A spectrum detection method and system for pour point depressant quality

Abstract

The present application relates to the field of spectral detection technology, and specifically to a spectral detection method and system for the quality of a pour point depressant, the method comprising: obtaining spectral data and quality data of each pour point depressant sample; obtaining a feature subset at each iteration of the CARS algorithm; obtaining the quality sensitivity of each spectral wavelength in the feature subset; obtaining the independent coefficient of each spectral wavelength in the feature subset, and obtaining the significant response coefficient of each spectral wavelength; based on the significant response coefficient of each spectral wavelength in the feature subset at each iteration and in combination with the CARS algorithm, the characteristic wavelength of the pour point depressant sample is screened, a regression model is trained, and the quality data of the pour point depressant sample to be detected is obtained. The present application aims to select spectral wavelengths that have a strong correlation with the prediction of the pour point depressant quality, and improve the accuracy and reliability of the quality detection of the pour point depressant sample.

Description

Technical Field

[0001] The present application relates to the field of spectral detection technology, and in particular to a spectral detection method and system for the quality of a pour point depressant. Background Art

[0002] Pour point depressant is an additive used to reduce the freezing point of petroleum products. It is widely used in lubricating oil, diesel, aerospace kerosene and other fields. It is an important means to ensure the economy and safety of pipeline transportation. In recent years, with the rapid development of the petroleum industry, the market demand for pour point depressants has continued to increase. However, due to factors such as raw material quality, production process and storage conditions, the quality of pour point depressants varies. Therefore, it is particularly important to conduct rapid, accurate and reliable quality testing of pour point depressants before using them.

[0003] With the development and progress of science and technology, the methods for quality detection of pour point depressants are becoming more and more mature. Common quality detection methods include spectroscopy, mass spectrometry, and chromatographic analysis technology. Among them, the quality detection method of pour point depressants based on spectroscopy technology is highly favored due to its rapid, non-destructive, and green detection advantages. At present, the quality detection of pour point depressants based on spectroscopy technology mainly adopts the modeling method. However, since spectral data usually contains a large amount of wavelength information, there is a problem of data redundancy. The interference of redundant data makes the original spectral data unable to be directly used for model training. It is necessary to screen the most relevant characteristic wavelengths from the massive wavelength information before modeling. However, the traditional competitive adaptive reweighted sampling (CARS) algorithm uses the least squares regression weight to evaluate the importance of wavelengths, which may lead to unsatisfactory wavelength selection effects, may lose important wavelengths, and affect the quality detection results of pour point depressants. Summary of the invention

[0004] In view of the above, it is necessary to provide a spectral detection method and system for the quality of the pour point depressant to solve the above problems.

[0005] The first aspect of the present application provides a spectral detection method for the quality of a pour point depressant, the method comprising:

[0006] Obtaining spectral data and mass data of each pour point depressant sample;

[0007] Obtain the characteristic subset of the spectral data of each pour point depressant sample at each iteration of the CARS algorithm; for each iteration, based on the distribution characteristics of the spectral data of each spectral wavelength of all pour point depressants in the characteristic subset and the response degree of the spectral data to the change of the mass data, obtain the mass sensitivity of each spectral wavelength in the characteristic subset;

[0008] The numerical differences and similarities between each spectral wavelength in the feature subset and its adjacent wavelengths were analyzed in each iteration of the CARS algorithm to obtain the independent coefficient of each spectral wavelength in the feature subset, which was combined with the mass sensitivity of each spectral wavelength to obtain the significant response coefficient of each spectral wavelength.

[0009] Based on the significant response coefficient of each spectral wavelength in the feature subset at each iteration and combined with the CARS algorithm, the characteristic wavelengths of the pour point depressant samples were screened, the regression model was trained, and the quality data of the pour point depressant samples to be tested were obtained.

[0010] The characteristic subset is specifically a set of all remaining spectral wavelengths except the characteristic wavelength in each iteration of the CARS algorithm.

[0011] The method of obtaining the mass sensitivity of each spectral wavelength in the feature subset is specifically as follows:

[0012] Perform straight line fitting on the mass data of all pour point depressant samples and the spectral data of each spectral wavelength in the characteristic subset at each iteration to obtain the slope of the fitted straight line;

[0013] Calculate the proportion of the slope of the fitted straight line of each spectral wavelength in all spectral wavelengths in the feature subset;

[0014] The first correlation between the mass data of all pour point depressant samples and all spectral data of each spectral wavelength in the feature subset at each iteration is analyzed; the straight line slope proportion is merged with the first correlation to obtain the mass sensitivity of each spectral wavelength in the feature subset.

[0015] The specific process of performing straight line fitting is as follows:

[0016] All spectral data of each spectral wavelength in the feature subset of all pour point depressant samples at each iteration are used as the abscissa, and the mass data of all pour point depressant samples are used as the ordinate, and a straight line fitting is performed.

[0017] The process of obtaining the independent coefficient of each spectral wavelength in the feature subset is as follows:

[0018] The sequence consisting of the spectral data of all the pour point depressant samples at each spectral wavelength is recorded as a spectral data sequence;

[0019] A classification algorithm is used to obtain all neighboring wavelengths of each spectral wavelength in the feature subset;

[0020] Calculate the difference between each spectral wavelength and each of its neighboring wavelengths to obtain a second correlation between the spectral data sequence of each spectral wavelength and the spectral data sequence of each of its neighboring wavelengths;

[0021] Calculate the fusion result of the negative correlation mapping result of the second correlation and the difference; and take the cumulative sum of the fusion results of each spectral wavelength and all its neighboring wavelengths as the independent coefficient of each spectral wavelength in the feature subset.

[0022] The formula for obtaining the significant response coefficient of each spectral wavelength is: ; In the formula, Indicates At the iteration Mass sensitivity for each spectral wavelength; Indicates At the iteration Independent coefficients for each wavelength; Indicates At the first iteration, the The significant response coefficient of each spectral wavelength; , Respectively represent a first preset value and a second preset value, wherein the sum of the first preset value and the second preset value is equal to 1, and the first preset value and the second preset value are both greater than or equal to 0.

[0023] The method of screening the characteristic wavelength of the pour point depressant sample based on the significant response coefficient of each spectral wavelength in the characteristic subset at each iteration and combining the CARS algorithm is specifically as follows:

[0024] A regression model is constructed based on all spectral wavelengths in the feature subset at each iteration of the CARS algorithm to obtain the regression coefficient of each spectral wavelength in the feature subset;

[0025] The significant response coefficient of each spectral wavelength in the feature subset at each iteration is merged with the regression coefficient to obtain the correction coefficient of each spectral wavelength in the feature subset at each iteration;

[0026] The spectral wavelength is screened according to the correction coefficient using an exponential decay function to obtain a characteristic wavelength and corresponding characteristic spectral data at each iteration.

[0027] The correction coefficient of each spectral wavelength in the feature subset obtained at each iteration is specifically:

[0028] The significant response coefficient of each spectral wavelength in the feature subset at each iteration is normalized and multiplied by the regression coefficient of the corresponding feature wavelength to obtain the correction coefficient of each spectral wavelength in the feature subset at each iteration.

[0029] The process of training the regression model and obtaining the quality data of the pour point depressant sample to be tested is specifically as follows:

[0030] The spectral data corresponding to the wavelength obtained by screening the pour point depressant samples are combined with the quality data of the pour point depressant samples for training to obtain a pour point depressant quality regression model;

[0031] The characteristic spectrum data of the pour point depressant sample to be tested is subjected to a pour point depressant mass regression model to obtain predicted mass data of the pour point depressant sample to be tested.

[0032] In a second aspect, an embodiment of the present application further provides a spectral detection system for the quality of a pour point depressant, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned methods when executing the computer program.

[0033] In the above scheme, the spectral data of the pour point depressant sample is obtained, and firstly, the mass sensitivity is constructed based on the response degree of different spectral wavelengths to the change of the pour point depressant mass. The beneficial effect is that the mass sensitivity can reflect the response degree of different wavelengths to the change of the pour point depressant mass based on the mass sensitivity by considering the correlation between the spectral data of wavelengths under different masses of the pour point depressant and the mass of the pour point depressant, as well as the degree of change of the spectral data at the same spectral wavelength under different masses of the pour point depressant; secondly, the independent coefficient is constructed based on the similarity between each wavelength and its adjacent wavelengths. The beneficial effect is that the independent coefficient takes into account the difference between adjacent characteristic wavelengths and the difference between corresponding spectral data, and the independence of each wavelength can be reflected based on the independent coefficient; in addition , a significant response coefficient is constructed based on quality sensitivity and independence coefficient. The beneficial effect is that the significant response coefficient comprehensively considers the response degree of spectral wavelength to the change of pour point depressant quality and the independence of spectral wavelength. Based on the significant response coefficient, the characteristic wavelength of pour point depressant can be better selected; finally, based on the significant response coefficient, combined with the CARS algorithm, the characteristic wavelength of each iteration of the pour point depressant sample is selected, which can effectively select spectral wavelengths that have a strong correlation with the prediction of pour point depressant quality, and gradually eliminate redundant or unimportant wavelengths, thereby improving the accuracy and computational efficiency of subsequent regression models; a pour point depressant quality regression model is constructed, and the quality detection of the pour point depressant sample to be tested is completed based on the pour point depressant quality regression model, thereby improving the accuracy and reliability of the detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flowchart of a method for spectral detection of pour point depressant quality provided in one embodiment of the present application;

[0035] Figure 2 A flow chart for screening characteristic wavelengths of a pour point depressant sample provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0036] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0038] It should also be noted that the terms "first" and "second" in this application and the accompanying drawings are used to distinguish similar objects, rather than to describe a specific order or sequence. The method disclosed in the embodiments of the present application or the method shown in the flow chart includes one or more steps for implementing the method. Without departing from the scope of protection of the present application, the execution order of multiple steps can be interchanged with each other, and some steps can also be deleted.

[0039] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0040] The specific scheme of a spectral detection method and system for the quality of a pour point depressant provided by the present application is described in detail below with reference to the accompanying drawings.

[0041] See also Figure 1 , which shows a flow chart of a method for spectral detection of pour point depressant quality provided by an embodiment of the present application, the method comprising the following steps:

[0042] The first step is to obtain the spectral data and mass data of each pour point depressant sample.

[0043] First, a preset number of pour point depressant samples of different masses are prepared, and the quality of the pour point depressant samples is evaluated by measuring the degree to which the diesel freezing point is reduced by different pour point depressants, that is, the degree of reduction of the diesel freezing point is used as the quality data of the pour point depressant in this application; the preset number in this embodiment is 50, and the implementer can adjust it according to the actual situation; in this embodiment, the spectral data of the pour point depressant samples are collected by Antaris II Fourier near-infrared spectrometer, and the collection range is set to , the spectral resolution is , select different positions to collect multiple spectral data for each pour point depressant sample. The average spectrum is used as the spectrum data of each pour point depressant sample. Finally, the collected spectrum data and the mass data of different pour point depressant samples are normalized.

[0044] So far, the spectral data of the pour point depressant sample after pretreatment is obtained.

[0045] The second step is to obtain the feature subset of the spectral data of each pour point depressant sample at each iteration of the CARS algorithm; for each iteration, based on the distribution characteristics of the spectral data of all pour point depressants at each spectral wavelength in the feature subset and the response degree of the spectral data to the change of mass data, the mass sensitivity of each spectral wavelength in the feature subset is obtained.

[0046] Since the near-infrared spectrum has a high spectral resolution, the spectral data of the collected pour point depressant often contains a high degree of data redundancy, that is, the spectral information of adjacent wavelengths has a high consistency. However, when the quality of the pour point depressant is detected based on spectral technology, in order to improve the generalization ability of the model and avoid overfitting, it is usually necessary to select the most representative characteristic wavelength from the original wavelength to better reflect the quality information of the pour point depressant. When selecting characteristic wavelengths, an important principle is that the greater the difference between wavelengths, the better. However, the traditional CARS algorithm uses the least squares regression weight to judge the importance of wavelengths, while ignoring the differences between characteristic wavelengths, resulting in poor wavelength selection, and may lose important wavelengths, affecting the quality detection results of the pour point depressant.

[0047] The present application adopts the CARS algorithm to select the characteristic wavelength according to the spectral data of each pour point depressant. In the first iteration, the spectral wavelengths in the feature subset are analyzed to optimize the feature wavelength selection process in the CARS algorithm. The feature subset represents the first feature wavelength selection process in the CARS algorithm. At the iteration, the spectral set composed of the remaining spectral wavelengths, that is, the set composed of all remaining spectral wavelengths except the characteristic wavelength at each iteration of the CARS algorithm, is recorded as the characteristic subset.

[0048] When the quality of the pour point depressant changes, the content of its internal chemical components will inevitably change, which will cause the collected spectral data of the pour point depressant to change. In addition, since different wavelengths have different sensitivities to the chemical components in the pour point depressant, the response changes at different wavelengths vary.

[0049] All spectral data of all pour point depressant samples at each spectral wavelength in the feature subset of each iteration are used as the horizontal coordinates, and the mass data of all pour point depressant samples are used as the vertical coordinates for straight line fitting to obtain the slope of the fitted straight line; the proportion of the straight line slope of each spectral wavelength in the feature subset among all spectral wavelengths is calculated; the first correlation between the mass data of all pour point depressant samples and all spectral data of each spectral wavelength in the feature subset at each iteration is analyzed; the straight line slope proportion is merged with the first correlation to obtain the mass sensitivity of each spectral wavelength in the feature subset.

[0050] Specifically, The first iteration in the feature subset For example, the mass data sequence of 50 samples of pour point depressant is composed of the mass data from small to large. ; The spectral data of 50 pour point depressant samples corresponding to the spectral wavelengths are organized into the order of the pour point depressant samples corresponding to the elements of the mass data sequence. Spectral data sequence of spectral wavelengths , the implementer can adjust the sorting order by himself; take the spectral data as the vertical axis and the quality data as the horizontal axis, and sort the spectral data sequence The quality data sequence is fitted with a least squares linear fitting algorithm to obtain the slope of the fitted line. , the slope of the fitted line The larger the The more sensitive the spectral wavelength is to the quality change of the pour point depressant, the The more likely the spectral wavelength is to be the characteristic wavelength for the quality detection of the spectral quality of the pour point depressant, the better. Among them, the least squares linear fitting algorithm belongs to the known technology, and the present application will not elaborate on its implementation process in detail.

[0051] In addition, in the spectral analysis of the quality detection of the pour point depressant, when the quality of the pour point depressant changes, the response of different spectral wavelengths to the quality change shows significant differences. Therefore, the response degree of the spectral data of the same spectral wavelength to the quality data change is further analyzed: the calculation of the first Spectral data sequence of spectral wavelengths Pour Point Depressant Quality Data Series The Pearson correlation coefficient between the sequences is used to describe the correlation characteristics between the sequences. The larger the Pearson correlation coefficient, the The closer the correlation between the wavelength of the spectrum and the quality change of the pour point depressant, the The more likely a wavelength is a characteristic wavelength, the more likely it is. The calculation of the Pearson correlation coefficient belongs to a well-known technique, and the present invention will not further elaborate on its implementation process.

[0052] Finally, calculate the At the iteration Mass sensitivity at each wavelength : ; In the formula , Respectively represent At the iteration , The slope of the spectrum data at each wavelength; Indicates The number of spectral wavelengths in the feature subset at the iteration; Indicates At the iteration Pearson correlation coefficient between the spectral data series of the spectral wavelengths and the quality data series of the pour point depressant; Indicates At the iteration The mass sensitivity of each spectral wavelength.

[0053] It should be understood that when the CARS algorithm is performed At the iteration, the When the first spectral wavelength is the characteristic wavelength of quality detection, the change of the internal chemical composition of the pour point depressant will cause the spectral data of the characteristic wavelength to show a large change, which will make the calculated The slope of the spectral data of the spectral wavelength The larger the The larger the value, the higher the Pearson correlation coefficient between the spectral data of the spectral wavelength and the quality change of the pour point depressant The larger the The closer the correlation between the wavelength of the spectrum and the quality change of the pour point depressant, the The more wavelengths, the better it can reflect the quality change of the pour point depressant, thus making the calculated Spectral wavelength quality sensitivity The larger the The smaller the wavelength, the more likely it is to be the characteristic wavelength for spectral detection of the quality change of the pour point depressant.

[0054] The third step: Analyze the numerical differences and similarities between each spectral wavelength in the feature subset and its adjacent wavelengths in each iteration of the CARS algorithm, obtain the independent coefficient of each spectral wavelength in the feature subset, and combine it with the mass sensitivity of each spectral wavelength to obtain the significant response coefficient of each spectral wavelength.

[0055] Since there is often a high degree of redundancy between adjacent wavelengths in spectral data, the traditional CARS algorithm ignores the independence of characteristic wavelengths during the characteristic wavelength selection process, which may cause the selected spectral characteristic wavelengths to aggregate. There is still a certain amount of information redundancy, and the potential spectral characteristic wavelengths are ignored, which affects the accuracy of spectral detection of pour point depressant quality.

[0056] A classification algorithm is used to obtain all neighboring wavelengths of each spectral wavelength in the feature subset; the difference between each spectral wavelength and each of its neighboring wavelengths is calculated to obtain the second correlation between the spectral data sequence of each spectral wavelength and the spectral data sequence of each of its neighboring wavelengths; the negative correlation mapping result of the second correlation and the fusion result of the difference are calculated; the cumulative sum of the fusion results of each spectral wavelength and all of its neighboring wavelengths is used as the independent coefficient of each spectral wavelength in the feature subset.

[0057] Specifically, the K nearest neighbor algorithm is used to process all spectral wavelengths in the feature subset. In this embodiment, the input parameters are The value is 8, and the K nearest neighbor wavelengths of each spectral wavelength in the feature subset are obtained. The K nearest neighbor algorithm is a widely used classification method, which is a well-known technology and will not be described in detail in this application. Calculate the spectral distance between each spectral wavelength and each of its nearest neighbor spectral wavelengths , where the spectral distance is expressed by the absolute value of the difference between wavelengths; the spectral data set for each spectral wavelength is calculated Its K nearest neighbor wavelength spectrum data set The Pearson correlation coefficient between them; the result of the numerical fusion is specifically the ratio of the difference to the second correlation. It should be noted that in order to prevent the denominator from being 0, a preset value needs to be added to the denominator. The value of this embodiment is 0.1; it should be further noted that the first The larger the spectral distance between a spectral wavelength and its K nearest neighbor wavelengths, the smaller the second correlation between the spectral data sequences, indicating that the The more independent the spectral wavelengths are, the less redundancy they have with their neighboring wavelengths.

[0058] Mass sensitivity is used to reflect the degree of response of different spectral wavelengths to the quality change of the pour point depressant. The spectral wavelength with higher mass sensitivity is more likely to contain information directly related to the quality of the pour point depressant; the independence coefficient can measure the independence of the spectral wavelengths in the characteristic subset. The wavelength with a larger independence coefficient is more likely to contain unique information related to the quality of the pour point depressant; however, in the process of characteristic wavelength selection, only considering mass sensitivity or independence coefficient is not enough to accurately select the characteristic wavelength for the quality detection of the pour point depressant. The wavelength with high mass sensitivity may lack uniqueness due to the high similarity with the adjacent wavelengths in the characteristic subset, which makes the selected characteristic wavelength have certain data redundancy; and the wavelength with a large independence coefficient may not directly reflect the quality change of the pour point depressant.

[0059] Based on the above analysis, the mass sensitivity and independent coefficient of each spectral wavelength in the feature subset at each iteration are combined to obtain the significant response coefficient of each spectral wavelength: ; In the formula, Indicates At the iteration Mass sensitivity for each spectral wavelength; Indicates At the iteration Independent coefficients for each wavelength; Indicates At the first iteration, the The significant response coefficient of each spectral wavelength; , Respectively represent the first preset value and the second preset value, and are used to represent the The weights of the mass sensitivity of the spectral wavelengths and the weights of the independent coefficients are , The average value is 0.5, wherein the sum of the first preset value and the second preset value is equal to 1, and the first preset value and the second preset value are both greater than or equal to 0.

[0060] The fourth step: based on the significant response coefficient of each spectral wavelength in the feature subset at each iteration and combined with the CARS algorithm, the characteristic wavelengths of the pour point depressant samples are screened, the regression model is trained, and the quality data of the pour point depressant samples to be tested are obtained.

[0061] The significant response coefficient measures the influence of each spectral wavelength on the quality change of the pour point depressant, while taking into account the independence of each spectral wavelength in the feature subset. The probability that a spectral wavelength is a characteristic wavelength. Therefore, based on the significant response coefficient of each spectral wavelength in the characteristic subset at each iteration and combined with the CARS algorithm, the characteristic wavelength of the pour point depressant sample is screened. The specific process is as follows:

[0062] First, based on the CARS algorithm The partial least squares regression (PLS) model is constructed for all spectral wavelengths in the feature subset at the iteration to obtain the regression coefficient of each spectral wavelength in the feature subset. The significant response coefficient of each spectral wavelength in the feature subset at each iteration is normalized and multiplied with the regression coefficient of the corresponding feature wavelength to obtain the correction coefficient of each spectral wavelength in the feature subset at each iteration.

[0063] All spectral wavelengths in the characteristic subset are sorted in descending order of the correction coefficient; an exponential decay function is used to gradually reduce the number of characteristic wavelengths, and characteristic wavelength selection is completed in the process of quality detection of pour point depressant by retaining characteristic wavelengths with larger correction coefficients to obtain characteristic spectral data. The PLS model and the exponential decay function are both well-known technologies.

[0064] Among them, the flowchart for screening the characteristic wavelength of the pour point depressant sample is as follows: Figure 2 shown.

[0065] The random forest regression model was used to train the characteristic wavelengths and corresponding mass data of the pour point depressant samples: the characteristic spectral data and corresponding mass data of 70% of the pour point depressant samples were used to train the pour point depressant mass regression model, and the characteristic spectral data and corresponding mass data of 30% of the pour point depressant samples were used to verify the effect of the pour point depressant mass regression model.

[0066] The characteristic spectrum data of the pour point depressant sample to be tested is used as the input of the pour point depressant quality regression model to obtain the predicted quality data of the pour point depressant sample to be tested.

[0067] Based on the same inventive concept as the above method, an embodiment of the present application also provides a spectral detection system for the quality of a pour point depressant, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of any one of the above-mentioned spectral detection methods for the quality of a pour point depressant are implemented.

[0068] In summary, the spectral data of the pour point depressant sample is obtained. First, the mass sensitivity is constructed based on the response degree of different spectral wavelengths to the change of the pour point depressant mass. The beneficial effect is that the mass sensitivity can reflect the response degree of different wavelengths to the change of the pour point depressant mass based on the mass sensitivity by considering the correlation between the spectral data of wavelengths under different masses of the pour point depressant and the mass of the pour point depressant, as well as the degree of change of the spectral data at the same spectral wavelength under different masses of the pour point depressant; secondly, the independent coefficient is constructed based on the similarity between each wavelength and its adjacent wavelengths. The beneficial effect is that the independent coefficient takes into account the difference between adjacent characteristic wavelengths and the difference between corresponding spectral data, and the independence of each wavelength can be reflected based on the independent coefficient; in addition , a significant response coefficient is constructed based on quality sensitivity and independence coefficient. The beneficial effect is that the significant response coefficient comprehensively considers the response degree of spectral wavelength to the change of pour point depressant quality and the independence of spectral wavelength. Based on the significant response coefficient, the characteristic wavelength of pour point depressant can be better selected; finally, based on the significant response coefficient, combined with the CARS algorithm, the characteristic wavelength of each iteration of the pour point depressant sample is selected, which can effectively select spectral wavelengths that have a strong correlation with the prediction of pour point depressant quality, and gradually eliminate redundant or unimportant wavelengths, thereby improving the accuracy and computational efficiency of subsequent regression models; a pour point depressant quality regression model is constructed, and the quality detection of the pour point depressant sample to be tested is completed based on the pour point depressant quality regression model, thereby improving the accuracy and reliability of the detection.

[0069] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two continuous operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0070] For those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the basic features of the present application. Therefore, no matter from which point of view, the above embodiments of the present application should be regarded as exemplary and non-restrictive; the technical solutions recorded in the above embodiments are modified, or some of the technical features are replaced by equivalents, which does not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A spectral detection method for the quality of a pour point depressant, characterized in that: The method comprises the following steps: Obtaining spectral data and mass data of each pour point depressant sample; Obtain the characteristic subset of the spectral data of each pour point depressant sample at each iteration of the CARS algorithm; for each iteration, based on the distribution characteristics of the spectral data of each spectral wavelength of all pour point depressants in the characteristic subset and the response degree of the spectral data to the change of the mass data, obtain the mass sensitivity of each spectral wavelength in the characteristic subset; The numerical differences and similarities between each spectral wavelength in the feature subset and its adjacent wavelengths were analyzed in each iteration of the CARS algorithm to obtain the independent coefficient of each spectral wavelength in the feature subset, which was combined with the mass sensitivity of each spectral wavelength to obtain the significant response coefficient of each spectral wavelength. Based on the significant response coefficient of each spectral wavelength in the feature subset at each iteration and combined with the CARS algorithm, the characteristic wavelengths of the pour point depressant samples are screened, the regression model is trained, and the quality data of the pour point depressant samples to be tested are obtained; The mass sensitivity of each spectral wavelength in the characteristic subset is obtained as follows: Perform straight line fitting on the mass data of all pour point depressant samples and the spectral data of each spectral wavelength in the characteristic subset at each iteration to obtain the slope of the fitted straight line; Calculate the proportion of the slope of the fitted straight line of each spectral wavelength in all spectral wavelengths in the feature subset; The first correlation between the mass data of all pour point depressant samples and all spectral data of each spectral wavelength in the feature subset at each iteration is analyzed; the straight line slope proportion is merged with the first correlation to obtain the mass sensitivity of each spectral wavelength in the feature subset.

2. The method for spectral detection of pour point depressant quality according to claim 1, characterized in that: The characteristic subset is specifically a set consisting of all remaining spectral wavelengths except the characteristic wavelength at each iteration of the CARS algorithm.

3. The spectral detection method for the quality of a pour point depressant according to claim 1, characterized in that: The specific process of performing straight line fitting is as follows: All spectral data of each spectral wavelength in the feature subset of all pour point depressant samples at each iteration are used as the abscissa, and the mass data of all pour point depressant samples are used as the ordinate, and a straight line fitting is performed.

4. The spectral detection method for the quality of a pour point depressant according to claim 1, characterized in that: The process of obtaining the independent coefficient of each spectral wavelength in the feature subset is: The sequence consisting of the spectral data of all the pour point depressant samples at each spectral wavelength is recorded as a spectral data sequence; A classification algorithm is used to obtain all neighboring wavelengths of each spectral wavelength in the feature subset; Calculate the difference between each spectral wavelength and each of its neighboring wavelengths to obtain a second correlation between the spectral data sequence of each spectral wavelength and the spectral data sequence of each of its neighboring wavelengths; Calculating a fusion result of a negative correlation mapping result of the second correlation and the difference; The cumulative sum of the fusion results of each spectral wavelength and all its neighboring wavelengths is used as the independent coefficient of each spectral wavelength in the feature subset.

5. The spectral detection method for the quality of a pour point depressant according to claim 1, characterized in that: The formula for obtaining the significant response coefficient of each spectral wavelength is: ; In the formula, Indicates At the iteration Mass sensitivity for each spectral wavelength; Indicates At the iteration Independent coefficients for each wavelength; Indicates At the first iteration, the The significant response coefficient of each spectral wavelength; , Respectively represent a first preset value and a second preset value, wherein the sum of the first preset value and the second preset value is equal to 1, and the first preset value and the second preset value are both greater than or equal to 0.

6. The method for spectral detection of pour point depressant quality according to claim 1, characterized in that: The characteristic wavelengths of the pour point depressant samples are screened based on the significant response coefficient of each spectral wavelength in the characteristic subset at each iteration in combination with the CARS algorithm, specifically: A regression model is constructed based on all spectral wavelengths in the feature subset at each iteration of the CARS algorithm to obtain the regression coefficient of each spectral wavelength in the feature subset; The significant response coefficient of each spectral wavelength in the feature subset at each iteration is merged with the regression coefficient to obtain the correction coefficient of each spectral wavelength in the feature subset at each iteration; The spectral wavelength is screened according to the correction coefficient using an exponential decay function to obtain a characteristic wavelength and corresponding characteristic spectral data at each iteration.

7. A spectral detection method for the quality of a pour point depressant according to claim 6, characterized in that: The correction coefficient of each spectral wavelength in the feature subset obtained at each iteration is specifically: The significant response coefficient of each spectral wavelength in the feature subset at each iteration is normalized and multiplied by the regression coefficient of the corresponding feature wavelength to obtain the correction coefficient of each spectral wavelength in the feature subset at each iteration.

8. The method for spectral detection of pour point depressant quality according to claim 1, characterized in that: The process of training the regression model and obtaining the quality data of the pour point depressant sample to be tested is specifically as follows: The spectral data corresponding to the wavelength obtained by screening the pour point depressant samples are combined with the quality data of the pour point depressant samples for training to obtain a pour point depressant quality regression model; The characteristic spectrum data of the pour point depressant sample to be tested is subjected to a pour point depressant mass regression model to obtain predicted mass data of the pour point depressant sample to be tested.

9. A spectral detection system for the quality of a pour point depressant, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images