrared spectrum data after the digitization processing and the ammonia nitrogen content; (6) Using the prediction deviation of the validation group in step (5) as a criterion for evaluating the quality of the model, a stable and reliable prediction model is obtained through multiple training and validation; (7) Select a wastewater sample to be tested, scan the sample using a mid-infrared spectrometer to obtain a mid-infrared spectrum of the wastewater sample to be tested, perform digitization on the mid-infrared spectrum of the wastewater sample to be tested, substitute it into the prediction model of step (6), and obtain a predicted value of the ammonia nitrogen content of the wastewater sample to be tested.

2. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: The number of sewage samples in step (1) is 150 to 250.

3. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: When filtering the sample in step (1), a 600-mesh filter is used to filter the sample.

4. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: In step (2), when each sample in step (1) is scanned by a mid-infrared spectrometer, the spectral range is 2.5 to 25 μm, the sampling interval is 10 nm, the filtered sample is statically scanned 8 times, the absorbance of the 8 times is averaged, and then the reference absorbance with pure water as the blank is deducted to obtain the wavenumber-absorbance characteristic spectrum of the sample.

5. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: In step (3), each wavenumber-absorbance information in the characteristic spectrum is converted into a 7-digit string of wavenumber value-absorbance.

6. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: In step (4), the ammonia nitrogen content of each sample in step (1) is measured according to the national standard method HJ 535-2009.

7. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: The ratio of the modeling group to the validation group in step (5) is 8:

2.

8. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: In step (5), the PLS algorithm is used to standardize the data first, and then the first principal components t1 and u1 are extracted through the NIPALS iterative algorithm. The second, third and more principal components are extracted repeatedly. The first 6-10 principal components are selected to establish a regression model, and the standard deviation of the validation group results is used to evaluate the effect of the regression model.

9. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: The prediction deviation in step (6) is within 10%, and the prediction deviation = (prediction value - national standard test value) × 100% / national standard test value.

10. The rapid detection method for ammonia nitrogen content in wastewater generated in the condiment production process according to claim 1, characterized in that: The prediction model described in step (6) includes the selected spectral segment, the number of principal component factors, the weight of each factor and the bias correction.

Images