A model construction method based on plug flow reactor

Abstract

The present invention relates to the field of reactor technology, and in particular to a model construction method based on a plug flow reactor, including: obtaining storage data information of the plug flow reactor of each data site; determining the data unit state according to the minimum matching coefficient, and determining the data unit desensitization mode according to the data unit state as desensitizing for data names or desensitizing for data samples; when desensitizing for data names, replacing the data names with hash characters; when desensitizing for data samples, determining the associated desensitization method as associated data desensitization or image data desensitization according to the reference matching coefficient; under preset conditions, selecting a data set according to the characteristic degree. The present invention can ensure the data security of model training samples and reduce data security risks.

Description

Technical Field

[0001] The invention relates to the technical field of reactors, and in particular to a model building method based on a plug flow reactor. Background Art

[0002] The plug flow reactor model is generally implemented by combining a plug flow reactor with a kinetic equation. The number and richness of the model training samples are of great significance to the model building effect. In order to improve the reliability of the model, the sample data provided by a single factory is currently difficult to meet the needs of model building. Therefore, sample data is often obtained through distributed data sources and the model is built through a model learning center. However, data security and privacy protection face multiple risks during transmission, including data leakage, illegal access or abuse, which poses serious security risks to enterprises. Therefore, how to ensure the secure transmission of model training sample data and reduce data security risks is a technical problem that technicians in this field need to solve urgently.

[0003] Chinese patent publication number CN115114845A discloses a method and device for modeling a plug flow reactor for a digital twin system, which is applied to a dynamic simulation system and includes a reactor mechanism model and an artificial neural network model; the method includes: calculation of the reactor intermediate value, calculation of the intermediate value through the reactor mechanism model, and transmission of the calculated intermediate value to the artificial neural network model; correction of the reactor intermediate value, correction of the reactor intermediate value through the artificial neural network model, and feedback of the corrected intermediate value to the reactor mechanism model, the reactor mechanism model performs calculations in combination with the corrected intermediate value, and then outputs the calculation results. It can be seen that the above technical solution has the following problems: the data security of the model training samples is not taken into account, resulting in greater data security risks. Summary of the invention

[0004] To this end, the present invention provides a model construction method based on a plug flow reactor to overcome the problem that the data security of model training samples is not taken into consideration in the prior art, resulting in greater data security risks.

[0005] To achieve the above object, the present invention provides a model construction method based on a plug flow reactor, comprising:

[0006] Acquire storage data information of the plug flow reactor at each data station, where the storage data information includes a plurality of data units consisting of a data name and a data sample;

[0007] Determine the data unit state according to the minimum matching coefficient, and determine the data unit desensitization method according to the data unit state, whether to desensitize the data name or the data sample;

[0008] When desensitizing data names, replace the data names with hash characters;

[0009] When desensitizing a data sample, the associated desensitization method is determined to be associated data desensitization or image data desensitization according to the reference matching coefficient;

[0010] In the associated data desensitization, the first reference sample and the second reference sample are determined according to the sample average value of the reference second-category data unit corresponding to the data sample, each data value in the first reference sample is replaced one by one with the corresponding character in the character library, and each data value in the second reference sample is replaced one by one with the first character and the second character;

[0011] Image data desensitization is to replace the data sample with a data combination of the number of feature points, feature segment length, and feature area;

[0012] Under the preset conditions, the data set is selected according to the characteristic degree;

[0013] The preset condition is that data name desensitization and data sample desensitization are completed.

[0014] Further, the data unit state is determined according to the minimum matching coefficient, the data unit state comprising a first data unit state in which the minimum matching coefficient is greater than a preset minimum matching coefficient, and a second data unit state in which the minimum matching coefficient is less than or equal to the preset minimum matching coefficient;

[0015] The minimum matching coefficient is the minimum value of the matching coefficients corresponding to the data units with the same data name in a single data site.

[0016] Further, determining a data unit desensitization method according to the data unit state;

[0017] If the data unit state is the first data unit state, the data unit desensitization method is to desensitize the data name;

[0018] If the data unit state is the second data unit state, the data unit desensitization method is to desensitize the data sample.

[0019] Furthermore, desensitizing the data name includes:

[0020] For a single data name, different hash functions are selected for processing according to the data discreteness of all the data units of the first category corresponding to the data name. For a single data unit of the first category, all the data in the data sample of the data unit of the first category are input into the hash function to generate the corresponding hash character, and the data name corresponding to the data unit of the first category is replaced with the hash character;

[0021] The type of data unit is a data unit whose data unit state is a first data unit state.

[0022] Furthermore, when desensitizing a data sample, for a single data sample, a data unit corresponding to the data sample is recorded as a target data unit, and an associated desensitization method is determined according to a reference matching coefficient corresponding to the target data unit;

[0023] If the reference matching coefficient is greater than or equal to the preset reference matching coefficient, the associated desensitization method is associated data desensitization;

[0024] If the reference matching coefficient is less than the preset reference matching coefficient, the associated desensitization method is image data desensitization;

[0025] The reference matching coefficient is a minimum value of matching coefficients corresponding to the target data unit and the reference data unit, and the reference data unit is a data unit with a data name different from that corresponding to the target data unit.

[0026] Furthermore, the associated data desensitization includes:

[0027] The second-category data units with the same data name are recorded as reference second-category data units, the data sample corresponding to the reference second-category data unit with the smallest sample average value is recorded as the first reference sample, and the data samples corresponding to other reference second-category data units are recorded as the second reference samples, and each data value in the first reference sample is replaced one by one with the corresponding character in the character library;

[0028] Record each data value in the first reference sample as the first data value, record each data value in the second reference sample as the second data value, record a single second data value as the second target data value, record the first data value in the same order as the second target data value as the first target data value, record the character corresponding to the first target data value in the character library as the first character, wherein the first character is recorded as A, the second character is recorded as B, and the second target data value is replaced by A+B or AB according to the first data value;

[0029] The second character is a character in the character library corresponding to a reference absolute value, and the reference absolute value is an absolute value of a difference between the first data value and the second data value;

[0030] The second type of data unit is a data unit whose data unit state is the second data unit state and whose matching coefficient is greater than or equal to the preset matching coefficient.

[0031] Furthermore, image data desensitization includes:

[0032] A line graph is drawn with the length position of the plug flow reactor as the horizontal coordinate and the data text corresponding to a single three-type data unit as the vertical coordinate, the number of characteristic points, the length of characteristic paragraphs and the characteristic area of ​​the line graph are calculated, and the data samples corresponding to the three-type data units are replaced with the data combination of the number of characteristic points, the length of characteristic paragraphs and the characteristic area;

[0033] The three types of data units are data units whose data unit state is the second data unit state and whose matching coefficients are smaller than the preset matching coefficients.

[0034] Further, selecting a data set according to the characteristic degree includes: recording data samples with the same name or the same hash function corresponding to the hash value as reference data samples, arranging the reference data samples in order from small to large according to the characteristic degree for a single reference data sample to obtain a reference sequence, dividing the reference sequence into n equal parts, selecting data at each equal part point and the initial position and the end position of the reference sequence to establish a data set as simulation data of the horizontal flow reactor model;

[0035] The feature degrees include name desensitization feature degrees, associated data desensitization feature degrees and image data desensitization feature degrees.

[0036] Furthermore, the calculation formula of the name desensitization characteristic τ1 is:

[0037]

[0038] Among them, α is the average value of a single type of desensitized data, β is the standard deviation of a single type of desensitized data, γ is the range of a single type of desensitized data, ω1 is the first coefficient, ω2 is the second coefficient, and ω3 is the third coefficient;

[0039] The type of desensitized data is data samples after desensitization is performed on a type of data unit.

[0040] Furthermore, the calculation formula of the desensitization characteristic degree τ2 of the associated data is:

[0041]

[0042] Among them, p is the maximum number of identical characters in a single second-category anonymized data, d is the character type of a single second-category anonymized data, ρ1 is the first weight coefficient, and ρ2 is the second weight coefficient;

[0043] The calculation formula of image data desensitization feature τ3 is:

[0044]

[0045] Among them, R is the total number of features of a single three-category desensitized data, is the relative abundance of the ith feature;

[0046] The second type of desensitized data is a data sample after the associated data of the second type of desensitized data unit is desensitized, and the third type of desensitized data is a data sample after the image data of the third type of desensitized data unit is desensitized.

[0047] Compared with the prior art, the beneficial effect of the present invention lies in that, in the technical solution of the present invention, the data unit state is determined according to the minimum matching coefficient, and the degree of correlation between the current data units is effectively reflected through the minimum matching coefficient, and then different data unit desensitization methods are selected according to actual conditions, so that the selection of data unit desensitization methods is more in line with actual working scenarios. The present invention reduces the risk of data leakage during the transmission process of each data site.

[0048] Furthermore, in the present invention, hash characters of different lengths are generated according to data discreteness when desensitizing data names. Data discreteness can be used to effectively distinguish data samples with different data names, and then different hash functions are selected for processing according to the data discreteness, thereby ensuring that the hash character lengths corresponding to different data names are different, thereby reducing data security risks while ensuring that the same data name remains consistent after desensitization, and then when the data set is selected according to the feature degree in the subsequent process, the data samples with the same data name can be accurately identified.

[0049] Furthermore, when desensitizing data samples in the present invention, an associated desensitization method is determined based on a reference matching coefficient, and the degree of association of data units is reflected by the reference matching coefficient, and different associated desensitization methods are adaptively selected, so that the selection of associated desensitization methods is more in line with actual working scenarios, so that the desensitized data can maintain high data quality and availability while meeting the privacy protection requirements, avoiding the problem of indistinguishable characteristic degrees of the desensitized data, and thus facilitating the selection of data sets based on characteristic degrees.

[0050] Furthermore, in the present invention, data sets are selected according to characteristic degrees, the characteristics of the data are reflected by the characteristic degrees, and data sets with different characteristic degrees are selected in a targeted manner, so that the selection of data sets is more in line with the user's usage requirements, avoiding the problems of poor data richness and poor practicality caused by a single data characteristic degree, and further improving the accuracy of plug flow reactor model construction and the applicability of the plug flow reactor model. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of a model building method based on a plug flow reactor of the present invention;

[0052] Figure 2 A flow chart of determining the state of a data unit according to a minimum matching coefficient according to the present invention;

[0053] Figure 3 This is a flow chart of the present invention for determining a data unit desensitization method according to a data unit state;

[0054] Figure 4 The present invention is a flow chart of a method for determining an associated desensitization method based on a reference matching coefficient. DETAILED DESCRIPTION

[0055] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0056] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0057] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0058] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0059] See also Figures 1 to 4 As shown, the present invention provides a model construction method based on a plug flow reactor, comprising:

[0060] Acquire storage data information of the plug flow reactor at each data station, where the storage data information includes a plurality of data units consisting of a data name and a data sample;

[0061] Determine the data unit state according to the minimum matching coefficient, and determine the data unit desensitization method according to the data unit state, whether to desensitize the data name or the data sample;

[0062] When desensitizing data names, replace the data names with hash characters;

[0063] When desensitizing a data sample, the associated desensitization method is determined to be associated data desensitization or image data desensitization according to the reference matching coefficient;

[0064] In the associated data desensitization, the first reference sample and the second reference sample are determined according to the sample average value of the reference second-category data unit corresponding to the data sample, each data value in the first reference sample is replaced one by one with the corresponding character in the character library, and each data value in the second reference sample is replaced one by one with the first character and the second character;

[0065] Image data desensitization is to replace the data sample with a data combination of the number of feature points, feature segment length, and feature area;

[0066] Under the preset conditions, the data set is selected according to the characteristic degree;

[0067] The preset condition is that data name desensitization and data sample desensitization are completed.

[0068] The application scenario of the present invention is the data processing of training samples in the process of building a plug flow reactor model. The data required for building the plug flow reactor model comes from multiple data sites. The data site contains several data sets, and the data set contains several data units. A single data unit includes a single data name and a data sample corresponding to the data name. The data sample is a monitoring parameter corresponding to a single site monitored in real time with the use time of a single data name during the historical use of the plug flow reactor. The monitoring parameters include but are not limited to product concentration, reaction temperature and reaction rate. The site is set by the user, and a site setting method is provided. The inlet of the plug flow reactor is recorded as a site, and every 10 cm is recorded as a site in the order from the inlet to the outlet of the plug flow reactor, that is, the monitoring parameters are recorded once every 10 cm.

[0069] Specifically, the data unit state is determined according to the minimum matching coefficient, the data unit state includes a first data unit state in which the minimum matching coefficient is greater than a preset minimum matching coefficient, and a second data unit state in which the minimum matching coefficient is less than or equal to the preset minimum matching coefficient;

[0070] The minimum matching coefficient is the minimum value of the matching coefficients corresponding to the data units with the same data name in a single data site.

[0071] For any two data units with the same data name, the calculation formula of the matching coefficient σ corresponding to the two data units is:

[0072]

[0073] The starting data of the data sample corresponding to a single data unit is taken as the starting point, and an interval point is set every 5 seconds in time sequence, and the starting point and all interval points are recorded as time points. If the number of time points of two data samples is different, data preprocessing is required to make the number of time points of the two data samples the same, and the data preprocessing includes: comparing the number of time points of the two data samples, recording the number of time points corresponding to the data sample with a smaller number of time points as the reference number of time points, and in the data sample with a larger number of time points, starting from the first time point, retaining the time points equal to the reference number of time points, deleting the data values ​​corresponding to all time points exceeding the reference number of time points, and recording the data sample after data preprocessing as the preprocessed data sample;

[0074] In the calculation formula of the matching coefficient σ, m is the number of time points in a single preprocessed data sample; and are the values ​​of the kth data point in the two preprocessed data samples, respectively. for The corresponding mean value of the preprocessed data samples, for The average value of the corresponding preprocessed data samples, k = 1, 2, 3, ..., m; The corresponding average value of the preprocessed data samples The calculation formula is: , The corresponding average value of the preprocessed data samples The calculation formula is: .

[0075] The value of the preset minimum matching coefficient can be set according to the actual application scenario. It can be understood that the larger the value of the preset matching coefficient is, the higher the correlation degree of the data samples corresponding to the data units for desensitizing the data names is. A value of the preset matching coefficient is provided, and the value of the preset matching coefficient is 0.5.

[0076] Specifically, the data unit desensitization method is determined according to the data unit status;

[0077] If the data unit state is the first data unit state, the data unit desensitization method is to desensitize the data name;

[0078] If the data unit state is the second data unit state, the data unit desensitization method is to desensitize the data sample.

[0079] Specifically, desensitizing data names includes:

[0080] For a single data name, different hash functions are selected for processing according to the data discreteness of all the data units of the first category corresponding to the data name. For a single data unit of the first category, all the data in the data sample of the data unit of the first category are input into the hash function to generate the corresponding hash character, and the data name corresponding to the data unit of the first category is replaced with the hash character;

[0081] The type of data unit is a data unit whose data unit state is a first data unit state.

[0082] When the present invention is processed through a hash function, different hash functions are selected for processing according to the data discreteness. The hash functions include but are not limited to SHA-1, SHA-256 and SHA-384. It can be understood that the lengths of hash characters generated by different hash functions are different, among which the SHA-1 hash character length is 20 bytes, the SHA-256 hash character length is 32 bytes, and the SHA-384 hash character length is 48 bytes. The relationship between the hash character length corresponding to the selected hash function and the data discreteness is a positive correlation. In the present invention, different hash functions are selected to make data units with different discreteness have hash characters of different lengths. While meeting the confidentiality requirements, the data units with the same data name can be determined according to the hash character length when the data set is selected according to the characteristic degree.

[0083] The hash character is a value generated by a hash function, and the hash character length is the number of binary bits occupied by the hash character generated by the hash function, in bytes; the calculation formula of the data dispersion δ is:

[0084]

[0085] Among them, Q is the number of data points contained in the data sample with the same data name, , sort the data samples with the same data name in descending order according to the sample average value, and record the data samples with the same data name after sorting as the reference sample combination, is the data value corresponding to the ath data point in the reference sample combination, and the sample average is recorded as , , is the data value corresponding to the cth data point in a data sample with the same single name, and q is the number of data points in a data sample with the same single name.

[0086] Specifically, when desensitizing a data sample, for a single data sample, the data unit corresponding to the data sample is recorded as the target data unit, and the associated desensitization method is determined according to the reference matching coefficient corresponding to the target data unit;

[0087] If the reference matching coefficient is greater than or equal to the preset reference matching coefficient, the associated desensitization method is associated data desensitization;

[0088] If the reference matching coefficient is less than the preset reference matching coefficient, the associated desensitization method is image data desensitization;

[0089] The reference matching coefficient is a minimum value of matching coefficients corresponding to the target data unit and the reference data unit, and the reference data unit is a data unit with a data name different from that corresponding to the target data unit.

[0090] Among them, the value of the preset reference matching coefficient can be set by the user according to the actual application scenario. It can be understood that the larger the value of the preset reference matching coefficient is, the higher the degree of association of the data samples corresponding to the data units for associated data desensitization is. A value of the preset reference matching coefficient is provided, and the value of the preset reference matching coefficient is 0.6.

[0091] Specifically, the associated data masking includes:

[0092] The second-category data units with the same data name are recorded as reference second-category data units, the data sample corresponding to the reference second-category data unit with the smallest sample average value is recorded as the first reference sample, and the data samples corresponding to other reference second-category data units are recorded as the second reference samples, and each data value in the first reference sample is replaced one by one with the corresponding character in the character library;

[0093] Record each data value in the first reference sample as the first data value, record each data value in the second reference sample as the second data value, record a single second data value as the second target data value, record the first data value in the same order as the second target data value as the first target data value, record the character corresponding to the first target data value in the character library as the first character, wherein the first character is recorded as A, the second character is recorded as B, and the second target data value is replaced by A+B or AB according to the first data value;

[0094] The second character is a character in the character library corresponding to a reference absolute value, and the reference absolute value is an absolute value of a difference between the first data value and the second data value;

[0095] The second type of data unit is a data unit whose data unit state is the second data unit state and whose matching coefficient is greater than or equal to the preset matching coefficient.

[0096] The character library of the present invention is used to store characters corresponding to data values ​​and characters corresponding to reference absolute values ​​set by the user, wherein when the data value and the reference absolute value are the same, the characters corresponding to the data value in the character library are different from the characters corresponding to the reference absolute value, and the characters include but are not limited to letters, symbols and graphics, etc.; the sample average value The calculation formula is: .

[0097] The method for confirming the first data value that is in the same order as the second target data value is as follows: if the second target data value is the Mth data value in the second reference sample, then the first data value that is in the same order as the second target data value is the Mth data value in the first reference sample, where M=1, 2, 3, ..., J, where J is the number of data values ​​in the second reference sample.

[0098] If the difference between the first target data value and the second target data value is greater than or equal to 0, the second target data value is replaced by A+B according to the first target data value;

[0099] If the difference between the first target data value and the second target data value is less than 0, the second target data value is replaced with AB according to the first target data value.

[0100] Specifically, image data desensitization includes:

[0101] A line graph is drawn with the length position of the plug flow reactor as the horizontal coordinate and the data text corresponding to a single three-type data unit as the vertical coordinate, the number of characteristic points, the length of characteristic paragraphs and the characteristic area of ​​the line graph are calculated, and the data samples corresponding to the three-type data units are replaced with the data combination of the number of characteristic points, the length of characteristic paragraphs and the characteristic area;

[0102] The three types of data units are data units whose data unit state is the second data unit state and whose matching coefficients are smaller than the preset matching coefficients.

[0103] In the present invention, a line graph is drawn with the length position of the plug flow reactor as the horizontal coordinate and the data text corresponding to the single three-category data unit as the vertical coordinate. It can be understood that the plug flow reactor is an idealized flow model, in which the material flows forward in the same direction at the same speed without axial mixing. Therefore, this flow characteristic makes the reactor length a key parameter for describing the material flow process. In a plug flow reactor, as the material advances, parameters such as product concentration, reaction temperature, and reaction rate will change. Therefore, with the length position of the reactor as the horizontal coordinate and the data text corresponding to the single three-category data unit as the vertical coordinate, the changes of each parameter in the reactor can be intuitively displayed.

[0104] The number of characteristic points is the number of points in the line graph that change from rising to falling or from falling to rising. The characteristic paragraph length is the sum of the lengths of all reference paragraphs, in cm. The line graph is divided into W line segments using each point as a segmentation point. It can be understood that W = the number of all line points in a single line graph - 1. The absolute value of the slope of each line segment is calculated, and the line segment with an absolute slope greater than the preset absolute slope value is recorded as a characteristic line segment. For a single characteristic line segment, the characteristic line segment is recorded as a reference segment. The calculation formula for the absolute value of the slope k is: , where (x1, y1) is the point on the left side of a single broken line paragraph, and (x2, y2) is the point on the right side of a single broken line paragraph; the calculation formula for the length of a single reference paragraph is: L=x2-x1; the value of the preset slope absolute value can be set according to the actual application scenario. The higher the user's demand for distinguishing feature paragraphs, the larger the value of the preset slope absolute value. A preset slope absolute value is provided, and the preset slope absolute value is equal to 0.7.

[0105] The characteristic area is the area of ​​the area enclosed by the line graph and the x-axis, in cm², and the characteristic area is calculated by a numerical integration method such as the trapezoidal method or Simpson's rule;

[0106] The data samples corresponding to the three types of data units are replaced with the data combination of the number of feature points, feature segment length and feature area. For example, when the number of feature points of the line graph corresponding to a single three-type data unit is 5, the feature segment length is 8cm and the feature area is 14cm², the data combination corresponding to the three types of data units is 5, 8, 14.

[0107] Specifically, the data set is selected according to the characteristic degree, including: recording data samples with the same name or the same hash function corresponding to the hash value as reference data samples, arranging the reference data samples in order from small to large according to the characteristic degree for a single reference data sample to obtain a reference sequence, dividing the reference sequence into n equal parts, selecting data at each equal division point and the initial position and the end position of the reference sequence to establish a data set as simulation data of the horizontal flow reactor model;

[0108] The feature degrees include name desensitization feature degrees, associated data desensitization feature degrees and image data desensitization feature degrees.

[0109] The value of n can be set by the user according to the needs. It can be understood that the larger the value of n is, the more simulation data of the plug flow reactor is selected, and the greater the accuracy of the plug flow reactor model construction is. A value of n is provided, n=1000.

[0110] In the present invention, the calculation formulas for the characteristic degrees of different reference data samples are different, wherein, for a single reference data sample, the reference data sample is recorded as a target reference data sample, if each data sample in the target reference data sample is a type of desensitized data, the characteristic degree is a name desensitization characteristic degree, and each type of desensitized data in the target reference data sample is arranged in ascending order according to the name desensitization characteristic degree to obtain a reference sequence;

[0111] If each data sample in the target reference data sample is a second-class desensitized data, the characteristic degree is the associated data desensitization characteristic degree, and each second-class desensitized data in the target reference data sample is arranged in ascending order according to the associated data desensitization characteristic degree to obtain a reference sequence;

[0112] If each data sample in the target reference data sample is three types of desensitized data, the characteristic degree is the image data desensitization characteristic degree, and the three types of desensitized data in the target reference data sample are arranged in ascending order according to the image data desensitization characteristic degree to obtain a reference sequence.

[0113] Specifically, the calculation formula of the name desensitization characteristic τ1 is:

[0114]

[0115] Among them, α is the average value of a single type of desensitized data, β is the standard deviation of a single type of desensitized data, γ is the range of a single type of desensitized data, ω1 is the first coefficient, ω2 is the second coefficient, and ω3 is the third coefficient;

[0116] The type of desensitized data is data samples after desensitization is performed on a type of data unit.

[0117] Among them, the calculation formula for the average value α of a single type of desensitized data is: , the calculation formula for the standard deviation β of a single type of desensitized data is: , the range γ of a single class of desensitized data = the maximum data value in the single class of desensitized data - the minimum data value in the single class of desensitized data, where E is the total data of the single class of desensitized data, is the data value of the sth data in the single class of desensitized data, and provides the values ​​of the first coefficient, the second coefficient and the third coefficient, ω1=0.3, ω2=0.5, ω3=0.2.

[0118] Specifically, the calculation formula of the desensitization characteristic degree τ2 of the associated data is:

[0119]

[0120] Among them, p is the maximum number of identical characters in a single second-category anonymized data, d is the character type of a single second-category anonymized data, ρ1 is the first weight coefficient, and ρ2 is the second weight coefficient;

[0121] The calculation formula of image data desensitization feature τ3 is:

[0122]

[0123] Among them, R is the total number of features of a single three-category desensitized data, is the relative abundance of the ith feature;

[0124] The second type of desensitized data is a data sample after the associated data of the second type of desensitized data unit is desensitized, and the third type of desensitized data is a data sample after the image data of the third type of desensitized data unit is desensitized.

[0125] The method for confirming the maximum number of identical characters in the single second-category desensitized data and the character types of the single second-category desensitized data is to record the identical characters in the single second-category desensitized data as a reference character combination, and record the number of characters in the reference character combination as a reference number, the maximum value of the reference number is the maximum number of identical characters in the single second-category desensitized data, and the number of reference character combinations in the single second-category desensitized data is the character type of the single second-category desensitized data;

[0126] It can be understood that the first weight coefficient and the second weight coefficient are both set according to the importance of the first weight coefficient and the second weight coefficient to the user in the historical records, where ρ1+ρ2=1, and one value of ρ1 and ρ2 is provided, ρ1=0.5, ρ2=0.5.

[0127] The features of the three types of desensitized data include feature points, feature segments, and feature areas. The value of R is 3. It can be understood that the relative abundance of different features The confirmation method is different. For a single three-category desensitized data, the three-category desensitized data is recorded as the target three-category desensitized data, and all three-category desensitized data with the same name as the target three-category desensitized data and including the target three-category desensitized data are recorded as reference three-category desensitized data. The relative abundance of feature points = the number of feature points in the target three-category desensitized data / the sum of the number of feature points of all reference three-category desensitized data, the relative abundance of feature paragraphs = the feature paragraph length of the target three-category desensitized data / the sum of the feature paragraph lengths of all reference three-category desensitized data, and the relative abundance of feature areas = the feature area of ​​the target three-category desensitized data / the sum of the feature areas of all reference three-category desensitized data.

[0128] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A model construction method based on a plug flow reactor, characterized in that: include: Acquire storage data information of the plug flow reactor at each data station, where the storage data information includes a plurality of data units consisting of a data name and a data sample; Determine the data unit state according to the minimum matching coefficient, and determine the data unit desensitization method according to the data unit state, whether to desensitize the data name or the data sample; When desensitizing data names, replace the data names with hash characters; When desensitizing a data sample, the associated desensitization method is determined to be associated data desensitization or image data desensitization according to the reference matching coefficient; In the associated data desensitization, the first reference sample and the second reference sample are determined according to the sample average value of the reference second-category data unit corresponding to the data sample, each data value in the first reference sample is replaced one by one with the corresponding character in the character library, and each data value in the second reference sample is replaced one by one with the first character and the second character; Image data desensitization is to replace the data sample with a data combination of the number of feature points, feature segment length, and feature area; Under the preset conditions, the data set is selected according to the characteristic degree; The preset condition is that the data name desensitization and data sample desensitization are completed; Determine a data unit state according to a minimum matching coefficient, the data unit state comprising a first data unit state in which the minimum matching coefficient is greater than a preset minimum matching coefficient, and a second data unit state in which the minimum matching coefficient is less than or equal to the preset minimum matching coefficient; The minimum matching coefficient is the minimum value of the matching coefficients corresponding to the data units with the same data name in a single data site; Related data masking includes: The second-category data units with the same data name are recorded as reference second-category data units, the data sample corresponding to the reference second-category data unit with the smallest sample average value is recorded as the first reference sample, and the data samples corresponding to other reference second-category data units are recorded as the second reference samples, and each data value in the first reference sample is replaced one by one with the corresponding character in the character library; Record each data value in the first reference sample as the first data value, record each data value in the second reference sample as the second data value, record a single second data value as the second target data value, record the first data value in the same order as the second target data value as the first target data value, record the character corresponding to the first target data value in the character library as the first character, wherein the first character is recorded as A, the second character is recorded as B, and the second target data value is replaced by A+B or AB according to the first data value; The second character is a character in the character library corresponding to a reference absolute value, and the reference absolute value is an absolute value of a difference between the first data value and the second data value; The second type of data unit is a data unit whose data unit state is the second data unit state and whose matching coefficient is greater than or equal to the preset matching coefficient; The data set is selected according to the characteristic degree, including: recording data samples with the same name or the same hash function corresponding to the hash value as reference data samples, arranging the reference data samples in order from small to large according to the characteristic degree for a single reference data sample to obtain a reference sequence, dividing the reference sequence into n equal parts, selecting data at each equal part point and the initial position and the end position of the reference sequence to establish a data set as simulation data of the horizontal flow reactor model; The feature degrees include name desensitization feature degrees, associated data desensitization feature degrees and image data desensitization feature degrees.

2. The model construction method based on the plug flow reactor according to claim 1, characterized in that: Determine the data unit desensitization method according to the data unit status; If the data unit state is the first data unit state, the data unit desensitization method is to desensitize the data name; If the data unit state is the second data unit state, the data unit desensitization method is to desensitize the data sample.

3. The method for constructing a model based on a plug flow reactor according to claim 2, characterized in that: Desensitization of data names includes: For a single data name, different hash functions are selected for processing according to the data discreteness of all the data units of the first category corresponding to the data name. For a single data unit of the first category, all the data in the data sample of the data unit of the first category are input into the hash function to generate the corresponding hash character, and the data name corresponding to the data unit of the first category is replaced with the hash character; The type of data unit is a data unit whose data unit state is a first data unit state.

4. The model construction method based on the plug flow reactor according to claim 3, characterized in that: When desensitizing a data sample, for a single data sample, the data unit corresponding to the data sample is recorded as the target data unit, and the associated desensitization method is determined according to the reference matching coefficient corresponding to the target data unit; If the reference matching coefficient is greater than or equal to the preset reference matching coefficient, the associated desensitization method is associated data desensitization; If the reference matching coefficient is less than the preset reference matching coefficient, the associated desensitization method is image data desensitization; The reference matching coefficient is a minimum value of matching coefficients corresponding to the target data unit and the reference data unit, and the reference data unit is each data unit with a different data name from the target data unit.

5. The method for constructing a model based on a plug flow reactor according to claim 4, characterized in that: Image data desensitization includes: A line graph is drawn with the length position of the plug flow reactor as the horizontal coordinate and the data text corresponding to a single three-type data unit as the vertical coordinate, the number of characteristic points, the length of characteristic paragraphs and the characteristic area of ​​the line graph are calculated, and the data samples corresponding to the three-type data units are replaced with the data combination of the number of characteristic points, the length of characteristic paragraphs and the characteristic area; The three types of data units are data units whose data unit state is the second data unit state and whose matching coefficients are smaller than the preset matching coefficients.

6. The method for constructing a model based on a plug flow reactor according to claim 5, characterized in that: The calculation formula of name desensitization characteristic τ1 is: Among them, α is the average value of a single type of desensitized data, β is the standard deviation of a single type of desensitized data, γ is the range of a single type of desensitized data, ω1 is the first coefficient, ω2 is the second coefficient, and ω3 is the third coefficient; the type of desensitized data is the data sample after desensitization of a type of data unit.

7. The method for constructing a model based on a plug flow reactor according to claim 6, characterized in that: The calculation formula of the desensitization characteristic degree τ2 of the associated data is: Among them, p is the maximum number of identical characters in a single second-category anonymized data, d is the character type of a single second-category anonymized data, ρ1 is the first weight coefficient, and ρ2 is the second weight coefficient; The calculation formula of image data desensitization feature τ3 is: Among them, R is the total number of features of a single three-category desensitized data, is the relative abundance of the ith feature; The second type of desensitized data is a data sample after the associated data of the second type of desensitized data unit is desensitized, and the third type of desensitized data is a data sample after the image data of the third type of desensitized data unit is desensitized.

Images