Method for establishing near infrared spectrum online quantitative model of trimethylolpropane synthesis product tmp and its by-products and application

Abstract

The application discloses a near-infrared spectrum online quantitative model establishing method and application of trimethylolpropane synthesis product TMP and by-products thereof. The model establishing method comprises the following steps: (1) determining the product TMP and by-products thereof in a TMP condensation reaction by using gas chromatography-mass spectrometry; (2) quantitatively analyzing the TMP and by-products thereof by using gas chromatography; (3) collecting spectrum information by using near-infrared spectrum analysis; and (4) establishing a quantitative model of the TMP and by-products thereof: pre-processing the spectrum information of the TMP and by-products thereof by using TQ Analyst software, removing abnormal points by using the Mahalanobis distance of the spectrum in the Spectrum Outlier option in TQ Analyst, removing abnormal samples in the near-infrared spectrum samples, combining the concentration measured by using the gas chromatography and the spectrum data collected by using the near-infrared spectrum analysis, establishing a database, and establishing a quantitative model of the TMP and by-products thereof by using a partial least square method. The model can be used for online monitoring of a trimethylolpropane synthesis process.

Description

Technical Field

[0001] This invention relates to the field of chemical synthesis technology, specifically to a method and application for establishing an online quantitative model of near-infrared spectroscopy for the synthesis product TMP of trimethylolpropane and its byproducts. Background Technology

[0002] Trimethylolpropane (TMP) is a triol. It is widely used in the production of high-grade alkyd resins and as a UV curing agent for automotive paints and inks. With the increasing quality of products in China's automotive and home appliance industries, the demand for high-grade coatings is growing rapidly. Simultaneously, due to rising environmental awareness, the demand for UV curing agents is also increasing, thus significantly increasing the demand for TMP. As a resin chain extender with extremely high application value, TMP participates in reactions that produce a variety of different results, resulting in a very high global market share. Even in the past, China remained a net importer. Since 2010, with the gradual increase in domestic production capacity, self-sufficiency has rapidly improved, reducing dependence on imports. Furthermore, with the continuous improvement in the quality of domestic products and the steady growth in downstream demand, TMP has been favored by investors for its good returns over the past three years.

[0003] In my country, the main method for producing trimethylolpropane (TMP) is the Cannizaro process. This method is mature, easy to master, and does not require high temperature, high pressure, or special catalysts. However, the overall yield is relatively low, and it produces many byproducts. These byproducts affect the color of TMP, thus impacting the purity and yield of the product. Therefore, a method is needed to monitor the TMP production process online, thereby monitoring changes in the content of TMP and its byproducts and achieving online monitoring of the entire TMP synthesis process.

[0004] Near-infrared spectroscopy (NIRS) is an easy-to-use, fast, and efficient analytical method that does not damage samples. It can determine the content of chemical components in a sample and simultaneously provide its physical parameters, with results comparable to those obtained by high-performance gas chromatography (HPLC) and liquid chromatography (LC). NIRS eliminates the need for sample pretreatment, does not use any organic solvents, and does not generate toxic waste, aligning with the current green production practices in scientific research. For online testing or on-site analysis, using statistical regression techniques, NIRS can provide real-time chemical information for controlling chemical production processes, solvent recovery processes, mixing, and extraction processes. The increasing demand for product quality improvement and production rationalization in the chemical, petrochemical, polymer, pharmaceutical, cosmetic, food, and agricultural industries has led to a pressing need to replace increasingly time-consuming and conservative analytical techniques (GC, HPLC, NMR, MS) and non-specific control procedures (temperature, pressure, pH, sample weight) with more specific and environmentally friendly analytical tools. In this regard, vibrational spectroscopy (mid-infrared [MIR], near-infrared [NIR], and Raman [ROMAN]) has begun to gain traction in scientific research. The main reason is that NIR technology has emerged in the past decade in combination with optical fibers, new online probe accessories, and chemometric evaluation procedures, serving as extremely powerful tools for industrial quality control and process monitoring.

[0005] Therefore, this invention aims to combine near-infrared spectroscopy with the production of trimethylolpropane (TMP) to establish a model capable of real-time online monitoring of TMP quality. By establishing a quantitative calibration model, the content of TMP and its byproducts during the synthesis process can be rapidly determined, allowing for quantitative monitoring of the TMP synthesis process and achieving online monitoring of the entire synthesis process. This provides experimental evidence for the quality control of TMP production and serves as a reference for the future industrial application of near-infrared spectroscopy in the field of TMP and other polyols. Summary of the Invention

[0006] The purpose of this invention is to provide a method for establishing an online quantitative model of near-infrared spectroscopy for the synthetic product TMP of trihydroxypropane and its byproducts.

[0007] Another objective of this invention is to provide an online quantitative model for near-infrared spectroscopy of the trihydroxypropane synthesis product TMP and its byproducts, and its application in the quality control of the products during the calcium synthesis of trihydroxypropane.

[0008] This invention is achieved through the following technical solution:

[0009] The present invention describes a method for establishing an online quantitative model of near-infrared spectroscopy for the synthesis product TMP and its byproducts of trihydroxypropane, comprising: (1) using gas chromatography-mass spectrometry and infrared spectroscopy to determine the final product TMP and its byproducts in the calcium method for the synthesis of trihydroxypropane; (2) using gas chromatography to quantitatively analyze TMP and its byproducts and determine their concentrations; (3) using near-infrared spectroscopy to collect spectral information of TMP and its byproducts; (4) establishing a quantitative model of TMP and its byproducts: using TQ Analyst software to preprocess the spectral information of TMP and its byproducts, using the Mahalanobis distance of the spectrum in the Spectrum Outlier option of TQ Analyst to remove outliers, and removing abnormal samples in the near-infrared spectral samples; using the concentration determined by gas chromatography and the spectral data collected by near-infrared spectroscopy to establish a database, and using partial least squares method to establish a quantitative model of TMP and its byproducts.

[0010] The byproducts of this invention are: 2-ethylpropenal, 2-ethyl-2-hexenal, and 2-ethyl-1,3-hexanediol.

[0011] Step (1) of the present invention uses Agilent 8860 5977b gas chromatography-mass spectrometry (GC-MS) for analysis; the GC-MS detection method is as follows: column: DB-5MS; column temperature: initial 40℃ for 5 min, rate 1℃ / min to 70℃ for 1 min, rate 3℃ / min to 160℃ for 1 min, rate 5℃ / min to 270℃ for 6 min; injection port temperature: 280℃; split ratio: 20:1.

[0012] The infrared spectrum was determined using an Antaris II Fourier transform near-infrared spectrometer under the following spectral conditions: resolution 8 cm⁻¹. –1 64 scans, spectral scanning range 4000~10000 cm⁻¹ –1 .

[0013] The specific steps of the gas chromatography quantitative analysis in step (2) of this invention include: ① Preparation of internal standard solution: accurately weigh an appropriate amount of 1,6-hexanediol, dissolve it in ethanol, prepare an internal standard solution of 11.57 mg / mL, and store it at low temperature for later use; ② Preparation of reference solution: accurately weigh an appropriate amount of trimethylolpropane reference standard, dissolve it in ethanol, prepare a trimethylolpropane reference standard solution of 12.24 mg / mL, and store it at low temperature for later use; ③ Preparation of mixed reference solution: accurately weigh 2-ethylpropenal and 2-ethyl-2-hexenal respectively. ① Dissolve appropriate amounts of 2-ethyl-1,3-hexanediol, internal standard 1,6-hexanediol, and trimethylolpropane reference standard in ethanol to prepare mixed reference solutions with mass concentrations of 0.43, 1.02, 2.42, 2.69, and 12.24 mg / mL, respectively, and store at low temperature for later use; ② Prepare the test solution: Take an appropriate amount of sample, dilute with anhydrous ethanol, add anhydrous sodium sulfate to remove water, centrifuge, and store at low temperature for later use; ③ Determine by gas chromatography. The gas chromatography conditions are as follows: Detection instrument: Agilent 7890B flame ionization detector; injection volume: 0.3µL; heater: 270℃; split ratio: 30:1, split flow rate: 6mL / min; column: DB-1701 (-20℃~280℃; 30m×250µm×0.25µm); flow rate: 0.2mL / min; pressure: 2.7033psi; initial temperature: 40℃ held for 5 min, v1=1℃ / min to 70℃ held for 1 min, v2=3℃ / min to 160℃ held for 1 min, v2=5℃ / min to 270℃ held for 6 min.

[0014] The ratio of the sample to anhydrous ethanol in this invention is: sample: anhydrous ethanol = 1:9.

[0015] The anhydrous ethanol described in this invention is 99.7% anhydrous ethanol.

[0016] The specific steps of the spectral acquisition and analysis described in step (3) of this invention are as follows: First, preheat the spectrometer for more than 1 hour, take samples at 15 min, 30 min, 45 min, 60 min, 75 min, 90 min, 105 min, 120 min, 135 min, 150 min, 165 min and 180 min during the TMP preparation process, process them according to the preparation method of the test solution, and use the Antaris II Fourier transform near-infrared spectrometer to acquire the near-infrared spectrum.

[0017] The spectral acquisition conditions described in this invention are: scanning mode using a transmission analysis module, resolution 8 cm⁻¹. –1 64 scans, spectral scanning range 4000~10000 cm⁻¹ –1The curing conditions described in step (4) of the present invention are: curing temperature 150-170℃, curing time 3-5 h.

[0018] The quantitative models of TMP and its byproducts in step (5) of this invention are: trimethylolpropane model with Rc=0.9759; 2-ethyl acrolein model with Rc=0.9412; 2-ethyl-2-hexenal model with Rc=0.9346; and 2-ethyl-1,3-hexanediol model with Rc=0.9221.

[0019] The application of the near-infrared spectroscopy online quantitative model for the synthesis product TMP and its byproducts of trimethylolpropane described in this invention in the quality control of the synthesis product of trimethylolpropane condensation reaction involves performing near-infrared spectroscopy on the sample to be tested, collecting spectral data, inputting it into the established model, rapidly determining the content of trimethylolpropane and its byproducts during the synthesis process, quantitatively monitoring the synthesis process of trimethylolpropane, and realizing online monitoring of the entire synthesis process of trimethylolpropane.

[0020] The method for synthesizing trimethylolpropane described above in this invention is as follows:

[0021] 39.0 g of calcium hydroxide was weighed and added to a three-necked flask, followed by 270.0 g of deionized water. The mixture was stirred for 5 minutes to ensure homogeneity. Then, 208.0 g of HCHO was weighed into the reactor and cooled to 25°C. Next, 60.0 g of n-butyraldehyde was weighed into a Longerpump constant-flow peristaltic pump and added dropwise at a rate of 10.0 g over a 25°C water bath for two hours. After the addition was complete, the temperature was raised to 40°C, and the reaction was allowed to proceed for 40 minutes. Excess HCHO was then oxidized with 27.5% hydrogen peroxide, and the mixture was neutralized to pH 6.7 with 88% formic acid solution to obtain the condensate.

[0022] The sample described in this invention is the condensation solution after the reaction is complete.

[0023] The beneficial effects of this invention are:

[0024] 1. This invention first uses gas chromatography-mass spectrometry and infrared spectroscopy to determine the formation of the product trimethylolpropane and its byproducts 2-ethylpropenal, 2-ethyl-2-hexenal, and 2-ethyl-1,3-hexanediol in the reaction.

[0025] 2. Based on preliminary investigations of solvents, internal standards, and detection methods, the results of this invention are as follows:

[0026] (1) It was determined that when anhydrous ethanol was selected as the solvent in the sample preparation and the dilution factor was 9 times, no crystals would precipitate in the sample over time;

[0027] (2) It was determined that 1,6-hexanediol was used as an internal standard. There was no overlap or adhesion between the peaks of the products and the standard, which can eliminate the influence of systematic error. After the sample was placed for one day, the peak area of ​​each component did not change. This indicates that 1,6-hexanediol does not react with the substances in the sample.

[0028] (3) The gas chromatography conditions were determined as follows: Detector: Agilent 7890B flame ionization detector; Injection volume: 0.3µL; Heater: 270℃; Split ratio: 30:1, split flow rate: 6mL / min; Column: DB-1701 (-20℃~280℃; 30m×250µm×0.25µm); Flow rate: 0.2mL / min; Pressure: 2.7033psi. The initial temperature was 40℃ and held for 5min, v1=1℃ / min was increased to 70℃ and held for 1min, v2=3℃ / min was increased to 160℃ and held for 1min, v2=5℃ / min was increased to 270℃ and held for 6min. Under this detection method, the resolution of each component peak was greater than 1.5, the resolution of each component was good, and the solvent peak had no effect on the sample.

[0029] (4) Under this detection method, the RSD values ​​of repeatability, linearity and detection limit, stability, precision and recovery rate of trimethylolpropane and its byproducts are all less than 3%, and the linear correlation coefficient is also above 0.999, which meets the requirements of gas chromatography. This indicates that the gas chromatography detection method determined above is stable and reliable and can be used for the quantitative analysis of trimethylolpropane and its byproducts.

[0030] 3. This invention determines that the near-infrared spectroscopy scanning method is a transmission analysis module, within the spectral range of 10000-4000 cm⁻¹. -1 Inside 8cm -1 Spectral acquisition was performed at a high resolution with 64 scans, acquiring three parallel sets of spectra. The average spectrum was selected as the modeling spectrum using OMIC software. Using pure trimethylolpropane as the experimental sample, partial least squares (PLS) was chosen from four different quantitative analysis models. The precision of this modeling method was RESEC = 0.0242, Rp = 0.9927; repeatability was RESEC = 0.0011, Rp = 0.9989; and stability was RMSEC = 0.0015, Rp = 0.9942. Since RMSEC was less than 0.1 and Rp was greater than 0.99, the quantitative model establishment method was accurate and reliable.

[0031] 4. Based on the RMSEC values ​​and Rc obtained from different spectral processing methods, this invention provides feedback on the results of partial least squares method model establishment in the experiment.

[0032] (1) Establishment of the trimethylolpropane model: The RMSEC of the smoothing method using standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norris derivative filter is 0.919 and Rc is 0.9730, which is the best. After removing outliers, the RMSEC of the model is 0.862 and Rc is 0.9759. It can be seen from the model diagram that the model established after removing outliers is accurate and reliable.

[0033] (2) 2-Ethylpropenal model establishment: The RMSEC value of the smoothing method using multivariate signal correction (MSC) + first derivative spectroscopy (FD) + Savitzky-Golay filter is 0.143 and the Rc value is 0.9417; after removing outliers, the RMSEC value of the model is 0.143 and the Rc value is 0.9417, indicating that the predicted value is well correlated with the measured value;

[0034] (3) Establishment of the 2-ethyl-2-hexenal model: The results of the smoothing method using standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norris derivative filter are as follows: RMSEC = 0.0429, Rc = 0.9307; after outlier removal, the RMSEC is 0.0420 and the Rc is 0.9346, indicating that the predicted values ​​are well correlated with the measured values.

[0035] (4) 2-Ethyl-1,3-hexanediol model establishment: The results of the smoothing method using standard canonical transformation (SNV) + first derivative spectroscopy (FD) + Norris derivative filter are: RMSEC=0.0212, Rc=0.9221; RMSEC and Rc after outlier removal have no effect, but the predicted RMSEC and Rp values ​​were adjusted from 0.0310 and 0.8159 to 0.0300 and 0.8432, which increased the accuracy of the predicted values ​​and improved the simulation performance.

[0036] 5. This invention employs single-factor and response surface methodology. Samples were collected at different time points, and a database was established using combined gas chromatography and near-infrared spectroscopy data. This database was analyzed, and quantitative analysis was performed using partial least squares method. Four quantitative models were established: a trimethylolpropane model with Rc=0.9759; a 2-ethylpropenal model with Rc=0.9412; a 2-ethyl-2-hexenal model with Rc=0.9346; and a 2-ethyl-1,3-hexanediol model with Rc=0.9221. These four models are accurate and reliable and can be used for quality control in the production process of trimethylolpropane.

[0037] 6. This invention first uses a model to calculate the spectral data to obtain the predicted concentration value, and then compares this concentration value with the actual value measured by gas chromatography with the internal standard method. The results show that the deviation rate between the predicted and measured values ​​of the four substances is less than 3%, indicating that the model established by sampling at different time points of the condensation reaction can predict the concentration at each point very well. Using the predicted value instead of the measured value saves the time of obtaining the gas chromatography results and greatly improves production efficiency.

[0038] 7. The model established in this invention was applied to verify the optimal experimental parameters obtained by the response surface methodology. The results showed that the errors between the actual and predicted values ​​of 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, and TMP were 0.84%, 2.23%, 2.67%, and 2.01%, respectively, all less than 3%. This indicates that the model can be correctly used to predict the laboratory production of trimethylolpropane and its byproducts. Attached Figure Description

[0039] Figure 1 Mechanism of Trimethylolpropane Synthesis

[0040] Figure 2 Gas chromatography-mass spectrometry (GC-MS) spectrum

[0041] Figure 3 Mass spectrum of byproduct 2-ethyl-1,3-hexanediol

[0042] Figure 4 Mass spectrum of byproduct 2-ethyl-2-hexenal

[0043] Figure 5 Mass spectrum of byproduct 2-ethyl acrolein

[0044] Figure 6 Mass spectrum of trimethylolpropane

[0045] Figure 7 Infrared spectrum of byproduct 2-ethyl-1,3-hexanediol

[0046] Figure 8 Infrared spectrum of byproduct 2-ethyl-2-hexenal

[0047] Figure 9 Infrared spectrum of byproduct 2-ethylpropenal

[0048] Figure 10 Infrared spectrum of trimethylolpropane

[0049] Figure 11 Trimethylolpropane production process

[0050] Figure 12 Solvent gas chromatogram

[0051] Figure 13 : Gas chromatogram of mixed standards (where 1 represents: 2-ethylpropenal; 2 represents: 2-ethyl-2-hexenal; 3 represents: 1,6-hexanediol; 4 represents: 2-ethyl-1,3-hexanediol; 5 represents: trimethylolpropane)

[0052] Figure 14 : Sample gas chromatogram (where 1 represents: 2-ethylpropenal; 2 represents: 2-ethyl-2-hexenal; 3 represents: 1,6-hexanediol; 4 represents: 2-ethyl-1,3-hexanediol; 5 represents: trimethylolpropane)

[0053] Figure 15 Near-infrared spectrum of pure trimethylolpropane solution

[0054] Figure 16 Spectroscopy of pure trimethylolpropane in solid state

[0055] Figure 17 Correlation between predicted and measured values

[0056] Figure 18 Deviation distribution

[0057] Figure 19 Precision testing model

[0058] Figure 20 Repeatable testing model

[0059] Figure 21 Stability assessment model

[0060] Figure 22 2-Ethyl-1,3-hexanediol variation diagram

[0061] Figure 23 Effect of raw material molar ratio on the concentration of each component

[0062] Figure 24 Concentration of trimethylolpropane at different aldol condensation temperatures

[0063] Figure 25 Internal student chemical residual

[0064] Figure 26 Residuals and rising predicted response values

[0065] Figure 27 Reaction time and molar ratio interact

[0066] Figure 28 Reaction time and molar ratio interact

[0067] Figure 29 The reaction time interacts with the aldol condensation temperature.

[0068] Figure 30 The reaction time interacts with the aldol condensation temperature.

[0069] Figure 31 Molar ratio and condensation temperature interaction

[0070] Figure 32 Molar ratio and condensation temperature interaction

[0071] Figure 33 Near-infrared spectral overlay

[0072] Figure 34 Markovnikov distance distribution map of trimethylolpropane

[0073] Figure 35 Correlation between measured and predicted values ​​of trimethylolpropane

[0074] Figure 36 Trimethylolpropane Deviation Distribution Chart

[0075] Figure 37 : Markovnikov distance distribution map of 2-ethylpropenal

[0076] Figure 38 Correlation between measured and predicted values ​​of 2-ethylpropionaldehyde

[0077] Figure 39 2-Ethylpropenal Deviation Distribution Chart

[0078] Figure 40 Markovnikov distance distribution diagram of 2-ethyl-2-hexenal

[0079] Figure 41 Correlation between measured and predicted values ​​of 2-ethyl-2-hexenal

[0080] Figure 42 2-Ethyl-2-hexenal Deviation Distribution Chart

[0081] Figure 43 Markovnikov distance distribution map of 2-ethyl-1,3-hexanediol

[0082] Figure 44 Correlation between measured and predicted values ​​of 2-ethyl-1,3-hexanediol

[0083] Figure 45 2-Ethyl-1,3-hexanediol Deviation Distribution Chart Detailed Implementation

[0084] The technical solution of the present invention will be further described in detail below through specific embodiments.

[0085] Example: Method for establishing an online near-infrared spectral quantitative model for the synthesis product TMP and its byproducts of trihydroxypropane.

[0086] Synthesis of trimethylolpropane: 39.0 g of calcium hydroxide was weighed and added to a three-necked flask, followed by 270.0 g of deionized water. The mixture was stirred for 5 minutes until homogeneous. Then, 208.0 g of HCHO was weighed into the reactor and cooled to 25°C. Next, 60.0 g of n-butyraldehyde was weighed into a Longerpump peristaltic pump and added dropwise at a rate of 10.0 g over a 25°C water bath for two hours. After the addition was complete, the temperature was raised to 40°C and the reaction was allowed to proceed for 40 minutes. Excess HCHO was then oxidized with 27.5% hydrogen peroxide, and the mixture was neutralized to pH 6.7 with 88% formic acid solution to obtain the condensate.

[0087] (1) Gas chromatography-mass spectrometry and infrared spectroscopy were used to determine the final product TMP and its byproducts in the calcium method for the synthesis of trihydroxypropane:

[0088] The condensate from the reaction was analyzed using an Agilent 8860 5977b gas chromatography-mass spectrometry (GC-MS) system. The analytical method was as follows: column: DB-5MS; initial column temperature: 40℃ for 5 min, ramped up to 70℃ at a rate of 1℃ / min and held for 1 min, ramped up to 160℃ at a rate of 3℃ / min and held for 1 min, ramped up to 270℃ at a rate of 5℃ / min and held for 6 min; injection port temperature: 280℃; split ratio: 20:1.

[0089] The condensate from the reaction was identified using an Antaris II Fourier transform near-infrared spectrometer under the following spectral conditions: resolution 8 cm⁻¹. –1 64 scans, spectral scanning range 4000~10000 cm⁻¹ –1 .

[0090] (2) TMP and its byproducts were quantitatively analyzed by gas chromatography to determine their concentrations:

[0091] ① Preparation of internal standard solution: Accurately weigh an appropriate amount of 1,6-hexanediol, dissolve it in 99.7% ethanol, prepare an internal standard solution of 11.57 mg / mL, and store it at low temperature for later use;

[0092] ② Preparation of reference solution: Accurately weigh an appropriate amount of trimethylolpropane reference standard, dissolve it in 99.7% ethanol to prepare a 12.24 mg / mL trimethylolpropane reference solution, and store it at low temperature for later use;

[0093] ③ Preparation of mixed reference solutions: Accurately weigh appropriate amounts of 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, internal standard 1,6-hexanediol, and trimethylolpropane reference standards, dissolve them in 99.7% ethanol, and prepare mixed reference solutions with mass concentrations of 0.43, 1.02, 2.42, 2.69, and 12.24 mg / mL, respectively. Store at low temperature for later use.

[0094] ④ Preparation of test solution: Take an appropriate amount of condensate, dilute with 9 times the amount of 99.7% ethanol, add anhydrous sodium sulfate to remove water, centrifuge to obtain the solution, and store at low temperature for later use.

[0095] ⑤ The determination was performed by gas chromatography. The gas chromatography parameters were as follows: Detector: Agilent 7890B flame ionization detector; Injection volume: 0.3µL; Heater: 270℃; Split ratio: 30:1; Split flow rate: 6mL / min; Column: DB-1701 (-20℃~280℃; 30m×250µm×0.25µm); Flow rate: 0.2mL / min; Pressure: 2.7033psi; Initial temperature: 40℃ held for 5 min, v1=1℃ / min increased to 70℃ held for 1 min, v2=3℃ / min increased to 160℃ held for 1 min, v2=5℃ / min increased to 270℃ held for 6 min.

[0096] (3) Near-infrared spectroscopy was used to collect and analyze the spectroscopic information of TMP and its byproducts.

[0097] Preheat the spectrometer for at least 1 hour. Take samples at 15 min, 30 min, 45 min, 60 min, 75 min, 90 min, 105 min, 120 min, 135 min, 150 min, 165 min, and 180 min during the TMP preparation process. Process the samples according to the above-described test solution preparation method. Use an Antaris II Fourier transform near-infrared spectrometer in transmission analysis mode with a resolution of 8 cm⁻¹. –1 64 scans, spectral scanning range 4000~10000 cm⁻¹ –1 Under these conditions, spectral data were acquired, including near-infrared spectra.

[0098] (4) Establishment of quantitative models for TMP and its byproducts:

[0099] The near-infrared spectra collected above were preprocessed using TQ Analyst software to analyze the spectral information of TMP and its byproducts. Outliers were removed using the Mahalanobis distance in the Spectrum Outlier option of TQ Analyst, thus eliminating abnormal samples from the near-infrared spectral samples. The concentrations determined by gas chromatography and the spectral data collected by near-infrared spectroscopy were coupled to establish a database. A quantitative model of TMP and its byproducts was then established using partial least squares method, wherein:

[0100] TMP model establishment: The smoothing method of standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norris derivative filter was adopted, and finally the trimethylolpropane model with Rc=0.9759 was obtained.

[0101] Model establishment of 2-ethylpropenal: The smoothing method of multivariate signal correction (MSC) + first derivative spectroscopy (FD) + Savitzky-Golay filter was adopted, and the final model of 2-ethylpropenal with Rc=0.9412 was obtained.

[0102] Model establishment of 2-ethyl-2-hexenal: The smoothing method of standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norris derivative filter was adopted, and finally the 2-ethyl-2-hexenal model with Rc=0.9346 was obtained.

[0103] Model establishment of 2-ethyl-1,3-hexanediol: The smoothing method of standard canonical transformation (SNV) + first derivative spectroscopy (FD) + Norris derivative filter was adopted, and the final model of 2-ethyl-1,3-hexanediol with RC=0.9221 was obtained.

[0104] To further verify the feasibility of the present invention, the inventors conducted a series of experiments, the steps of which are as follows:

[0105] I. Synthesis of Trimethylolpropane

[0106] The raw materials for the synthesis of trimethylolpropane are formaldehyde and n-butyraldehyde. The main synthetic methods include the Cannizzaro process and the catalytic hydrogenation process. The synthesis mechanism is as follows: Figure 1 As shown.

[0107] II. Gas Chromatographic Detection Method for Trimethylolpropane

[0108] To achieve quality control of trimethylolpropane, the first step is to establish a detection method that can be used for coupled analysis. This method combines concentration and spectral data. Gas chromatography, as the name suggests, is a substance separation method that uses a gas (also called a carrier gas, commonly hydrogen / helium and nitrogen) as the mobile phase. Chromatography has extremely high separation capabilities and excels at separating substances containing large molecular weights. GC includes flame ionization detectors (FID), thermal conductivity detectors (TCD), nitrogen-phosphorus detectors (NPD), and electron capture detectors (ECD), among others.

[0109] Since the raw materials for this experiment are formaldehyde and n-butyraldehyde, the final product obtained after the reaction is trimethylolpropane, and the byproducts are 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, and the internal standard 1,6-hexanediol, these are easily converted into CH4 and CO by programmed temperature rise in a flame ionization detector (FID). Therefore, an FID is selected. Method validation should focus on 1) repeatability; 2) linearity and detection limit; 3) stability; 4) precision; and 5) recovery rate.

[0110] 2.1 Trimethylolpropane condensation reaction

[0111] 2.1.1 Experimental Equipment and Reagents

[0112]

[0113] 2.1.2 Experiment

[0114] 39.0 g of calcium hydroxide was weighed and added to a three-necked flask, followed by 270.0 g of deionized water. The mixture was stirred for 5 minutes until homogeneous. Then, 208.0 g of HCHO was weighed into the reactor and the temperature was lowered to 25°C. Next, 60.0 g of n-butyraldehyde was weighed into a Longerpump constant-flow peristaltic pump and added dropwise at a rate of 10.0 g over a 25°C water bath for two hours. After the addition was complete, the temperature was raised to 40°C and the reaction was allowed to proceed for 40 minutes. Excess HCHO was then oxidized with 27.5% hydrogen peroxide, followed by neutralization with 88% formic acid solution to pH 6.7. The mixture was filtered, extracted, and dried to obtain a white, flaky solid.

[0115] 2.1.3 Product Characterization

[0116] The condensate from the reaction was analyzed using an Agilent 8860 5977b gas chromatography-mass spectrometry (GC-MS) system and identified by scanning an Antaris II Fourier transform near-infrared spectrometer. The products and byproducts of the reaction were analyzed. GC-MS detection method: Column: DB-5MS; initial column temperature 40℃, held for 5 min; ramped up to 70℃ at 1℃ / min, held for 1 min; ramped up to 160℃ at 3℃ / min, held for 1 min; ramped up to 270℃ at 5℃ / min, held for 6 min; injection port temperature: 280℃; split ratio 20:1. See [link to details] Figures 2-10 .

[0117] The peaks with the largest peak areas (elevation times of 4.891 min, 10.040 min, 18.876 min, 20.794 min, 40.910 min, 41.892 min, 43.822 min, 45.820 min, and 53.058 min) were selected for mass spectrometry analysis. Spectral analysis revealed that the peak at 4.891 min was 2-ethylpropenal; the peak at 10.040 min was n-butyraldehyde; the peak at 18.876 min was 2-ethyl-2-propenal; the peak at 43.822 min was 2-ethyl-1,3-hexanediol; and the peak at 45.820 min was trimethylolpropane. (3201.95 cm⁻¹) -1 The characteristic absorption of hydroxyl groups in water is 2964.82 cm⁻¹. -1 2928.33cm -1 2878.17cm -1 The absorption peaks are for the stretching vibrations of -CH3 and -CH2-, at 1375.57 cm⁻¹. -1 1215.96cm -1 1149.85cm -1 The absorption peak for CH on alkyl groups is 1054.07 cm⁻¹. -1 and 1008cm -1 The absorption peak at 2972 ​​cm⁻¹ corresponds to the CO stretching vibration of the alcohol hydroxyl group, indicating the synthesis of TMP. Simultaneously, an absorption peak at 2972 ​​cm⁻¹ is observed. -1 2928cm -1 2878cm -1 1418cm -1 1379cm -1 1327cm -1 1087cm -1 1045cm -1 879cm -1 803cm -1 629cm -1 The peaks observed indicate the presence of 2-ethyl-1,3-hexanediol; at 2972 ​​cm⁻¹ -12928cm -1 1693cm -1 1397cm -1 1330cm -1 1274cm -1 1087cm -1 1274cm -1 1087cm -1 948cm -1 879cm -1 It was also proven that 2-ethylpropenal was present; at 2972 ​​cm⁻¹ -1 2930cm -1 2878cm -1 1679cm -1 1379cm -1 1087cm -1 1045cm -1 879cm -1 800cm -1 631cm -1 The peaks indicate the presence of 2-ethyl-2-hexenal.

[0118] 2.1.4 In the experiment: the reaction process for the production of trimethylolpropane is as follows... Figure 11 As shown.

[0119] 2.2 Sample processing methods

[0120] 2.2.1 Solvent Investigation

[0121] In the initial stage of this invention, the solubility of trimethylolpropane and its byproducts in four solvents—CH3OH, CH3CH2OH, CH3CN, and H2O—was investigated. Because the calcium formate produced in the reaction is readily soluble in water, and the gas chromatography sample cannot contain water, anhydrous sodium sulfate must be used to remove the water first. However, calcium formate still remains in the sample, resulting in crystal precipitation, which impairs the use of the chromatography. While trimethylolpropane and other byproducts are soluble in acetonitrile, prolonged reaction time leads to crystal precipitation, violating the stability requirements of gas chromatography detection. A trace amount of bis(trimethylolpropane) is generated during the reaction, which is insoluble in CH3OH. Ultimately, it was found that although crystals precipitate after more than 24 hours when ethanol is used as a solvent and the sample is prepared at a 1:1 ratio, the amount precipitated is less than that in acetonitrile, and calcium formate is insoluble in ethanol; after mixing and centrifugation, the calcium formate is removed. To reduce the amount of precipitated crystals, the sample was diluted. Results showed that at a sample:solvent ratio of 1:9, no crystals precipitated over time, ensuring stability in gas chromatography detection. Finally, 99.7% ethanol was selected as the solvent and diluted 9 times.

[0122] 2.2.2 Selection and Configuration of Internal Standards

[0123] To achieve quantitative analysis of trimethylolpropane and its byproducts, the internal standard method was selected. The internal standard method involves accurately weighing a certain amount of sample, adding a certain standard substance (internal standard), and calculating the content of the analyte based on the mass of the internal standard and the sample, as well as the corresponding peak area on the chromatogram. Previous investigations of 1,6-hexanediol, dodecane, and pentadecane revealed that dodecane peaks overlapped with the sample peaks; pentadecane's retention time differed significantly from that of 2-ethylacrolein, potentially introducing errors in the experimental results. 1,6-hexanediol, with its moderate retention time and lack of overlap or adhesion with the product peaks, eliminated the influence of systematic errors. Furthermore, re-injection of the sample after one day showed no change in the peak areas of each component, indicating that 1,6-hexanediol does not react with the substances in the sample. Therefore, 1,6-hexanediol was chosen as the internal standard.

[0124] An internal standard solution with a concentration of 11.57 mg / mL was prepared using anhydrous ethanol as a solvent and stored at low temperature for later use.

[0125] 2.2.3 Preparation of TMP and Mixed Reference Standards

[0126] After investigating the solvent and internal standard compound, 99.7% anhydrous ethanol was ultimately chosen as the solvent, and 1,6-hexanediol as the internal standard. A mixed standard was prepared by adding 0.43 mg / mL 2-ethylpropenal, 1.02 mg / mL 2-ethyl-2-hexenal, 2.42 mg / mL 2-ethyl-1,3-hexanediol, 2.69 mg / mL 1,6-hexanediol (internal standard), and 12.24 mg / mL trimethylolpropane, and stored at low temperature for later use.

[0127] 2.3 Gas Chromatography Detection of Trimethylolpropane

[0128] The characterization results above indicate that 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, and trimethylolpropane are present in this condensation reaction. The previously validated gas chromatographic conditions were as follows: Instrument: Agilent 7890B flame ionization detector; Injection volume: 0.3 µL; Heater: 270 °C; Split ratio: 30:1; Flow rate: 6 mL / min; Column: DB-1701 (-20 °C to 280 °C; 30 m × 250 µm × 0.25 µm); Flow rate: 0.2 mL / min; Pressure: 2.7033 psi. The initial temperature was 40 °C held for 5 min, then increased to 70 °C at a rate of 1 °C / min and held for 1 min, then increased to 160 °C at a rate of 3 °C / min and held for 1 min, and finally increased to 270 °C at a rate of 5 °C / min and held for 6 min. The detection results are shown below. Figures 12-14 :

[0129] As can be seen from the solvent peak, mixed reference standard and trimethylolpropane experimental sample above, the resolution of each component chromatographic peak is greater than 1.5, the resolution of each component is good, and the solvent peak has no effect on the sample.

[0130] 2.3.1 Repeatability Experiment

[0131] Six samples were prepared from the product obtained in section 2.1.2, and injected continuously six times under the above chromatographic conditions. The peak time and peak area were recorded for each sample. The repeatability test results showed that the RSDs for trimethylolpropane, 2-ethylpropenal, 2-ethylpropenal, 2-ethyl-2-hexenal, and 2-ethyl-1,3-hexanediol were all less than 3%, indicating good repeatability and meeting the detection requirements. See Tables 3 and 4.

[0132]

[0133]

[0134] 2.3.2 Linearity and Detection Limit

[0135] The prepared reference standards were diluted 2, 4, 8, and 16 times, respectively, and detected under the same chromatographic conditions as described above. A function was fitted with concentration on the x-axis and peak area on the y-axis. The results are shown in Table 5 below. It can be seen that the correlation coefficients of all components are above 0.999, indicating that the method linearly meets the detection requirements.

[0136]

[0137] 2.3.3 Precision Experiment

[0138] To ensure minimal variation in injection results, the above-prepared mixed reference standard was used, and the same sample was injected six times consecutively under the same chromatographic conditions. The mass concentrations of each component are shown in Table 6 below:

[0139]

[0140] As shown in Table 6, the RSD of 2-ethylpropenal is 1.68%; the RSD of 2-ethyl-2-hexenal is 2.22%; the RSD of 2-ethyl-1,3-hexanediol is 2.94%; and the RSD of trimethylolpropane is 0.99%, which meets the RSD standard (RSD should be less than 3%), proving that this gas phase detection method meets the detection requirements.

[0141] 2.3.4 Stability Test

[0142] To ensure that the sample would not change again at any time, repeated injections were performed under the same chromatographic conditions at 0h, 2h, 4h, 8h, 16h, and 24h after sample preparation. See Table 7.

[0143]

[0144] As shown in Table 7, the RSD of 2-ethylpropenal is 2.07%; the RSD of 2-ethyl-2-hexenal is 2.88%; the RSD of 2-ethyl-1,3-hexanediol is 2.93%; and the RSD of trimethylolpropane is 1.38%, which meets the RSD standard (RSD should be less than 3%), proving that this gas phase detection method meets the detection requirements.

[0145] 2.3.5 Recovery rate

[0146] Nine samples of known concentration were taken, and 80%, 100%, and 120% of the contents of 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, and TMP were added to each sample, respectively. The results showed that the RSDs for 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, and TMP were all below 3%, meeting the methodological requirements. See Table 8.

[0147]

[0148] 2.4 Chapter Summary

[0149] (1) First, gas chromatography-mass spectrometry and infrared spectroscopy were used to confirm the formation of 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol and trimethylolpropane in the reaction.

[0150] (2) Based on preliminary investigations, anhydrous ethanol was selected as the solvent and 1,6-hexanediol as the internal standard in sample preparation. The RSD values ​​of repeatability, linearity, detection limit, stability, precision, and recovery rate of trimethylolpropane and its byproducts were all less than 3%, and the linear correlation coefficients were also above 0.999. This indicates that the gas phase detection method determined above can be used for the quantitative analysis of trimethylolpropane and its byproducts, laying the foundation for the subsequent near-infrared spectroscopy modeling.

[0151] III. Near-infrared detection method

[0152] Trimethylolpropane (TMP) and its byproducts mainly include C, CO, -CH3, and -CH2-, which exhibit absorption bands in the near-infrared spectral region. This allows for the detection and analysis of compounds at various stages of TMP production using near-infrared spectroscopy. Sample preparation is rapid and simple, avoiding the influence of changes in internal substances during sample preparation. This chapter primarily focuses on determining the near-infrared detection method, the modeling method, and examining the stability, repeatability, and precision of the instrument used in the detection process. This ensures that this method can be used to establish a quantitative analysis model for trimethylolpropane and its byproducts.

[0153] 3.1 Determining the Scanning Method

[0154] This thesis utilizes the Thermo Fisher Analytic Fourier Transform Near-Infrared Spectrometer, which offers sampling modes including an integrating sphere module, a transmission analyzer, a fiber optic module, and a tablet analyzer. Since the samples used in this study were primarily liquid, the transmission analyzer was selected. In near-infrared spectroscopy, besides the spectral information of the sample, some noise peaks may appear. Increasing the number of spectral acquisitions reduces the signal-to-noise ratio. Therefore, through examining 16, 32, 64, and 128 scans, the results showed that 64 scans provided the best repeatability; thus, 64 scans were chosen as the optimal number of scans. This was achieved by analyzing 8cm... -1 16cm -1 32cm -1 The study examined three different resolutions, using PLS as the modeling method, and found that at a resolution of 8cm... -1 The model at that time was 16cm -1 and 32cm -1 The R-value is higher and more accurate, and no better model emerged when the resolution was further increased, so 8cm was chosen. -1 As the scan resolution.

[0155] Therefore, a transmission analysis module was used, with 64 8cm [times / times] [times / times]. -1 The samples were scanned at a high resolution, with each sample scanned three times in parallel. The average value was taken as the sample spectrum using OMIC software.

[0156] 3.2 Determination of Spectral Preprocessing Method

[0157] In TQ Analyst software, when acquiring NIR spectra of liquid samples using cuvettes, the optical path is generally constant. Alternatively, multivariate signal correction (MSC) or standard canonical transform (SNV) can be used for spectral preprocessing. Other options include selecting the original spectrum, first derivative spectrum, or second derivative spectrum, and applying noise filtering (Filter to Smooth Date: 1. No smoothing filter 2. Savitzky-Golay filter 3. Norris derivative filter), resulting in various signal processing combinations. For the results of different processing methods, the mean square error (RMSEC) and correlation coefficient R are used as evaluation indicators.

[0158] The TQ Analyst spectral preprocessing method is as follows:

[0159] ;

[0160] ;

[0161] ;

[0162]

[0163] 3.2.1 Experiment

[0164] Using TMP produced by Chifeng Ruiyang Chemical Co., Ltd. as the sample, and 99.7% anhydrous ethanol as the solvent, 28 pure trimethylolpropane solutions of different concentrations were prepared. Three spectra were acquired for each solution using the acquisition method determined in section 3.1, and the average value was taken. The measured concentrations of the samples were also recorded. Simultaneously, solid-state near-infrared spectra of these samples were acquired using an integrating sphere acquisition module. See Table 9. Figures 15-16 .

[0165]

[0166] 3.2.2 Experimental Results

[0167] A quantitative analysis model for pure trimethylolpropane was established using partial least squares (PLS) in TQ Analyst. The results for different sample processing methods are shown in the table below. Table 10 shows that the mean square error (RESEC) and Rc of the constant optical path length (COP) + Spectrum + NG processing method are closest to 1, indicating that this method provides the best model prediction.

[0168]

[0169] 3.3 Determination of Modeling Method

[0170] Near-infrared spectroscopy (NIR) modeling methods include four types: multiple linear regression, stepwise multiple linear regression, principal component regression, and partial least squares regression. Each has its own advantages, disadvantages, and applicable samples. PCR and PLS can be used for more complex samples. The optimal modeling method was determined using the root mean square error (RMSEC) and correlation coefficient (Rc) as reference indicators. Table 11 below shows that partial least squares regression yielded the best modeling results. The modeling results are as follows: Figure 17-18 .

[0171]

[0172] 3.4 Model Examination

[0173] 3.4.1 Precision Examination

[0174] Using a transmission analysis module, 64 times at 8cm -1 The scanning method with high resolution was used. The mixed control standard configured in 2.2.2 was scanned 18 times, with each scan consisting of three samples corresponding to one concentration value. Spectral preprocessing was performed using (Constant) + Spectrum + NG, and partial least squares regression modeling was used. The TMP instrument showed RMSEC = 0.242 and Rc = 0.9927, indicating that its stability with respect to samples met the requirements. The results are as follows... Figure 19 As shown.

[0175] 3.4.2 Repeatability Test

[0176] Using a transmission analysis module, 64 times at 8cm -1 A high-resolution scanning method was used, scanning six experimental samples three times per sample, with each set of three scans corresponding to a single concentration value. Spectral preprocessing was performed using (Constant) + Spectrum + NG, and partial least squares regression was used for modeling. Figure 20 As shown, RMSEC=0.011 and R=0.9989, indicating that this modeling method has good repeatability.

[0177] 3.4.3 Stability Assessment

[0178] Using a transmission analysis module, 64 times at 8cm -1 The resolution scanning method involved taking three sets of data in parallel at sample intervals of 0h, 2h, 4h, 8h, 16h, and 24h, using the same spectral preprocessing and modeling methods, such as... Figure 21As shown, the results indicate that RMSEC=0.0151 and R=0.9942, which means that the detection method is stable and meets the test requirements.

[0179] 3.5 Conclusion

[0180] This chapter defines the near-infrared spectroscopy scanning method as the transmission analysis module, within the spectral range of 10000-4000 cm⁻¹. -1 Inside 8cm -1 Spectral acquisition was performed at a high resolution with 64 scans, acquiring three parallel sets of spectra. The average spectrum was selected as the modeling spectrum using OMIC software. The RMSEC and Rc values ​​of different modeling methods were compared, and partial least squares (PLS) was ultimately chosen as the modeling method. The stability, repeatability, and precision of this sample processing method were validated. The R values ​​for all three parameters exceeded 0.99, indicating that this method can be used for modeling large databases.

[0181] IV. Quality Control Research in Trimethylolpropane Production

[0182] Gas chromatography-mass spectrometry (GC-MS) results revealed that the synthesis of trimethylolpropane (TMP) using 37% formaldehyde and n-butyraldehyde as raw materials via the calcium-based reaction principle of the Cannizaro process generates some byproducts, thus affecting product quality. Therefore, a near-infrared spectroscopy (NIRS) model can be used to control the quality of TMP. This chapter first optimizes the factory's production parameters and then selects samples at different times for GC-MS analysis to obtain real-time concentration data. By combining the sample GC-MS data with NIRS spectral data, a widely used NIRS model is constructed, providing theoretical support for its subsequent application in factory monitoring.

[0183] 4.1 Single-factor experiment

[0184] Based on the existing production parameters of Chifeng Ruiyang Chemical Co., Ltd., we re-examined the single-factor effects of the Connizaro reaction temperature, reaction time, feeding rate, aldol condensation reaction temperature, and molar ratio in its production.

[0185] 4.1.1 Effect of reaction time on trimethylolpropane concentration

[0186] 39.0 g of calcium hydroxide was weighed into a three-necked flask and dissolved in 270.0 g of water. Then, 207 g of 37% formaldehyde and 60.0 g of n-butyraldehyde were weighed and placed in a longer pump peristaltic pump. The mixture was added dropwise at a constant flow rate of 10.4 g / mL at 25°C for 2 hours. The temperature was then raised to 40°C, and the reaction times were 140 min, 160 min, 180 min, 200 min, and 220 min, respectively. Samples were taken, neutralized with formic acid, and 1 mL of the supernatant was added. Two mL of internal standard were added, followed by dilution with 7 mL of anhydrous ethanol. Gas chromatography and near-infrared spectroscopy were performed simultaneously. The concentrations of each component are shown in Table 12 below.

[0187]

[0188] Appropriately increasing the reaction time can increase the yield of trimethylolpropane (TMP). However, excessively prolonged time will cause the reaction to proceed in reverse, leading to an increase in 2-ethyl-2-hexenal. In the early stages of the reaction, 2-hydroxymethylbutyraldehyde, formed from the reaction of formaldehyde and n-butyraldehyde, is prone to self-dehydration condensation to 2-ethylpropenal. However, as 2-hydroxymethylbutyraldehyde continues to undergo aldol condensation to 2,2-dihydroxymethylbutyraldehyde, the yield decreases, causing 2-ethylpropenal to oxidize back to 2-hydroxymethylbutyraldehyde. The yield of TMP reaches an equilibrium state with increasing reaction time. Considering the mass concentration of TMP, 140 min, 160 min, and 180 min were selected.

[0189] 4.1.2 The effect of the Connizaro reaction temperature on trimethylolpropane

[0190] 39.0 g of calcium hydroxide was weighed into a three-necked flask and dissolved in 270.0 g of water. Then, 207 g of 37% formaldehyde and 60.0 g of n-butyraldehyde were weighed into a longer pump and added dropwise at a constant flow rate of 10.4 g / mL at 20°C for 2 hours. The temperature was then adjusted to 20°C, 30°C, 40°C, 50°C, and 60°C, and the reaction was allowed to proceed for 40 minutes. A sample was taken, neutralized with formic acid, and 1 mL of supernatant was added. 2 mL of internal standard was added, followed by dilution with 7 mL of anhydrous ethanol. Gas chromatography and near-infrared spectroscopy were performed simultaneously. The concentrations of each component are shown in Table 13 below.

[0191]

[0192] The table shows that when the reaction temperature reaches 60℃, the concentration of byproducts increases significantly, while the concentration of products decreases significantly. At lower temperatures, the TMP concentration is very low, only 7.8630 mg / mL. Increasing the temperature promotes the self-condensation of trimethylolpropane to bis(trimethylolpropane), thus reducing the yield of trimethylolpropane. At lower temperatures, the carbon atom activity in the disproportionation reaction of 2,2-dimethylolbutyraldehyde decreases, with the highest trimethylolpropane concentration only observed at 40℃. Therefore, 40℃ was chosen as the reaction temperature for the second stage.

[0193] 4.1.3 Effect of Feeding Rate on Trimethylolpropane Product Concentration

[0194] Weigh 39.0 g of calcium hydroxide into a three-necked flask and dissolve it in 270.0 g of water. Weigh 60.0 g of n-butyraldehyde and place it in a longer pump. Add the solution at a constant flow rate of 10, 10.1, 10.2, 10.3, and 10.4 h at 20 °C for 2 h. Then adjust the temperature to 40 °C and react for 40 min. Take a sample, neutralize with formic acid, take 1 ml of the sample, add 2 ml of internal standard, and then dilute with 7 ml of anhydrous ethanol. Perform simultaneous gas chromatography and near-infrared spectroscopy to determine the concentrations of each component. The concentrations are shown in Table 14 below.

[0195]

[0196] Depend on Figure 22 It can be seen that the feeding rate has little effect on the concentration of trimethylolpropane, but a faster dropping rate of n-butyraldehyde will increase the amount of free n-butyraldehyde in the reactor, thereby initiating the self-condensation reaction of n-butyraldehyde; with the increase of the feeding rate, the increase rate of 2-ethyl-1,3-hexanediol is accelerated. Therefore, the minimum feeding rate is selected as the subsequent reaction condition.

[0197] 4.1.4 Raw material molar ratio for trimethylolpropane

[0198] Weigh 39.0 g of calcium hydroxide into a three-necked flask and dissolve it in 270.0 g of water. Weigh an appropriate amount of formaldehyde, and weigh 60.0 g of n-butyraldehyde into a longer pump. Add the formaldehyde dropwise at a constant flow rate of 10 g / L at 20 °C for 2 hours. Then adjust the temperature to 40 °C and react for 40 minutes. Take a sample, neutralize with formic acid, and take 1 mL of the sample. Add 2 mL of internal standard, and then dilute with 7 mL of anhydrous ethanol. Perform simultaneous determination by gas chromatography and near-infrared spectroscopy. See Tables 15-16; results are shown in Table 16. Figure 23 .

[0199] ;

[0200]

[0201] from Figure 23 It can be seen that the concentration of trimethylolpropane initially increases with time, then tends to stabilize, with the fastest increase rate observed at a molar ratio of 3.08. From the perspective of chemical reaction equilibrium, in organic reactions, increasing the amount of inexpensive raw materials increases the conversion rate. At molar ratios of 3.06–3.1, the concentrations of 2-ethylpropenal and 2-ethyl-1,3-hexanediol are relatively low, while the concentration of trimethylolpropane is significantly higher compared to molar ratios of 3.02 and 3.04.

[0202] 4.1.5 The effect of aldol condensation reaction temperature on trimethylolpropane

[0203] Weigh 39.0 g of calcium hydroxide into a three-necked flask and dissolve it in 270.0 g of water. Weigh 208 g of 37% formaldehyde and 60.0 g of n-butyraldehyde into a longer pump. Add the solution dropwise at a constant flow rate of 10 mL at 20°C, 25°C, 30°C, 35°C, and 40°C for 2 hours. Then adjust the temperature to 40°C and react for 40 minutes. Take a sample, neutralize with formic acid, and take 1 mL of the sample. Add 2 mL of internal standard, then dilute with 7 mL of anhydrous ethanol. Perform simultaneous gas chromatography and near-infrared spectroscopy. The mass concentrations of each component are shown in Table 17.

[0204]

[0205] from Figure 24 It can be seen that after determining the optimal feeding rate and Cannizaro reaction temperature, the concentration of trimethylolpropane increases, but both the concentration and the rate of increase are relatively low at lower temperatures. However, increasing the temperature also increases the concentration of byproducts. Using the concentration of trimethylolpropane as a reference standard, the optimal temperature range is 30–40°C.

[0206] 4.2 Response Surface Optimization Experiment

[0207] Based on the above single-factor experimental results, with product concentration as the reference standard, the reaction time of 140~180min, the raw material molar ratio of 3.06~3.1, and the aldol condensation reaction temperature of 30~40℃ were selected as the response surface experimental level values.

[0208] 4.2.1 Experimental Design

[0209] The experiment was designed using DesignExpert 12.0 software. Details of the design are shown in Table 18 below:

[0210]

[0211] 4.2.2 Response Surface Experiment Results

[0212] Based on the above experimental table, design-expert 12.0 software designed the following 17 experiments, and the experimental results are shown in Tables 19-21 below.

[0213]

[0214] By fitting the experimental data, the regression equation of the model is Y=12.82+0.1663X1-1.16X2+0.3561X3-0.7465X1X2+0.8075X1X3+0.3367X2X3-1.48X1 2 -1.09X2 2 - 0.8555X3 2 .

[0215] ;

[0216]

[0217] As shown in Tables 18-21, the model's F-value is 66.04, and the P-value is <0.0001, indicating that the model is significant. The P-value for the lack-of-fit term is 0.2926 > 0.05, indicating that the model error is within the acceptable range and the lack-of-fit term is not significant. Experimental R... 2 =0.9884, compared with the correction value R 2 The difference of 0.9737 is less than 0.2, indicating high model reliability. The signal-to-noise ratio (SNR) of 22.674 > 4, and the model's CV% (coefficient of variation) of 2.21% (less than 10%), indicate high precision and accuracy. Furthermore, the p-values ​​of the linear terms X2, X1, and X2 (<0.0001) are shown in the table above, indicating they are highly significant. The interaction between the linear term X1-cannizzaro reaction time and the molar ratio of formaldehyde to n-butyraldehyde and condensation temperature (p-value <0.05) has a significant effect on TMP concentration. The p-value of the linear term X1-Cannizzaro reaction time is >0.05, indicating its insignificant effect on TMP concentration.

[0218] 4.2.3 Response Surface Analysis

[0219] 4.2.3.1 Residual Analysis

[0220] The correlation and reliability of the model and data can be tested using residuals.

[0221] As a result, by Figure 25 The internal biochemical residual plot shows that the experimental data points are evenly distributed on both sides of the line, indicating a normal distribution pattern and a small difference between the standard deviation and the actual value. The model follows a normal distribution. Figure 26 To illustrate the relationship between the residuals and the rising predicted response values, the residuals show a random distribution and are uncorrelated with the fitted values, indicating that the model accurately describes the experiment. The difference between the actual and predicted values ​​can be used to verify the correlation and reliability of the model and the data. The graph shows that the points representing the experimental values ​​are evenly distributed along a straight line, indicating that the measured values ​​deviate little from the predicted values, and the model exhibits a normal distribution. Outliers in the numerical regression equation can be examined using externally derived residuals. All residuals do not exceed ±2.0, indicating that the experimental data meets the requirements and there are no outliers. In the residual-run graph, all residuals show a random distribution without any specific pattern, indicating that the experimental assumptions are independent. These results demonstrate the validity of the regression model.

[0222] 4.2.3.2 Response Surface Analysis

[0223] In 3D plots, steeper surface slopes indicate a stronger impact on TMP concentration. The more circular the contour lines, the less significant the interaction between the two phases on TMP concentration; conversely, the more elliptical the contour lines, the more significant the interaction's effect on TMP. Figure 27-28 As can be seen, when the aldol condensation temperature is constant, the contour plot of the interaction between the Cannizzaro reaction time and the molar ratio of formaldehyde to n-butyraldehyde is elliptical, and the three-dimensional plot has a steep slope, indicating that the interaction has a large and significant effect on the TMP concentration. With increasing Cannizzaro reaction time, the intensity of the effect of the molar ratio on the TMP mass concentration first increases and then decreases. This is because when the Cannizzaro reaction time is short, the disproportionation reaction between HCHO and the intermediate 2,2-dihydroxymethylbutyraldehyde has not reached equilibrium. However, when equilibrium is reached, TMP will self-condense to form the byproduct bis(trimethylolpropane). Figures 29-30 The contour plot and 3D plot show the interaction between the Cannizzaro reaction time and the aldol condensation temperature, respectively elliptical and steep, indicating a strong and significant effect of this interaction on the TMP concentration. When HCHO:CH3CH2CH2CHO is fixed, as the Cannizzaro reaction time increases, the aldol condensation temperature increases, causing the TMP concentration to initially increase and then decrease. This is because higher temperatures shift the reaction towards the formation of the target compound. Figures 31-32 It can be seen that the contour lines of the interaction between the formaldehyde-n-butyraldehyde molar ratio and the aldol condensation temperature are elliptical, but not as strong as the two interactions mentioned above; the steep slope of the three-dimensional graph indicates that the interaction effect is strong and relatively significant. Under a fixed Cannizzaro reaction time, as the molar ratio of formaldehyde to n-butyraldehyde increases, the effect of aldol condensation temperature on TMP first increases and then decreases, and the aldol condensation temperature is optimal at 34~38℃. This may be because at this temperature, the aldol molar ratio of n-butyraldehyde to formaldehyde and the aldol condensation lock temperature reach a relatively suitable state, which can better synthesize the intermediate product CH3CH2C(CH2OH)2CHO.

[0224] 4.2.3.3 Optimal Synthesis Conditions

[0225] The optimal synthesis conditions recommended by Design-Expert 12.0 were: Cannizzaro reaction time: 169 min, formaldehyde-n-butyraldehyde molar ratio: 3.061, and aldol condensation temperature: 40℃. The optimal TMP concentration was 12.744 mg / mL. Three parallel experiments were conducted under these parameters, yielding results of 12.418 mg / mL, 12.410 mg / mL, and 12.391 mg / mL, respectively. The average of the three experiments was 12.406 mg / mL, the RSD of the parallel experiments was 0.22% (<2%), and the deviation rate from the predicted value was 2.65%, indicating that the established model has high reliability.

[0226] 4.3 Quality Control Study of Trimethylolpropane and its Byproducts

[0227] From the 27 single-factor and response surface methodology experiments mentioned above, samples were collected at different time points (0 min, 15 min, 30 min, 45 min, 60 min, 75 min, 90 min, 105 min, 120 min, 135 min, 140 min, 150 min, 160 min, 165 min, 180 min, and 220 min), totaling 316 data points. This data was used as the modeling database. The partial least squares method was used to build the model, determining the optimal spectral processing method and the impact of outlier screening on the model's R-value. Because some data points deviated from extreme values ​​due to inconsistent experimental procedures or machine usage, the calibration set could not be well-matched. To address this issue, the spectral distribution differences in the SpectrumOutlier option of TQ Analyst were used to remove unreasonable data, resulting in a more accurate model. See [link to TQ Analyst documentation]. Figure 33 .

[0228] 4.3.1 Establishment of the trimethylolpropane model

[0229] 4.3.1.1 Selection of Spectral Processing Method

[0230] Based on the RMSEC values ​​and Rc values ​​obtained from different spectral processing methods, and the feedback from the partial least squares method in the model establishment experiment, it can be seen that the smoothing method using standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norris derivative filter yields an RMSEC of 0.919 and an Rc of 0.9730. The prediction results are shown in Table 22.

[0231]

[0232] 4.3.1.2 Anomaly Removal

[0233] Outliers were removed using Mahalanobis distance from the spectra in the Spectrum Outlier option of TQ Analyst. The Mahalanobis distance results for outliers during the synthesis of trimethylolpropane are shown in the figure below. The results show three outliers. With outliers present, the model's RMSEC was 0.919 and Rc was 0.9730. After outlier removal, the model's RMSEC was 0.862 and Rc was 0.9759. The model diagram shows that the model established after outlier removal is accurate and reliable. The trimethylolpropane model is shown below. Figures 34-36 .

[0234] 4.3.2 Establishment of the 2-Ethylpropenal Model

[0235] 4.3.2.1 Selection of Spectral Processing Method

[0236] According to the feedback table 4-11 on the results of partial least squares method model establishment in the experiment, based on the RMSEC value and Rc of different spectral processing methods, the smoothing method using multivariate signal correction (MSC) + first derivative spectroscopy (FD) + Savitzky-Golayfilter has an RMSEC value of 0.143 and an Rc value of 0.9417.

[0237]

[0238] 4.3.2.2 Anomaly Removal

[0239] Outliers were removed using Mahalanobis distance from the spectra in the Spectrum Outlier option of TQ Analyst. The Mahalanobis distance results for outlier samples during the synthesis of 2-ethylpropenal are shown in the figure below. When outliers appear, they are separated by dashed lines in the software. Since no outliers were found on the Mahalanobis distance map, there were no outlier samples for 2-ethylpropenal. The model has an RMSEC of 0.413 and an Rc of 0.9417, indicating good correlation between predicted and measured values. The model can be found here. Figures 37-39 .

[0240] 4.3.3 Establishment of the 2-Ethyl-2-hexenal model

[0241] 4.3.3.1 Selection of Spectral Processing Method

[0242] According to the feedback table 24 on the results of partial least squares method for model establishment in the experiment, based on the RMSEC value and Rc of different spectral processing methods, the results of the smoothing method using standard canonical transformation (SNV) + second derivative spectroscopy (SD) + Norrisderivative filter are: RMSEC=0.0429, Rc=0.9307.

[0243]

[0244] 4.3.3.2 Outlier Removal

[0245] Outliers were removed using the Mahalanobis distance of the spectra in the Spectrum Outlier option of TQ Analyst. Outliers were marked with dashed lines in the software. The Mahalanobis distance results for outlier samples during the synthesis of 2-ethyl-2-hexenal are shown in the figure below. Four outliers were present; with these outliers, the RMSEC was 0.0429 and Rc was 0.9307. After removing these four outliers, the RMSEC was 0.0420 and Rc was 0.9346. Removing the outliers resulted in a better correlation between the predicted and measured values, and higher simulation accuracy. The model results are shown below. Figures 40-42 .

[0246] 4.3.4 Establishment of the 2-Ethyl-1,3-Hexanediol Model

[0247] 4.3.4.1 Selection of Spectral Processing Method

[0248] According to the feedback table 25 on the results of partial least squares method model establishment in the experiment, based on the RMSEC value and Rc of different spectral processing methods, the results of the smoothing method using standard canonical transformation (SNV) + first derivative spectroscopy (FD) + Norrisderivative filter are: RMSEC=0.0212, Rc=0.9221.

[0249]

[0250] 4.3.4.2 Outlier Removal

[0251] Outliers were removed using the Mahalanobis distance of the spectra in the Spectrum Outlier option of TQ Analyst. Outliers are indicated by dashed lines in the software. The Mahalanobis distance results for outlier samples during the synthesis of 2-ethyl-1,3-hexanediol are shown in the figure below. One outlier is present; when this outlier is present, the RMSEC is 0.0212 and Rc is 0.9221. Removing these four outliers did not affect the RMSEC and Rc values, but it adjusted the predicted RMSEC and Rp values ​​from 0.0310 and 0.8159 to 0.0300 and 0.8432, respectively, increasing the accuracy of the predictions and improving the simulation accuracy. (The model is shown below.) Figures 43-45 .

[0252] 4.4 Summary

[0253] Based on the established detection and model building methods, and using the trihydroxypropane production method at Chifeng Ruiyang Chemical Co., Ltd., the effects of reaction time, reaction temperature, Cannizaro reaction temperature, n-butyraldehyde dropping rate, and the molar ratio of formaldehyde to n-butyraldehyde on the concentrations of each component were re-examined. The influence of these factors was in the following order: raw material molar ratio > aldol condensation temperature > reaction time > reaction temperature > feed rate. Then, the raw material molar ratio (3.06-3.1), aldol condensation temperature (30-40℃), and reaction time (140-160 min) were selected as factor levels for response surface methodology (RSM) experiments. The RSM analysis showed a model F-value of 66.04 and an experimental R-value of [missing value]. 2 =0.9884, compared with the correction value R 2 =0.9734, the difference is less than 0.2, indicating that the model has high credibility. The signal-to-noise ratio is 22.674>4, and the CV% (coefficient of variation) of the model is 2.21% and less than 10%, indicating that the model has high precision and accuracy.

[0254] Samples were collected at different time points using both single-factor and response surface methodology. A database was established using combined gas chromatography and near-infrared spectroscopy data. This database was then analyzed using partial least squares (PLS) quantitative analysis, resulting in four quantitative models: Rc=0.9759 for trimethylolpropane; Rc=0.9412 for 2-ethylpropenal; Rc=0.9346 for 2-ethyl-2-hexenal; and Rc=0.9221 for 2-ethyl-1,3-hexanediol. These four models are accurate and reliable, and their application to quality control studies in the trimethylolpropane production process can be considered.

[0255] V. Application of models for trimethylolpropane and its byproducts

[0256] 5.1 Application of the model to optimal experimental parameters

[0257] 5.1.1 Experiment

[0258] Weigh 208.0 g of 37% formaldehyde, 60.0 g of n-butyraldehyde, 39.0 g of calcium hydroxide, and 270.0 g of water. Add the solution dropwise at a constant flow rate of 10.0 g using a long-range peristaltic pump at 40℃ for 2 hours, maintaining the reaction for 49 minutes. Take samples at 30, 60, 90, 120, 140, and 169 minutes. Adjust the pH to approximately 6.7 with 88% formic acid. Take 1 ml of the sample, add 2 mL of 12.17 mg / mL internal standard solution, dilute with 7 mL of 99.7% anhydrous ethanol, centrifuge, and collect the supernatant. Analyze the solution using the previously established method.

[0259] 5.1.2 Experimental Results

[0260] First, the spectral data were used to calculate the predicted concentration values ​​using a model. These predicted concentration values ​​were then compared with the actual values ​​measured by gas chromatography using the internal standard method. The results are shown in Table 26 below. The table shows that the deviation rate between the predicted and measured values ​​for all four substances is less than 3%, indicating that the model established by sampling at different time points in the condensation reaction can accurately predict the concentration at each point. Using predicted values ​​instead of measured values ​​saves time compared to gas chromatography results, greatly improving production efficiency.

[0261]

[0262] 5.2 Applicability of Near-Infrared Model to Actual Production

[0263] 5.2.1 Experiment

[0264] Subsequent experiments were conducted using samples obtained from the production workshop of Chifeng Ruiyang Chemical Co., Ltd. Three batches of samples were collected from various production processes, including the condensation liquid after reaction, the bottom of the extraction tower, the extraction feed tower, the top of the extraction tower, the reflux tank of the solvent recovery tower, the bottom of the solvent recovery tower, the TMP refining raw material tank, the TMP recovery tower reflux tank, and the distillation tower reflux tank. Sudden reductions in concentration were considered during the experiments. For each sample, 1 mL was taken, 2 mL of internal standard was added, anhydrous sodium sulfate was added to remove moisture, and 7 mL of 99.7% anhydrous ethanol was added for dilution. Near-infrared spectroscopy and gas chromatography measurements were performed using the same methods as in section 5.1.

[0265] 5.2.2 Comparison of Model Predictions and Measured Values

[0266] The established model was used in TQ Analyst's Quantify function to analyze samples of unknown concentrations, identify the spectra to be analyzed, and generate a sample analysis report. Data from three batches of samples are shown in Tables 27, 28, and 29.

[0267] ;

[0268] ;

[0269]

[0270] The predicted and measured values ​​from the three batches show that, except for the condensate after the reaction, where the deviation was less than 3%, the deviations for all other processes were greater than 3%. This indicates that the established model can only accurately predict the concentrations of each component after the reaction, but not the changes in concentrations during other processes. It is speculated that the model might have used data from the entire condensation reaction, excluding post-processing steps. To address this issue, multiple batches of data should be collected separately for each process, and individual models should be developed.

[0271] Although the present invention has been described in detail above with general descriptions, specific embodiments, and experiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for establishing an online near-infrared spectral quantitative model for the synthetic product TMP of trimethylolpropane and its byproducts, characterized in that, include: (1) Gas chromatography-mass spectrometry and infrared spectroscopy were used to determine the final product TMP and its byproducts in the calcium method for the synthesis of trimethylolpropane; (2) Gas chromatography was used to quantitatively analyze TMP and its byproducts and determine their concentrations; (3) Near-infrared spectroscopy was used to collect and analyze the spectroscopic information of TMP and its byproducts; the conditions for the spectroscopic collection and analysis were: the scanning mode was a transmission analysis module, and the resolution was 8 cm⁻¹. –1 64 scans, spectral scanning range 4000~10000 cm⁻¹ –1 (4) Establishment of quantitative models for TMP and its byproducts: TQ Analyst software was used to preprocess the spectral information of TMP and its byproducts. The Mahalanobis distance of the spectrum in the Spectrum Outlier option of TQ Analyst was used to remove outliers and abnormal samples in the near-infrared spectral samples. The concentrations determined by gas chromatography and the spectral data collected by near-infrared spectroscopy were combined to establish a database. The partial least squares method was used to establish quantitative models for TMP and its byproducts. The byproducts are: 2-ethylpropenal, 2-ethyl-2-hexenal, and 2-ethyl-1,3-hexanediol. The method for synthesizing trimethylolpropane is as follows: 39.0 g of calcium hydroxide is weighed and added to a three-necked flask, followed by 270.0 g of deionized water, and stirred for 5 min to mix evenly; then 208.0 g of HCHO is weighed into the same three-necked flask and cooled to 25°C; then 60.0 g of n-butyraldehyde is weighed into a Longerpump constant flow peristaltic pump, and added dropwise at a rate of 10.0 g over a 25°C water bath for two hours; after the addition is complete, the temperature is raised to 40°C and the reaction is allowed to proceed for 40 min; then 27.5% hydrogen peroxide is added to oxidize excess HCHO, and then neutralized to pH 6.7 with 88% formic acid solution to obtain the condensate; The sample is the condensate after the reaction is complete; The specific steps of the gas chromatography quantitative analysis described in step (2) include: ① Preparation of internal standard solution: Accurately weigh an appropriate amount of 1,6-hexanediol, dissolve it in ethanol, prepare an internal standard solution of 11.57 mg / mL, and store it at low temperature for later use; ② Preparation of reference solution: Accurately weigh an appropriate amount of trimethylolpropane reference standard, dissolve it in ethanol to prepare a 12.24 mg / mL trimethylolpropane reference solution, and store it at low temperature for later use; ③ Preparation of mixed reference solutions: Accurately weigh appropriate amounts of 2-ethylpropenal, 2-ethyl-2-hexenal, 2-ethyl-1,3-hexanediol, internal standard 1,6-hexanediol, and trimethylolpropane reference standards, dissolve them in ethanol, and prepare mixed reference solutions with mass concentrations of 0.43, 1.02, 2.42, 2.69, and 12.24 mg / mL, respectively. Store at low temperature for later use. ④ Preparation of the test solution: Take an appropriate amount of sample, dilute with anhydrous ethanol, add anhydrous sodium sulfate to remove water, centrifuge to obtain the solution, and store at low temperature for later use. The ratio of the sample to anhydrous ethanol is sample:anhydrous ethanol = 1:9; ⑤ The determination was performed by gas chromatography. The gas chromatography parameters were as follows: Detector: Agilent 7890B flame ionization detector; Injection volume: 0.3µL; Heater: 270℃; Split ratio: 30:1; Split flow rate: 6mL / min; Column: DB-1701, -20℃~280℃, 30m×250µm×0.25µm; Flow rate: 0.2mL / min; Pressure: 2.7033psi; Initial temperature: 40℃ held for 5min, v1=1℃ / min increased to 70℃ held for 1min, v2=3℃ / min increased to 160℃ held for 1min, v2=5℃ / min increased to 270℃ held for 6min.

2. The method for establishing an online near-infrared spectral quantitative model for the trimethylolpropane synthesis product TMP and its byproducts according to claim 1, characterized in that, Step (1) uses Agilent 8860 5977b gas chromatography-mass spectrometry (GC-MS) for analysis; the GC-MS detection method is as follows: column: DB-5MS; column temperature: initial 40℃ for 5 min, rate 1℃ / min to 70℃ for 1 min, rate 3℃ / min to 160℃ for 1 min, rate 5℃ / min to 270℃ for 6 min; injection port temperature: 280℃; split ratio: 20:

1.

3. The method for establishing an online near-infrared spectral quantitative model for the trimethylolpropane synthesis product TMP and its byproducts according to claim 1, characterized in that, The specific steps of the spectral acquisition and analysis described in step (3) are as follows: First, preheat the spectrometer for more than 1 hour, and take samples at 15 min, 30 min, 45 min, 60 min, 75 min, 90 min, 105 min, 120 min, 135 min, 150 min, 165 min and 180 min during the TMP preparation process. Process the samples according to the preparation method of the test solution described above, and use the Antaris II Fourier transform near-infrared spectrometer to acquire the near-infrared spectrum.

4. The method for establishing an online quantitative model of the near-infrared spectra of the trimethylolpropane synthesis product TMP and its byproducts according to claim 1, characterized in that, The quantitative models of TMP and its byproducts in step (5) are: trimethylolpropane model with Rc=0.9759; 2-ethylpropenal model with Rc=0.9412; 2-ethyl-2-hexenal model with Rc=0.9346; and 2-ethyl-1,3-hexanediol model with Rc=0.9221.

5. An online quantitative model for near-infrared spectroscopy of the trimethylolpropane synthesis product TMP and its byproducts as described in any one of claims 1-4, applied to the quality control of the trimethylolpropane condensation reaction synthesis product, wherein the sample to be tested is subjected to near-infrared spectroscopy, spectral data is collected, input into the established model, the content of trimethylolpropane and its byproducts during the process synthesis is rapidly determined, the process synthesis of trimethylolpropane is quantitatively monitored, and online monitoring of the entire trimethylolpropane synthesis process is achieved.

Images