Establishment method and application of multi-substance on-line analysis model and method for simultaneous on-line determination of multiple substances in 1, 4-DMN production process

Abstract

The invention provides an establishment method and application of a multi-substance online analysis model and a method for simultaneous online determination of multiple substances in a 1, 4-DMN production process, and belongs to the field of near infrared spectrum analysis. The establishment method provided by the invention comprises the following steps: (1) collecting a sample; (2) sample analysis; (3) establishing a model to be verified; and (4) model verification. According to the method, the near-infrared scanning spectrogram is associated with the contents of the multiple target compounds of each sample to establish multiple models, and the models are verified, so that the multiple established models can quickly detect the contents of the multiple target compounds, and the contents of the multiple target compounds can be monitored on line in real time at the same time. The result of the embodiment shows that the RMSECV value of the model obtained by the establishment method provided by the invention is 0.292, and the contents of various substances in the 1, 4-dimethylnaphthalene production process can be simultaneously and accurately measured.

Description

Technical Field

[0001] The invention relates to the field of near-infrared spectral analysis, and in particular to a method for establishing a multi-substance online analysis model and an application thereof, and a method for simultaneously determining multiple substances online in a 1,4-DMN production process. Background Art

[0002] 1,4-Dimethylnaphthalene (1,4-DMN) is very important in industry as an intermediate raw material for 1,4-naphthalenedicarboxylic acid, a raw material for resins and dyes. Generally, industrial applications expect the isomer content in 1,4-DMN to be less than 1.0%, or even less than 0.5%, or further reduced to less than 0.4%. In DMN production, it was found that 1,4-DMN produced by cyclodehydrogenation of 5-phenyl-2-hexene contained a large amount of 1,3-DMN, which is an isomer. The boiling point difference between the isomers 1,3-DMN and 1,4-DMN is small, so the proportion of 1,3-DMN in the product and the control of the pre-distillation process parameters during the distillation of crude 1,4-DMN are particularly important in industrial production.

[0003] The existing analytical technology for detecting the content of each substance in the production process of 1,4-DMN is usually gas chromatography analysis, and the process parameters are adjusted based on the content data of each substance obtained by gas chromatography analysis. However, gas chromatography analysis takes a long time to analyze, and it is necessary to take samples from the reactor and distillation device in the workshop and send them to the analysis and detection center for analysis. The time for taking and sending samples and measuring a sample is about 1 hour. During the long sampling and analysis process, the intermediate products and products have undergone significant changes, and it is impossible to accurately provide real-time and accurate data for the regulation of production parameters. In addition, during the sampling and measurement process, the measurement results may be inaccurate due to sampling contamination, resulting in incorrect workshop operation and product quality fluctuations. In addition, the sampling process involves high temperatures, and there are also major safety issues. Summary of the invention

[0004] The purpose of the present invention is to provide a method for establishing a multi-substance online analysis model and its application and a method for simultaneously online determination of multiple substances in a 1,4-DMN production process. When the model obtained by the establishment method provided by the present invention is used to determine the content of multiple substances in a 1,4-DMN production process, the safety factor is high, the content of multiple substances in the 1,4-DMN production process can be quickly detected at the same time, the analysis results are fast and time-effective, and the changes in the content of multiple substances in the 1,4-DMN production process can be monitored in real time.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0006] The present invention provides a method for establishing a multi-substance online analysis model, comprising the following steps:

[0007] (1) Sample collection: Samples are collected at sampling points to obtain samples; the sampling points are different time points in the same reaction stage during the production process;

[0008] (2) Sample analysis: performing gas chromatography analysis and near infrared scanning on each sample in step (1) to obtain measured values ​​of the contents of multiple target compounds in each sample and a near infrared spectrum of each sample;

[0009] (3) Establishing the model to be verified: Correlating the measured values ​​of the content of each target compound in some samples in step (2) with the near-infrared spectra of the corresponding samples and optimizing the model to obtain multiple models to be verified;

[0010] (4) Model verification: importing the near infrared spectra of the remaining samples in step (2) into the multiple models to be verified obtained in step (3) to predict the content of the corresponding target compounds, and obtain predicted values, and then comparing the obtained predicted values ​​with the measured values ​​of the content of the multiple target compounds in the corresponding samples obtained in step (2); when the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, a multi-substance online analysis model is obtained; when the samples whose relative deviations between the predicted values ​​and the measured values ​​do not meet the detection accuracy requirements account for more than 10% of the remaining sample amount, changing the association method and / or model optimization method in step (3) until the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, and a multi-substance online analysis model is obtained.

[0011] Preferably, the number of samples in step (1) is ≥ 70.

[0012] Preferably, the number of partial samples in step (3) is 85-90% of the total number of samples.

[0013] Preferably, the step (3) of establishing the model to be verified comprises the following steps:

[0014] 1) Input the measured values ​​of multiple target compounds of some samples and the near infrared spectra of the corresponding samples into the software;

[0015] 2) Preprocessing each near-infrared spectrum input in step 1) and then selecting a band interface to obtain a processed near-infrared spectrum;

[0016] 3) correlating the measured values ​​of the multiple target compounds input in step 1) with the processed near infrared spectra obtained in step 2) to obtain multiple initial models;

[0017] 4) Optimizing the multiple initial models obtained in step 3) respectively to obtain multiple models to be verified.

[0018] Preferably, the preprocessing in step 2) includes one or more of first-order derivative processing, second-order derivative processing and vector normalization.

[0019] Preferably, the associated method in step 3) includes partial least squares regression quantitative method, multivariate linear regression or principal component regression.

[0020] The present invention also provides the application of the model obtained by the establishment method described in the above technical solution in the online detection of multiple substances in chemical production.

[0021] The present invention also provides a method for simultaneously online determination of multiple substances in a 1,4-dimethylnaphthalene production process, comprising:

[0022] Performing near infrared detection on the reaction liquid of the reaction stage to be detected in the production process of 1,4-dimethylnaphthalene to obtain a near infrared spectrum;

[0023] Substituting the near infrared spectrum into a multi-substance online analysis model of a corresponding reaction stage to obtain the contents of multiple target compounds in the reaction solution of the corresponding reaction stage;

[0024] The multi-substance online analysis model of the corresponding reaction stage is established by the establishment method described in the above technical solution, and the sample used when establishing the model is the reaction liquid of the corresponding reaction stage in the 1,4-dimethylnaphthalene production process.

[0025] Preferably, the reaction stages in the 1,4-dimethylnaphthalene production process include an ethylbenzene recovery stage, a 5-phenyl-2-hexene distillation stage, a 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene distillation stage and a 1,4-dimethylnaphthalene product purification stage.

[0026] Preferably, when the multi-substance online analysis model of the ethylbenzene recovery stage is established, the sampling points are different reaction times in the ethylbenzene recovery stage, and the target compound in the sample is ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene;

[0027] When establishing the multi-substance online analysis model of the 5-phenyl-2-hexene distillation stage, the sampling points are at different times during the 5-phenyl-2-hexene distillation process, and the target compound in the sample is ethylbenzene or 5-phenyl-2-hexene;

[0028] When the multi-substance online analysis model of the 1,4-dimethyl-1,2,3,4-tetralin distillation stage is established, the sampling points are different times during the 1,4-dimethyl-1,2,3,4-tetralin distillation process, and the target compound in the sample is ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetralin;

[0029] When the multi-substance online analysis model of the 1,4-dimethylnaphthalene product purification stage is established, the sampling points are at different times during the purification process of the 1,4-dimethylnaphthalene product, and the target compound in the sample is 1,4-dimethylnaphthalene or impurities; the impurities include 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene or 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene.

[0030] The present invention provides a method for establishing a multi-substance online analysis model, comprising the following steps: (1) sample collection: collecting samples at sampling points; the sampling points are different time points in the same reaction stage during the production process; (2) sample analysis: performing gas chromatography analysis and near-infrared scanning on each sample in the step (1) to obtain the measured values ​​of the contents of multiple target compounds in each sample and the near-infrared spectrum of each sample; (3) establishing a model to be verified: correlating the measured values ​​of the contents of each target compound in some samples in the step (2) with the near-infrared spectrum of the corresponding samples and optimizing the model to obtain multiple models to be verified; (4) model verification: importing the near-infrared spectrum of the remaining samples in the step (2) into the near-infrared spectrum of the remaining samples; The corresponding target compound contents are predicted in the multiple models to be verified obtained in step (3) to obtain predicted values, and then the predicted values ​​are compared with the measured values ​​of the target compound contents in the corresponding samples obtained in step (2). When the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, a multi-substance online analysis model is obtained; when the samples whose relative deviations between the predicted values ​​and the measured values ​​do not meet the detection accuracy requirements account for more than 10% of the remaining sample amount, the association method and / or model optimization method in step (3) are changed until the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, and a multi-substance online analysis model is obtained. The present invention associates a near-infrared scanning spectrum with the contents of multiple target compounds of each sample to establish multiple models, and verifies the multiple models, so that the relative deviations between the predicted values ​​and the measured values ​​of the contents of multiple target compounds in the near-infrared spectrum by the established multiple models meet the demand for detection accuracy in production. By utilizing the model, near-infrared online detection can be adopted without manual sample collection and delivery, with a high safety factor, and the corresponding target compound contents can be quickly detected. The analysis results are fast and timely, and simultaneous online real-time monitoring of the contents of multiple target compounds can be achieved.The results of the embodiment show that the RMSECV value of the model obtained by the establishment method provided by the present invention can reach 0.292. The model is used to determine the content of 1,4-dimethylnaphthalene and impurities in the purification stage of 1,4-dimethylnaphthalene product in the production process of 1,4-dimethylnaphthalene. The samples whose predicted 1,4-dimethylnaphthalene content value has a relative deviation of less than 0.5% from the measured value account for 100% of the remaining sample amount, the samples whose predicted 1,3-dimethylnaphthalene content value has a relative deviation of less than 3% from the measured value account for 100% of the remaining sample amount, and the predicted 5,8-dimethyl-1,2,3 , the samples whose relative deviation between the content of 4-tetralin and the measured value is less than 3% account for 100% of the remaining samples, the samples whose relative deviation between the content of 1,4-dimethyl-1,2,3,4-tetralin and the measured value is less than 3% account for 100% of the remaining samples, and the samples whose relative deviation between the content of each other impurity (impurity A, impurity B, impurity C and impurity D) and the measured value is less than 15% account for 100% of the remaining samples, which meets the production requirements for detection accuracy and can simultaneously and accurately determine the contents of multiple substances in the production process of 1,4-dimethylnaphthalene. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The near infrared spectra of 70 samples used for establishing the model to be verified in Application Example 1 of the present invention;

[0032] Figure 2 In Application Example 1 of the present invention, a model to be verified after cross-validation is used in establishing the model to be verified. DETAILED DESCRIPTION

[0033] The present invention provides a method for establishing a multi-substance online analysis model, comprising the following steps:

[0034] (1) Sample collection: Samples are collected at sampling points to obtain samples; the sampling points are different time points in the same reaction stage during the production process;

[0035] (2) Sample analysis: performing gas chromatography analysis and near infrared scanning on each sample in step (1) to obtain measured values ​​of the contents of multiple target compounds in each sample and a near infrared spectrum of each sample;

[0036] (3) Establishing the model to be verified: Correlating the measured values ​​of the content of each target compound in some samples in step (2) with the near-infrared spectra of the corresponding samples and optimizing the model to obtain multiple models to be verified;

[0037] (4) Model verification: importing the near infrared spectra of the remaining samples in step (2) into the multiple models to be verified obtained in step (3) to predict the content of the corresponding target compounds, and obtain predicted values, and then comparing the obtained predicted values ​​with the measured values ​​of the content of the multiple target compounds in the corresponding samples obtained in step (2); when the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, a multi-substance online analysis model is obtained; when the samples whose relative deviations between the predicted values ​​and the measured values ​​do not meet the detection accuracy requirements account for more than 10% of the remaining sample amount, changing the association method and / or model optimization method in step (3) until the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, and a multi-substance online analysis model is obtained.

[0038] The model building method of the present invention comprises sample collection; the sample collection comprises: collecting at sampling points to obtain samples; the sampling points are different time points in the same reaction stage during the production process.

[0039] As an embodiment of the present invention, the interval time between adjacent sampling points in the same reaction stage can be 2 to 3 hours. The present invention sets the sampling points as different time points in the same reaction stage in the production process and limits the interval time between adjacent sampling points in the same reaction stage to the above range, which can cover the entire process of the same reaction stage in the production process as much as possible.

[0040] In the present invention, the number of samples is preferably ≥ 70, more preferably 70 to 100, and even more preferably 70 to 80. The present invention limits the number of samples to the above range so that a better model can be established.

[0041] The present invention performs sample analysis after sample collection.

[0042] After obtaining the samples, the present invention performs gas chromatography analysis and near infrared scanning on each sample to obtain the measured values ​​of the contents of multiple target compounds in each sample and the near infrared spectrum of each sample.

[0043] As an embodiment of the present invention, the gas chromatography analysis can be carried out in a gas chromatograph; the chromatographic column model used in the gas chromatograph can be a DB series chromatographic column; the size of the chromatographic column can be 30m×0.25mm×0.25μm; the column temperature of the chromatographic column when the gas chromatograph is in use can be 100°C; the vaporization chamber temperature of the high performance liquid chromatograph can be 200°C; the detection device used in the gas chromatograph can be a FID detector; the temperature of the detector can be 200°C; the carrier gas flow rate of the gas chromatograph can be 3mL / min; the split ratio of the gas chromatograph can be 1:42; the injection volume of the gas chromatograph can be 1μL; the ratio of the sample mass to the volume of methanol during the gas chromatography analysis can be 0.05g:10ml.

[0044] As an embodiment of the present invention, the liquid tested by the near-infrared scanning test can be the liquid in the near-infrared liquid transmission flow cell at each reaction stage; the main material of the near-infrared liquid transmission flow cell can be 316L stainless steel; the optical path in the near-infrared scanning can be 2 mm; in the near-infrared scanning, after the liquid to be tested flows through the near-infrared liquid transmission flow cell, the number of scans can be 32 times; the resolution of the near-infrared scanning can be 8.0 cm -1 The scanning spectrum range of the near infrared scanning can be 12800~4000cm -1 .

[0045] The present invention establishes a model to be verified after sample analysis.

[0046] After obtaining the measured values ​​of the contents of multiple target compounds in each sample and the near-infrared spectrum of each sample, the present invention associates and optimizes the model by respectively correlating the measured values ​​of the contents of each target compound in the partial samples with the near-infrared spectrum of the corresponding samples to obtain multiple models to be verified.

[0047] In the present invention, the software used for model building is preferably OPUS software. As a specific embodiment of the present invention, the OPUS software is selected from Bruker.

[0048] In the present invention, the amount of the partial samples is preferably 85-90% of the total amount of samples, more preferably 85-88%, and further preferably 85.7%. The present invention uses partial sample data for modeling, and a portion of the data can be retained for model verification.

[0049] In the present invention, the establishment of the model to be verified preferably includes the following steps:

[0050] 1) Input the measured value of each target compound in some samples and the near infrared spectrum of the corresponding samples into the software;

[0051] 2) Preprocessing each near-infrared spectrum input in step 1) and then selecting a band interface to obtain a processed near-infrared spectrum;

[0052] 3) correlating the measured value of each target compound input in step 1) with the processed near infrared spectrum obtained in step 2) to obtain a plurality of initial models;

[0053] 4) Optimizing the multiple initial models obtained in step 3) respectively to obtain multiple models to be verified.

[0054] The present invention preferably inputs the measured values ​​of multiple target compounds of some samples and the near infrared spectra of the corresponding samples into the software.

[0055] As an embodiment of the present invention, the number of the partial samples may be 70.

[0056] The present invention preferably pre-processes each input near-infrared spectrum, and then selects a band interface to obtain a processed near-infrared spectrum.

[0057] In the present invention, the preprocessing preferably includes one or more of first-order derivative processing, second-order derivative processing and vector normalization, and more preferably second-order derivative processing. The present invention defines the preprocessing as the above-mentioned method can effectively eliminate the systematic error of the near-infrared spectrum.

[0058] In the present invention, the band interface is preferably selected based on the band interface where the characteristic peak of the corresponding target compound is located. The present invention selects the band interface where the characteristic peak of the corresponding target compound is located to fully extract effective information in the spectrum.

[0059] After obtaining the processed near-infrared spectrum, the present invention preferably associates the input measured values ​​of multiple target compounds with the processed near-infrared spectrum to obtain multiple initial models.

[0060] In the present invention, the association method is preferably a partial least squares regression quantitative method, a multivariate linear regression or a principal component regression. The present invention adopts the above association method to establish a more accurate model.

[0061] After obtaining the initial model, the present invention preferably optimizes the multiple initial models separately to obtain multiple models to be verified.

[0062] As an implementation of the present invention, the optimization method may be a cross-validation method. The present invention can make the established model more accurate through optimization processing.

[0063] As an embodiment of the present invention, the RMSECV value of the model to be verified can be 0.292. In the present invention, the smaller the RMSECV value of the model, the higher the prediction accuracy of the model to be verified.

[0064] The present invention performs model verification after the model to be verified is established.

[0065] After obtaining multiple models to be verified, the present invention imports the near-infrared spectra of the remaining samples into the multiple models to be verified to predict the corresponding target compound contents, obtain predicted values, and then compare the obtained predicted values ​​with the actual measured values ​​of the contents of multiple target compounds in the corresponding samples obtained during sample analysis. When the samples whose relative deviations from the predicted values ​​and the actual measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample volume, a multi-substance online analysis model is obtained.

[0066] In the present invention, when the number of samples whose relative deviations between the predicted values ​​and the measured values ​​do not meet the detection accuracy requirements exceeds 10% of the remaining sample amount, the association method and / or model optimization method in establishing the model to be verified is changed until the number of samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements accounts for more than 90% of the remaining sample amount, thereby obtaining a multi-substance online analysis model.

[0067] The present invention associates a near-infrared scanning spectrum with the contents of multiple target compounds of each sample to establish multiple models, and verifies the multiple models, so that the relative deviations between the predicted values ​​and the measured values ​​of the contents of multiple target compounds in the near-infrared spectrum by the established multiple models meet the demand for detection accuracy in production. By utilizing the model, near-infrared online detection can be adopted without manual sample collection and delivery, with a high safety factor, and the corresponding target compound contents can be quickly detected. The analysis results are fast and timely, and simultaneous online real-time monitoring of the contents of multiple target compounds can be realized.

[0068] The present invention also provides the application of the model obtained by the establishment method described in the above technical solution in the online detection of multiple substances in chemical production.

[0069] The present invention also provides a method for simultaneously online determination of multiple substances in a 1,4-dimethylnaphthalene production process, comprising:

[0070] Performing near infrared detection on the reaction liquid of the reaction stage to be detected in the production process of 1,4-dimethylnaphthalene to obtain a near infrared spectrum;

[0071] Substituting the near infrared spectrum into a multi-substance online analysis model of a corresponding reaction stage to obtain the contents of multiple target compounds in the reaction solution of the corresponding reaction stage;

[0072] The multi-substance online analysis model of the corresponding reaction stage is established by the establishment method described in the above technical solution, and the sample used when establishing the model is the reaction liquid of the corresponding reaction stage in the 1,4-dimethylnaphthalene production process.

[0073] In the present invention, the reaction stages in the 1,4-dimethylnaphthalene production process include an ethylbenzene recovery stage, a 5-phenyl-2-hexene distillation stage, a 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene distillation stage and a 1,4-dimethylnaphthalene product purification stage.

[0074] In the present invention, when the multi-substance online analysis model of the ethylbenzene recovery stage is established, the sampling points are different reaction times in the ethylbenzene recovery stage, and the target compounds in the samples are ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene.

[0075] As an embodiment of the present invention, when the model of the multi-substance online analysis model of the ethylbenzene recovery stage is established, the time of the first sampling point of the sampling points can be 2 hours after the start of ethylbenzene distillation in the ethylbenzene recovery stage, and the remaining sampling points can be sequentially spaced 2 hours from the previous sampling point.

[0076] As an embodiment of the present invention, in the ethylbenzene recovery stage, when the target compound in the sample is ethylbenzene, the relative deviation between the predicted value and the measured value of the ethylbenzene content by the established model is less than 3%; when the target compound in the sample is 5-phenyl-2-hexene, the relative deviation between the predicted value and the measured value of the 5-phenyl-2-hexene content by the established model is less than 3%; when the target compound in the sample is 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, the relative deviation between the predicted value and the measured value of the 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content by the established model is less than 3%.

[0077] In the present invention, when establishing the multi-substance online analysis model of the 5-phenyl-2-hexene distillation stage, the sampling points are at different times during the 5-phenyl-2-hexene distillation process, and the target compound in the sample is ethylbenzene or 5-phenyl-2-hexene.

[0078] As an embodiment of the present invention, when the model of the multi-substance online analysis model of the 5-phenyl-2-hexene distillation stage is established, the time of the first sampling point of the sampling points can be 2 hours after the start of distillation of 5-phenyl-2-hexene in the 5-phenyl-2-hexene distillation stage, and the remaining sampling points can be sequentially spaced 2 hours from the previous sampling point.

[0079] As an embodiment of the present invention, in the 5-phenyl-2-hexene distillation stage, when the target compound in the sample is ethylbenzene, the relative deviation between the predicted value and the measured value of the ethylbenzene content by the established model is less than 3%; when the target compound in the sample is 5-phenyl-2-hexene, the relative deviation between the predicted value and the measured value of the 5-phenyl-2-hexene content by the established model is less than 3%.

[0080] In the present invention, when the multi-substance online analysis model of the 1,4-dimethyl-1,2,3,4-tetralin distillation stage is established, the sampling points are different times in the 1,4-dimethyl-1,2,3,4-tetralin distillation process, and the target compound in the sample is ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetralin.

[0081] As an embodiment of the present invention, when the model of the multi-substance online analysis model of the 1,4-dimethyl-1,2,3,4-tetralin distillation stage is established, the time of the first sampling point of the sampling points can be 2 hours after the distillation of 1,4-dimethyl-1,2,3,4-tetralin starts in the distillation stage of 1,4-dimethyl-1,2,3,4-tetralin, and the remaining sampling points can be sequentially spaced 2 hours from the previous sampling point.

[0082] As an embodiment of the present invention, in the distillation stage of 1,4-dimethyl-1,2,3,4-tetralin, when the target compound in the sample is ethylbenzene, the relative deviation between the predicted value and the measured value of the ethylbenzene content by the established model is less than 3%; when the target compound in the sample is 5-phenyl-2-hexene, the relative deviation between the predicted value and the measured value of the 5-phenyl-2-hexene content by the established model is less than 3%; when the target compound in the sample is 1,4-dimethyl-1,2,3,4-tetralin, the relative deviation between the predicted value and the measured value of the 1,4-dimethyl-1,2,3,4-tetralin content by the established model is less than 3%.

[0083] In the present invention, when the multi-substance online analysis model of the 1,4-dimethylnaphthalene product purification stage is established, the sampling points are different times in the 1,4-dimethylnaphthalene product purification process, and the target compound in the sample is 1,4-dimethylnaphthalene or impurities.

[0084] In the present invention, the impurities include 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene or 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene.

[0085] As an embodiment of the present invention, the impurities may also include various impurities other than 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene and 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene.

[0086] In a specific embodiment of the present invention, the plurality of impurities other than 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene and 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene may be four kinds, respectively named as impurity A, impurity B, impurity C and impurity D.

[0087] As an embodiment of the present invention, when the model of the multi-substance online analysis model of the 1,4-dimethylnaphthalene product purification stage is established, the time of the first sampling point of the sampling points can be 2 hours after the start of distillation of 1,4-dimethylnaphthalene in the 1,4-dimethylnaphthalene product purification stage, and the remaining sampling points can be sequentially spaced 2 hours from the previous sampling point.

[0088] As an embodiment of the present invention, in the purification stage of 1,4-dimethylnaphthalene product, when the target compound in the sample is 1,4-dimethylnaphthalene, the relative deviation between the predicted value and the measured value of the 1,4-dimethylnaphthalene content by the established model is less than 0.5%; when the target compound in the sample is 1,3-dimethylnaphthalene, the relative deviation between the predicted value and the measured value of the 1,3-dimethylnaphthalene content by the established model is less than 3%; when the target compound in the sample is 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, the relative deviation between the predicted value and the measured value of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene by the established model is less than 1.5%. The relative deviation between the predicted value and the measured value of the content is less than 3%; when the target compound in the sample is 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, the relative deviation between the predicted value and the measured value of the content of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene by the established model is less than 3%; when the target compound in the sample is 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene and multiple impurities other than 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, the relative deviation between the predicted value and the measured value of the content of each impurity by each established model is independently less than 15%.

[0089] In a specific embodiment of the present invention, when the target compound in the sample is impurity A, impurity B, impurity C or impurity D, the relative deviation between the predicted value and the measured value of each impurity content of each established model is independently <15%.

[0090] The present invention preferably establishes an online analysis model for each target compound in the sample separately. The present invention establishes a model for each target compound in the sample separately, which can make the model establishment more accurate and is conducive to more accurate acquisition of the content of each target compound.

[0091] The technical solutions in the present invention will be described clearly and completely below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0092] Example 1

[0093] A method for establishing a multi-substance online analysis model is:

[0094] (1) Sample collection: Samples are collected at sampling points; the sampling points are the 1,4-dimethylnaphthalene product purification stage in the 1,4-dimethylnaphthalene production process; the first sampling point is taken 2 hours after the distillation of 1,4-dimethylnaphthalene begins in the 1,4-dimethylnaphthalene product purification stage, and the remaining sampling points are taken at intervals of 2 hours from the previous sampling point. The number of samples taken from the sampling points is 70;

[0095] (2) Sample analysis: The 70 samples in step (1) are subjected to gas chromatography analysis and near infrared scanning respectively to obtain the measured values ​​and near infrared spectra of the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D of the 70 samples; the gas chromatography analysis is carried out in a gas chromatograph; the chromatographic column model used in the gas chromatograph is a DB series chromatographic column; the size of the chromatographic column is 30m×0.25mm×0.25μm; the column temperature of the chromatographic column when the gas chromatograph is in use is 100°C; the vaporization chamber temperature of the high performance liquid chromatograph is 200°C; the gas chromatograph The detection device used in the instrument is a FID detector; the temperature of the detector is 200°C; the carrier gas flow rate of the gas chromatograph is 3mL / min; the split ratio of the gas chromatograph is 1:42; the injection volume of the gas chromatograph is 1μL; the ratio of the sample mass to the volume of methanol during the gas chromatography analysis is 0.05g:10ml; the liquid tested by the near-infrared scanning is the liquid in the near-infrared liquid transmission flow cell in the purification stage of 1,4-dimethylnaphthalene product; the main material of the near-infrared liquid transmission flow cell is 316L stainless steel; the optical path in the near-infrared scanning is 2mm; in the near-infrared scanning, after the liquid to be tested flows through the near-infrared liquid transmission flow cell, the number of scans is 32 times; the resolution of the near-infrared scanning is 8.0cm -1 The near infrared scanning spectrum range is 12800~4000cm -1 .

[0096] (3) Establishment of the model to be verified: The measured values ​​of the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D of the 60 samples in the step (2) are respectively associated with the corresponding near-infrared spectra and the model is optimized to obtain eight models to be verified (1,4-dimethylnaphthalene to be verified model, 1,3-dimethylnaphthalene to be verified model, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene to be verified model, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene to be verified model, impurity A to be verified model, impurity B to be verified model, impurity C to be verified model and impurity D to be verified model);

[0097] (4) Model verification: The near infrared spectra of the remaining 10 samples in step (2) are introduced into the eight to-be-verified models obtained in step (3) to predict the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D to obtain predicted values, and the eight predicted values ​​obtained for each sample are compared with the corresponding measured values ​​of the sample. , when the relative deviation between the predicted value and the measured value of 1,4-DMN content is less than 0.5%, the samples account for more than 90% of the remaining samples; the relative deviation between the predicted value and the measured value of 1,3-DMN content is less than 3%, the samples account for more than 90% of the remaining samples; the relative deviation between the predicted value and the measured value of 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content is less than 3%, the samples account for more than 90% of the remaining samples; The samples whose relative deviation between the predicted value and the measured value is less than 3% account for more than 90% of the remaining samples, the samples whose relative deviation between the predicted value and the measured value of the impurity A content is less than 15% account for more than 90% of the remaining samples, the samples whose relative deviation between the predicted value and the measured value of the impurity B content is less than 15% account for more than 90% of the remaining samples, the samples whose relative deviation between the predicted value and the measured value of the impurity C content is less than 15% account for more than 90% of the remaining samples, and the samples whose relative deviation between the predicted value and the measured value of the impurity D content is less than 15% account for more than 90% of the remaining samples, and the multi-substance online analysis model (1,4-dimethylnaphthalene online analysis model, 1,3-dimethylnaphthalene online analysis model, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene online analysis model, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene online analysis model, impurity A online analysis model, impurity B online analysis model, impurity C online analysis model and impurity D online analysis model) are obtained;

[0098] The model to be verified (1,4-dimethylnaphthalene model to be verified) is established as:

[0099] 1) Input the measured values ​​of 1,4-DMN content and the corresponding near-infrared spectra of 60 samples into OPUS software;

[0100] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of 1,4-DMN is located to obtain a processed near-infrared spectrum;

[0101] 3) correlating the measured value of the 1,4-DMN content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0102] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0103] The model to be verified (1,3-dimethylnaphthalene model to be verified) is established as:

[0104] 1) Input the measured values ​​of 1,3-DMN content and the corresponding near-infrared spectra of 60 samples into OPUS software;

[0105] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of 1,3-DMN is located to obtain a processed near-infrared spectrum;

[0106] 3) correlating the measured value of the 1,3-DMN content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0107] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0108] The model to be verified (5,8-dimethyl-1,2,3,4-tetrahydronaphthalene model to be verified) is established as:

[0109] 1) Input the measured values ​​of 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content and the corresponding near-infrared spectra of 60 samples into OPUS software;

[0110] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of 5,8-dimethyl-1,2,3,4-tetralin is located to obtain a processed near-infrared spectrum;

[0111] 3) correlating the measured value of the 5,8-dimethyl-1,2,3,4-tetralin content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0112] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0113] The model to be verified (1,4-dimethyl-1,2,3,4-tetrahydronaphthalene model to be verified) is established as:

[0114] 1) Input the measured values ​​of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content and the corresponding near-infrared spectra of 60 samples into OPUS software;

[0115] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of 1,4-dimethyl-1,2,3,4-tetralin is located to obtain a processed near-infrared spectrum;

[0116] 3) correlating the measured value of the 1,4-dimethyl-1,2,3,4-tetralin content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0117] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0118] The model to be verified (impurity A model to be verified) is established as:

[0119] 1) Input the measured values ​​of impurity A content and the corresponding near infrared spectra of 60 samples into OPUS software;

[0120] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of impurity A is located to obtain a processed near-infrared spectrum;

[0121] 3) correlating the measured value of the impurity A content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0122] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0123] The model to be verified (impurity B model to be verified) is established as:

[0124] 1) Input the measured values ​​of impurity B content and the corresponding near infrared spectra of 60 samples into OPUS software;

[0125] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of impurity B is located to obtain a processed near-infrared spectrum;

[0126] 3) correlating the measured value of the impurity B content input in step 1) and the processed near infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0127] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0128] The model to be verified (impurity C model to be verified) is established as:

[0129] 1) Input the measured values ​​of impurity C content and the corresponding near-infrared spectra of 60 samples into OPUS software;

[0130] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of impurity C is located to obtain a processed near-infrared spectrum;

[0131] 3) correlating the measured value of the impurity C content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0132] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method;

[0133] The model to be verified (impurity D model to be verified) is established as:

[0134] 1) Input the measured values ​​of impurity D content and the corresponding near infrared spectra of 60 samples into OPUS software;

[0135] 2) performing second-order derivative processing on each near-infrared spectrum input in step 1), and then selecting the band interface where the characteristic peak of impurity D is located to obtain a processed near-infrared spectrum;

[0136] 3) correlating the measured value of the impurity D content input in step 1) and the processed near-infrared spectrum obtained in step 2) using a partial least squares regression quantitative method to obtain an initial model;

[0137] 4) Optimizing the initial model obtained in step 3) to obtain a model to be verified; the optimization method is a cross-validation method.

[0138] Application Example 1

[0139] A method for simultaneous online determination of multiple substances in the production process of 1,4-dimethylnaphthalene:

[0140] The reaction liquid in the 1,4-dimethylnaphthalene product purification stage in the 1,4-dimethylnaphthalene production process is subjected to near infrared detection to obtain a near infrared spectrum;

[0141] The near infrared spectrum is substituted into the 1,4-dimethylnaphthalene online analysis model, 1,3-dimethylnaphthalene online analysis model, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene online analysis model, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene online analysis model, impurity A online analysis model, impurity B online analysis model, impurity C online analysis model and impurity D online analysis model obtained in Example 1 to obtain the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D.

[0142] The near infrared spectra of the 60 samples used to establish the model to be verified in Application Example 1 are as follows: Figure 1 shown.

[0143] Application Example 1: In establishing the model to be verified, the model to be verified after cross-validation (1,4-dimethylnaphthalene model to be verified) is used. Figure 2 As shown, from Figure 2 It can be seen that the RMSECV value of the established model to be verified (1,4-dimethylnaphthalene model to be verified) is 0.292.

[0144] Verification of the eight models established in Application Example 1. The verification process is as follows:

[0145] The near-infrared spectra of the remaining 10 samples after the establishment of the model to be verified are input into the model to be verified in Application Example 1 (1,4-DMN model to be verified), and the 1,4-dimethylnaphthalene content is predicted for the near-infrared spectra of the 10 samples. The predicted value of the 1,4-dimethylnaphthalene content in the reaction solution is compared with the data of the actual value of the 1,4-dimethylnaphthalene content (the 1,4-dimethylnaphthalene content data of the 10 samples obtained by gas chromatograph analysis). The comparison results are shown in Table 1.

[0146] Table 1 Comparison of the predicted values ​​of the 1,4-dimethylnaphthalene content in the reaction solution by the model to be verified and the measured values ​​of the 1,4-dimethylnaphthalene content

[0147] Serial number batch number Near infrared prediction value Gas phase measurement value Relative deviation 1 240601-1 88.74 88.68 0.03％ 2 240602-2 93.21 93.15 0.03％ 3 240603-4 92.19 92.1 0.05％ 4 240604-8 95.59 95.53 0.03％ 5 240605-9 89.14 89.22 -0.04％ 6 240606-6 90.64 90.72 -0.04％ 7 240607-5 96.35 96.27 0.04％ 8 240608-10 87.44 87.49 -0.03％ 9 240609-11 91.87 91.76 0.06％ 10 240610-12 93.43 93.21 0.12％

[0148] As can be seen from Table 1, the model in Application Example 1 has high prediction accuracy for the samples with a relative deviation of less than 0.5% between the predicted value and the measured value of the 1,4-dimethylnaphthalene content in the reaction liquid in the purification stage of the 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the 1,4-dimethylnaphthalene content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0149] The near-infrared spectra of the remaining 10 samples after the model to be verified is established are input into the model to be verified in Application Example 1 (1,3-dimethylnaphthalene model to be verified), and the 1,3-dimethylnaphthalene content is predicted for the near-infrared spectra of the 10 samples. The predicted value of the 1,3-dimethylnaphthalene content in the reaction solution and the data of the actual measured value of the 1,3-dimethylnaphthalene content (the 1,3-dimethylnaphthalene content data of the 10 samples obtained by gas chromatograph analysis) are compared. The comparison results are shown in Table 2.

[0150] Table 2 Comparison of the predicted values ​​of the 1,3-dimethylnaphthalene content in the reaction solution by the model to be verified and the measured values ​​of the 1,3-dimethylnaphthalene content

[0151] Serial number batch number Near infrared prediction value Gas phase measurement value Relative deviation 1 240601-1 6.31 6.35 -0.32％ 2 240602-2 0.93 0.91 1.09％ 3 240603-4 0.87 0.86 0.58％ 4 240604-8 0.41 0.4 1.23％ 5 240605-9 4.12 4.09 0.37％ 6 240606-6 1.37 1.4 -1.08％ 7 240607-5 0.61 0.63 -1.61％ 8 240608-10 5.92 5.97 -0.42％ 9 240609-11 1.19 1.21 -0.83％ 10 240610-12 0.87 0.84 1.75％

[0152] As can be seen from Table 2, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 3% between the predicted value and the measured value of the 1,3-dimethylnaphthalene content in the reaction liquid in the purification stage of the 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the 1,3-dimethylnaphthalene content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0153] The near-infrared spectra of the remaining 10 samples after the model to be verified is established are input into the model to be verified in Application Example 1 (5,8-dimethyl-1,2,3,4-tetrahydronaphthalene model to be verified), and the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content is predicted for the near-infrared spectra of the 10 samples. The predicted value of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content in the reaction solution and the data of the actual measured value of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content (5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content data of 10 samples obtained by gas chromatograph analysis) are compared. The comparison results are shown in Table 3.

[0154] Table 3 Comparison of the predicted values ​​of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content in the reaction solution by the model to be verified and the measured values ​​of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content

[0155]

[0156]

[0157] As can be seen from Table 3, the model in Application Example 1 has high prediction accuracy for the samples with a relative deviation of less than 3% between the predicted value and the measured value of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content in the reaction liquid in the purification stage of 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0158] The near-infrared spectra of the remaining 10 samples after the model to be verified is established are input into the model to be verified in Application Example 1 (1,4-dimethyl-1,2,3,4-tetrahydronaphthalene model to be verified), and the content of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene is predicted for the near-infrared spectra of the 10 samples. The predicted value of the content of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene in the reaction solution and the data of the measured value of the content of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene (the content data of 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene of 10 samples obtained by gas chromatograph analysis) are compared. The comparison results are shown in Table 4.

[0159] Table 4 Comparison of the predicted values ​​of the 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content in the reaction solution by the model to be verified and the measured values ​​of the 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content

[0160]

[0161]

[0162] As can be seen from Table 4, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 3% between the predicted value and the measured value of the 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content in the reaction liquid in the purification stage of the 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0163] The near-infrared spectra of the remaining 10 samples after the establishment of the model to be verified are input into the model to be verified in Application Example 1 (impurity A model to be verified), and the impurity A content of the near-infrared spectra of the 10 samples is predicted. The predicted value of the impurity A content in the reaction liquid is compared with the data of the actual measured value of the impurity A content (the impurity A content data of the 10 samples obtained by gas chromatograph analysis). The comparison results are shown in Table 5.

[0164] Table 5 Data comparison of the predicted value of the impurity A content in the reaction solution by the model to be verified and the measured value of the impurity A content

[0165] Serial number batch number Near infrared prediction value Gas phase measurement value Relative deviation 1 240601-1 0.07 0.08 -6.67％ 2 240602-2 0.07 0.09 -12.50％ 3 240603-4 0.09 0.09 0.00％ 4 240604-8 0.06 0.07 -7.69％ 5 240605-9 0.07 0.09 -12.50％ 6 240606-6 0.07 0.08 -6.67％ 7 240607-5 0.07 0.06 7.69％ 8 240608-10 0.08 0.06 14.29％ 9 240609-11 0.09 0.08 5.88％ 10 240610-12 0.11 0.12 -4.35％

[0166] As can be seen from Table 5, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 15% between the predicted value and the measured value of the impurity A content in the reaction liquid in the purification stage of 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the impurity A content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0167] The near-infrared spectra of the remaining 10 samples after the establishment of the model to be verified are input into the model to be verified in Application Example 1 (impurity B model to be verified), and the impurity B content of the near-infrared spectra of the 10 samples is predicted. The predicted value of the impurity B content in the reaction liquid is compared with the data of the actual measured value of the impurity B content (the impurity B content data of the 10 samples obtained by gas chromatograph analysis). The comparison results are shown in Table 6.

[0168] Table 6 Data comparison of the predicted value of the impurity B content in the reaction solution by the model to be verified and the measured value of the impurity B content

[0169] Serial number batch number Near infrared prediction value Gas phase measurement value Relative deviation 1 240601-1 0.12 0.11 4.35％ 2 240602-2 0.14 0.12 7.69％ 3 240603-4 0.15 0.14 3.45％ 4 240604-8 0.16 0.14 6.67％ 5 240605-9 0.15 0.17 -6.25％ 6 240606-6 0.22 0.23 -2.22％ 7 240607-5 0.15 0.17 -6.25％ 8 240608-10 0.12 0.14 -7.69％ 9 240609-11 0.13 0.14 -3.70％ 10 240610-12 0.12 0.12 0.00％

[0170] As can be seen from Table 6, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 15% between the predicted value and the measured value of the impurity B content in the reaction liquid in the purification stage of 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the impurity B content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0171] The near-infrared spectra of the remaining 10 samples after the establishment of the model to be verified are input into the model to be verified in Application Example 1 (impurity C model to be verified), and the impurity C content of the near-infrared spectra of the 10 samples is predicted. The predicted value of the impurity C content in the reaction liquid is compared with the data of the actual value of the impurity C content (the impurity C content data of the 10 samples obtained by gas chromatograph analysis). The comparison results are shown in Table 7.

[0172] Table 7 Data comparison of the predicted value of the impurity C content in the reaction solution by the model to be verified and the measured value of the impurity C content

[0173] Serial number batch number Near infrared prediction value Gas phase measurement value Relative deviation 1 240601-1 0.09 0.09 0.00％ 2 240602-2 0.13 0.14 -3.70％ 3 240603-4 0.13 0.14 -3.70％ 4 240604-8 0.15 0.17 -6.25％ 5 240605-9 0.09 0.09 0.00％ 6 240606-6 0.07 0.06 7.69％ 7 240607-5 0.09 0.1 -5.26％ 8 240608-10 0.09 0.11 -10.00％ 9 240609-11 0.13 0.12 4.00％ 10 240610-12 0.1 0.12 -9.09％

[0174] As can be seen from Table 7, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 15% between the predicted value and the measured value of the impurity C content in the reaction liquid in the purification stage of 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the impurity C content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0175] The near-infrared spectra of the remaining 10 samples after the establishment of the model to be verified are input into the model to be verified in Application Example 1 (impurity D model to be verified), and the impurity D content of the near-infrared spectra of the 10 samples is predicted. The predicted value of the impurity D content in the reaction liquid is compared with the data of the actual measured value of the impurity D content (the impurity D content data of the 10 samples obtained by gas chromatograph analysis). The comparison results are shown in Table 6.

[0176] Table 8 Data comparison of the predicted value of the impurity D content in the reaction solution by the model to be verified and the measured value of the impurity D content

[0177]

[0178]

[0179] As can be seen from Table 8, the model in Application Example 1 has high prediction accuracy for samples with a relative deviation of less than 15% between the predicted value and the measured value of the impurity D content in the reaction liquid in the purification stage of 1,4-dimethylnaphthalene product, accounting for 100% of the remaining sample volume, and can be used for the analysis of the impurity D content in the 1,4-dimethylnaphthalene production process by near-infrared spectroscopy.

[0180] Test Case

[0181] Sample stability test

[0182] The reaction liquid from the purification stage of the 1,4-dimethylnaphthalene product retrieved from the sample collection was subjected to near-infrared analysis every 10 minutes for 1 hour to measure the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D. A total of 6 measurements were made, and the relative standard deviation data obtained are shown in Table 9.

[0183] Table 9 Near infrared measurement data of the reaction solution in the purification stage of 1,4-dimethylnaphthalene product

[0184]

[0185]

[0186] As can be seen from Table 9, the relative standard deviation of 1,4-dimethylnaphthalene is 0.02%, and the relative standard deviation of impurities is less than 10%, indicating that the storage solution has good stability within 1 hour, and the test is completed within 1 hour after the sample is taken without affecting the result.

[0187] The model (1,4-dimethylnaphthalene online analysis model) obtained by the establishment method provided by the present invention has an RMSECV value of up to 0.292. The multi-substance online analysis model is used to determine the contents of 1,4-dimethylnaphthalene, 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene, 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene, impurity A, impurity B, impurity C and impurity D in the 1,4-dimethylnaphthalene production process. The samples whose predicted 1,4-dimethylnaphthalene content value has a relative deviation of less than 0.5% from the measured value account for 100% of the remaining sample amount, and the samples whose predicted 1,3-dimethylnaphthalene content value has a relative deviation of less than 3% from the measured value account for 100% of the remaining sample amount. The samples whose predicted 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene content value and the measured value have a relative deviation of less than 3% account for 100% of the remaining samples, the samples whose predicted 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene content value and the measured value have a relative deviation of less than 3% account for 100% of the remaining samples, and the samples whose predicted content value of each other impurity (impurity A, impurity B, impurity C and impurity D) and the measured value have a relative deviation of less than 15% account for 100% of the remaining samples, which meets the production demand for detection accuracy and can simultaneously and accurately determine the contents of multiple substances in the 1,4-dimethylnaphthalene production process.

[0188] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for establishing a multi-substance online analysis model comprises the following steps: (1) Sample collection: Samples are collected at sampling points to obtain samples; the sampling points are different time points in the same reaction stage during the production process; (2) Sample analysis: performing gas chromatography analysis and near infrared scanning on each sample in step (1) to obtain measured values ​​of the contents of multiple target compounds in each sample and a near infrared spectrum of each sample; (3) Establishing the model to be verified: Correlating the measured values ​​of the content of each target compound in some samples in step (2) with the near-infrared spectra of the corresponding samples and optimizing the model to obtain multiple models to be verified; (4) Model verification: importing the near infrared spectra of the remaining samples in step (2) into the multiple models to be verified obtained in step (3) to predict the content of the corresponding target compounds, and obtain predicted values, and then comparing the obtained predicted values ​​with the measured values ​​of the content of the multiple target compounds in the corresponding samples obtained in step (2); when the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, a multi-substance online analysis model is obtained; when the samples whose relative deviations between the predicted values ​​and the measured values ​​do not meet the detection accuracy requirements account for more than 10% of the remaining sample amount, changing the association method and / or model optimization method in step (3) until the samples whose relative deviations between the predicted values ​​and the measured values ​​meet the detection accuracy requirements account for more than 90% of the remaining sample amount, and a multi-substance online analysis model is obtained.

2. The model building method according to claim 1, characterized in that: The number of samples in step (1) is ≥70.

3. The establishment method according to claim 1, characterized in that: The quantity of the partial samples in step (3) is 85-90% of the total quantity of the samples.

4. The establishment method according to claim 1 or 3, characterized in that: The step (3) of establishing the model to be verified includes the following steps: 1) Input the measured values ​​of multiple target compounds of some samples and the near infrared spectra of the corresponding samples into the software; 2) Preprocessing each near-infrared spectrum input in step 1) and then selecting a band interface to obtain a processed near-infrared spectrum; 3) correlating the measured values ​​of the multiple target compounds input in step 1) with the processed near infrared spectra obtained in step 2) to obtain multiple initial models; 4) Optimizing the multiple initial models obtained in step 3) respectively to obtain multiple models to be verified.

5. The establishment method according to claim 4, characterized in that: The preprocessing in step 2) includes one or more of first-order derivative processing, second-order derivative processing and vector normalization.

6. The establishment method according to claim 4, characterized in that: The associated methods in step 3) include partial least squares regression quantitative method, multiple linear regression or principal component regression.

7. Application of the model obtained by the establishment method according to any one of claims 1 to 6 in online detection of multiple substances in chemical production.

8. A method for simultaneous online determination of multiple substances in a 1,4-dimethylnaphthalene production process, characterized in that: include: Performing near infrared detection on the reaction liquid of the reaction stage to be detected in the production process of 1,4-dimethylnaphthalene to obtain a near infrared spectrum; Substituting the near infrared spectrum into a multi-substance online analysis model of a corresponding reaction stage to obtain the contents of multiple target compounds in the reaction solution of the corresponding reaction stage; The multi-substance online analysis model of the corresponding reaction stage is established by the establishment method described in any one of claims 1 to 6, and the sample used when establishing the model is the reaction liquid of the corresponding reaction stage in the 1,4-dimethylnaphthalene production process.

9. The online determination method according to claim 8, characterized in that: The reaction stages in the 1,4-dimethylnaphthalene production process include an ethylbenzene recovery stage, a 5-phenyl-2-hexene distillation stage, a 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene distillation stage and a 1,4-dimethylnaphthalene product purification stage.

10. The method according to claim 9, characterized in that When the multi-substance online analysis model of the ethylbenzene recovery stage is established, the sampling points are different reaction times in the ethylbenzene recovery stage, and the target compounds in the samples are ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetralin; When establishing the multi-substance online analysis model of the 5-phenyl-2-hexene distillation stage, the sampling points are at different times during the 5-phenyl-2-hexene distillation process, and the target compound in the sample is ethylbenzene or 5-phenyl-2-hexene; When the multi-substance online analysis model of the 1,4-dimethyl-1,2,3,4-tetralin distillation stage is established, the sampling points are different times during the 1,4-dimethyl-1,2,3,4-tetralin distillation process, and the target compound in the sample is ethylbenzene, 5-phenyl-2-hexene or 1,4-dimethyl-1,2,3,4-tetralin; When the multi-substance online analysis model of the 1,4-dimethylnaphthalene product purification stage is established, the sampling points are at different times during the purification process of the 1,4-dimethylnaphthalene product, and the target compound in the sample is 1,4-dimethylnaphthalene or impurities; the impurities include 1,3-dimethylnaphthalene, 5,8-dimethyl-1,2,3,4-tetrahydronaphthalene or 1,4-dimethyl-1,2,3,4-tetrahydronaphthalene.

Images