Data analysis method for the preparation of polyethylene glycol active ester chitosan composite membrane

Abstract

The present invention relates to the field of data processing technology, and specifically to a data analysis method for the preparation of polyethylene glycol active ester chitosan composite membranes; decomposing the preparation feature monitoring sequence to obtain a residual feature time series; obtaining outlier residual points according to the residual feature time series; determining the suspected abnormal preparation moment according to the quantity characteristics and distribution characteristics of the outlier residual points and the change characteristics of the residual points at the recent historical moments of the outlier residual points; obtaining the data change degree according to the data difference characteristics of adjacent moments; obtaining a multidimensional difference coefficient according to the data change degrees of two dimensions at the suspected abnormal preparation moment; obtaining a neighborhood difference coefficient according to the residual data at the suspected abnormal preparation moment at any position and within a preset neighborhood range. The present invention obtains the final abnormal degree of preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient and detects the preparation process, improves the accuracy of the preparation process analysis, and reduces the interference of sensor errors.

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a data analysis method for preparing a polyethylene glycol active ester chitosan composite film. Background Art

[0002] Polyethylene glycol active ester chitosan composite membrane is a functional composite membrane material formed by combining polyethylene glycol and chitosan, and is widely used in biomedicine, food packaging, tissue engineering and other fields. In the preparation process of polyethylene glycol active ester chitosan composite membrane, it is of great significance to analyze the data of the preparation of polyethylene glycol active ester chitosan composite membrane. By analyzing the data, the influence of key parameters (such as temperature, pH, reaction time, ratio, etc.) on the performance of the composite membrane can be identified, thereby optimizing the preparation process and ensuring the stability of the performance of the composite membrane.

[0003] In the large-scale preparation process, the main stage is the process of polyethylene glycol molecules reacting with disuccinimidyl carbonate to form active esters, which requires continuous stirring for 8 to 24 hours; this process is an esterification reaction, which will cause the temperature to continue to rise, and the heat dissipation will cause the temperature to drop, so the abnormal data analysis can be used to determine whether the preparation process is abnormal. However, due to long-term continuous stirring, the sensor is easily pulled by the colloid flow, resulting in data errors at some moments in the sensor, making the data analysis of the preparation process inaccurate, and ultimately resulting in low accuracy in abnormal detection of the preparation process. Summary of the invention

[0004] In order to solve the above-mentioned technical problem that the reliability of data analysis in the preparation process is affected by sensor factors, resulting in low accuracy of abnormal detection in the preparation process, the object of the present invention is to provide a data analysis method for the preparation of polyethylene glycol active ester chitosan composite membrane, and the technical scheme adopted is as follows:

[0005] Obtain the preparation characteristic monitoring sequence of two dimensions at different positions in the container during the preparation of polyethylene glycol active ester chitosan composite film;

[0006] Perform time series decomposition on the preparation feature monitoring sequence of any dimension at any position to obtain the residual feature time series; obtain the outlier residual point according to the data distribution characteristics of the residual feature time series; obtain the initial abnormality degree of preparation according to the outlier residual point, the quantity characteristics and distribution characteristics of other outlier residual points at the neighboring moments of the outlier residual point, and the change characteristics of the residual points at the recent historical moments of the outlier residual point; determine the suspected abnormal moment of preparation in any dimension at the arbitrary position according to the initial abnormality degree of preparation;

[0007] The data variation degree at any moment is obtained according to the data difference characteristics at any moment and adjacent moments in the residual feature time series; the multidimensional difference coefficient is obtained according to the similar characteristics of the data variation degree at the suspected abnormal preparation moment in two dimensions at the arbitrary position; the neighborhood difference coefficient is obtained according to the residual data difference characteristics of the same dimension at the suspected abnormal preparation moment within the arbitrary position and the preset neighborhood range;

[0008] The final abnormality degree of the preparation at the moment of the suspected abnormal preparation is obtained according to the multidimensional difference coefficient and the neighborhood difference coefficient; and the preparation process is detected according to the final abnormality degree of the preparation.

[0009] Furthermore, the step of obtaining outlier residual points according to the data distribution characteristics of the residual feature time series includes:

[0010] The numerical absolute value of the residual point in the residual feature time series is used as the mapping value; the average value of the mapping value is used as the first value, and the product of the standard deviation of the mapping value and the constant 3 is used as the second value; when the absolute value of the difference between the mapping value of any residual point and the first value exceeds the second value, the arbitrary residual point is an outlier residual point.

[0011] Furthermore, the step of preparing the initial abnormality degree according to the outlier residual point, the quantity characteristics and distribution characteristics of other outlier residual points at the nearest moments of the outlier residual point, and the change characteristics of the residual points at the recent historical moments of the outlier residual point includes:

[0012] Calculate the inverse of the mapping value of the outlier residual point to obtain a first factor; calculate the average of the time distances between the outlier residual point and other outlier residual points closest to the previous and next moments to obtain an interval characteristic value; calculate the product of the inverse of the interval characteristic value and the number of other outlier residual points in a preset adjacent time period of the outlier residual point and perform positive correlation mapping to obtain a second factor; perform linear fitting based on the numerical values ​​of the residual points in the preset historical time period of the outlier residual point to obtain a fitting straight line; calculate the slope of the fitting straight line and perform positive correlation mapping to obtain a third factor; calculate the sum of the first factor, the second factor and the third factor to obtain the initial abnormality degree of the preparation of the outlier residual point.

[0013] Furthermore, the step of determining the suspected preparation abnormality time of the arbitrary dimension at the arbitrary position according to the initial preparation abnormality degree includes:

[0014] When the initial abnormality of the preparation of an outlier residual point of any dimension at any position exceeds a preset initial abnormality threshold, the time at which the outlier residual point is located is used as the suspected abnormal time of the preparation of the arbitrary dimension at the arbitrary position.

[0015] Furthermore, the step of obtaining the data change degree at any moment according to the data difference feature at any moment and the adjacent moments in the residual feature time series includes:

[0016] The difference between the residual values ​​at any moment and any adjacent moment is calculated and normalized by the maximum and minimum normalization method to obtain the data change degree between the any moment and any adjacent moment.

[0017] Furthermore, the step of obtaining a multidimensional difference coefficient at the arbitrary position according to similar characteristics of the data change degree in two dimensions at the time of the suspected abnormal preparation includes:

[0018] The absolute values ​​of the differences of the data change degrees of the two dimensions at the moment of the suspected preparation abnormality are calculated to obtain the difference degree; the average value of the difference degree corresponding to the moment of the suspected preparation abnormality and two adjacent moments is calculated to obtain the multidimensional difference coefficient.

[0019] Furthermore, the step of obtaining a neighborhood difference coefficient according to the residual data difference characteristics of the same dimension at the moment of the suspected abnormal preparation within the arbitrary position and the preset neighborhood range comprises:

[0020] The absolute value of the difference between the residual values ​​of the same dimension of the arbitrary position and other positions within the preset neighborhood range at the time of suspected preparation abnormality is calculated to obtain the neighborhood difference; the average value of the neighborhood difference between the arbitrary position and all other positions within the preset neighborhood range is calculated and positively correlated to obtain the neighborhood difference coefficient.

[0021] Furthermore, the step of obtaining the final abnormality degree of the preparation at the moment of the suspected abnormal preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient includes:

[0022] The average values ​​of the multidimensional difference coefficient and the neighborhood difference coefficient are calculated and negatively correlated to obtain the final abnormality degree of the preparation at the moment of the suspected abnormal preparation.

[0023] Furthermore, the step of detecting the preparation process according to the final abnormality degree of the preparation includes:

[0024] When the final abnormality degree of the preparation in the residual feature time series of the arbitrary dimension at the arbitrary position exceeds the preset final abnormality threshold, the preparation process of the polyethylene glycol active ester chitosan composite membrane is abnormal.

[0025] Furthermore, the two-dimensional preparation feature monitoring sequences are respectively a temperature sequence and a pH sequence.

[0026] The present invention has the following beneficial effects:

[0027] In the present invention, obtaining the residual feature time series can characterize the data abnormality characteristics of the preparation feature monitoring sequence, and obtaining the outlier residual point can characterize the moment when the data in the preparation feature monitoring sequence is relatively abnormal. Obtaining the initial abnormality degree of preparation can distinguish the data feature differences caused by sensor errors and uneven preparation stirring, so as to quantify and judge the data abnormalities caused by different factors according to the initial abnormality degree of preparation; obtaining the suspected abnormal moment of preparation can preliminarily screen out the outlier residual points caused by sensor errors, and preliminarily improve the accuracy of the preparation process analysis. Obtaining the data change degree can characterize the data change degree of adjacent moments; obtaining the multidimensional difference coefficient can further judge the abnormal cause of the suspected abnormal moment of preparation according to the correlation characteristics of the changes in temperature and pH during the production of active esters. Obtaining the neighborhood difference coefficient can judge the abnormal cause of the suspected abnormal moment of preparation according to the data difference characteristics of adjacent sensors; finally, obtaining the final abnormality degree of preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient can accurately judge the cause of data abnormality. Detecting the preparation process according to the final abnormality degree of preparation can improve the accuracy of the preparation process analysis and reduce the interference of sensor errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 A flow chart of a data analysis method for preparing a polyethylene glycol active ester chitosan composite membrane provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a data analysis method for preparing a polyethylene glycol active ester chitosan composite film according to the present invention, its specific implementation method, structure, characteristics and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0031] 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 invention belongs.

[0032] The specific scheme of the data analysis method for preparing polyethylene glycol active ester chitosan composite membrane provided by the present invention is described in detail below with reference to the accompanying drawings.

[0033] See also Figure 1 , which shows a flow chart of a data analysis method for preparing a polyethylene glycol active ester chitosan composite film provided by an embodiment of the present invention, the method comprising the following steps:

[0034] Step S1, obtaining a two-dimensional preparation feature monitoring sequence at different positions in a container during the preparation process of a polyethylene glycol active ester chitosan composite film.

[0035] In an embodiment of the present invention, the implementation scenario is to perform data analysis on the preparation process of polyethylene glycol active ester chitosan composite membrane to determine whether the preparation process is normal; first, obtain the preparation feature monitoring sequence of two dimensions at different positions in the container during the preparation process of polyethylene glycol active ester chitosan composite membrane. The main stage of the preparation process is the reaction of polyethylene glycol molecules with disuccinimidyl carbonate to generate active esters; the reaction in this stage is an esterification reaction, which releases heat during the process, and acidic byproducts are released during this stage. Therefore, the preparation effect can be judged by analyzing the temperature and pH during the preparation process; therefore, the preparation feature monitoring sequences are temperature sequence and pH sequence respectively. Since large-scale preparation requires stirring reaction in a large container, in order to be able to analyze the preparation process more comprehensively, sensors need to be installed at multiple positions in the container. In the embodiment of the present invention, multiple sensors are evenly installed in a matrix form around the container, and the sensors at each position simultaneously collect temperature and pH; the implementer can determine the installation position of the sensor according to the implementation scenario. Since the stirring process for preparing active ester takes 8 to 24 hours, each sensor in the present invention acquires a temperature sequence and a pH sequence every 30 minutes, and the acquisition frequency is once per second, which can be determined by the implementer according to the implementation scenario.

[0036] Furthermore, the change characteristics of temperature and pH during the stirring process can be used to determine whether the stirring is sufficient. However, during the stirring process, the sensor may be affected by the flow of colloid, which may cause collection errors and lead to deviations in some data. Therefore, it is difficult to accurately determine the cause of data deviation when analyzing the preparation process, making the accuracy of the preparation process analysis low. Therefore, in order to improve the accuracy of the preparation process analysis, when relatively abnormal data is found, it is necessary to first analyze the cause of the data abnormality.

[0037] Step S2, performing time series decomposition on the preparation feature monitoring sequence of any dimension at any position to obtain the residual feature time series; obtaining the outlier residual point according to the data distribution characteristics of the residual feature time series; obtaining the initial abnormality degree of the preparation according to the outlier residual point, the quantity characteristics and distribution characteristics of other outlier residual points at the neighboring moments of the outlier residual point, and the change characteristics of the residual points at the recent historical moments of the outlier residual point; determining the suspected preparation abnormality moment of any dimension at any position according to the initial abnormality degree of the preparation.

[0038] The process of producing active esters is an exothermic esterification reaction. During this period, the temperature in the container will continue to rise, and the temperature will drop when the heat is dissipated. The pH will first increase and then decrease with stirring; therefore, the data collected during the entire stage will show slight periodic changes. For periodic data, the existing STL time series decomposition algorithm is often used to decompose seasonal terms, trend terms and residual terms, and analyze abnormal characteristics through residual terms; therefore, the preparation feature monitoring sequence of any dimension at any position is decomposed in time series to obtain the residual feature time series; it should be noted that the STL time series decomposition algorithm belongs to the existing technology, and the specific steps will not be repeated. The residual feature time series characterizes the degree of abnormality at each acquisition moment. When the value of the residual point in the residual feature time series is less close to 0, the original data at that moment is less consistent with the distribution of trend terms and seasonal terms, and the moment is more likely to be abnormal; therefore, the more abnormal residual points can be obtained according to the existing Laida criterion, so the outlier residual points are obtained according to the data distribution characteristics of the residual feature time series.

[0039] Preferably, in an embodiment of the present invention, the step of obtaining outlier residual points includes: using the absolute value of the residual point in the residual feature time series as the mapping value; using the average value of the mapping value as the first value, and using the product of the standard deviation of the mapping value and the constant 3 as the second value; when the absolute value of the difference between the mapping value of any residual point and the first value exceeds the second value, according to the Laida criterion, it can be known that the difference between the mapping value of the arbitrary residual point and the mean value has exceeded 3 standard deviations, so the arbitrary residual point is an outlier residual point; the outlier residual point means that the data obtained by the sensor at this moment is relatively abnormal. If there are no outlier residual points in the preparation feature monitoring sequence of any dimension at any position in the container, it means that the stirring preparation process of this period is normal; when an outlier residual point appears, it is necessary to judge that the preparation stirring process is uneven or the sensor is abnormal, so as to improve the accuracy of the preparation analysis.

[0040] The process of producing active esters is an esterification reaction that releases heat. Due to the presence of colloids in the raw materials, if there is uneven stirring during the stirring process, the local esterification reaction exothermic accumulation temperature will be too high, resulting in outlier residual points in the temperature dimension. At the same time, acidic by-products will be generated during the production of active esters. If the stirring is uneven in a local area, the concentration of the acidic by-product in that area will be too high, resulting in outlier residual points in the pH dimension. Long-term stirring may cause sensor errors, and outlier residual points of temperature and pH will also appear. Since sensor errors may cause large deviations in the data, the upper limit of the corresponding outlier residual points is relatively high, while the numerical deviation range of the outlier residual points of temperature and pH caused by abnormal preparation stirring is relatively small, and the upper limit of the corresponding outlier residual points is relatively low. Moreover, the temperature and pH anomalies caused by uneven stirring are cumulative. The outlier residual points within a period of time are distributed densely and have a trend, while the outlier residual points caused by sensor errors are relatively random. Therefore, the initial abnormality degree of the preparation can be obtained based on the outlier residual points, the quantity characteristics and distribution characteristics of other outlier residual points in the vicinity of the outlier residual points, and the change characteristics of the residual points in the recent historical moments of the outlier residual points.

[0041] Preferably, in an embodiment of the present invention, the step of obtaining the initial abnormality of the preparation includes: calculating the inverse of the mapping value of the outlier residual point to obtain a first factor; when the mapping value is larger, the first factor is smaller, which means that the outlier residual point may be caused by a sensor error. Calculate the average value of the time distance between the outlier residual point and other outlier residual points closest to the previous and next moments to obtain an interval characteristic value; when the interval characteristic value is larger, it means that the appearance of the outlier residual point is more random, and it is more likely to be caused by a sensor error; when the interval characteristic value is smaller, it means that the outlier residual point appears more densely, and it is more likely to be caused by uneven preparation and stirring. Calculate the product of the inverse of the interval characteristic value and the number of other outlier residual points in the preset adjacent time period of the outlier residual point and positively map to obtain a second factor; in the embodiment of the present invention, the preset adjacent time period is the 20 closest acquisition moments to the outlier residual point, and the implementer can determine it according to the implementation scenario. When the interval characteristic value is larger and the number of other outlier residual points in the preset adjacent time period is smaller, the second factor is smaller, and the outlier residual point is more likely to be caused by sensor abnormality. According to the numerical value of the residual point in the preset historical period of the outlier residual point, a linear fit is performed to obtain a fitting straight line; in the embodiment of the present invention, the preset historical period is the 10 adjacent acquisition moments before the outlier residual point, and the outlier residual points in the first 10 acquisition moments in the residual feature time series do not participate in the calculation of the third factor. The implementer can determine the length of the preset historical period according to the implementation scenario. Calculate the slope of the fitting straight line and positively map it to obtain the third factor; the larger the slope, the larger the third factor, which means that the numerical value of the residual point is gradually increasing, and it is more likely that the temperature or pH value gradually increases due to abnormal preparation stirring. Calculate the sum of the first factor, the second factor and the third factor to obtain the initial abnormality degree of the preparation of the outlier residual point; when the initial abnormality degree of preparation is smaller, it means that the outlier residual point is more likely to be caused by sensor error, and when the initial abnormality degree of preparation is larger, it means that the outlier residual point is more likely to be caused by uneven preparation stirring. The formula for obtaining the initial abnormality degree of preparation includes:

[0042]

[0043] In the formula, R represents the initial abnormality of the outlier residual point, D represents the mapping value of the outlier residual point, represents the first factor, represents the normalized function for positive correlation mapping, N represents the number of other outlier residual points in the preset adjacent period of the outlier residual point, W represents the interval eigenvalue, represents the second factor, F represents the slope of the fitting line, Represents the third factor.

[0044] Furthermore, after obtaining the initial abnormality degree of the preparation of the outlier residual point, the abnormal moment caused by uneven preparation stirring can be preliminarily determined, so the suspected abnormal moment of preparation in any dimension at any position is determined according to the initial abnormality degree of preparation; preferably, in the embodiment of the present invention, the step of obtaining the suspected abnormal moment of preparation includes: when the initial abnormality degree of the preparation of the outlier residual point in any dimension at any position exceeds the preset initial abnormality threshold, the moment of the outlier residual point is used as the suspected abnormal moment of preparation in any dimension at the arbitrary position. In the embodiment of the present invention, the preset initial abnormality threshold is 0.8, which can be determined by the implementer according to the implementation scenario; the data abnormality that may be caused by sensor error is screened out by the initial abnormality degree of preparation, which preliminarily improves the accuracy of judging abnormal conditions in the preparation process.

[0045] Step S3, in the residual feature time series, obtain the data change degree at any moment according to the data difference characteristics between any moment and adjacent moments; obtain the multidimensional difference coefficient at any position according to the similar characteristics of the data change degree in two dimensions at the moment of suspected preparation abnormality; obtain the neighborhood difference coefficient at any position and within a preset neighborhood range according to the residual data difference characteristics of the same dimension at the moment of suspected preparation abnormality.

[0046] Judging uneven preparation stirring only based on the degree of initial abnormality in preparation has certain limitations. If the deviation data caused by a sensor error is continuously triggered or appears at subsequent moments of an outlier residual point of uneven stirring, the moment of the outlier residual point caused by such sensor error may be considered as a suspected moment of preparation abnormality. Therefore, in order to further improve the accuracy of the analysis of the preparation stirring process, it is necessary to continue analyzing the abnormal conditions at the moment of suspected preparation abnormality.

[0047] Due to the uneven stirring in the process of producing active ester, the temperature and pH value will change abnormally at the same time, and the changes in temperature and pH value have a certain correlation; if the sensor abnormality causes an error in a certain dimension, such as temperature, and the pH value is collected normally, the data changes in the two dimensions are not correlated. Therefore, the correlation of the data change characteristics of the two dimensions at the same time can be analyzed to determine the cause of the abnormality, so the data change degree at any time is obtained according to the data difference characteristics of any time and the adjacent time in the residual feature time series; preferably, in an embodiment of the present invention, the step of obtaining the data change degree includes: calculating the difference between the residual values ​​of any time and any adjacent time and normalizing them by the maximum and minimum normalization method to obtain the data change degree of the arbitrary time and any adjacent time. Any adjacent time is the adjacent time before and after the arbitrary time. It should be noted that the maximum and minimum normalization method belongs to the prior art, in which the maximum value is the maximum value of the difference of the residual values ​​of the adjacent time in the residual feature time series, and the minimum value is the minimum value of the difference of the residual values ​​of the adjacent time in the residual feature time series; when there are previous and next moments at the arbitrary time, there are two corresponding data change degrees.

[0048] Further, when the data variation of the two dimensions at the same position at the same time is more similar, it means that the suspected abnormal preparation moment is more likely to be caused by uneven preparation stirring. Therefore, at any position, the multidimensional difference coefficient is obtained according to the similar characteristics of the data variation of the two dimensions at the suspected abnormal preparation moment; preferably, in an embodiment of the present invention, the step of obtaining the multidimensional difference coefficient includes: calculating the absolute value of the difference between the data variation of the two dimensions at the suspected abnormal preparation moment to obtain the difference; calculating the average value of the difference corresponding to the suspected abnormal preparation moment and two adjacent moments to obtain the multidimensional difference coefficient. When the data variation of the two dimensions at the same moment is more similar, the difference is smaller, the multidimensional difference coefficient is smaller, and the suspected abnormal preparation moment is more likely to be caused by uneven preparation stirring; for example, in the temperature dimension, a certain moment is a suspected abnormal preparation moment, then the two data variation degrees before and after the temperature dimension and the two data variation degrees before and after the pH dimension are calculated at this moment, and finally two differences are obtained. When the average value of the two differences is smaller, it means that the change characteristics of the two dimensions are more related, and the multidimensional difference coefficient is smaller.

[0049] Since large-scale preparation is carried out in large containers, evenly distributed sensors collect data at the same time. If uneven preparation stirring occurs, the data change characteristics collected by multiple sensors in the area are similar; if a sensor is abnormal, there is a significant deviation between the data of this sensor and other surrounding sensors. Therefore, the neighborhood difference coefficient is obtained based on the difference characteristics of the residual data of the same dimension at the time of the suspected abnormal preparation at any position and within the preset neighborhood range.

[0050] Preferably, in an embodiment of the present invention, the step of obtaining a neighborhood difference coefficient includes: calculating the absolute value of the difference between the residual values ​​of the same dimension of the arbitrary position and other positions within the preset neighborhood range at the suspected moment of preparation abnormality to obtain a neighborhood difference; in an embodiment of the present invention, the other positions within the preset neighborhood range are the eight nearest sensor positions to the arbitrary position, which can be determined by the implementer according to the implementation scenario. If a suspected moment of preparation abnormality appears in the temperature dimension at the arbitrary position, the residual point value of the temperature dimension at the arbitrary position and other positions at that moment is calculated; the smaller the neighborhood difference, the more likely it is that the suspected moment of preparation abnormality is caused by uneven preparation stirring. Calculate the average value of the neighborhood difference between the arbitrary position and all other positions within the preset neighborhood range and perform positive correlation mapping to obtain a neighborhood difference coefficient; the smaller the neighborhood difference coefficient, the more likely it is that the suspected moment of preparation abnormality is caused by uneven preparation stirring.

[0051] Step S4, obtaining the final abnormality degree of the preparation at the moment of suspected abnormal preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient; and detecting the preparation process according to the final abnormality degree of the preparation.

[0052] After obtaining the multidimensional difference coefficient and the neighborhood difference coefficient of the suspected preparation abnormality moment, it is possible to accurately determine whether the suspected preparation abnormality moment is caused by uneven preparation stirring, so the final abnormality degree of preparation at the suspected preparation abnormality moment is obtained according to the multidimensional difference coefficient and the neighborhood difference coefficient; preferably, in an embodiment of the present invention, the step of obtaining the final abnormality degree of preparation includes: calculating the average value of the multidimensional difference coefficient and the neighborhood difference coefficient and negatively correlating and mapping them to obtain the final abnormality degree of preparation at the suspected preparation abnormality moment. The greater the final abnormality degree of preparation, it means that the moment is more likely to be caused by uneven preparation stirring, and the possibility of sensor error is smaller. The formula for obtaining the final abnormality degree of preparation includes:

[0053]

[0054] In the formula, H represents the final abnormality degree of preparation, represents an exponential function with a natural constant as the base, Q represents the multidimensional difference coefficient, and S represents the neighborhood difference coefficient.

[0055] After obtaining the final abnormality degree of the preparation at the moment of suspected preparation abnormality at any position and any dimension in the process of producing active ester, the preparation process can be detected according to the final abnormality degree of the preparation; preferably, in the embodiment of the present invention, when the final abnormality degree of the preparation appears in the residual feature time series of any dimension at any position and exceeds the preset final abnormality threshold, the preparation process of the polyethylene glycol active ester chitosan composite film is abnormal, the preparation stirring is uneven, and the reaction is insufficient. In the embodiment of the present invention, the preset final abnormality threshold is 0.5, which can be determined by the implementer according to the implementation scenario; when the preparation process is detected to be abnormal, the implementer can increase the stirring time or other measures according to the implementation scenario, which is not limited here; when the final abnormality degree at the moment of suspected preparation abnormality does not exceed the preset final abnormality threshold, it means that the preparation process is normal and the subsequent production process can be carried out. By analyzing the outlier residual points, the suspected moments of preparation anomalies are obtained, and the outlier residual points that may be caused by sensor errors are preliminarily screened out; after obtaining the suspected moments of preparation anomalies, the final degree of preparation anomaly is obtained based on the correlation of data changes in two dimensions and the correlation of adjacent sensor data; the final degree of preparation anomaly can be used to accurately determine whether there is uneven preparation stirring in the process of producing active esters, thereby improving the accuracy of preparation process analysis and reducing the influence of sensor errors on preparation process analysis.

[0056] In summary, the embodiment of the present invention provides a data analysis method for the preparation of polyethylene glycol active ester chitosan composite membrane; decomposes the preparation feature monitoring sequence to obtain the residual feature time series; obtains the outlier residual point according to the residual feature time series; determines the suspected abnormal preparation moment according to the quantity characteristics and distribution characteristics of the outlier residual point and the change characteristics of the residual point in the recent historical moment of the outlier residual point; obtains the data change degree according to the data difference characteristics of adjacent moments; obtains the multidimensional difference coefficient according to the data change degree of two dimensions at the suspected abnormal preparation moment; obtains the neighborhood difference coefficient according to the residual data at the suspected abnormal preparation moment at any position and within the preset neighborhood range. The present invention obtains the final abnormal degree of preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient and detects the preparation process, improves the accuracy of the preparation process analysis, and reduces the interference of sensor errors.

[0057] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0058] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A data analysis method for preparing polyethylene glycol active ester chitosan composite membrane, characterized in that: The method comprises the following steps: Obtain the preparation characteristic monitoring sequence of two dimensions at different positions in the container during the preparation of polyethylene glycol active ester chitosan composite film; Perform time series decomposition on the preparation feature monitoring sequence of any dimension at any position to obtain the residual feature time series; obtain the outlier residual point according to the data distribution characteristics of the residual feature time series; obtain the initial abnormality degree of preparation according to the outlier residual point, the quantity characteristics and distribution characteristics of other outlier residual points at the neighboring moments of the outlier residual point, and the change characteristics of the residual points at the recent historical moments of the outlier residual point; determine the suspected abnormal moment of preparation in any dimension at the arbitrary position according to the initial abnormality degree of preparation; The data variation degree at any moment is obtained according to the data difference characteristics at any moment and adjacent moments in the residual feature time series; the multidimensional difference coefficient is obtained according to the similar characteristics of the data variation degree at the suspected abnormal preparation moment in two dimensions at the arbitrary position; the neighborhood difference coefficient is obtained according to the residual data difference characteristics of the same dimension at the suspected abnormal preparation moment within the arbitrary position and the preset neighborhood range; Obtaining the final abnormality degree of the preparation at the moment of the suspected abnormal preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient; detecting the preparation process according to the final abnormality degree of the preparation; The step of obtaining outlier residual points according to the data distribution characteristics of the residual feature time series comprises: The absolute value of the residual point in the residual feature time series is used as the mapping value; the average value of the mapping value is used as the first value, and the product of the standard deviation of the mapping value and the constant 3 is used as the second value; when the absolute value of the difference between the mapping value of any residual point and the first value exceeds the second value, the any residual point is an outlier residual point; The step of obtaining and preparing the initial abnormality degree according to the outlier residual point, the quantity characteristics and distribution characteristics of other outlier residual points at the nearest moments of the outlier residual point, and the change characteristics of the residual points at the recent historical moments of the outlier residual point comprises: Calculate the inverse of the mapping value of the outlier residual point to obtain a first factor; calculate the average of the time distance between the outlier residual point and other outlier residual points closest to the previous and next moments to obtain an interval characteristic value; calculate the product of the inverse of the interval characteristic value and the number of other outlier residual points in a preset adjacent time period of the outlier residual point and perform positive correlation mapping to obtain a second factor; perform linear fitting based on the numerical values ​​of the residual points in the preset historical time period of the outlier residual point to obtain a fitting straight line; calculate the slope of the fitting straight line and perform positive correlation mapping to obtain a third factor; calculate the sum of the first factor, the second factor and the third factor to obtain the initial abnormality degree of the preparation of the outlier residual point; The step of determining the suspected preparation abnormality moment of the arbitrary dimension at the arbitrary position according to the initial preparation abnormality degree comprises: When the initial abnormality of the preparation of an outlier residual point of any dimension at any position exceeds a preset initial abnormality threshold, the time at which the outlier residual point is located is used as the suspected abnormal time of the preparation of the arbitrary dimension at the arbitrary position; The step of obtaining a multidimensional difference coefficient at the arbitrary position according to similar characteristics of the data change degree in two dimensions at the time of the suspected abnormal preparation includes: Calculate the absolute value of the difference between the data change degrees of the two dimensions at the suspected abnormal preparation moment to obtain the difference degree; calculate the average value of the difference degree corresponding to the suspected abnormal preparation moment and two adjacent moments to obtain the multidimensional difference coefficient; The step of obtaining a neighborhood difference coefficient according to the residual data difference characteristics of the same dimension at the time of the suspected abnormal preparation within the arbitrary position and the preset neighborhood range comprises: Calculate the absolute value of the difference between the residual value of the same dimension of the arbitrary position and other positions within the preset neighborhood range at the time of suspected preparation abnormality to obtain the neighborhood difference; calculate the average value of the neighborhood difference between the arbitrary position and all other positions within the preset neighborhood range and perform positive correlation mapping to obtain the neighborhood difference coefficient; The step of obtaining the final abnormality degree of the preparation at the moment of suspected abnormal preparation according to the multidimensional difference coefficient and the neighborhood difference coefficient comprises: The average values ​​of the multidimensional difference coefficient and the neighborhood difference coefficient are calculated and negatively correlated to obtain the final abnormality degree of the preparation at the moment of the suspected abnormal preparation.

2. A data analysis method for preparing polyethylene glycol active ester chitosan composite membrane according to claim 1, characterized in that: The step of obtaining the data change degree at any moment according to the data difference feature at any moment and the adjacent moments in the residual feature time series comprises: The difference between the residual values ​​at any moment and any adjacent moment is calculated and normalized by the maximum and minimum normalization method to obtain the data change degree between the any moment and any adjacent moment.

3. A data analysis method for preparing polyethylene glycol active ester chitosan composite membrane according to claim 1, characterized in that: The step of detecting the preparation process according to the final abnormality degree of the preparation comprises: When the final abnormality degree of the preparation in the residual feature time series of the arbitrary dimension at the arbitrary position exceeds the preset final abnormality threshold, the preparation process of the polyethylene glycol active ester chitosan composite membrane is abnormal.

4. A data analysis method for preparing polyethylene glycol active ester chitosan composite membrane according to claim 1, characterized in that: The two-dimensional preparation feature monitoring sequences are respectively a temperature sequence and a pH sequence.

Images