Near-infrared spectroscopy quickly determines the content of free toluene isocyanate in coating curing agents

Near-infrared spectroscopy to quickly determine the content of free toluene isocyanate in coating curing agents
Chen Jiwen, Liu Fujian, Shen Honglin, Li Jun, Chen Junping
(Shunde 528300, Guangdong Product Quality Supervision and Inspection Research Institute
Toluene isocyanate (TDI) is a colorless or pale yellow transparent and irritating odor of the liquid, is a major raw material of the polyurethane industry, widely used in automotive coatings, wood coatings, building exterior coatings, metal anti-corrosion coatings and aerospace coatings. TDI is a highly toxic inhalable poison, which accumulates and latently in the human body, and long-term contact and use can cause damage to the body's respiratory system, immune system, endocrine system and skin, and is one of the main causes of occupational asthma. At present, there are more reports of detection of diocyanate residues in polyurethane coatings, such as fill column gas chromatography (GC) method, capillary GC method and GC/MS method, national standard GB/T 18446 -2009 paint free toluene isocyanate determination, its operating procedures are locked, the use of a variety of organic solvents, there is a certain harm to the human body, it is difficult to meet the import and export of coatings batch rapid detection needsTherefore, it is urgent to establish a solvent-based coating free toluene isocyanate rapid detection technology.
near-infrared (NIR) spectroscopic analysis technology is a convenient, fast and lossless green analysis technology. NIR light can be analyzed deeply inside the sample, and its main spectral peaks are the multi-frequency and joint-frequency vibration absorption of hydrogen-containing groups such as C-H, N-H and O-H in organic molecules, with stable spectral characteristics and large amount of information, which can reflect the comprehensive information of the sample and can be used for qualitative and quantitative analysis of organic matter. In this paper, near-infrared transmission technology is used to study the curing agent, niR spectroscopy technology is used to quickly determine the content of free toluene dicycyanate (TDI), the correction model is established by the partial least-multiplication method (PLS), and the content of free toluene isocyanate in the sample is investigated in real time.
1 Experimental part
1. 1 Instruments and reagents
Thermo Antaris II. Fourier near-infrared spectrometer with transmission analysis module (Thermo-Fisher, USA, including TQAnalyst 8 data analysis software and Result 3 operating software). Agilent 7890A gas chromatography with hydrogen flame detector (FID, Agilent, USA). Chromatography conditions: detector temperature of 250 degrees C, hydrogen flow of 35 mL /min, air flow of 400 mL /min; Column: HP-5 (30 m × 0. 25 mm × 0. 25 m); Carrier H2 gas, flow rate of 1 mL /min; The temperature of the infringing port is 125 degrees C; Column box heating procedure: 130 degrees C (5 min) → 20 degrees C /min→ 200 degrees C (3 min); Sample-in mode: se through the sample, separation ratio: 40:1. Solvent ethyl acetate (pure analysis, drying with pre-use dehydration), orthane, toluene dicycyanate (TDI) purchased in Aladdin. Curing agent is a solvent-based polyurethane wood coating, samples provided by the National Paint Product Quality Supervision and Inspection Center (Guangdong).
1. 2 Near-infrared spectral analysis method
samples are randomly divided into correction sets and verification sets, and ensure that the concentration range of correction sets samples is greater than the verification set, and establish a NIR spectral correction model. The correction set sample is used to correct the establishment of the model. Select the spectral interval for modeling, use PLS method as the chemical econometric method to establish the correction model, and use internal cross-validation mean variance (RMSECV) to select model parameters for the indicator to optimize the model structure to predict mean variance (RMSEP) to measure the predictive performance of the model, all data processing is carried out in the instrument's own T AnalystQ 8 software.2 Results and Discussions
2. 1 Spectral Acquisition
Using the near-infrared spectrometer's transmission sampling module to capture the sample spectrum for continuous scanning, the wave number range of 10000 to 4000 cm - 1, with the instrument's built-in background as the reference, the number of scans 32 times, the resolution of 8 cm - 1, the near-infrared spectrometer and spectral acquisition are in the air-conditioned constant temperature (23 ± 2) room, spectral map 1 shows.
2. 2 TDI content gas chromatography in the sample
sample free diocyanate TDI content determination, according to GB / T 18446-2009 "color paint and varnish with paint-based isocyanate resin determination" method, the samples were measured in parallel, taking its arithmetic average as the result of the determination, its content distribution as shown in Table 1. The TDI content in the correction set and verification set samples is evenly distributed from low to high, and the linearity is good.
2. 3 Model interval selection
curing agent sample composition is mainly ethyl acetate, butyl acetate, free toluene diocyanate, resin and a small amount of toluene, ethyl benzene, xylene, etc. , the main official groups are - COOR, - CH3, - CH2, CH and - NCO. Because the characteristics of different official groups absorb wavelengths differently, the same multi-frequency and frequency characteristics absorb wavelengths are also different. The near-infrared absorption of the curing agent sample mainly comes from the first-level multi-frequency absorption of 5930 to 5860 cm - 1 interval - CH3, 5797 to 5652 cm - 1 interval - CH2, and about 4551 cm - 1 place - CH of the first-level multi frequency absorption. By measuring the near-infrared spectral absorption of pure free toluene isocyanate, so as to avoid the error caused by the serious peak overlap and common line phenomenon usually existing in the near-infrared spectral data, the model interval of 7320 to 7250 cm - 1, 8485 to 8370 cm - 1, the best performance of the model.
2. 4 Spectral pre-processing method
in order to improve the predictive accuracy of the model, before analyzing the near-infrared spectrum, the spectral data should generally be pre-processed. In this study, after removing singular samples by the Mars distance method, first-order guide processing in the near-infrared spectrum can effectively eliminate baseline offset, reduce the overlap between peaks and peaks, and obtain more effective information, but at the same time, the calculation of the conductor will introduce noise, reduce the signal-to-noise ratio, and smoothing can reduce high-frequency random noise. Therefore, after using a constant light range, the first derivative method is combined with Norris Derivative smooth filtering with a segment length of 5 and a interval of 5 to preprocess the spectral data.
2. 5 The number of model main factors determines
when using PLS to establish quantitative correction model, there are great differences in model prediction ability between the number of different main factors. In the case of a correction set sample, too many factors can cause the model to be too complex and may have been fitted in validation. Therefore, a reasonable selection of the number of main factors used is required. In this paper, the effect of the number of main factors on the value of RMSECV is examined by cross-validation method. Figure 2 is a graph of the RMSECV value changing with the number of main factors, with the lowest point in the figure referring to the optimal value. The optimal number of main factors used for free toluene isocyanate in the correction model is 1.
2. 6 Correction model established
free toluene isocyanate in the selected spectral interval (7320 to 7250 cm - 1,8485 to 8370 cm - 1), using a constant light range, and then a first-order After using the derivative method in combination with Norris Derivative smooth filtering with a segment length of 5 and an interval of 5, a correction model was established using the PLS method, with the correction set average variance (RMSEC) of 0. 0.0815 and the correlation coefficient of 0. 9492. Using the established calibration model, 26 samples of the prediction verification set can be seen from Figure 3, the near-infrared spectral prediction value is basically the same as the actual value, there is a good linear correlation between the two, and the verification set mean variance (RMSEP) is 0. 0715, correlation coefficient is 0. 9767, so it can be seen that the model prediction performance is good. The forecast value and the actual value are further confirmed by the t-test, and the t-test value is 0. 95 (P < 0. 05), indicating that at 0. 05 Within the level of significantness, the difference in results is not significant. It is further explained that the corrected model has good predictive ability and can meet the accuracy requirements of conventional analysis methods.
3 Conclusion
The NIR analysis technology is used in the study of the content of toluene isocyanate in the curing agent, and the content of harmful substances in the curing agent is analyzed in a timely manner, so that the curing agent can be determined quickly and accurately, non-destructively and effectively. This method uses transmission fiber technology, no sample processing is required, and the correction model has good performance. With the development of the model in the actual application of production, the actual monitoring data will constantly re-correct the model, which will improve the applicability and stability of the model curing agent sample testing.