CN113759834A - Chaos multi-scale intelligent optimal propylene polymerization process measuring instrument - Google Patents

Abstract

The invention discloses a chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument, which is used for quality inspection of polypropylene production products. The chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument comprises a chaotic reconstruction module , Gabor multi-scale analysis module, extreme random tree measurement model module, system update module, chaotic artificial bee colony optimization module. The invention measures the melt index, an important quality index of the propylene polymerization process, on-line, and overcomes the shortcomings of large measurement time lag and low measurement accuracy of traditional chemical instruments, and the chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument realizes the online measurement of important quality indexes. , Taking into account the multi-scale characteristics and chaotic characteristics of the aggregation process, the system has strong application and promotion ability and high precision.

Description

A Chaos Multi-scale Intelligent Optimal Propylene Polymerization Process Measuring Instrument

technical field

The invention relates to the fields of polymerization process measuring instruments, machine learning and intelligent optimization, in particular to a chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument.

Background technique

Polypropylene is a thermoplastic resin obtained by the polymerization of propylene. The most important downstream product of propylene, 50% of the world's propylene and 65% of our country's propylene are used to make polypropylene. It is one of the five general-purpose plastics. daily life is closely related. Polypropylene is the fastest growing general-purpose thermoplastic resin in the world, second only to polyethylene and polyvinyl chloride in total. In order to make our polypropylene products have market competitiveness, develop impact copolymer products, random copolymer products, BOPP and CPP film materials, fibers and non-woven materials with good rigidity, toughness and fluidity balance, and develop polypropylene in automobiles. It is an important research topic in the future.

Melt index is one of the important quality indicators for determining the grade of polypropylene products. It determines the different uses of the product. The measurement of melt index is an important part of product quality control in polypropylene production. It is very important for production and scientific research. important role and guiding significance.

However, online analysis and measurement of melt index is currently difficult to achieve. On the one hand, there is a lack of online melt index analyzers. Difficulty in use. Therefore, at present, the measurement of MI in industrial production is mainly obtained by manual sampling and off-line laboratory analysis, and generally can only be analyzed once every 2-4 hours, and the time lag is large, which brings difficulties to the quality control of propylene polymerization production. A bottleneck problem that needs to be solved urgently in production.

SUMMARY OF THE INVENTION

In order to overcome the shortcomings of the existing propylene polymerization production process that the measurement accuracy is not high and is easily affected by human factors, the purpose of the present invention is to provide an online measurement, taking into account the multi-scale characteristics and chaotic characteristics of the polymerization process, and the ability of system application and promotion. A chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument with high strength and high precision.

The purpose of this invention is to realize through the following technical solutions:

The invention provides a polypropylene quality detection system for online detection of polypropylene product quality, which is characterized in that it includes a chaos reconstruction module, a Gabor multi-scale analysis module and an extreme random tree measurement model module.

In the polypropylene quality detection system provided by the present invention, the chaotic reconstruction module is used to reconstruct the input variables of the model from the DCS database into dynamic chaotic system signals according to its chaotic characteristics.

In the polypropylene quality detection system provided by the present invention, the Gabor multi-scale analysis module is used to analyze the multi-scale characteristics of the dynamic chaotic system signal based on the frequency, and perform multi-scale analysis on the dynamic chaotic system signal through the Gabor kernel function. Reconstruction to obtain the local texture feature information of each scale at different frequencies.

In the polypropylene quality detection system provided by the present invention, the extreme random tree measurement model module is used to establish a mapping relationship between chaotic multi-scale feature information and melt index, establish a propylene polymerization process measurement system, and then predict polypropylene melt index.

The polypropylene quality detection system provided by the present invention, wherein, preferably, the input variables are the first propylene feed flow rate, the second propylene feed flow rate, the third propylene feed flow rate, the main catalyst Flow rate, co-catalyst flow rate, temperature in the stirred tank, pressure in the tank, liquid level in the tank and volume concentration of hydrogen in the tank.

In the polypropylene quality detection system provided by the present invention, preferably, the chaotic system of input variables is expressed as z(n)=[s(n), s(n+T ₁ ), s(n+T ₂ ), ...,s(n+T _d-1 )], where s(n) is the signal at the nth sampling point in the propylene polymerization process, and T ₁ , T ₂ ,...,T _d-1 are the nth signal, respectively sampling time after sampling points.

The polypropylene quality detection system provided by the invention, wherein, preferably, the expression of Gabor kernel function is:

Among them, z represents the coordinate information of the reconstructed variable, u represents the direction of the Gabor filter, v represents the scale of the Gabor filter, i is the complex symbol, exp(ik _{u, v} z) is the oscillatory function in complex exponential form, σ ² is the kernel function width, and k _{u, v} represents the response of the Gabor filter in all directions at each scale.

The polypropylene quality detection system provided by the present invention, wherein, preferably, the Gabor feature is obtained by convolution of the kernel function, and the expression is as follows:

G _u,v (z)=f(z)*ψ _u,v (z)

Among them, _Gu,v (z) represents the convolution function of the corresponding scale v and direction u near the coordinate z, and ψ is the Gabor kernel function; the input variable is analyzed by the Gabor function to obtain the complex input feature signal:

Gu _,v (z)=Re(Gu _,v (z))+jIm(Gu _,v (z))

The amplitude and phase of the Gabor characteristic signal are:

The polypropylene quality detection system provided by the present invention, wherein, preferably, the establishment of the mapping relationship takes the input signal of a group of extreme random trees as the local texture feature information under a single scale, and the multiple groups of extreme random trees are based on ensemble learning. A framework to model chaotic multiscale mapping from input to output.

The polypropylene quality detection system provided by the present invention, preferably, further includes a chaotic artificial bee colony optimization module, which uses a chaotic artificial bee colony algorithm to optimize the bifurcation threshold parameter of the extreme random tree measurement model module.

The polypropylene quality detection system provided by the present invention, wherein, preferably, the optimization comprises the following steps:

(1) Initialize the parameters of the artificial bee colony algorithm, set the number of nectar sources p, the maximum number of iterations iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the nectar source represents the feasible solution of the problem, measured in the correlation vector The dimension of p _i in the system is 2 dimensions, and the initial iteration number iter=0;

(2) assign a leading bee to the nectar source p _i , and generate a new nectar source V _i based on the chaotic mapping;

为混沌因子，由混沌映射动态系统产生。in,

is the chaotic factor, which is generated by the chaotic mapping dynamic system.

(3) Calculate the fitness value of _Vi , and determine the retained nectar source according to the method of greedy selection;

(4) Calculate the probability that the nectar source found by the leading bee is followed up;

(5) The following bees search in the same way as the leading bees, and determine the reserved nectar source according to the method of greedy selection;

(6) judging whether the nectar source V _i satisfies the abandoned condition, if satisfied, the corresponding leading bee role becomes the scout bee, otherwise directly go to step (8);

(7) Scout bees randomly generate new nectar sources;

(8) iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, and select the solution with the optimal fitness value as the optimal solution of the algorithm, otherwise go to step (2);

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

The polypropylene quality inspection system provided by the present invention, preferably, further includes a system update module for inputting the offline assay data into the training set to update the extreme random tree measurement model online.

According to some embodiments of the present invention, the present invention can also be stated as follows:

The invention provides a chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument, which is used for quality inspection of polypropylene production products. The chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument comprises a chaotic reconstruction module, Gabor multi-scale analysis module, extreme random tree measurement model module, system update module, chaotic artificial bee colony optimization module. in:

(1) Chaos reconstruction module: It is used to reconstruct the model input variables input from the DCS database according to its chaotic characteristics. The input signals of the propylene polymerization process measuring instrument are 9 operational variables of the industrial propylene polymerization process, which are the first Propylene feed flow rate, second propylene feed flow rate, third propylene feed flow rate, main catalyst flow rate, auxiliary catalyst flow rate, temperature in the stirred tank, pressure in the tank, liquid level in the tank, and tank The volume concentration of hydrogen inside. The chaotic system of its input signal is expressed as z(n)=[s(n),s(n+T ₁ ),s(n+T ₂ ),...,s(n+T _d-1 )], Wherein s(n) is the signal of the nth sampling point in the propylene polymerization process, and T ₁ , T ₂ , . . . , T _d-1 are the sampling moments after the nth sampling point, respectively. In the chaotic system, the delay time satisfies the condition of T _m =mτ, where τ is the delay time, and T _m represents the mth sampling time. Therefore, the input signal of the propylene process can be reconstructed into a dynamic chaotic system signal by the embedded dimension and delay time z(n)=[s(n),s(n+τ),s(n+2τ),...,s(n+(d-1)τ)], where z(n) is the nth is the chaotic reconstruction signal at time, τ is the delay time, and d is the embedding dimension of the input signal. The delay time and embedding dimension of chaotic reconstruction are obtained by mutual information method and pseudo-nearest neighbor method, respectively.

(2) Gabor multi-scale analysis module: It is used to analyze the multi-scale characteristics of the input variables after chaotic reconstruction based on the frequency, and perform multi-scale reconstruction of the variables through the Gabor kernel function, so as to extract the input variables at different frequencies. The local texture feature information in each direction of the scale, the Gabor kernel function is defined as follows:

Among them, z represents the coordinate information of the reconstructed variable, u represents the direction of the Gabor filter, v represents the scale of the Gabor filter, i is the complex symbol, exp(ik _{u, v} z) is the oscillatory function in complex exponential form, σ ² is the kernel function width, and k _{u, v} represents the response of the Gabor filter in all directions at each scale. The partial functions of the Gabor kernel function are as follows: _ku,v ² z ² /2σ ² is a Gaussian envelope function, _ku,v ² /σ ² is used to compensate for the weakening of the energy spectrum, and the envelope function is usually windowed The range of the oscillation function can be limited, the locality of the wave can be maintained, and the feature information near the coordinates can be extracted. exp(ik _{u, v} z) is an oscillatory function whose real part is a cosine function that is even symmetric, and whose imaginary part is a sine function that is odd symmetric. exp(-σ ² /2) represents the DC component of the filter. The purpose of the [exp(ik _u,v z)-exp(-σ ² /2)] operation is to eliminate the influence of the DC component on the filtering effect. The width of the kernel function is σ ² is used to determine the bandwidth size of the Gabor filter. k _{u, v} represents the response of the Gabor filter in all directions at various scales, each k _{u, v} represents a Gabor filter, so when multiple different k _{u, v} are selected, multiple different filters can be obtained device group.

The Gabor feature is obtained by convolution of the kernel function, and the expression is as follows:

G _u,v (z)=f(z)*ψ _u,v (z)

Among them, _Gu,v (z) represents the convolution function of the corresponding scale v and direction u near the coordinate z, and ψ is the Gabor kernel function. Use the Gabor function to analyze the input variables to obtain the complex input feature signal:

Gu _,v (z)=Re(Gu _,v (z))+jIm(Gu _,v (z))

The amplitude and phase of the Gabor characteristic signal are:

(3) Extreme random tree measurement model module: used to establish a measurement system for the propylene polymerization process, using extreme random trees and an integrated learning framework to complete the input-to-output mapping modeling. The extreme random tree training splitting rule ensures the sparsity of the model by introducing a hyperparameter to assign a Gaussian prior distribution with zero mean to the weight vector. The hyperparameter can be estimated by maximizing the edge likelihood function. The purpose of the whole model is to design a system based on the sample set and prior knowledge, so that the system can predict the polypropylene melt index output on new data.

(4) Chaos artificial bee colony optimization module: The chaotic artificial bee colony algorithm is used to optimize the bifurcation threshold parameters of the extreme random measuring instrument, and the following process is used to complete:

(4.1) Initialize the parameters of the artificial bee colony algorithm, set the number of nectar sources P, the maximum number of iterations iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the nectar source represents the feasible solution of the problem, measured in the correlation vector The dimension of p _i in the system is 2 dimensions, and the initial iteration number iter=0;

(4.2) assign a leading bee to the nectar source p _i , and generate a new nectar source V _i based on the chaotic map;

为混沌因子，由混沌映射动态系统产生。in,

is the chaotic factor, which is generated by the chaotic mapping dynamic system.

(4.3) Calculate the fitness value of _Vi , and determine the retained nectar source according to the method of greedy selection;

(4.4) Calculate the probability that the nectar source found by the leading bee will be followed;

(4.5) The follower bee searches in the same way as the leader bee, and determines the reserved nectar source according to the method of greedy selection;

(4.6) Judging whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding leading bee role becomes a scout bee, otherwise, go directly to (4.8);

(4.7) Scout bees randomly generate new nectar sources;

(4.8) iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, and select the solution with the optimal fitness value as the optimal solution of the algorithm, otherwise go to step (4.2).

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

(5) System update module, the chaotic multi-scale intelligent optimal propylene polymerization process measurement instrument also includes a system update module, which is used for online update of the system, regularly inputs offline test data into the training set, and updates extreme random tree measurements Model.

The technical idea of the present invention is as follows: online prediction of the melt index, an important quality index of the propylene polymerization process, is introduced in order to overcome the low measurement accuracy and insufficient extraction of nonlinear features of the existing polypropylene melt index measuring instruments. The multi-scale analysis method is used for feature extraction and sequence reconstruction, and the introduction of intelligent methods does not require human experience or multiple tests to adjust the system parameters, thereby obtaining the chaotic multi-scale intelligent optimal measuring instrument in the propylene polymerization production process. In order to overcome the problem that the measurement accuracy of the existing propylene polymerization production process is not high, the purpose of the present invention is to provide a chaotic multi-scale intelligent optimal propylene polymerization process measurement instrument.

The beneficial effects of the present invention are mainly manifested in:

1. The chaos reconstruction analysis and Gabor multi-scale analysis of the propylene polymerization production process effectively characterize the dynamic characteristics and nonlinearity at different scales in the actual industrial polymerization process, and realize the high-precision measurement of the product quality index melt index;

2. The chaotic artificial bee colony optimization correlation vector system realizes the online measurement and matching of the detection system, and improves the application and promotion ability of the system.

Description of drawings

FIG. 1 is an overall structural diagram of a polypropylene quality inspection system according to an embodiment of the present invention;

FIG. 2 is a functional structural diagram of a polypropylene quality detection system according to an embodiment of the present invention.

Detailed ways

The present invention will be specifically described below with reference to the accompanying drawings.

Example

Referring to FIG. 1, the overall structure diagram of a polypropylene quality inspection system of an embodiment involves a propylene polymerization production process 1, an on-site intelligent instrument for measuring easily measurable variables 2, a control station for measuring operational variables 3, and a data storage device. The DCS database 4 and the melt index soft-measurement value display device 6, the on-site intelligent instrument 2 and the control station 3 are connected with the propylene polymerization production process 1, and the on-site intelligent instrument 2 and the control station 3 are connected with the DCS database 4, and also involve A chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument 5, the DCS database 4 is connected with the input end of the chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument 5, the chaotic multi-scale intelligent The output end of the optimal propylene polymerization process measuring instrument 5 is connected to the melt index soft-measurement value display instrument 6 . Both the manipulated variables and the measurable variables mentioned above can be used as input variables.

According to the analysis of the reaction mechanism and process technology, taking into account various factors that affect the melt index in the production process of polypropylene, the nine operational variables and easily measurable variables commonly used in the actual production process are taken as modeling variables, which are: Dilute feed flow rate, main catalyst flow rate, auxiliary catalyst flow rate, temperature, pressure, liquid level in the kettle, and volume concentration of hydrogen in the kettle. Table 1 lists the 9 modeling variables required as a chaotic multi-scale intelligent optimal propylene polymerization process measurement instrument, which are the temperature (T) in the kettle, pressure (p) in the kettle, liquid level (L) in the kettle, Hydrogen volume concentration in the kettle (Xv), 3 propylene feed flow rates (the first propylene feed flow rate f1, the second propylene feed flow rate f2, the third propylene feed flow rate f3) , 2 streams of catalyst feed flow rate (main catalyst flow rate f4, auxiliary catalyst flow rate f5). The polymerization reaction in the reaction kettle is that the reaction materials participate in the reaction after repeated mixing, so the process variables of the model input variables involving materials use the average value of the previous several moments. The data in this example are averaged over the previous hour. The offline assay value of the melt index is used as the output variable of the chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument 5 . Obtained by manual sampling, offline assay analysis, and analysis and collection every 4 hours.

Table 1. Modeling variables required for the chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument

2, a functional structure diagram of a polypropylene quality inspection system according to an embodiment, including:

(1) Chaos reconstruction module 7, which is used to reconstruct the model input variables input from the DCS database according to its chaotic characteristics. The chaotic system of the input signal in the propylene polymerization production process is expressed as z(n)=[s(n), s(n+T ₁ ),s(n+T ₂ ),...,s(n+T _d-1 )], where s(n) is the signal of the nth sampling point in the propylene polymerization process, T ₁ , T ₂ ,...,T _d-1 are the sampling moments after the nth sampling point, respectively. In the chaotic system, the delay time satisfies the condition of T _m =mτ, where τ is the delay time, and T _m represents the mth sampling time. Therefore, the input signal of the propylene process can be reconstructed into a dynamic chaotic system signal by the embedded dimension and delay time z(n)=[s(n),s(n+τ),s(n+2τ),...,s(n+(d-1)τ)], where z(n) is the nth is the chaotic reconstruction signal at time, τ is the delay time, and d is the embedding dimension of the input signal. The delay time and embedding dimension of chaotic reconstruction are obtained by mutual information method and pseudo-nearest neighbor method, respectively.

(2) Gabor multi-scale analysis module 8, which is used to analyze the multi-scale characteristics of the input variables based on the frequency, and realize the multi-scale reconstruction of the variables through the Gabor kernel function to characterize the input variables at different frequencies. For local texture information, the Gabor kernel function is defined as follows:

Among them, z represents the coordinate information of the reconstructed variable, u represents the direction of the Gabor filter, v represents the scale of the Gabor filter, i is the complex symbol, exp(ik _{u, v} z) is the oscillatory function in complex exponential form, σ ² is the kernel function width, and k _{u, v} represents the response of the Gabor filter in all directions at each scale. The partial functions of the Gabor kernel function are as follows: _ku,v ² z ² /2σ ² is a Gaussian envelope function, _ku,v ² /σ ² is used to compensate for the weakening of the energy spectrum, and the envelope function is usually windowed The range of the oscillation function can be limited, the locality of the wave can be maintained, and the feature information near the coordinates can be extracted. exp(ik _{u, v} z) is an oscillatory function whose real part is a cosine function for even symmetry, and its imaginary part is a sine function for odd symmetry. exp(-σ ² /2) represents the DC component of the filter. The purpose of the [exp(ik _u,v z)-exp(-σ ² /2)] operation is to eliminate the influence of the DC component on the filtering effect. The width of the kernel function is σ ² is used to determine the bandwidth size of the Gabor filter. k _{u, v} represents the response of the Gabor filter in all directions at various scales, each k _{u, v} represents a Gabor filter, so when multiple different k _{u, v} are selected, multiple different filters can be obtained device group.

The Gabor feature is obtained by convolution of the kernel function, and the expression is as follows:

G _u,v (z)=f(z)*ψ _u,v (z)

Among them, _Gu,v (z) represents the convolution function of the corresponding scale v and direction u near the coordinate z, and ψ is the Gabor kernel function. Use the Gabor function to analyze the input variables to obtain the complex input feature signal:

Gu _,v (z)=Re(Gu _,v (z))+jIm(Gu _,v (z))

The amplitude and phase of the Gabor characteristic signal are:

(3) The extreme random tree measurement model module 9 is used to complete the input-to-output mapping modeling based on the ensemble learning framework by using the extreme random tree. The extreme random tree training splitting rule ensures the sparsity of the model by introducing a hyperparameter to assign a Gaussian prior distribution with zero mean to the weight vector. The hyperparameter can be estimated by maximizing the edge likelihood function. The purpose of the whole model is to design a system based on the sample set and prior knowledge, so that the system can predict the polypropylene melt index output on new data.

(4) The chaotic artificial bee colony optimization module 10 adopts the chaotic artificial bee colony algorithm to optimize the bifurcation threshold parameter of the measurement system, and adopts the following process to complete:

(4.1) Initialize the parameters of the artificial bee colony algorithm, set the number of nectar sources P, the maximum number of iterations iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the nectar source represents the feasible solution of the problem, measured in the correlation vector The dimension of p _i in the system is 2 dimensions, and the initial iteration number iter=0;

(4.2) assign a leading bee to the nectar source p _i , and generate a new nectar source V _i based on the chaotic map;

为混沌因子，由混沌映射动态系统产生。in,

is the chaotic factor, which is generated by the chaotic mapping dynamic system.

(4.3) Calculate the fitness value of _Vi , and determine the retained nectar source according to the method of greedy selection;

(4.4) Calculate the probability that the nectar source found by the leading bee will be followed;

(4.5) The follower bee searches in the same way as the leader bee, and determines the reserved nectar source according to the method of greedy selection;

(4.6) Judging whether the nectar source V _i satisfies the condition of being abandoned, if so, the corresponding leading bee role becomes a scout bee, otherwise, go directly to (4.8);

(4.7) Scout bees randomly generate new nectar sources;

(4.8) iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, and select the solution with the optimal fitness value as the optimal solution of the algorithm, otherwise go to step (4.2).

Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100.

(5) System update module 11, the chaotic multi-scale intelligent optimal propylene polymerization process measuring instrument also includes a system update module, which is used for online update of the system, regularly inputs offline test data into the training set, and updates the extreme random tree measurement model.

Taking the specific data as an example: the present embodiment extracts the required 9 modeling variables collected in the DCS system, and obtains the variable input matrix:

其中，

为丙烯聚合过程测量仪表5输出值,y_i为熔融指数离线化验值，则丙烯聚合过程测量仪表5的预报偏差为[0.0239,0.0311,-0.0075,-0.0004,-0.0228]，其均方根误差为0.0239，得到测量系统的熔融指数预报值与其预报精度。Input the data into the propylene polymerization process measuring instrument 5, and the chaotic multi-scale detection module obtains the predicted value of the melt index [2.4439, 2.4011, 2.4525, 2.4796, 2.4872]. The melt index offline assay value [2.42, 2.37, 2.46, 2.48, 2.51] is used as the calibration value of the propylene polymerization process measuring instrument 5, which is used to calculate the forecast error to evaluate the forecast accuracy of the polypropylene production quality measurement system 5. Root Mean Square Error (RMSE), its calculation formula is

in,

is the output value of the propylene polymerization process measuring instrument 5, and y _i is the offline assay value of the melt index, then the prediction deviation of the propylene polymerization process measuring instrument 5 is [0.0239, 0.0311, -0.0075, -0.0004, -0.0228], and its root mean square error is 0.0239, and the predicted value of the melt index of the measurement system and its prediction accuracy are obtained.

The above-mentioned embodiments are used to explain the present invention, rather than limit the present invention. Within the spirit of the present invention and the protection scope of the claims, any modifications and changes made to the present invention all fall into the protection scope of the present invention.

Claims (9)

1. a polypropylene quality detection system is used to carry out on-line detection to polypropylene product quality, it is characterised in that comprising: The chaotic reconstruction module is used to reconstruct the input variables of the model from the DCS database into dynamic chaotic system signals according to their chaotic characteristics; The Gabor multi-scale analysis module is used to analyze the multi-scale characteristics of the dynamic chaotic system signal based on the frequency, and perform multi-scale reconstruction of the dynamic chaotic system signal through the Gabor kernel function to obtain the local parts of each scale at different frequencies. texture feature information; The extreme random tree measurement model module is used to establish the mapping relationship between the chaotic multi-scale feature information and the melt index, establish a measurement system for the propylene polymerization process, and then predict the polypropylene melt index. 2. The polypropylene quality inspection system according to claim 1, wherein the input variable is the first propylene feed flow rate, the second propylene feed flow rate, and the third propylene feed flow rate , the flow rate of the main catalyst, the flow rate of the auxiliary catalyst, the temperature in the stirred tank, the pressure in the tank, the liquid level in the tank and the volume concentration of hydrogen in the tank. 3. The polypropylene quality detection system according to claim 1, wherein the chaotic system of the input variable is expressed as z(n)=[s(n), s(n+T ₁ ), s(n+T ₂ ),...,s(n+T _d-1 )], where s(n) is the signal of the nth sampling point in the propylene polymerization process, T ₁ , T ₂ ,..., T _d-1 are respectively is the sampling time after the nth sampling point. 4. polypropylene quality detection system according to claim 1 is characterized in that: the expression of Gabor kernel function is:

Among them, z represents the coordinate information of the reconstructed variable, u represents the direction of the Gabor filter, v represents the scale of the Gabor filter, i is the complex symbol, exp(ik _{u, v} z) is the oscillatory function in complex exponential form, σ ² is the kernel function width, and k _{u, v} represents the response of the Gabor filter in all directions at each scale. 5. polypropylene quality detection system according to claim 4 is characterized in that: Gabor feature is obtained by kernel function convolution, and expression is as follows: G _u,v (z)=f(z)*ψ _u,v (z) Among them, _Gu,v (z) represents the convolution function of the corresponding scale v and direction u near the coordinate z, and ψ is the Gabor kernel function; the input variable is analyzed by the Gabor function to obtain the complex input feature signal: Gu _,v (z)=Re(Gu _,v (z))+jIm(Gu _,v (z)) The amplitude and phase of the Gabor characteristic signal are:

6. The polypropylene quality detection system according to claim 1, wherein the establishment of the mapping relationship is that the input signal of a group of extreme random trees is the local texture feature information under a single scale, and multiple groups of extreme random trees are The chaotic multi-scale mapping modeling of input to output is completed based on the ensemble learning framework. 7. The polypropylene quality detection system according to claim 1, further comprising a chaotic artificial bee colony optimization module, which adopts a chaotic artificial bee colony algorithm to perform the bifurcation threshold parameter of the extreme random tree measurement model module. optimization. 8. The polypropylene quality detection system according to claim 7, wherein the optimization comprises the following steps: (1) Initialize the parameters of the artificial bee colony algorithm, set the number of nectar sources p, the maximum number of iterations iter _max , the minimum and maximum values of the initial search space L _d and U _d ; the position of the nectar source represents the feasible solution of the problem, measured in the correlation vector The dimension of p _i in the system is 2 dimensions, and the initial iteration number iter=0; (2) assign a leading bee to the nectar source p _i , and generate a new nectar source V _i based on the chaotic mapping;

in,

is the chaotic factor, which is generated by the chaotic mapping dynamic system. (3) Calculate the fitness value of _Vi , and determine the retained nectar source according to the method of greedy selection; (4) Calculate the probability that the nectar source found by the leading bee is followed up; (5) The following bees search in the same way as the leading bees, and determine the reserved nectar source according to the method of greedy selection; (6) judging whether the nectar source V _i satisfies the abandoned condition, if satisfied, the corresponding leading bee role becomes the scout bee, otherwise directly go to step (8); (7) Scout bees randomly generate new nectar sources; (8) iter=iter+1, judge whether the maximum number of iterations has been reached, if so, output the optimal parameters, and select the solution with the optimal fitness value as the optimal solution of the algorithm, otherwise go to step (2); Among them, the number of nectar sources is 100, the minimum and maximum values of the initial search space are 0 and 100, and the maximum number of iterations is 100. 9 . The polypropylene quality inspection system according to claim 1 , further comprising a system update module for inputting offline assay data into the training set to update the extreme random tree measurement model online. 10 .

Images