A method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy

Abstract

The present invention belongs to the technical field of measuring nitrate nitrogen concentration in water bodies, and specifically relates to a method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy. Fourier transform infrared attenuated total reflection (FTIR-ATR) is used as a technical research method, and the partial least squares method is used to model the infrared spectrum after deconvolution. The application potential of the model is determined by the model evaluation index, thereby indicating whether rapid detection of nitrate nitrogen in industrial wastewater can be achieved. The method can quickly and accurately determine the content of nitrate nitrogen in polluted water bodies, monitor in real time, and has good application prospects.

Description

Technical Field

[0001] The invention belongs to the technical field of measuring nitrate nitrogen concentration in water bodies, and particularly relates to a method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy. Background Art

[0002] Industrial wastewater has complex components, high concentrations of pollutants, and contains a large amount of salts and organic matter, among which nitrate is one type of pollutant. The discharge of wastewater containing nitrate will cause eutrophication of water bodies, reduce the oxygen content in water, affect the growth of aquatic organisms, and thus damage the ecological environment of water bodies. Therefore, effectively removing nitrates from polluted wastewater is the key to governance, and real-time monitoring of nitrate content in wastewater is a prerequisite.

[0003] The main methods for determining nitrate in water bodies include wet chemical method, high performance liquid chromatography, chemiluminescence method, etc. The common methods for determining nitrate nitrogen ion concentration mainly include ion-specific electrode method, reduction distillation method and colorimetry, among which colorimetry is more widely used. Phenol disulfonic acid method was once widely used, but this method takes a long time to analyze and is affected by organic matter and chloride ions (Cl - ) interference, it is impossible to analyze a large number of samples in a short period of time, and it is now rarely used. - -N results are low, due to the interference of metal ions, the preparation and aging process of cadmium columns are cumbersome, and each batch of cadmium columns needs to be calibrated and activated, which is not suitable for the determination of large quantities of samples. In addition, the use of cadmium powder can easily cause environmental pollution. Ion chromatography can determine the content of several anions at the same time, but when the ion concentration is high, the sample needs to be diluted to an appropriate multiple, which takes a long time. Gas phase molecular absorption spectroscopy is relatively simple and fast, but it is easily affected by NO2 - 、SO2 3- and S2O2 3- Positive interference, high-valent cations and organic matter will also affect the detection. In the thymol (thymol) spectrophotometry, nitrite and chloride will interfere. At the same time, when thymol solution is prepared with anhydrous ethanol, the result is very likely to be low because ethanol is volatile. Ultraviolet spectrophotometry has strong anti-interference ability, short analysis cycle, and low pollution of chemical reagents. It is suitable for the determination of large quantities of samples, but nitrite, bromide, carbonate, and bicarbonate will affect the determination and require appropriate pretreatment; and when twice the absorbance at 275nm is greater than 10% of the absorbance at 220nm, the method will no longer be applicable.

[0004] In recent years, flow injection analysis has been widely used and promoted due to its advantages such as fast analysis speed, high accuracy and precision, strong versatility, high efficiency, low consumption, and the ability to be combined with a variety of detection methods, but the instrument is expensive. In summary, in view of the characteristics of the above various analysis methods, in order to respond to environmental problems in a timely manner, Fourier transform infrared spectroscopy (FTIR-ATR) is used to achieve non-destructive and efficient monitoring and determination of nitrate nitrogen in industrial wastewater, providing support for the treatment of nitrate industrial wastewater.

[0005] In recent years, with the intensification of human activities, more nitrogen has entered the environment in the form of reactive nitrogen, nitrate (NO3 - -N) is an important form of active nitrogen that causes eutrophication of water bodies and harms human health. Therefore, rapid and timely monitoring of nitrate in water bodies is an important basis for water environmental management. The determination of nitrate by traditional chemical analysis methods requires pretreatment, the use of chemical reagents and easy destruction of samples, while spectral technology is not only simple to operate and fast to measure, but also can achieve in-situ measurement and real-time tracking, which is very beneficial for the study of the dynamic process and mechanism of physical or chemical changes of environmental substances. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides a method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy. The method can quickly and accurately determine the content of nitrate nitrogen in polluted water and monitor it in real time, and has good application prospects.

[0007] In order to solve the problems of the prior art, the technical solution adopted by the present invention is:

[0008] The method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy comprises the following steps:

[0009] Step 1: Collect water samples and add them directly to the ATR crystal tank. Set the spectral scanning range to 4000-500cm -1 , directly added to the ATR crystal tank, and the spectrum scanning range was set to 4000~650cm -1 , 32 repeated scans were continuously recorded to record the FTIR-ATR spectra;

[0010] Step 2: Import the original spectral data into Peakfit 4.12 software and perform smoothing preprocessing on the data. In order to maximize the elimination of the effect of water on the nitrate spectral absorption, the 1500-1200cm -1 The band spectrum was subjected to Gaussian deconvolution, baseline correction was performed, and the spectrum peak fitting parameters were optimized to improve the goodness of fit.

[0011] in,

[0012]

[0013] During the deconvolution process, the goodness of fit R 2 All of them are above 0.97, and the error sum of squares is between 0.002183-0.005632, and the fitting effect is good; after deconvolution, 1360cm -1 An obvious characteristic peak can be observed at , which is the characteristic absorption peak of nitrate;

[0014] Step 3, using partial least squares method to model the infrared spectrum after deconvolution, and using cross-validation method to determine the optimal number of principal components;

[0015] Step 4: Use the model established in step 3 to divide the entire data set into a training set and a test set. According to the goodness of fit R of the PLS model training set and test set, 2 The RPD value is used to evaluate the prediction effect of the model, where the training set accounts for 75% of the entire data set and the test set accounts for 25% of the entire data set.

[0016] As an improvement, the scanning parameters in step 1 are 4 cm resolution. -1 , the mirror speed is 0.4747cm· -1 s.

[0017] As an improvement, the specific steps of deconvolution of the spectral curve described in step 2 are as follows:

[0018] Assuming that the experimental spectrum Y(x) is composed of several single-peak spectral bands superimposed on each other, the purpose of fitting is to find the single-peak spectral band F i (x)(i=1,2,…,n), the principle is as follows:

[0019] Y(x)=ΣFi(x). (1)

[0020] Where Y is the spectrum, x is the wave number, i is the number of independent peaks, and F is the kernel function of Gaussian deconvolution:

[0021]

[0022] where a0, a1 and a2 represent the peak amplitude, position and width respectively, and x and y are the wave number and absorption intensity respectively.

[0023] As an improvement, the modeling model described in step 3 is a bilinear model, including a matrix X of variables (spectral wavenumbers) and a function matrix Y of variables X, which is used to predict the minimum number of potential variables, and the model evaluation uses R 2 , RMSE and RPD evaluate the predictive ability of the model, and high R 2 and RPD values, as well as low RMSE indicate a robust and accurate model.

[0024] A further improvement is that R 2 The value, RPD value, and RMSE are specifically defined by the following formulas:

[0025]

[0026]

[0027] Among them, y i and are the measured value and predicted value of nitrate in the ith wastewater sample, is the average value of nitrate determination, n is the number of samples, SD is the standard deviation, RMSE is C and RMSE V They represent the root mean square error in the calibration and validation data set models, respectively. When RPD<1.8, it means that the model cannot be used for quantitative determination. When the RPD value is between 2-2.5, it means that the model prediction performance is good. When RPD>3, it means that the model prediction performance reaches an excellent level.

[0028] Beneficial effects:

[0029] Compared with the prior art, the present invention discloses a method for online detection of nitrate-nitrogen in sewage based on mid-infrared total reflectance spectroscopy, which takes Fourier transform infrared attenuated total reflectance (FTIR-ATR) as a technical research means, adopts partial least squares method to model the infrared spectrum after deconvolution, and determines the application potential of the model through model evaluation index, thereby indicating whether rapid detection of nitrate-nitrogen in industrial wastewater can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is the concentration of nitrate nitrogen in polluted water;

[0031] Figure 2 The mid-infrared attenuated total reflection spectrum of industrial wastewater and the second-order derivative spectrum of the characteristic absorption area of ​​nitrate nitrogen, including: (a) the mid-infrared attenuated total reflection spectrum of wastewater from an electronic chip factory, (b) the mid-infrared attenuated total reflection spectrum of wastewater from a printing and dyeing factory, (c) the second-order derivative spectrum of wastewater from an electronic chip factory, and (d) the second-order derivative spectrum of wastewater from a printing and dyeing factory;

[0032] Figure 3 The infrared absorption spectra of nitrate nitrogen in different wastewaters after deconvolution, including (a) wastewater from an electronic chip factory; (b) wastewater from a printing and dyeing factory;

[0033] Figure 4Principal component screening results (a), (d), distribution of training sets (b), (e) and test sets (c), (f) and model evaluation for establishing PLS prediction model based on deconvolution spectral data of electronic chip wastewater (a), (b), (c) and printing and dyeing wastewater (d), (e), (f). DETAILED DESCRIPTION

[0034] The following examples will enable those skilled in the art to more fully understand the present invention, but are not intended to limit the present invention in any way.

[0035] Wastewater from the electronic chip factory and the printing and dyeing factory was collected separately, and water samples were taken every 20 days (June-December 2020). Among them, 8 wastewater samples were taken from the electronic chip factory and the printing and dyeing factory respectively.

[0036] The water samples were measured using a Nicolet 6700 Fourier transform infrared spectrometer (ThermoFisher Scientific, USA) with a 45°C ZnSe accessory. During the measurement, an appropriate amount of water sample was placed in the sample cell. After each scan, the sample cell was completely dried with nitrogen.

[0037] The nitrate nitrogen content of water samples was determined using a martchem200 fully automatic discontinuous chemical analyzer (AMS Alliance, Frepillon, France). The specific spectrometer measurement parameters were as follows: the scanning range was 4000-650 cm -1 , the mirror moving speed is 0.32cm·s -1 , resolution 4cm -1 Each spectrum was scanned 32 times continuously, each water sample was measured 3 times, an empty ATR sample cell and deionized water were measured once every hour for calibration, and each sample was repeated 4 times. Figure 1 .

[0038] From June to December 2020, the concentration of nitrate nitrogen in the wastewater of electronic chip factories and printing and dyeing factories showed a certain degree of volatility, with the highest concentration in August and September. This may be related to the cyclical production. Among them, the concentration of wastewater from electronic chip factories was the highest in September, at 5.29 mg / L; the concentration of wastewater from printing and dyeing factories was the highest in August, at 7.32 mg / L. From June to December 2020, the concentration of carbonate in the wastewater from electronic chip factories and printing and dyeing factories showed greater differences than the concentration of nitrate nitrogen. The carbonate concentration detected in the wastewater from electronic chip factories was 1.27-34.2 mg / L, and the carbonate concentration detected in the wastewater from printing and dyeing factories was 1.27-41.8 mg / L.

[0039] The mid-infrared attenuated total reflection spectra of two different wastewaters from electronic chip factories and printing and dyeing factories are shown as follows: Figure 2 (a) and Figure 2 As shown in (b), the spectral morphology of wastewater in different time periods is basically the same. It can be seen from the figure that in the range of 3000-3800cm -1 and 1500-1800cm -1 There are two obvious peaks in the band, both of which are characteristic absorption of water, but the characteristic peak of nitrate nitrogen (1200-1500cm -1 ), and the second-order derivatives of the characteristic absorption areas of nitrate nitrogen in the two wastewaters were calculated.

[0040] Finally, the second-order derivative spectra are obtained as follows Figure 2 (c) and Figure 2 As shown in (d), it can be seen that the second-order derivative spectrum is relatively mixed and it is difficult to observe the characteristic absorption peak. After analysis, it is found that the complexity of the second-order derivative spectrum should be caused by water interference. Therefore, it is necessary to further process this band to obtain the characteristic absorption of nitrate nitrogen.

[0041] Peakfit 4.12 software was used to analyze the 1500-1200cm -1 The spectrum of the band is deconvolved. The specific steps are as follows:

[0042] Assuming that the experimental spectrum Y(x) is composed of several single-peak spectral bands superimposed on each other, the purpose of fitting is to find the single-peak spectral band F i (x)(i=1,2,…,n), the principle is as follows:

[0043] Y(x)=ΣFi(x). (1)

[0044] Where Y is the spectrum, x is the wave number, i is the number of independent peaks, and F is the kernel function of Gaussian deconvolution:

[0045]

[0046] Where a0, a1, and a2 represent the peak amplitude, position, and width, respectively, and x and y represent the wave number and absorption intensity, respectively;

[0047] During the deconvolution process, the goodness of fit R 2 All of them are above 0.97, and the sum of square errors is between 0.002183 and 0.005632, and the fitting effect is good.

[0048]

[0049] from Figure 3 As can be seen from (a) and 3(b), the characteristic absorption peak of the water nitrate nitrogen spectrum after deconvolution is approximately at 1363 cm -1The intensity of the peak increases with the increase of nitrate nitrogen concentration. Therefore, this characteristic peak can be used for the rapid quantitative determination of nitrate nitrogen in water.

[0050] The partial least squares method is used to model the infrared spectrum after deconvolution. The model is a bilinear model, including a matrix X of variables (spectral wavenumbers) and a function matrix Y of variables X, which is used to predict the minimum number of potential variables, and the model evaluation is performed using R 2 , RMSE and RPD evaluate the predictive ability of the model, and high R 2 and RPD values, as well as low RMSE indicate a robust and accurate model.

[0051] The optimal number of principal components is determined by cross-validation method. The wastewater from electronic chip factory and printing and dyeing factory are as follows: Figure 4 (a) Figure 4 As shown in (d), the optimal number of principal components for both models is 4, so the first 4 principal components are selected for modeling.

[0052] The results show that the correlation coefficients between the measured and predicted values ​​of the two model training sets are 0.992 and 0.849 ( Figure 4 (b) Figure 4 (e)), the correlation reaches a significant level.

[0053] The correlation coefficients between the measured and predicted values ​​of the two models in the test set are 0.868 and 0.956 respectively ( Figure 4 (c) Figure 4 (f)), are all significantly correlated.

[0054] The RPD value and RMSE are specifically defined by the following formula:

[0055]

[0056]

[0057] Among them, y i and are the measured value and predicted value of nitrate in the ith wastewater sample, is the average value of nitrate determination, n is the number of samples, SD is the standard deviation, RMSE is C and RMSE V They represent the root mean square error in the calibration and validation data set models, respectively. When RPD < 1.8, it means that the model cannot be used for quantitative determination. When the RPD value is between 2 and 2.5, it means that the model has good prediction performance. When RPD > 3, it means that the model prediction performance reaches an excellent level.

[0058] Specifically, for electronic chip wastewater, the RPD value of its model was 2.865, while the RPD value of the printing and dyeing wastewater model was 3.578 (Table 1). The detection limit (LOD) 3σ / m (σ is the standard deviation of the predicted concentration, which can be replaced by RMSEP, and m is the slope of the model's fitting curve) was 0.652 mg / L and 0.713 mg / L, respectively. Both prediction models reached an excellent level.

[0059] Table 1 The best principal components and model evaluation results of the PLS regression model established based on mid-infrared attenuated total reflectance spectra of different types of wastewater

[0060]

[0061] Compared with natural water bodies, the detection limit of the deconvolution infrared spectroscopy model of industrial wastewater is much higher than that of natural water bodies. This should be related to the higher carbonate concentration in industrial wastewater and the more complex composition of industrial wastewater.

[0062] In summary, the present invention provides a method for online detection of nitrate-nitrogen in sewage based on mid-infrared total reflectance spectroscopy. Fourier transform infrared attenuated total reflectance (FTIR-ATR) is used as a research method, and the partial least squares method is used to model the infrared spectrum after deconvolution. The application potential of the model is determined by the model evaluation index, thereby indicating whether rapid detection of nitrate-nitrogen in industrial wastewater can be achieved.

Claims

1. A method for online detection of nitrate nitrogen in sewage based on mid-infrared total reflection spectroscopy, characterized in that: The following steps are involved: Step 1: Collect water sample and add it directly into the ATR crystal tank. Set the spectrum scanning range to 650-4000cm -1 , the mirror moving speed is 0.32cm·s -1 , resolution 4cm -1 , 32 repeated scans were continuously recorded to record the FTIR-ATR spectra; Step 2: Import the original spectral data into Peakfit 4.12 software, perform smoothing preprocessing on the data, and perform a -1 The band spectrum is subjected to Gaussian deconvolution, baseline correction is performed, and the spectrum peak fitting parameters are optimized to improve the goodness of fit; (5) During the deconvolution process, the goodness of fit of all spectra R 2 All of them are above 0.97, and the error sum of squares is between 0.002183-0.005632, and the fitting effect is good; after deconvolution, 1360cm -1 An obvious characteristic peak can be observed at , which is the characteristic absorption of nitrate; The specific steps of deconvolution of the spectral curve are as follows: Assuming experimental spectrum Y(x) It is composed of several single-peak spectral bands superimposed on each other. The purpose of fitting is to find the single-peak spectral band Fj (x) (j=1, 2, ..., n), the principle is as follows: Y(x)=ΣFj(x) (1) In the formula, Y is the spectrum, x is the wave number, j is the number of independent peaks, F is the kernel function of Gaussian deconvolution: （2） Where a0, a1 and a2 represent the peak amplitude, position and width respectively. x and y are the wave number and absorption intensity, respectively; Step 3, using partial least squares method to model the infrared spectrum after deconvolution, and using cross-validation method to determine the optimal number of principal components, that is, the modeling model is a bilinear model, including the matrix X of variable spectral wavenumber and the matrix Y of function of matrix X as variable, which is used to predict the minimum number of potential variables, and the model evaluation is used R 2 , RMSE and RPD Evaluate the predictive ability of the model, high R 2 and RPD values, as well as low RMSE indicate a robust and accurate model. The RPD value and RMSE are specifically defined by the following formula: (3) (4) Among them, y i and are the measured value and predicted value of nitrate in the ith wastewater sample, is the average value of nitrate determination, n is the number of samples, SD is the standard deviation, RMSE represents the root mean square error, when RPD<1.8, it means that the model cannot be used for quantitative determination, when the RPD value is between 2-2.5, it means that the model prediction performance is good, when RPD>3, it means that the model prediction performance reaches an excellent level; Step 4: Use the model established in step 3 to divide the entire data set into a training set and a test set. R 2 The prediction effect of the model was evaluated by RPD value, where the training set accounted for 75% of the entire data set and the test set accounted for 25% of the entire data set.

Images