Online monitoring and distinguishing method for material leakage of catalytic cracking water cooler

Through online purge and trap-gas chromatography analyzer and partial least squares discriminant analysis technology, the problem of difficult-to-discover material leakage in catalytic cracked water cooler is solved, and the online monitoring and traceability of catalytic cracked water cooler materials is realized, and monitoring is improved. The sensitivity and automation level of the problem of micro leakage detection in the prior art.

CN119023824APending Publication Date: 2024-11-26CHINA PETROLEUM & CHEMICAL CORP +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310601728.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-26
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing water-cooled material leakage monitoring technology is difficult to detect trace leakage in a timely manner, especially the leakage of liquefied petroleum gas, gasoline, light diesel, and lean-absorbing oil in catalytic cracking devices, and online monitoring and traceability cannot be achieved.

Method used

The online purge trap-gas chromatography analyzer combined with partial least squares discriminant analysis (PLS-DA) method is used to analyze the leakage components in circulating water, establish a material standard sample characteristic database, and use the PLS-DA model to perform the Discrimination and traceability of leaked materials.

Benefits of technology

It realizes online monitoring of catalytic cracked water-cooler materials and traceability identification of leaked materials, without manual sampling and organic solvents, improves the sensitivity and automation of monitoring, saves labor costs, and is green and economical.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119023824A_ABST
    Figure CN119023824A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of water cooler material leakage monitoring and distinguishing, and particularly relates to a catalytic cracking water cooler material leakage online monitoring and distinguishing method. According to the method, online monitoring is adopted, a purging and trapping-gas chromatographic analyzer and a partial least square discriminant analysis method are adopted for discriminating and tracing leaked materials, and the method comprises the following steps: determining retention time and peak area data of a plurality of water cooler material standard samples by using the online purging and trapping-gas chromatographic analyzer; the method comprises the following steps: obtaining a standard sample characteristic database of leaked materials in circulating water through processing, establishing a PLS-DA prediction model, substituting online monitored circulating water sample detection result data into the established PLS-DA prediction model, calculating a predicted value of a sample, and automatically judging the category of the leaked materials. The method can realize online monitoring of water cooler material leakage and traceability identification of leaked material types, is high in sensitivity and automation degree, does not need to add an organic solvent in the analysis process, and is green and economical.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of water cooler material leakage monitoring and identification, and in particular relates to an online monitoring and identification method for catalytic cracking water cooler material leakage. Background Art

[0002] Water coolers are important equipment in the production and operation of petrochemical enterprises. Water materials leaking into the circulating water due to water cooler corrosion is a common problem in refining and chemical enterprises. Once the process medium leaks into the circulating water system, the water quality of the entire circulating water system will change abnormally, leading to water quality deterioration, which will not only aggravate the corrosion of the water cooler, shorten the service life of the water cooler, but also aggravate the scaling of the water cooler, resulting in reduced heat exchange efficiency of the system and increased energy consumption. In severe cases, it may even cause the circulation water system to shut down and the production equipment to stop working unplanned, causing great economic losses.

[0003] Existing leak detection in circulating water mostly adopts traditional manual leak detection methods such as observing the color change of circulating water, smelling odor, or analyzing the changes of indicators such as turbidity, petroleum, chemical oxygen demand, etc., but only large leaks can be found. For trace leaks, it is difficult to find because the changes in the color, odor, turbidity, etc. of circulating water are not obvious. Relying on manual methods to check usually takes a lot of energy and time. The water cooler materials of catalytic cracking units are mainly gasoline, light diesel, lean absorption oil, liquefied petroleum gas, etc. It is extremely difficult to find trace leaks of such materials. When trace leaks cause abnormal water quality in the circulating water field, it takes several days or even weeks to find the leaking water cooler. Since the source of the leak cannot be found in time, it has a huge impact on the normal operation of the circulating water field and the long-term normal operation of the device. The online oil-in-water monitor based on infrared, ultraviolet and other technologies usually has an alarm value above 10ppm. Leakage can only be warned when the concentration of leaked materials in the water is greater than 10ppm.

[0004] CN 206496874 U uses optical fiber, infrared and ultraviolet fluorescence methods to analyze liquid leakage in circulating water; patent CN 101509822A uses gas stripping to extract combustible gas leakage in circulating water; patent CN 102279082 B combines multiple detection technologies such as ultraviolet spectroscopy, ultraviolet fluorescence, optical fiber, infrared, gas detector, etc., and selects different detection instruments for detection according to different leakages; CN206161244U, CN106595982A, CN102175403A use optical fiber and ultraviolet detection to monitor the leakage of materials in circulating water, and CN206161244U and CN109752150A use color developing materials to monitor the leakage of materials in circulating water; CN101509822B gas strips the combustible gas components in the water sample, and then uses a highly sensitive combustible gas detector to detect the concentration of the combustible gas, and the gas stripping method measures the total amount of volatile hydrocarbons. All of the above technologies can only obtain the total amount of leaked material, but cannot determine what type of material is leaking.

[0005] CN1769851A establishes a corresponding characteristic chromatogram for the heat exchanger in the refinery circulating water system according to the medium in the heat exchange section through solid phase microextraction; then the circulating water of the refinery circulating water system is tested and the analysis result spectrum of the circulating water is obtained by analysis, which requires offline water sample collection for analysis and manual comparison of the spectrum.

[0006] CN114428125A performs chromatographic analysis on the oil phase in the liquid sample from the circulating water system to obtain a measurement spectrum, and then determines the similarity S between the measurement spectrum and the characteristic spectrum of the oil medium in at least one oil-water heat exchanger in the circulating water system, and determines the oil-water heat exchanger that has leaked in the circulating water system based on the similarity S. This technology is an offline water sample collection and analysis, which has poor timeliness, and the similarity algorithm is difficult to distinguish between lean absorption oil and light diesel oil with very similar compositions. This method cannot measure leaked liquefied gas in water.

[0007] Jiang Huiyun et al. used solid phase extraction-gasoline chromatography-mass spectrometry to identify diesel and gasoline leaks in water. Liquid-liquid extraction or solid phase extraction-gasoline chromatography-mass spectrometry was used, and the pre-treatment was complex and required the use of organic solvents. The concentration of the water sample studied was 50 ppm. Xiao Ke et al. used solid phase microextraction to determine circulating water to determine the leakage of water coolers in aromatics and other devices. This method requires the collection of water samples for laboratory analysis. Due to the large randomness of offline sampling, the SPME method of 20 mL small volume water samples is prone to large errors. In addition, the disadvantage of offline water sampling is that it is difficult to capture trace amounts of gaseous leaked materials, such as liquefied petroleum gas.

[0008] The main water cooler materials for catalytic cracking are liquefied petroleum gas, gasoline, light diesel and lean absorption oil. The components of light diesel and lean absorption oil are highly similar. It is difficult to obtain their difference characteristics by using solid phase microextraction, solid phase extraction and liquid-liquid extraction methods commonly used to determine oil in water, and it is difficult to capture the leakage of liquefied petroleum gas by manually collecting water samples offline. In addition, there is currently no technology that can simultaneously monitor and trace the leakage of gas (liquefied gas) and liquid oil water cooler materials in catalytic cracking units; in the field of petrochemical water cooler material leakage monitoring and tracing, there is no report on the use of online monitoring technology combined with statistical algorithms to monitor and identify trace material leakage.

[0009] Partial least squares discriminant analysis (PLS-DA) is a new multivariate statistical analysis method that has been developed and widely used in recent years due to practical needs. It is an organic combination of multivariate linear regression, principal component analysis and canonical correlation analysis. PLS-DA uses the method of partial least squares regression to reduce the dimension of data, establish a regression model, and perform discriminant analysis on the regression results.

[0010] The process of the purge and trap method is to use an inert gas such as nitrogen to purge through a liquid or solid at a certain flow rate. After the trace volatile components to be analyzed are blown out, they are adsorbed by the adsorbent in the cold trap, and then heated and desorbed into the gas chromatography system for analysis. The purge and trap method can measure lower trace components than the static headspace method. Therefore, in general, the principle of the purge and trap-gas chromatography method is: dynamic headspace extraction-adsorption capture-thermal desorption-gas chromatography analysis. The purge and trap method does not require the use of organic solvents for sample pretreatment, does not cause secondary pollution to the environment, and has the advantages of small sampling volume, high enrichment efficiency, little matrix interference, and easy online detection.

[0011] At present, there is no related report on the use of online purge and trap analysis technology combined with statistical model algorithms in the field of water cooler material leakage monitoring and tracing. The purpose of the present invention is to use an online purge and trap-gas chromatography monitor to analyze the leakage components in the circulating water and combine it with the partial least squares discriminant analysis method to solve the problem of difficulty in detecting and tracing the leakage of water cooler materials in the circulating water of a catalytic cracking unit. Summary of the invention

[0012] The technical problem to be solved by the present invention is to provide an online monitoring and identification method for material leakage in a catalytic cracking water cooler. The technical solution adopted is:

[0013] An online monitoring and identification method for material leakage in a catalytic cracking water cooler is disclosed. The online monitoring adopts a purge-trap-gas chromatograph analyzer installed on the circulating water return pipe. The partial least squares discriminant analysis method is used to identify and trace the leaked material. The steps adopted include:

[0014] (1) An online purge-and-trap gas chromatograph was used to measure the retention time and peak area data of the organic components that could be purged in circulating water of several standard samples of water cooler materials. Compounds with peak areas greater than 3% of the total peak area were retained as characteristic compounds, and a characteristic database of standard samples of leaked materials was obtained.

[0015] (2) All samples were divided into test and validation sets, and data samples from some databases were used as training samples. The compound peak area matrix was used as the independent variable, and the material category information was used as the observed variable. UV scaling was performed on the observed variables. The model was established by PLS-DA, and Q 2 The latent variable is retained at the maximum value of Y, and a correction model based on partial least squares is established, and the model is used to predict the material category.

[0016] (3) Collect circulating water return samples online at regular intervals and measure them using an online purge-and-trap gas chromatograph. When there is material leakage, obtain the peak area data of each component in the leaked material;

[0017] (4) Substitute the data of the unknown sample obtained in step (3) into the PLS-DA prediction model established in step (2), calculate the predicted category Y value through the chromatographic peak area matrix X of the unknown sample, and determine the category of the leaked material based on the predicted value.

[0018] The liquid leakage material is gasoline, light diesel, lean absorption oil, etc. These materials contain at least one of C5-C12 alkanes, C6-12 cycloalkanes, C6-12 olefins and C6-12 aromatic hydrocarbons, and the gaseous leakage is C3-C5 liquefied gas. The water cooler material is used as the standard spiked sample, and the standard spiked and actual water sample monitoring adopt the same online monitoring method.

[0019] Preferably, the online purge and trap-gas chromatograph comprises a sampling unit, a purge unit, an adsorption unit and a detection unit.

[0020] Preferably, the sampling unit includes a controller, a water sampling system, a washer and a liquid level sensor or a flow meter. The controller can set the sampling time and the cleaning time regularly to control the sampling water volume. The purging unit uses an inert gas for purging. The adsorption unit includes an adsorption tube. The detection unit uses a gas chromatography-hydrogen flame ionization detector for analysis. The inert gas includes nitrogen, argon, etc.

[0021] Preferably, the step (1) of obtaining a known leak material standard sample feature database comprises the following specific steps:

[0022] Several liquid materials were added to the circulating water of the water cooler without leakage, and sample spike tests were conducted with concentrations of 0.1-10 μL / L liquid materials and 0.02-5 mL / L gas materials. The volatile organic matter that could be purged was determined by online purge and trap-gas chromatography, and the retention time and peak area data of 30-50 compounds were obtained;

[0023] Arrange by retention time, extract common peaks to establish common patterns, group by material name, obtain different standard samples and their compound peak area data; summarize the obtained compound retention time and common characteristic compound peak area detection data, and use all data to establish a known material standard sample characteristic database.

[0024] Preferably, in step (2), the sub-step of establishing a PLS-DA prediction model for several known materials includes:

[0025] a. Divide the data samples of all standard samples tested into two subsets, namely a training set and a validation set; the training set is used for modeling, and the training set data samples cover all material types, and the validation set is used to verify the accuracy of the model;

[0026] b. Classify the data samples in the training set by material name;

[0027] c. For the observed variables, UV scaling is performed (subtract the mean of each column variable from the original data, and then divide it by the variance of the column variable);

[0028] d. The compound peak area matrix corresponding to each known material is set as the independent variable X, the water cooler material classification information is set as the observed variable Y, and a PLS-DA prediction model is established;

[0029] e. Optimize the obtained PLS-DA prediction model. If a data sample is not within the 95% confidence interval, it means that the sample is not representative enough. Discard the sample and refit the calibration set.

[0030] f. Perform k-fold cross validation on the PLS-DA prediction model of the established validation set, with a k value of 7-10, and perform random validation 50-100 times to verify the accuracy of the model; if the correct discrimination rate of the source of the standard samples in the validation set reaches more than 90%, it proves that the PLS-DA prediction model has high accuracy and stability.

[0031] Preferably, the training set includes 50%-90% of the data samples, and the rest are validation set data samples; that is, most of the data samples are used as training sets.

[0032] Preferably, the fitting parameters include the cumulative explanation rate of the model through the Y-axis, the cumulative prediction rate of the model, the corrected root mean square error, and the cross-validation root mean square error; wherein, the closer the cumulative explanation rate of the model through the Y-axis and the cumulative prediction rate of the model are to 1, the stronger the predictive ability of the model is.

[0033] Preferably, in step (3), the specific steps of collecting circulating water samples in the water cooler for measurement include:

[0034] The controller controls the sampling unit of the online purge and trap-gas chromatograph to perform sampling at regular intervals. The water sampling system of the sampling unit collects water samples from the water inlet pipeline to the water cup, and then accurately controls the water volume in the water cup through a liquid level sensor or flow meter to keep the water volume consistent each time.

[0035] After the collection is completed, the instrument first starts the pre-pumping function to remove the remaining air above the water cup. After the pre-pumping is completed, the purge unit is purged with high-purity inert gas for 3-12 minutes, and the purge flow rate is 80-100mL / min; while purging the water cup, the instrument starts the sampling pump to extract the purge gas above the sampling water cup to the adsorption tube of the adsorption unit at a flow rate of 50-200mL / min for enrichment; after the purge and enrichment are completed, the adsorption tube is heated to desorb the sample to the gas chromatography-hydrogen flame ionization detector for analysis. When there is material leakage, the volatile organic compound components that can be purged are determined to obtain the retention time and peak area data of multiple compounds;

[0036] Arrange the compounds detected from different water cooler materials by retention time and extract the common peaks; group them according to the material names to obtain the peak area data of different water cooler materials and their compounds;

[0037] The instrument is an online purge and trap gas chromatograph.

[0038] Preferably, in step (3), the circulating water sample in the water cooler is collected online at an interval of 1-2 hours.

[0039] Preferably, in step (3), the volume of the circulating water sample collected each time is the same, and the collected volume is 500-1500 ml.

[0040] Preferably, after the instrument stops detecting and analyzing, the controller is used to set a cleaner to clean the water sampling system regularly to ensure that leaked materials in the water sampling system are not deposited in the sampling pipeline, and then the instrument pipeline is purged for 2-5 minutes through a purge unit.

[0041] Preferably, when the instrument is used, a 6% cyanopropyl / phenyl, 94% polydimethylsiloxane or 100% dimethylpolysiloxane chromatographic column is used.

[0042] The water cooler leakage collection system provided by the present invention can simultaneously collect volatile components in gaseous leakages above C3, liquefied gas, and liquid water cooler materials such as gasoline, diesel, and lean absorption oil, obtain the leakage material characteristics, and identify the leakage material according to the calculation results of the prediction model;

[0043] In particular, the technology provided by the present invention can realize online monitoring of water cooler materials and traceability identification of leaked materials, without the need for manual sampling in the laboratory for analysis. It has a high degree of automation and sensitivity, and can greatly save labor costs. In addition, no organic solvents need to be added during the analysis process, which is green and economical. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 The chromatograms of the standard samples of catalytic gasoline, catalytic light diesel oil, and catalytic lean absorption oil circulating water in Example 1 of the present invention (from top to bottom);

[0045] Figure 2 It is a fitting curve diagram established based on the PLS-DA method in Example 1;

[0046] Figure 3 It is a score diagram of the material identification model in circulating water established based on the PLS-DA method in Example 1, that is, a score diagram of the PLS-DA prediction model (Figure a is a 3D score diagram of LV1-LV3, and Figure b is a score diagram of LV1 and LV3);

[0047] Figure 4 This is a diagram showing the classification results of different leaked materials using the PLS-DA model in Example 1;

[0048] Figure 5 The chromatogram of liquefied petroleum gas leaked in the circulating water in Example 4;

[0049] Figure 6 The chromatograms of 0.1 μL / L catalytic gasoline and catalytic light diesel in the circulating water in Example 4 (from top to bottom);

[0050] Figure 7 This is the result diagram of gasoline, diesel and lean absorption oil of a catalytic cracking unit calculated by the similarity algorithm in Comparative Example 1;

[0051] Figure 8 The mass spectrum of gasoline in circulating water measured by SPME method in Comparative Example 1;

[0052] Fig. 9 This is the mass spectrum of lean absorption oil and light diesel oil in circulating water determined by solid phase extraction-gas chromatography-mass spectrometry in Comparative Example 2. DETAILED DESCRIPTION

[0053] The drawings are for illustrative purposes only. To better illustrate the present embodiment, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, some well-known structures in the drawings and their descriptions may be omitted.

[0054] The technical solution of this specification is clearly and completely described below in conjunction with the embodiments and drawings.

[0055] Example 1

[0056] An online purge and trap-gas chromatograph is installed at the circulating water return monitoring point of the catalytic cracking unit of the refinery to monitor the circulating water return online. The main heat exchange materials of the water cooler of the catalytic cracking unit are gasoline, light diesel, lean absorption oil and liquefied petroleum gas.

[0057] The online purge and trap-gas chromatograph comprises a sampling unit, a purge unit, an adsorption unit and a detection unit.

[0058] The sampling unit includes a controller, a water sampling system, a washer and a liquid level sensor or flow meter, and a water cup for measuring water samples. The washer can clean the pipe for collecting water samples. The controller can set the sampling time and cleaning time regularly, and control the sampling water volume at the same time. In order to make the sample representative and have high sensitivity, it is necessary to ensure that the volume of each sample is equal.

[0059] The purge unit is purged with nitrogen;

[0060] The adsorption unit includes an adsorption tube, which can enrich the volatile organic components in the adsorption unit;

[0061] The detection unit uses a gas chromatography-hydrogen flame ionization detector for analysis.

[0062] The online purge and trap gas chromatograph was set to run once every 2 hours, and 1.5L water sample was collected each time. The purge lasted for 3 minutes. The sampling pump was started at the same time to extract the purge gas above the sampling cup to the adsorption tube at a flow rate of 50mL / min for enrichment. After the purge and enrichment was completed, the sample was desorbed by heating the adsorption tube to the gas chromatography-hydrogen flame ionization detector (GC-FID) for analysis. Chromatographic column: DB-624 (30m*0.25mm*1.4μm); carrier gas type: N2; program temperature rise: initial temperature 40℃ for 2 minutes, 5℃ / min to 80℃ for 2 minutes, 20℃ / min to 180℃ for 2 minutes; column flow (constant flow mode): 1mL / min; split ratio: 10:1; hydrogen flow: 30ml / min; air flow: 300ml / min; tail flow: 35mL / min; desorption flow: 10ml / min.

[0063] An online monitoring and identification method for catalytic cracking water cooler material leakage adopts online monitoring and purge-trap-gas chromatograph to monitor the water cooler, and adopts partial least squares discriminant analysis method to identify and trace the leaked material. The steps adopted include:

[0064] (1) Using an online purge-and-trap gas chromatograph to measure the retention time and peak area data of standard spiked samples of water cooler materials that may leak, more than 20 samples of data are obtained for each material, and a characteristic database of standard samples of known materials is established; the specific steps include:

[0065] Gasoline, light diesel, lean absorption oil and liquefied petroleum gas were added to circulating water without leakage, and standard water samples of gasoline, light diesel and lean absorption oil with concentrations of 1, 2, 3, 4, 5, 7 and 10 μL / L were measured online. Liquefied petroleum gas was added to circulating water without leakage, and standard water samples with concentrations of 0.04, 0.08, 1, 2, 3, 4 and 5 mL / L were measured online. Each concentration was repeated 2-4 times, and 16-25 groups of results were obtained for each material, with a total of 84 samples. The online monitoring characteristic chromatogram was obtained by online purge and trap-gas chromatograph analysis, see Figure 1 Each leaked material has dozens of chromatographic peaks, and the purge and trap-gas chromatography method can roughly distinguish the material type from the spectrum.

[0066] First, the gasoline sample was integrated, and the compounds with peak areas greater than 3% of the total peak area were retained. The peak integration was named gasoline 1-25. Then the diesel peak was integrated, and the compounds with peak areas greater than 3% of the total peak area were retained, named diesel 1-16. The LPG peak was integrated and named LPG1. The above compounds have covered the main compounds in the lean absorption oil sample, so the compounds in the lean absorption oil sample were not added. A total of 42 compounds were collected, and the peak areas of the compounds in all samples were summarized to establish a known material standard sample feature database.

[0067] (2) A common pattern was established for all the integrated peaks of four materials, namely gasoline, light diesel, lean absorption oil and liquefied petroleum gas, with a total of 42 peaks. The data samples in 90% of the database were used as training samples, the compound peak area matrix was used as the independent variable, and the material type information was used as the observed variable. UV scaling was performed on the observed variables (the original data was subtracted from the mean of each column variable and then divided by the variance of the column variable); the preprocessed chromatographic matrix X and category vector Y were orthogonally decomposed to obtain the score matrices T and U of the chromatographic matrix X and category vector Y. Then, linear regression was performed on the score matrices T and U to obtain the regression coefficient B. Among them, P and Q are the loading matrices of the chromatographic matrix X and category vector Y respectively; E and F are the PLS fitting residual matrices of the chromatographic matrix X and category vector Y respectively. The partial least squares regression latent variable loading data are shown in Table 1, the regression latent variable score data are shown in Table 2, and the regression fitting value and residual analysis are shown in Table 3.

[0068] The PLS-DA prediction model of gasoline, diesel, lean absorption oil and liquefied petroleum gas was established. Four data samples were not within the 95% confidence interval. These four outlier samples were discarded and the calibration set was refitted to establish the regression equation. The optimized regression equation is shown in Table 4 and the fitting curve is shown in Figure 2 , coefficient of determination R 2 0.97. The closer the determination coefficient value is to 1, the better the correlation between the predicted value of the established model and the true value.

[0069] The projection score diagram of the established recognition model is as follows: Figure 3 As shown, through Figure 3 It can be seen from Figure a and Figure b that after optimization, the samples participating in modeling are all distributed within the 95% confidence interval, and different categories of materials can obtain good classification results.

[0070] Table 1 Partial least squares regression latent variable loadings (first three latent variables)

[0071]

[0072]

[0073]

[0074] Table 2 Partial least squares regression latent variable scores (first three latent variables)

[0075]

[0076]

[0077]

[0078] Table 3 Regression fitting values ​​and residual analysis

[0079]

[0080]

[0081]

[0082] Table 4 Partial least squares regression equation

[0083]

[0084]

[0085] The fitted parameters include the cumulative explanation rate (R 2 Y), model cumulative prediction rate (Q 2 Y), corrected root mean square error, and cross-validation root mean square error; among them, the closer the model's cumulative explanation rate on the Y axis and the model's cumulative prediction rate are to 1, the stronger the model's predictive ability is.

[0086] The hypothesis test of the partial least squares regression equation is shown in Table 5. Among them, at the test level α = 0.05, H0 is rejected (H0 is the null hypothesis), and the difference is statistically significant. That is, there is a significant linear regression relationship between the data. The quality evaluation of the partial least squares regression model shown in Table 6 shows that when three latent variables are selected, the model Y prediction rate Q2Y is 1.00, indicating that this model has strong prediction ability.

[0087] Table 5 Hypothesis test of partial least squares regression equation

[0088]

[0089] Table 6 Quality evaluation of partial least squares regression model

[0090]

[0091] 10% of the samples were selected as the validation set, and k-fold cross validation was used, k = 7, and random validation was performed 100 times. When the prediction rate Q2Y of model Y is 1, the latent variable is 3, so 3 latent variables are selected, as shown in Table 7. At this time, the cross recognition rate reaches 99.71%, proving that the PLS-DA prediction model has high accuracy and stability.

[0092] Table 7 Model cross-validation recognition rate

[0093]

[0094] The classification probability distribution is predicted by partial least squares discriminant analysis (PLS-DA), as shown in Table 8.

[0095] Table 8 Partial least squares discriminant analysis (PLS-DA) predicted classification probability distribution

[0096]

[0097] (3) The online purge and trap gas chromatograph was operated once every 2 hours. The water sampling system of the sampling unit collected water samples from the water inlet pipeline into the water cup. The liquid level sensor was used to accurately control the amount of water in the water cup. 1.5 L of water sample was collected each time.

[0098] The instrument first starts the pre-pumping function to remove the remaining air above the water cup. After the pre-pumping is completed, the purge unit is purged with nitrogen for 3 minutes at a purge flow rate of 100 mL / min. While purging the water cup, the instrument starts the sampling pump to extract the purge gas above the sampling water cup at a flow rate of 200 mL / min to the adsorption tube of the adsorption unit for enrichment. After the purge and enrichment are completed, the sample is desorbed by heating the adsorption tube to the gas chromatography-hydrogen flame ionization detector for analysis to obtain the retention time and peak area data.

[0099] In order to ensure accurate sampling results each time, after the instrument stops detecting and analyzing, the controller is used to set the cleaner to clean the water sampling system regularly to ensure that leaked materials in the water sampling system are not deposited in the sampling pipeline. Then, the pipelines of the instrument are purged for 2 minutes through the purge unit to prevent the leakage materials from being deposited in the pipelines of other units, which may cause inaccurate sampling and analysis next time.

[0100] The compounds detected in the samples were arranged according to the common peaks to obtain the peak area data, see Table 9.

[0101] Table 9 Sample Peak Area

[0102]

[0103]

[0104] (2) Substitute the chromatographic peak data of the unknown material obtained in step (3) into the PLS-DA prediction model established in step (2) to calculate the predicted value of the leaked material in the water sample measured this time. The predicted value is 1.18, the confidence level is 0.97, and the prediction is gasoline leakage, as shown in Table 10. After investigation, it was found that the gasoline water cooler E309A was leaking. The method of the present invention timely discovered the leakage of the water cooler material, narrowed the scope of investigation, and saved the investigation time.

[0105] Table 10 Prediction results of samples identified by PLS-DA prediction model

[0106]

[0107] Example 2

[0108] This embodiment is used to illustrate the detection of leakage of lean absorption oil in circulating water using the online purge and trap-gas chromatograph.

[0109] An online purge and trap-gas chromatograph is installed at the circulating water return monitoring point of the catalytic cracking unit of the refinery to monitor the circulating water return online. The main heat exchange materials of the water cooler of the catalytic cracking unit are gasoline, light diesel, lean absorption oil and liquefied petroleum gas.

[0110] The online monitoring system was run once every 2 hours, and 1.0L of water sample was collected online each time, and purged for 5 minutes. During the purging, the instrument started the sampling pump and extracted the purge gas above the sampling water cup to the adsorption tube at a flow rate of 50mL / min for enrichment. After the purge and enrichment were completed, the adsorption tube was heated to desorb the sample to the gas chromatography-hydrogen flame ionization detector (GC-FID) for analysis. Chromatographic conditions: HP-PONA capillary column (50m×0.2mm×0.25μm); injection port temperature 250℃, split ratio 5:1, column flow rate 0.7mL / min, column initial temperature 40℃, maintained for 10min, 5℃·min -1 The temperature was raised to 310°C at a rate of 1.5 °C. Hydrogen flow rate: 30 ml / min; air flow rate: 300 ml / min; tail gas flow rate: 35 mL / min; desorption flow rate: 12 ml / min.

[0111] (1) Using an online purge-and-trap gas chromatograph to measure the retention time and peak area data of several known material standard samples that may be detected in water coolers, and obtaining a known material standard sample feature database through processing; specifically including:

[0112] Gasoline, light diesel oil, and lean absorption oil were added to the circulating water without leakage. The spiked concentrations of gasoline, light diesel oil, and lean absorption oil were 0.2, 1, 3, 5, and 7 μL / L, respectively. The spiked concentrations of liquefied gas were 0.02, 1, 2, 3, and 5 mL / L. Through online analysis, each concentration was paralleled 6 times to obtain 30 groups of chromatograms and integrated peak areas for each material. The online monitoring characteristic chromatogram was obtained by online purge and trap-gas chromatograph analysis. In summary, a characteristic database of known material standard samples was established, and 45 common peaks were obtained, with specific reference to Example 1.

[0113] (2) 90% of the data samples in the database were used as training samples, the compound peak area matrix was used as the independent variable, and the material type information was used as the observed variable. UV scaling was used for data processing for the observed variable. A partial least squares regression equation was established through PLS-DA to establish a prediction model for identifying leaked materials, namely the PLS-DA prediction model. The remaining data samples were substituted into the PLS-DA prediction model to test its discrimination accuracy.

[0114] The obtained PLS-DA prediction model was optimized. If a data sample was not within the 95% confidence interval, it indicated that the sample was not representative enough. The sample was discarded and the calibration set was refitted. The fitting parameters included the cumulative explanation rate (R 2 Y), model cumulative prediction rate (Q 2 Y), corrected root mean square error, and cross-validation root mean square error; among them, the closer the model's cumulative explanation rate on the Y axis and the model's cumulative prediction rate are to 1, the stronger the model's predictive ability is.

[0115] The obtained PLS-DA prediction model was optimized and validated using k-fold cross validation, k=10, 80 random validations, and 3 latent variables.

[0116] The model has been verified to be highly reliable, with a cross-validation accuracy of 98.6%. The distribution of the classification probability predicted by partial least squares discriminant analysis (PLS-DA) is shown in Table 11.

[0117] Table 11 Partial least squares discriminant analysis (PLS-DA) predicted classification probability distribution

[0118]

[0119] (3) The online purge and trap gas chromatograph was operated once every 2 hours. The water sampling system of the sampling unit collected water samples from the water inlet pipeline into the water cup. The liquid level sensor was used to accurately control the amount of water in the water cup. 1.0 L of water sample was collected each time.

[0120] The instrument first starts the pre-pumping function to remove the remaining air above the water cup. After the pre-pumping is completed, the purge unit is purged with high-purity nitrogen for 5 minutes at a purge flow rate of 80 mL / min. While purging the water cup, the instrument starts the sampling pump to extract the purge gas above the sampling water cup to the adsorption tube of the adsorption unit at a flow rate of 100 mL / min for enrichment. After the purge and enrichment are completed, the sample is desorbed by heating the adsorption tube to the gas chromatography-hydrogen flame ionization detector for analysis to obtain the peak area data of each compound.

[0121] (4) Substitute the data of the unknown leaked material obtained in step (3) into the PLS-DA prediction model established in step (2), and calculate the predicted value of the leaked material to be 1.88, and the predicted classification is lean absorption oil with a confidence level of 0.96. The method of the present invention timely discovered the leakage of the water cooler material, and determined the category of the leaking water cooler. After checking the lean absorption oil water cooler, it was confirmed that the E216A lean absorption oil water cooler had leaked. See Table 12.

[0122] Table 12 Prediction results of samples identified by PLS-DA prediction model

[0123]

[0124] Other unmentioned places are the same as those in Example 1.

[0125] Example 3

[0126] This embodiment is used to illustrate the detection of liquefied petroleum gas leakage in circulating water using the method of the present invention.

[0127] An online purge and trap-gas chromatograph is installed at the circulating water return monitoring point of the refinery catalytic cracking unit to monitor the circulating water return online. The main heat exchange materials of the water cooler of the unit are gasoline, light diesel, lean absorption oil and liquefied petroleum gas.

[0128] The online monitoring system was run once every 2 hours, and 1.5L of water sample was collected online each time. The purge gas above the sampling cup was drawn to the adsorption tube for enrichment at a flow rate of 50mL / min. After the purge and enrichment was completed, the sample was desorbed by heating the adsorption tube to the gas chromatography-hydrogen flame ionization detector (GC-FID) for analysis. Chromatographic conditions: DB-624 capillary column (30m×0.25mm×1.4μm); injection port temperature 250℃, split ratio 5:1, column flow rate 1.0mL / min, column initial temperature 40℃, maintained for 2min, at 5℃·min -1 The temperature was raised to 80℃ at a rate of 10℃·min -1 The temperature was raised to 200°C at a rate of 100 °C and maintained for 8 min. Hydrogen flow rate: 30 ml / min; air flow rate: 300 ml / min; tail gas flow rate: 35 mL / min; desorption flow rate: 10 ml / min.

[0129] 0.4, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 μL of catalytic gasoline, diesel, lean absorption oil, and 0.04, 1, 2, 3, 4, 5 mL of liquefied petroleum gas were added online respectively, and each concentration was sampled 3 times. After analysis by the online monitor, 33 characteristic spectra were obtained for each material. The leakage identification models of liquefied petroleum gas, diesel, gasoline, and lean absorption oil were established by the PLS-DA algorithm, and the mean values ​​were respectively When 3 latent variables were selected, the model explanation rate Q2Y reached 1.0, and the model cross-validation recognition rate reached 98.1%. The recognition rates of the latent variables are shown in Table 13.

[0130] Table 13 Model cross-validation recognition rate

[0131]

[0132] The classification probability distribution was predicted by partial least squares discriminant analysis (PLS-DA), as shown in Table 14. The classification means of liquefied petroleum gas, diesel, gasoline, and lean absorption oil were 0.03, 1.05, 1.98, and 2.97, respectively.

[0133] Table 14 Partial least squares discriminant analysis (PLS-DA) predicted classification probability distribution

[0134]

[0135] The sample classification was predicted by partial least squares discriminant analysis (PLS-DA), as shown in Table 15. The prediction value of the online monitoring system was 0.02, which was close to the classification value of liquefied petroleum gas. It was determined to be a liquefied petroleum gas leakage alarm. The online gas chromatogram of liquefied petroleum gas is shown in Figure 4 , the retention time of C3-C5 components is less than 5 minutes. On-site inspection confirmed that the E310A liquefied gas water cooler had a leak, and the liquefied petroleum gas leak was discovered early through online monitoring.

[0136] Table 15 Prediction results of samples identified by PLS-DA prediction model

[0137]

[0138] Other unmentioned places are the same as those in Example 1.

[0139] Example 4

[0140] This example is used to illustrate the actual measurement of the detection limit (0.1 uL / L) of trace gasoline leakage in circulating water using the method of the present invention.

[0141] An online purge and trap-gas chromatograph is installed at the circulating water return monitoring point of the refinery catalytic cracking unit to monitor the circulating water return online. The main heat exchange materials of the water cooler of the unit are gasoline, light diesel and liquefied petroleum gas.

[0142] The online monitoring system was run once every 2 hours, and 1.5L of water sample was collected online each time. The purge gas above the sampling cup was drawn to the adsorption tube for enrichment at a flow rate of 50mL / min. After the purge and enrichment was completed, the sample was desorbed by heating the adsorption tube to the gas chromatography-hydrogen flame ionization detector (GC-FID) for analysis. Chromatographic conditions: DB-624 capillary column (30m×0.25mm×1.4μm); injection port temperature 250℃, split ratio 5:1, column flow rate 1.0mL / min, column initial temperature 40℃, maintained for 2min, at 5℃·min -1 The temperature was raised to 80℃ at a rate of 10℃·min -1 The temperature was raised to 200°C at a rate of 100 °C and maintained for 8 min. Hydrogen flow rate: 30 ml / min; air flow rate: 300 ml / min; tail gas flow rate: 35 mL / min; desorption flow rate: 10 ml / min.

[0143] Figure 5The chromatogram of the circulating water sample of catalytic gasoline and light diesel at a concentration of 0.1μL / L is the result of the analysis by the online monitor. It can be seen from the figure that the response of the main chromatographic peak is greater than 3 times the signal-to-noise ratio, indicating that the actual detection limit of catalytic gasoline and light diesel in circulating water is less than 0.1μL / L.

[0144] Comparative Example 1:

[0145] The same method as in Example 1 was used to obtain the online leakage spectral library of different materials, but the PLS-DA algorithm was changed to the similarity algorithm (angle cosine) to identify diesel and lean absorption oil in the circulating water of a catalytic device. Figure 6 The similarity values ​​between diesel and lean absorption oil are between 0.5 and 0.65, and it is difficult to distinguish lean absorption oil from light diesel using the similarity algorithm.

[0146] Comparative Example 2

[0147] Laboratory solid phase microextraction (SPME)-GCMS method for determination of catalytic cracking water samples

[0148] Add 10 μL of catalytic cracking oil into 500 mL of circulating water without interferences. After shaking thoroughly, take 2 portions of 19 mL of water sample and add them into 20 mL sample bottles. After shaking extraction at 40 °C for 40 min, desorb at 260 °C for 3 min.

[0149] Chromatographic conditions: HP-PONA capillary column (50 m × 0.2 mm × 0.25 μm); injection port temperature 260 °C; split ratio 5:1; column temperature program: initial column temperature 40 °C, maintained for 5 min, at 5 °C min -1 The rate was raised to 320℃. CTC injector was used for SPME mode injection, extraction temperature was 40℃, and probe aging temperature was 260℃.

[0150] Mass spectrometry conditions: EI source, electron energy 70 eV, ion source temperature 230 °C; MS quadrupole 150 °C; transfer line temperature: 280 °C; acquisition mode: full scan; scanning mass range: (5-300) amu.

[0151] SPME is suitable for the analysis of homogeneous water samples. Figure 6 It can be seen that for diesel water samples that cannot be completely dispersed and dissolved, the small contact area between the SPME probe or stirring rod and the water sample can easily lead to large differences in the test results of parallel samples from the same batch, making it difficult to obtain representative data.

[0152] Comparative Example 3

[0153] A standard water sample of catalytic cracking oil product with a concentration of 1-10 μL / L was prepared, and a spectral library of leaked materials was established by solid phase extraction-gasoline chromatography-mass spectrometry. A similarity algorithm was used to classify gasoline, diesel, and lean absorption oil. The solid phase extraction method can classify gasoline and diesel, but the diesel and lean absorption oil components and peak characteristics obtained by this method are extremely similar. The mass spectra are shown in Figure 7 The accuracy of distinguishing the two types of oils by solid phase extraction is poor. In addition, the water sample of the solid phase extraction method needs to pass through a C18 column, eluted with dichloromethane, and then concentrated and purified. Not only are the pre-treatment steps complicated, but the leakage of liquefied petroleum gas cannot be detected.

[0154] Comparative Example 4

[0155] A refinery catalytic cracking unit circulating water return water is equipped with an ultraviolet fluorescence online water monitor. The main water cooler heat exchange materials of the unit are gasoline, light diesel, lean absorption oil, and liquefied petroleum gas. One day, the online oil in water alarm was triggered, and the detection result was 12.5mg / L (the threshold for judging leakage is 5-10mg / L), indicating that there was an oil leak, but it was impossible to determine what kind of oil was leaking, and this method had no response to liquefied petroleum gas leaks.

[0156] Comparative Example 5

[0157] The return water of a catalytic cracking unit is equipped with an online monitor with a stripping-photoionization detector. When oil or liquefied gas leaks, the combustible gas component is stripped and the concentration of the combustible gas is displayed. The total amount of volatile hydrocarbons can be measured, but the specific type of material leaked cannot be obtained.

[0158] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for online monitoring and identification of material leakage in a catalytic cracking water cooler, characterized in that: Online monitoring uses a purge-trap-gas chromatograph, which is installed on the circulating water return pipe. The partial least squares discriminant analysis method is used to identify and trace the leaked materials. The steps used include: (1) Using an online purge-and-trap gas chromatograph to measure several water cooler material standard samples in circulating water, the retention time and peak area data of the compounds were obtained, and a characteristic database of leakage material standard samples was obtained through processing; (2) Using data samples in part of the database as training samples, compound peak area matrix as independent variable, material type information as observation variable, UV scaling processing is performed on the material type information observation variable, and a prediction model for identifying leaked materials is established through PLS-DA, namely, the PLS-DA prediction model. The remaining data samples are substituted into the PLS-DA prediction model to test its discrimination accuracy; (3) Collect circulating water samples from the water cooler online at regular intervals and measure them using an online purge-and-trap gas chromatograph. When there is material leakage, the peak area data of the compound consistent with the common peak in the model may be obtained; (4) Substitute the data of the unknown material obtained in step (3) into the PLS-DA prediction model established in step (2), calculate the predicted value of the leaked material, and determine the category of the leaked material.

2. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 1, characterized in that: The online purge and trap-gas chromatograph comprises a sampling unit, a purge unit, an adsorption unit and a detection unit.

3. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 2, characterized in that: The sampling unit includes a controller, a water sampling system, a washer and a liquid level sensor or a flow meter. The controller can set the sampling time and the cleaning time regularly to control the sampling water volume; the purging unit uses an inert gas for purging; the adsorption unit includes an adsorption tube; and the detection unit uses a gas chromatography-hydrogen flame ionization detector for analysis.

4. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 1, characterized in that: The step (1) of obtaining a known leak material standard sample feature database comprises the following specific steps: Several water cooler liquid materials and gas materials were added to the circulating water without leakage to prepare 0.1-10μL / L circulating water standard samples, and the volatile organic compounds that could be purged were determined by online purge and trap-gas chromatography, and 30-50 compounds whose peak areas accounted for more than 3% of the total peak area were retained; the materials were grouped according to their names, different standard samples and their compound peak area data were obtained, and a characteristic database of standard samples of known materials was established.

5. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 1, characterized in that: In step (2), the sub-steps of establishing a PLS-DA prediction model for several known materials include: a. Divide the data samples of all standard samples tested into two subsets, namely a training set and a validation set; the training set is used for modeling, and the training set data samples cover all material types, and the validation set is used to verify the accuracy of the model; b. Classify the data samples in the training set by material name; c. For the observed variables, UV scaling is performed (subtract the mean of each column variable from the original data, and then divide it by the variance of the column variable); d. The compound peak area matrix corresponding to each known material is set as the independent variable X, the water cooler material classification information is set as the observed variable Y, and a PLS-DA prediction model is established; e. Optimize the obtained PLS-DA prediction model. If a data sample is not within the 95% confidence interval, it means that the sample is not representative enough. Discard the sample and refit the calibration set. f. Perform k-fold cross validation on the PLS-DA prediction model of the established validation set, with a k value of 7-10, and perform random validation 50-100 times to verify the accuracy of the model; if the correct discrimination rate of the source of the standard samples in the validation set reaches more than 90%, it proves that the PLS-DA prediction model has high accuracy and stability.

6. A method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 5, characterized in that: The training set includes 50%-90% of the data samples, and the rest are validation set data samples.

7. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 5, characterized in that: The fitting parameters include the cumulative explanation rate of the model through the Y-axis, the cumulative prediction rate of the model, the corrected root mean square error, and the cross-validation root mean square error; among them, the closer the cumulative explanation rate of the model through the Y-axis and the cumulative prediction rate of the model are to 1, the stronger the prediction ability of the model is.

8. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 3, characterized in that: In step (3), the specific steps of collecting circulating water samples in the water cooler for measurement include: The controller controls the sampling unit of the online purge and trap gas chromatograph to perform sampling at regular intervals. The water sampling system of the sampling unit collects water samples from the water inlet pipeline to the water cup, and then accurately controls the water volume in the water cup through a liquid level sensor or flow meter to keep the water volume consistent each time. The instrument first starts the pre-pumping function to remove the remaining air above the water cup. After the pre-pumping is completed, the purge unit is purged with high-purity inert gas for 3-12 minutes, and the purge flow rate is 80-100mL / min; while purging the water cup, the instrument starts the sampling pump to extract the purge gas above the sampling water cup to the adsorption tube of the adsorption unit at a flow rate of 50-200mL / min for enrichment; after the purge and enrichment are completed, the adsorption tube is heated to desorb the sample to the gas chromatography-hydrogen flame ionization detector for analysis. When there is material leakage, the volatile organic compounds that can be purged are measured to obtain the retention time and peak area data of each compound; The instrument is an online purge and trap gas chromatograph.

9. A method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 8, characterized in that: In the step (3), the circulating water sample in the water cooler is collected online at regular intervals of 1-2 hours.

10. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 8, characterized in that: In the step (3), the volume of the circulating water sample collected each time is the same, and the collected volume is 500-1500 ml.

11. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 8, characterized in that: After the instrument stops detecting and analyzing, the controller is used to set the washer to clean the water sampling system regularly to ensure that the leaked materials in the water sampling system are not deposited in the sampling pipeline, and then the instrument pipeline is purged for 2-5 minutes through the purge unit.

12. The method for online monitoring and distinguishing material leakage in a catalytic cracking water cooler according to claim 8, characterized in that: When the instrument is used, a 6% cyanopropyl / phenyl, 94% polydimethylsiloxane or 100% dimethylpolysiloxane chromatographic column is adopted.