Method and system for predicting melt index during polypropylene grade transition, and storage medium

Abstract

The present invention provides a method for predicting a melt index during polypropylene grade transition, comprising the following steps: acquiring a plurality of related process variables of a sample to be tested during polypropylene grade transition; inputting an input vector composed of the plurality of related process variables into a pre-constructed melt index prediction model, so as to respectively output, via a first steady-state sub-model and a second steady-state sub-model of the melt index prediction model, a first melt index of said sample corresponding to a first steady state before polypropylene grade transition and a second melt index of said sample corresponding to a second steady state after polypropylene grade transition; and determining a first membership degree of the input vector to the first steady-state sub-model and a second membership degree of the input vector to the second steady-state sub-model, and on the basis of the first membership degree and the second membership degree, performing weighted summation on the first melt index and the second melt index to determine a melt index of said sample during polypropylene grade transition.

Description

Method, system and storage medium for predicting melt index during polypropylene grade switching TECHNICAL FIELD

[0001] The present application relates to the technical field of polypropylene product quality prediction, and in particular to a method for predicting melt index during polypropylene grade switching, a system for predicting melt index during polypropylene grade switching, and a computer-readable storage medium. BACKGROUND

[0002] The definition of the grade of polypropylene, the naming and the processing method of polypropylene material, the melt index (MI), the polymerization type and the specific use or performance index are closely related to the information, and the grade switching is an optimized production process according to the change of market demand, which involves the conversion from the currently produced polypropylene grade to another grade. The melt index during the polypropylene grade switching process is a key indicator for evaluating the flowability and processing performance. The melt index reflects the flowability of polypropylene in the molten state, and has an important influence on the melting, injection molding and extrusion processes in the plastic processing process. However, in the current actual production, there is a lack of effective online melt index analyzer, and the method of obtaining the melt index through offline analysis has the problems of large time lag and high cost. Moreover, in the polymerization reaction, the residence time of the polymer in the reactor is relatively long, resulting in a long grade switching process and a certain time is needed to make the quality indicators at the outlet of the reactor meet the product requirements. This switching process also produces a large amount of unqualified transition material, causing economic losses to the enterprise. In addition, in many polymer plants, the grade switching is manually operated, resulting in relatively long switching time, product quality fluctuation and even device shutdown problems.

[0003] In order to overcome the above-mentioned defects existing in the prior art, there is an urgent need in the field for an improved method for predicting melt index during polypropylene grade switching, which is used to realize real-time estimation of the melt index during polypropylene grade switching, so as to realize real-time monitoring and control of the quality of polypropylene products. SUMMARY

[0004] The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements of all aspects nor delineate the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.

[0005] In order to overcome the above-mentioned defects existing in the prior art, the application provides a melt index prediction method in a polypropylene grade switching process, a melt index prediction system in a polypropylene grade switching process and a computer readable storage medium, which can construct a melt index prediction model through a sliding window method, and respectively establish a steady-state model for a front and a rear steady-state production process in the grade switching process, so as to realize real-time estimation of the melt index in the polypropylene grade switching process, and realize real-time monitoring and control of the quality of the polypropylene product.

[0006] Specifically, the above-mentioned melt index prediction method in a polypropylene grade switching process according to the first aspect of the application comprises the following steps: obtaining a plurality of related process variables of a to-be-detected sample in a polypropylene grade switching process; inputting an input vector composed of the plurality of related process variables into a pre-constructed melt index prediction model, so as to respectively output a first melt index corresponding to a first steady state before the polypropylene grade switching and a second melt index corresponding to a second steady state after the polypropylene grade switching of the to-be-detected sample via a first steady-state sub-model and a second steady-state sub-model of the melt index prediction model; and determining a first membership degree of the input vector to the first steady-state sub-model and a second membership degree of the input vector to the second steady-state sub-model, and then performing weighted summation on the first melt index and the second melt index according to the first membership degree and the second membership degree, so as to determine the melt index of the to-be-detected sample in the polypropylene grade switching process.

[0007] Further, in some embodiments of the application, the step of determining the first membership degree comprises: determining a first kernel function similarity of the input vector to the first steady-state sub-model according to a first Euclidean distance between the input vector and each first support vector in the first steady-state sub-model; performing weighted summation on the first kernel function similarity and each first support vector according to a first weight coefficient of each first support vector, so as to determine a first weighted similarity of each first support vector; and determining the first membership degree of the input vector to the first steady-state sub-model according to a sum of each first weighted similarity. The first weight coefficient is output via the first steady-state sub-model of the melt index prediction model. The step of determining the second membership degree comprises: determining a second kernel function similarity of the input vector to the second steady-state sub-model according to a second Euclidean distance between the input vector and each second support vector in the second steady-state sub-model; performing weighted summation on the second kernel function similarity and each second support vector according to a second weight coefficient of each second support vector, so as to determine a second weighted similarity of each second support vector; and determining the second membership degree of the input vector to the second steady-state sub-model according to a sum of each second weighted similarity. The second weight coefficient is output via the second steady-state sub-model of the melt index prediction model.

[0008] Further, in some embodiments of the present invention, the step of constructing the melt index prediction model includes: acquiring multiple relevant process variables, true melt index values, and true polypropylene grades of multiple historical samples during the polypropylene grade switching process to construct a historical sample dataset; using a sliding window method, selecting multiple relevant process variables of multiple historical samples within multiple time periods from the historical sample dataset to form multiple training sample vectors; training a classification model for predicting the first membership degree and the second membership degree based on multiple first training sample vectors of the transition state between the first steady state and the second steady state and their corresponding true polypropylene grades; training a first steady-state sub-model for predicting the first melt index and a second steady-state sub-model for predicting the second melt index based on multiple second training sample vectors of the first steady state and the second steady state and their corresponding true melt index values; and combining the classification model, the first steady-state sub-model, and the second steady-state sub-model to construct the melt index prediction model.

[0009] Furthermore, in some embodiments of the present invention, the step of obtaining multiple relevant process variables for multiple historical samples during the polypropylene grade switching process includes: obtaining multiple candidate process variables for the multiple historical samples during the polypropylene grade switching process. The multiple candidate process variables are selected from at least one of catalyst flow rate, hydrogen concentration inside multiple reactors, and propylene concentration inside multiple reactors; and performing sensitivity analysis on the multiple candidate process variables based on the true melt index value of each historical sample to screen the multiple relevant process variables.

[0010] Furthermore, in some embodiments of the present invention, the step of performing sensitivity analysis on the plurality of candidate process variables based on the true melt index values ​​of each of the historical samples to screen the plurality of relevant process variables includes: calculating the mean μ of each of the candidate process variables with respect to each of the historical samples. X And the mean μ of each of the said melt index true values ​​with respect to each of the said historical samples. Y ; Calculate the candidate process variable X and its corresponding mean μ for each of the historical samples. X The difference between them is used to determine the standard deviation σ of each candidate process variable for each of the historical samples. X ; Calculate the true melt flow index Y and the mean μ for each of the historical samples. Y The difference between them is used to determine the standard deviation σ of the true melt index values ​​for each of the historical samples. Y ; based on the mean μ of each of the candidate process variables X and standard deviation σ X and the mean value μ of the true melt index.Y and standard deviation σ Y , respectively, calculate the Pearson correlation coefficient p X,Y of each of the candidate process variables

[0011] wherein the output range of p X,Y is (-1, +1). 0 represents no correlation. Negative values represent negative correlation. Positive values represent positive correlation. E is the covariance of each of the candidate process variables X and the melt index true value Y; and according to the input dimension of the melt index prediction model, select multiple candidate process variables with the largest absolute value |p X，Y | of the Pearson correlation coefficient as the relevant process variables.

[0012] Further, in some embodiments of the present application, the training sample vector is represented as:

[0013] wherein t-h is the starting time of the time period corresponding to the historical sample. t is the termination time of the time period corresponding to the historical sample. h is the length of the sliding window. N ca t is the catalyst flow rate. N donor is the electron donor flow rate. is the concentration ratio of hydrogen to propylene inside the first reactor. is the concentration ratio of hydrogen to propylene inside the second reactor.

[0014] Further, in some embodiments of the present application, the step of training a classification model for predicting the first membership degree and the second membership degree according to the first training sample vectors and their corresponding true polypropylene grades in the transition state between the first steady state and the second steady state comprises: establishing a first training sample data set according to the true polypropylene grades corresponding to each of the training sample vectors:

[0015] T = {(x1, y1), (x2, y2), …, (x N , y N )}

[0016] wherein is a set composed of all training sample vectors. N is the number of elements in the first training sample data set T. y i ∈{+1, -1}. +1 indicates the polypropylene grade before switching. -1 indicates the polypropylene grade after switching; using the support vector machine method, construct and solve the convex quadratic programming problem of the hyperplane wx+b=0 of the first training sample data set T about the classification model to determine the hyperplane parameters w and b of the classification model.

[0017] Further, in some embodiments of the present application, the step of training the first steady state sub-model for predicting the first melt index and the second steady state sub-model for predicting the second melt index according to the plurality of first training sample vectors and their corresponding melt index true values of the first steady state and the second steady state comprises: using support vector regression method to respectively establish the first steady state sub-model and the second steady state sub-model to be trained; according to the melt index true values MI of each of the training sample vectors, establishing a second training sample data set T A corresponding to the first steady state sub-model and a third training sample data set T B :

[0018] wherein, is a set composed of all the training sample vectors. N A is the number of elements in the second training sample data set T A . NB is the number of elements in the third training sample data set T B ; the slack variables ξ and are introduced, and the kernel function is selected to respectively construct the optimization problems for training each of the steady state sub-models:

[0019] wherein,

[0020] According to the constraint conditions of each of the steady state sub-models, the partial derivatives of the Lagrange function are constructed to respectively represent the hyperplane parameters w, b and the slack variables ξ i , of each of the steady state sub-models; and according to the KKT conditions, the hyperplane parameters w, b of each of the steady state sub-models are solved to respectively determine the first steady state sub-model SVR A and the second steady state sub-model SVR B .

[0021] In addition, the prediction system for melt index in the polypropylene grade switching process according to the second aspect of the present application comprises a memory and a processor. The memory has computer instructions stored thereon. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the prediction method for melt index in the polypropylene grade switching process according to the first aspect of the present application.

[0022] In addition, the computer readable storage medium according to the third aspect of the present application has computer instructions stored thereon. When the computer instructions are executed by a processor, the prediction method for melt index in the polypropylene grade switching process according to the first aspect of the present application is implemented. BRIEF DESCRIPTION OF DRAWINGS

[0023] The above features and advantages of the present application will be better understood by reading the following detailed description of the embodiments of the present application in conjunction with the drawings, in which:

[0024] FIG. 1 shows a flow chart of building a melt index prediction model according to some embodiments of the present application.

[0025] FIG. 2 is a schematic diagram of training a classification model based on support vector machine;

[0026] FIG. 3 is a schematic diagram of building a steady state sub-model based on sliding window method and support vector regression;

[0027] FIG. 4 is a schematic diagram of fusing steady state sub-models with membership as weight;

[0028] FIG. 5 is a result plot of predicting the melt index of polymer product during grade transition by the present method. DETAILED DESCRIPTION

[0029] The present application is described in detail below by specific embodiments, and other advantages and effects of the present application can be easily understood by those skilled in the art from the content disclosed in the specification. Although the description of the present application will be introduced in combination with preferred embodiments, this does not mean that the features of the present application are limited to the embodiments. On the contrary, the purpose of introducing the present application in combination with the embodiments is to cover other options or modifications that can be extended based on the claims of the present application. In order to provide a deep understanding of the present application, many specific details will be included in the following description. The present application can also be implemented without using these details. In addition, in order to avoid confusion or obscure the focus of the present application, some specific details will be omitted in the description.

[0030] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0031] In addition, "upper", "lower", "left", "right", "top", "bottom", "horizontal", "vertical" used in the following description and shown in the related drawings should be understood as the orientation shown in the section and the related drawings. The relative terms are only for the convenience of description, and do not mean that the device described thereby should be manufactured or operated in a specific orientation, and therefore should not be understood as a limitation to the present application.

[0032] It can be understood that although the terms "first", "second", "third" and the like can be used herein to describe various components, regions, layers and / or parts, these components, regions, layers and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers and / or parts. Therefore, the first components, regions, layers and / or parts discussed below can be referred to as the second components, regions, layers and / or parts without departing from some embodiments of the present application.

[0033] As described above, in the current actual production, there is a lack of effective online melt index analyzer, and the method of obtaining the melt index through offline analysis has problems of large time lag and high cost. In addition, in the polymerization reaction, the residence time of the polymer in the reactor is relatively long, resulting in a long grade switching process and a certain time is needed to make the quality index at the outlet of the reactor meet the product requirements. This switching process also produces a large amount of unqualified transition material, causing economic losses to the enterprise. In addition, in many polymer plants, the grade switching is manually operated, resulting in relatively long switching time, product quality fluctuation and even device shutdown problems.

[0034] In order to overcome the above-mentioned defects existing in the prior art, the present application provides a melt index prediction method in a polypropylene grade switching process, a melt index prediction system in a polypropylene grade switching process and a computer readable storage medium, which can construct a melt index prediction model by a sliding window method, and establish a steady-state model for the front and rear two steady-state production processes in the grade switching process, for realizing real-time estimation of the melt index in the polypropylene grade switching process, so as to realize real-time monitoring and control of the quality of the polypropylene product.

[0035] In some non-limiting embodiments, the melt index prediction method in a polypropylene grade switching process provided by the first aspect of the present application can be implemented based on the melt index prediction system in a polypropylene grade switching process provided by the second aspect of the present application. Specifically, the melt index prediction system in a polypropylene grade switching process can be configured with a memory and a processor. The memory includes but is not limited to the above-mentioned computer readable storage medium provided by the third aspect of the present application, and the computer instructions are stored on the memory. The processor is connected to the memory and is configured to execute the computer instructions stored on the memory to implement the above-mentioned melt index prediction method in a polypropylene grade switching process provided by the first aspect of the present application.

[0036] Please refer to Figure 1. Figure 1 shows a schematic flowchart of constructing a melt flow index prediction model according to some embodiments of the present invention.

[0037] As shown in Figure 1, the processor can pre-build a melt flow index prediction model offline. Specifically, the processor can first use the DCS system to obtain multiple relevant process variables, the true melt flow index value, and the true polypropylene grade from multiple historical samples during the polypropylene grade switching process, in order to construct a historical sample dataset.

[0038] Furthermore, in acquiring multiple relevant process variables, the processor can first acquire multiple candidate process variables from multiple historical samples during the polypropylene grade switching process. Here, the multiple candidate process variables are selected from at least one of catalyst flow rate, hydrogen concentration inside multiple reactors, and propylene concentration inside multiple reactors.

[0039] Then, the processor can perform sensitivity analysis on multiple candidate process variables based on the true melt index values ​​of each historical sample, in order to screen out multiple relevant process variables.

[0040] Specifically, in the process of sensitivity analysis of multiple candidate process variables, the processor can calculate each candidate process variable X: {X1, X2, ..., X...} N Regarding the mean μ of each historical sample x And the true values ​​of each melt index Y: {Y1, Y2, ..., Y} N Regarding the mean μ of each historical sample Y :

[0041] Then, the processor can calculate the candidate process variable X and its corresponding mean μ for each historical sample. x The difference between them is used to determine the standard deviation σ of each candidate process variable for each historical sample. X :

[0042] Similarly, the processor can calculate the true melt flow index Y and the mean μ for each historical sample separately. Y The difference between them is used to determine the standard deviation σ of the true melt index value for each historical sample. Y :

[0043] Then, the processor can determine the mean μ of each candidate process variable. X and standard deviation σ X And the mean value of the true melt index μ Y and standard deviation σ Y Calculate the Pearson correlation coefficient ρ for each candidate process variable.X,Y :

[0044] wherein, p X,Y The output range of p X,Y is (-1, +1), 0 represents no correlation, negative value represents negative correlation, and positive value represents positive correlation. E is the covariance of each candidate process variable X and the true value of melt index Y.

[0045] After that, the processor can select multiple candidate process variables with the maximum absolute value of Pearson correlation coefficient |p X,Y | as the relevant process variables according to the input dimension of the melt index prediction model.

[0046] For example, the processor can determine that multiple candidate process variables with the absolute value of Pearson correlation coefficient |p X,Y |>0.7 have strong correlation with the melt index, and select them as the relevant process variables. In this case, the multiple candidate process variables are the catalyst flow, the promoter flow, the hydrogen-to-propylene concentration ratio inside the first reactor, and the hydrogen-to-propylene concentration ratio inside the second reactor.

[0047] After that, the processor can use the sliding window method to select multiple relevant process variables of multiple historical samples in multiple time periods from the historical sample data set to form multiple training sample vectors x.

[0048] Further, the above training sample vector is expressed as:

[0049] wherein, t-h is the starting time of the time period corresponding to the historical sample, t is the ending time of the time period corresponding to the historical sample, h is the length of the sliding window, N cat is the catalyst flow, N donor is the promoter flow, is the hydrogen-to-propylene concentration ratio inside the first reactor, is the hydrogen-to-propylene concentration ratio inside the second reactor.

[0050] Please refer to FIG. 2. FIG. 2 shows a flowchart of training a classification model according to some embodiments of the present application.

[0051] As shown in FIG. 2, the processor can train a classification model for predicting the first membership degree and the second membership degree according to multiple first training sample vectors in the transition state between the first steady state and the second steady state and their corresponding true polypropylene grades.

[0052] Specifically, the processor can first establish a first training sample data set according to the true polypropylene grades corresponding to each training sample vector:

[0053] T = {(x1, y1), (x2, y2),..., (xN, yN)} N N}

[0054] wherein is a set of all training sample vectors, N is the number of elements in the first training sample data set T, y i ∈{+1, -1}, +1 indicates the polypropylene grade before switching, and -1 indicates the polypropylene grade after switching.

[0055] Afterwards, the processor can use the support vector machine method to construct and solve a convex quadratic programming problem of the hyperplane wx+b=0 of the classification model with respect to the first training sample data set T to determine the hyperplane parameters w and b of the classification model.

[0056] Specifically, for the above first training sample data set T and the hyperplane wx+b=0, the processor can define the geometric interval of the hyperplane with respect to the training sample vector (x i , y i ) as:

[0057] wherein γ i is the geometric interval of the ith training sample vector to the hyperplane.

[0058] Afterwards, the processor can define the minimum value in all geometric intervals as the distance of the support vector of the training sample vector to the hyperplane:

[0059] Further afterwards, the processor can express the above solving problem as a constrained optimization problem:

[0060] Further afterwards, the processor can divide both sides of the constraint condition by γ to obtain:

[0061] Further afterwards, the processor converts the original objective function with constraints into a new unconstrained Lagrange objective function:

[0062] wherein α i is a Lagrange multiplier, and α i ≥0.

[0063] Further afterwards, the processor can set when the training sample vector does not satisfy the constraint condition, i.e. outside the feasible solution region, y i (w·x i +b)<1, α i ​is set to infinity, which leads to θ(w) also being infinity. When the training sample vector satisfies the constraint condition, i.e., in the feasible solution region, y i (w·x i +b)≥1), θ(w) is equal to the original function itself. The processor can combine the two cases to obtain a new objective function:

[0064] Further, the processor can equivalently transform the original constraint problem to:

[0065] Further, the processor can use the duality of the Lagrange function to exchange the positions of the minimum and maximum, and obtain:

[0066] Further, the processor can make the above formula satisfy the KKT (Karush Kuhn Tucker) condition, which is a necessary condition for judging whether a point is an extreme point, and then p * =d * , that is:

[0067] Further, the processor can first select a suitable kernel function K(x, z) and a penalty parameter C > 0, and construct a convex quadratic programming problem:

[0068] Where N is the number of data sets.

[0069] Then, the processor can solve the above convex quadratic programming problem to obtain the optimal solution

[0070] Further, the processor can select an α * one component of α to satisfy the condition and calculate:

[0071] Further, the processor can calculate the classification decision function:

[0072] Please refer to FIG. 3. FIG. 3 shows a flowchart of training a first steady-state sub-model and a second steady-state sub-model according to some embodiments of the present application.

[0073] As shown in FIG. 3, the processor can train a first steady-state sub-model for predicting a first melt index and a second steady-state sub-model for predicting a second melt index according to a plurality of second training sample vectors of the first steady state and the second steady state and their corresponding melt index true values.

[0074] Specifically, the processor can use a support vector regression (SVR) to respectively establish the first steady-state sub-model and the second steady-state sub-model based on the training sample vectors.

[0075] Then, the processor can establish a second training sample data set T A corresponding to the first steady-state sub-model according to the melt index true values MII corresponding to the training sample vectors. B

[0076] wherein, is a set composed of all the training sample vectors, N A is the number of elements in the second training sample data set T A , N B is the number of elements in the third training sample data set T B .

[0077] Then, the processor can calculate the minimized regularized ε-insensitive error function of the SVR as follows:

[0078] Then, the processor can replace the above quadratic error function with an ε-insensitive error function, wherein ε represents a non-negative slack variable, so as to better adapt to the existence of outliers or noise and improve the robustness of the regression model:

[0079] Then, the processor can obtain the minimized regularized ε-insensitive error function:

[0080] wherein C is a regularization coefficient, and ξ and are slack variables, and ξ i ≥ 0,

[0081] Then, the processor can calculate the updated SVR loss function:

[0082] Then, the processor can respectively construct an optimization problem for training each steady-state sub-model:

[0083] wherein,

[0084] Then, the processor can construct a Lagrange function according to the constraint conditions of each steady-state sub-model, so as to respectively represent the partial derivatives of the hyperplane parameters w, b and the slack variable ξ i , of each steady-state sub-model.​

[0085] Further, according to the partial derivative being 0, the following formula can be derived:

[0086] Further, the processor can solve the hyperplane parameters w, b of each steady-state sub-model according to the KKT condition, to determine the first steady-state sub-model SVR A and the second steady-state sub-model SVR B .

[0087] In this way, the processor can use the support vector regression method to find the optimal hyperplane, so that the distance between all training sample vectors and the hyperplane is minimized, thereby better fitting the data.

[0088] Further, the processor can combine the classification model, the first steady-state sub-model and the second steady-state sub-model to construct a melt index prediction model.

[0089] Please refer to FIG. 4. FIG. 4 shows a flowchart of a melt index prediction method in a polypropylene grade switching process according to some embodiments of the present application.

[0090] As shown in FIG. 4, in the process of online prediction of melt index in a polypropylene grade switching process, the processor can first obtain a plurality of related process variables of a sample to be tested in a polypropylene grade switching process. Then, an input vector is formed by the plurality of related process variables. The input vector is input into a pre-constructed melt index prediction model, so as to output, via the first steady-state sub-model and the second steady-state sub-model of the melt index prediction model, a first melt index corresponding to a first steady state before the polypropylene grade switching and a second melt index corresponding to a second steady state after the polypropylene grade switching.

[0091] Further, the processor can determine a first membership degree of the input vector to the first steady-state sub-model and a second membership degree of the input vector to the second steady-state sub-model.

[0092] The step of determining the first membership degree includes: determining a first kernel function similarity between the input vector and each first support vector in the first steady-state sub-model according to a first Euclidean distance between the input vector and each first support vector; performing weighted summation on the first kernel function similarity and each first support vector according to a first weight coefficient of each first support vector, to determine a first weighted similarity of each first support vector; and determining the first membership degree of the input vector to the first steady-state sub-model according to a sum of each first weighted similarity, wherein the first weight coefficient is output by the first steady-state sub-model of the melt index prediction model,

[0093] The step of determining the second membership degree comprises: determining a second kernel function similarity between the input vector and the second steady-state sub-model according to a second Euclidean distance between the input vector and each second support vector in the second steady-state sub-model; performing weighted summation on the second kernel function similarity and each second support vector according to a second weight coefficient of each second support vector to determine a second weighted similarity of each second support vector; and determining a second membership degree of the input vector to the second steady-state sub-model according to a sum of the second weighted similarities, wherein the second weight coefficient is output by the second steady-state sub-model of the melt index prediction model.

[0094] Specifically, in the process of determining the first membership degree, the processor can first determine a first kernel function similarity between the input vector and the first steady-state sub-model according to a first Euclidean distance between the input vector and each first support vector in the first steady-state sub-model, and then perform weighted summation on the first kernel function similarity and each first support vector according to a first weight coefficient of each first support vector to determine a first weighted similarity of each first support vector. Here, the first weight coefficient is output by the first steady-state sub-model of the melt index prediction model.

[0095] Then, the processor can determine a first membership degree w1 of the input vector to the first steady-state sub-model according to a sum of the first weighted similarities.

[0096] Similarly, in the process of determining the second membership degree, the processor can determine a second kernel function similarity between the input vector and the second steady-state sub-model according to a second Euclidean distance between the input vector and each second support vector in the second steady-state sub-model, and then perform weighted summation on the second kernel function similarity and each second support vector according to a second weight coefficient of each second support vector to determine a second weighted similarity of each second support vector. Here, the second weight coefficient is output by the second steady-state sub-model of the melt index prediction model.

[0097] Then, the processor can determine a second membership degree w2 of the input vector to the second steady-state sub-model according to a sum of the second weighted similarities.

[0098] Then, the processor can perform weighted summation on the first melt index and the second melt index to determine a melt index of the sample to be tested in the polypropylene grade switching process:

[0099] Please refer to FIG. 5. FIG. 5 shows a schematic diagram of a melt index prediction result provided according to some embodiments of the present application.

[0100] As shown in Figure 5, the melt index prediction method for polypropylene grade switching provided in the first aspect of the present invention is used for prediction, and compared with the actual melt index value. It can be found that the melt index prediction method for polypropylene grade switching proposed in this method has excellent prediction accuracy for the melt index of polymer during grade switching.

[0101] In summary, the above-mentioned method, system, and computer-readable storage medium for predicting the melt index during polypropylene grade switching provided by this invention can all construct a melt index prediction model using the sliding window method. Furthermore, by establishing steady-state models for the two steady-state production processes before and after the grade switching process, they can achieve real-time estimation of the melt index during polypropylene grade switching, thereby enabling real-time monitoring and control of polypropylene product quality.

[0102] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments.

[0103] Those skilled in the art will understand that information, signals, and data can be represented using any of a variety of different techniques and arts. For example, the data, instructions, commands, information, signals, bits, symbols, and chips described throughout the above description can be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or optical particles, or any combination thereof.

[0104] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in a generalized manner in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in different ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the invention.

[0105] Although the processor described in the above embodiments can be implemented by a combination of software and hardware, it is understood that the processor can be implemented in software or hardware. For hardware implementation, the controller 40 can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, micro-controllers, microprocessors, other electronic units, or a selective combination thereof implementing the functions described in the above embodiments. For software implementation, the processor can be implemented by separate software modules such as procedures and functions, each of which performs one or more functions and operations described in the above embodiments, running on a general purpose chip.

[0106] The various illustrative logical blocks, circuits, and circuitry described in connection with the embodiments disclosed herein can be implemented or performed with a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor can be a microprocessor, but in the alternative, the processor can be any conventional processor, controller, microcontroller, or state machine. A processor can also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0107] The steps of a method or algorithm described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium can be integral to the processor. The processor and the storage medium can reside in an ASIC. The ASIC can reside in a user terminal. In the alternative, the processor and the storage medium can reside as discrete components in a user terminal.

[0108] In one or more exemplary embodiments, the functions described can be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Also, any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk and disc, as used herein, includes compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0109] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting the melt index during a polypropylene grade switch, characterized by, The method comprises the following steps: obtaining a plurality of relevant process variables of a to-be-tested sample in a polypropylene grade switching process; inputting an input vector composed of the plurality of relevant process variables into a pre-constructed melt index prediction model to respectively output, via a first steady-state sub-model and a second steady-state sub-model of the melt index prediction model, a first melt index corresponding to a first steady state before the polypropylene grade switching of the to-be-tested sample and a second melt index corresponding to a second steady state after the polypropylene grade switching of the to-be-tested sample; and determining a first membership degree of the input vector to the first steady-state sub-model and a second membership degree of the input vector to the second steady-state sub-model, and performing weighted summation on the first melt index and the second melt index according to the first membership degree and the second membership degree to determine the melt index of the to-be-tested sample in the polypropylene grade switching process.

2. The prediction method of claim 1, wherein, The step of determining the first membership degree comprises: determining a first kernel function similarity between the input vector and the first steady-state sub-model according to a first Euclidean distance between the input vector and each first support vector in the first steady-state sub-model; performing weighted summation on the first kernel function similarity and each first support vector according to a first weight coefficient of each first support vector to determine a first weighted similarity of each first support vector; and determining the first membership degree of the input vector to the first steady-state sub-model according to a sum of the first weighted similarities, wherein the first weight coefficient is output via the first steady-state sub-model of the melt index prediction model, The step of determining the second membership degree comprises: determining a second kernel function similarity between the input vector and the second steady-state sub-model according to a second Euclidean distance between the input vector and each second support vector in the second steady-state sub-model; performing weighted summation on the second kernel function similarity and each second support vector according to a second weight coefficient of each second support vector to determine a second weighted similarity of each second support vector; and determining the second membership degree of the input vector to the second steady-state sub-model according to a sum of the second weighted similarities, wherein the second weight coefficient is output via the second steady-state sub-model of the melt index prediction model.

3. The prediction method of claim 1, wherein, The step of constructing the melt index prediction model comprises: respectively obtaining a plurality of relevant process variables, a melt index true value and a true polypropylene grade of a plurality of historical samples in the polypropylene grade switching process to construct a historical sample data set; using a sliding window method to select a plurality of relevant process variables of a plurality of historical samples in a plurality of time periods from the historical sample data set to form a plurality of training sample vectors; training a classification model for predicting the first membership degree and the second membership degree according to a plurality of first training sample vectors of a transition state between the first steady state and the second steady state and the corresponding true polypropylene grade; and training a classification model for predicting the first membership degree and the second membership degree according to a plurality of first training sample vectors of a transition state between the first steady state and the second steady state and the corresponding true polypropylene grade. training a first steady state sub-model for predicting the first melt index and a second steady state sub-model for predicting the second melt index according to a plurality of first training sample vectors and their corresponding melt index true values between the first steady state and the second steady state; and combining the classification model, the first steady state sub-model and the second steady state sub-model to build the melt index prediction model.

4. The prediction method of claim 3, wherein, The step of obtaining the plurality of relevant process variables of the plurality of historical samples in the polypropylene grade switching process respectively comprises: obtaining a plurality of candidate process variables of the plurality of historical samples in the polypropylene grade switching process respectively, wherein the plurality of candidate process variables are selected from at least one of a catalyst flow, hydrogen concentrations inside a plurality of reactors, propylene concentrations inside a plurality of reactors; and performing sensitivity analysis on the plurality of candidate process variables according to melt index true values of each of the historical samples to screen the plurality of relevant process variables therefrom.

5. The prediction method of claim 4, wherein, The step of performing sensitivity analysis on the plurality of candidate process variables according to melt index true values of each of the historical samples to screen the plurality of relevant process variables therefrom comprises: respectively calculating a mean μ of each of the candidate process variables with respect to each of the historical samples X and a mean μ of each of the melt index true values with respect to each of the historical samples Y ; respectively, to determine a standard deviation σ of each of the candidate process variables X of each of the historical samples X respectively, to determine a standard deviation σ of each of the candidate process variables X of each of the historical samples X ; calculating the difference between the melt index true value Y of each of the historical samples and the mean value μ Y to determine the standard deviation σ Y of the melt index true value of each of the historical samples. According to the mean μ X and standard deviation σ X of each of the candidate process variables, and the mean μ Y and standard deviation σ Y of the melt index true value, the Pearson correlation coefficient ρ X,Y is calculated for each of the candidate process variables, respectively: wherein, p X,Y The output range of is (-1, +1), 0 represents no correlation, negative value represents negative correlation, positive value represents positive correlation, E is the covariance of each candidate process variable X and the true value of melt index Y; and According to the input dimension of the melt index prediction model, the absolute value |p X，Y | the maximum number of candidate process variables as the relevant process variable.

6. The prediction method of claim 4, wherein, The training sample vector is represented as: wherein t-h is the starting time of the time period corresponding to the historical sample, t is the ending time of the time period corresponding to the historical sample, h is the length of the sliding window, N cat is the flow rate of the catalyst, N donor is the flow rate of the electron donor, the concentration ratio of hydrogen to propylene inside the first reactor, a concentration ratio of hydrogen to propylene inside the second reactor.

7. The prediction method of claim 3, wherein, The step of training a classification model for predicting the first membership and the second membership according to a plurality of first training sample vectors and their corresponding true polypropylene grades between the transition state and the first steady state and the second steady state comprises: establishing a first training sample data set according to the true polypropylene grades corresponding to each of the training sample vectors: T = {(xl,yl), (x2,y2),..., (xn,yn)} N , y N} wherein for all training sample vectors in the set, N is the number of elements in the first training sample dataset T, y i ∈ {+1, -1}, +1 indicates the polypropylene grade before the switch, -1 indicates the polypropylene grade after the switch; and using a support vector machine method, constructing and solving a convex quadratic programming problem of a hyperplane wx+b=0 of the classification model with respect to the first training sample data set T to determine hyperplane parameters w and b of the classification model.

8. The prediction method of claim 3, wherein, The step of training a first steady state sub-model for predicting the first melt index and a second steady state sub-model for predicting the second melt index according to a plurality of first training sample vectors and their corresponding melt index true values between the first steady state and the second steady state comprises: using a support vector regression method, respectively establishing a first steady state sub-model and a second steady state sub-model to be trained based on; According to the melt index true value MI corresponding to each of the training sample vectors, a second training sample data set T corresponding to the first steady-state sub-model is established A , and a third training sample data set T corresponding to the second steady-state sub-model B : wherein N is the set of all training sample vectors. A For the second training sample dataset T A The number of elements in, N B For the third training sample dataset T B The number of elements in; Introducing the slack variable ξ and and a kernel function is selected to construct an optimization problem for training each of the steady-state sub-models: wherein, According to constraint conditions of each of the steady-state sub-models, a Lagrange function is constructed to represent a hyperplane parameter w, b and a relaxation variable ξ of each of the steady-state sub-models respectively i 、 partial derivatives of; and According to the KKT condition, the hyperplane parameters w, b of each of the steady-state sub-models are solved to determine the first steady-state sub-model SVR A and the second steady-state sub-model SVR B , respectively.

9. A system for predicting the melt flow index during polypropylene grade switching, characterized in that, comprise: a memory having computer instructions stored thereon; and a processor connected to the memory and configured to execute the computer instructions stored on the memory to implement the method for predicting a melt index in a polypropylene grade switching process according to any one of claims 1-8.

10. A computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are executed by the processor to implement the method for predicting a melt index in a polypropylene grade switching process according to any one of claims 1-8.

Images