Evaluation method of polymer polymerization products based on quantum support vector machine

Abstract

The present invention provides a method for evaluating polymer polymerization products based on a quantum support vector machine. The method collects and preprocesses raw data of reaction variables such as the amounts of different reaction monomers and the ratio of initiators involved in the polymerization reaction of adsorption-type polymer materials, as well as absorbance data of solutions after adsorption, and amplifies the data set using a quantum genetic optimization algorithm. A polymer polymerization product evaluation model is constructed based on the quantum support vector machine, and the model is trained and tested and evaluated using the amplified data set, so as to achieve the purpose of efficiently and accurately evaluating the adsorption performance of the polymerization product.

Description

Technical Field

[0001] The invention relates to a method for evaluating a polymer polymerization reaction product, and in particular to a method for evaluating a polymer polymerization reaction product based on a quantum support vector machine. Background Art

[0002] The polymer chemical polymerization process is complex and has many influencing factors, including different reaction monomer concentrations and their interaction relationships, the ratio of initiator to monomer dosage, reaction temperature, reaction time, stirring speed, drying time and temperature, etc. For adsorption-type polymer polymerization products, the particle size value, temperature and solution concentration in the post-processing procedure, as well as the interaction relationship between functional groups and adsorption media have an important influence on the adsorption performance of the polymerization products. Traditional research methods are based on a large number of comparative observations of experimental processes. There are direct or implicit complex interrelationships between variable factors and data, which have a certain impact on the analysis and optimization of test results. There are problems such as long analysis cycle, cumbersome experimental process, manual operation errors, chemical consumption and environmental pollution.

[0003] The use of machine learning algorithms to evaluate the performance of polymer polymerization reaction products can fully consider the direct or indirect interactions between data variables, reduce the impact of interfering factors such as operational errors on the results to a certain extent, and improve the efficiency of related tests and evaluations of polymerization reaction products.

[0004] Different from traditional research methods, applying machine learning algorithms to the performance evaluation of polymer polymerization products is a new and fast and efficient research model. The mechanism of polymer polymerization is relatively clear, and its product performance is closely related to factors such as monomer concentration ratio, reaction temperature and time, so the theoretical foundation is solid.

[0005] Among machine learning algorithms, property prediction using support vector machines is one of them. Although there are many real-life scenarios, such as rice taste quality analysis, water source category judgment, water chemical composition or gas concentration analysis, sewage COD value prediction, etc., there has not yet been any application of quantum support vector machines in the performance evaluation of polymer polymerization reaction products.

[0006] In order to solve the above problems, people have been seeking an ideal technical solution. Summary of the invention

[0007] The purpose of the present invention is to address the deficiencies of the prior art and thus provide a method for evaluating polymer polymerization products based on a quantum support vector machine.

[0008] In order to achieve the above object, the technical solution adopted by the present invention is: a method for evaluating polymer polymerization reaction products based on quantum support vector machine, comprising the following steps:

[0009] Step 1, obtaining the original data set related to the polymer polymerization reaction process;

[0010] Different reaction conditions were set, different reaction products were prepared by polymer polymerization reaction and classified and numbered, and they were in white powder after treatment. Under the same solubility, temperature and time, static adsorption capacity test was carried out with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers were taken, and absorbance was tested by ultraviolet spectrophotometer, and then the concentration value was calculated according to the standard curve, and finally the static adsorption capacity of the product was output;

[0011] Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size;

[0012] The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for evaluating polymer polymerization reaction products;

[0013] Step 2, preprocessing and amplification of the raw data set for evaluation of polymer polymerization reaction products;

[0014] The raw data set for evaluating the polymer polymerization reaction product is sequentially standardized to obtain a pre-amplification data set;

[0015] Based on the pre-amplified dataset, a quantum genetic optimization algorithm is used to generate the optimal artificial dataset in the sample space;

[0016] Repeat the above steps multiple times to generate multiple sets of different optimal artificial data sets, merge them with the pre-amplification data set, and generate a new raw data set for evaluating polymer polymerization reaction products;

[0017] The new artificial sample data set is divided into a training sample set and a test sample set in a ratio of 80 / 20%;

[0018] Step 3, constructing a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtaining a polymer polymerization reaction product evaluation model according to the constructed training sample set;

[0019] Step 4, verifying the correctness of the polymer polymerization reaction product evaluation model based on the polymer polymerization reaction product evaluation model trained in step 3 and the test sample set in step 2;

[0020] Step 5, in a real polymer polymerization reaction experimental environment, use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables, and calibrate the evaluation model according to the feedback deviation.

[0021] The second aspect of the present invention provides a polymer polymerization reaction product evaluation system based on quantum support vector machine, wherein the data acquisition module is configured to acquire the original data set related to the polymer polymerization reaction process.

[0022] Different reaction conditions were set, different reaction products were prepared by polymer polymerization reaction and classified and numbered, and they were in white powder after treatment. Under the same solubility, temperature and time, static adsorption capacity test was carried out with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers were taken, and absorbance was tested by ultraviolet spectrophotometer, and then the concentration value was calculated according to the standard curve, and finally the static adsorption capacity of the product was output;

[0023] Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size;

[0024] The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for evaluating polymer polymerization reaction products;

[0025] The original data set processing and classification module is configured to perform standardization processing on the original data set for evaluating the polymer polymerization reaction product in sequence to obtain a pre-amplification data set; based on the pre-amplification data set, a quantum genetic optimization algorithm is used to generate an optimal artificial data set in the sample space; the above steps are repeated multiple times to generate multiple groups of different optimal artificial data sets, which are merged with the pre-amplification data set to generate a new original data set for evaluating the polymer polymerization reaction product; the new artificial sample data set is divided into a training sample set and a test sample set at a ratio of 80 / 20%;

[0026] A polymer polymerization reaction product evaluation model training module is configured to construct a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtain a polymer polymerization reaction product evaluation model based on a constructed training sample set;

[0027] A model testing module is configured to verify the correctness of the polymer polymerization reaction product evaluation model based on the trained polymer polymerization reaction product evaluation model and the test sample set;

[0028] The prediction and correction module is configured to use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables in a real polymer polymerization reaction experimental environment, and to correct the evaluation model according to the feedback deviation.

[0029] Based on the above, based on the pre-amplified data set, the quantum genetic optimization algorithm is used to generate the optimal artificial data set in the sample space, including:

[0030] Set the population size, maximum number of genetic iterations, and chromosome length. Each individual in the population corresponds to a model input vector, and the chromosome length is the total length of the binary strings of all elements in the model input vector.

[0031] Initialize the population and randomly generate chromosomes encoded with a certain number of quantum bits;

[0032] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0033] The binary array obtained by measurement is converted into decimal according to the set range of each reaction variable, and is substituted into the support vector machine model to realize fitness evaluation for the measurement results, wherein the root mean square error is used as the fitness value function;

[0034] Record the original optimal individual and the corresponding optimal fitness value;

[0035] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0036] Conduct fitness assessment based on the measurement results;

[0037] Update the best fitness and related index information;

[0038] Use quantum rotating gates to update the probability amplitude of quantum bits to achieve genetic variation of population individuals and obtain new population individuals;

[0039] The number of iterations is increased by 1, and the original optimal individual and the corresponding optimal fitness value of the execution record are returned until all iterations are completed;

[0040] Finally, all the recorded optimal individuals and optimal fitness values ​​are output as the optimal artificial data set.

[0041] The present invention has outstanding substantive features and significant progress compared to the prior art. Specifically, the present invention collects original data and preprocesses data on reaction variables such as different reaction monomer amounts, initiator ratios, and absorbance data of the solution after adsorption involved in a polymerization reaction of an adsorption-type polymer material, and based on a quantum support vector machine algorithm, uses the preprocessed polymerization reaction data to train and test and evaluate the model, so as to achieve the purpose of efficiently and accurately evaluating the adsorption performance of the polymerization reaction product; in view of the defects that small sample data sets often have loose data structures, discrete distributions, information gaps between sample points, and inability to obtain effective information, a quantum genetic optimization algorithm is used to generate an optimal artificial data set that meets the requirements in the sample space based on the small sample data set, so as to achieve effective amplification of the data set, thereby meeting the training and testing requirements of subsequent evaluation models. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a flow chart of the method for evaluating a polymer polymerization reaction product of the present invention.

[0043] Figure 2 It is the SWAP-test quantum circuit diagram.

[0044] Figure 3 It is the quantum circuit diagram of the HHL algorithm for solving linear equations. DETAILED DESCRIPTION

[0045] The technical solution of the present invention is further described in detail below through specific implementation methods.

[0046] like Figure 1 As shown, a method for evaluating polymer polymerization reaction products based on quantum support vector machine comprises the following steps:

[0047] Step 1, obtaining the original data set related to the polymer polymerization reaction process;

[0048] The original test data collection step is the focus of the present invention. The quality of the test data is related to the training and testing of the subsequent model and is of great significance.

[0049] In this step, different reaction conditions are set for the polymer material polymerization process, different reaction products are prepared by polymer polymerization reaction and classified and numbered, and the products are in white powder form after treatment. Under the same solubility, temperature and time, the static adsorption amount is tested with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers are taken, and the absorbance is tested with an ultraviolet spectrophotometer at a wavelength of 590nm, and then the concentration value is calculated according to the standard curve, and finally the static adsorption amount of the product is output; the whole data collection process relatively comprehensively constructs the original data affecting the performance of the polymerization reaction product, which can be effectively used for the training and testing of the performance evaluation model of the polymerization reaction product;

[0050] The static adsorption amount is calculated by the formula: Γ=V(c o -c e ) / m s , where Γ is the static adsorption capacity, V is the volume of the test solution, m s is the mass of quartz sand, c o and c e are the effective concentrations of the solution before and after static adsorption, respectively;

[0051] Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size;

[0052] The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for evaluating polymer polymerization reaction products;

[0053] Step 2, preprocessing and amplification of the raw data set for evaluation of polymer polymerization reaction products;

[0054] The raw data related to the polymer polymerization reaction process obtained in step 1 has many problems such as different dimensions and large dimensions, which will lead to slow calculation speed and low calculation accuracy, which is not conducive to the subsequent training and testing of the support vector machine model. Therefore, it is necessary to perform standardization processing on the raw data set of the polymer polymerization reaction product evaluation obtained in step 1 in turn to obtain a pre-amplification data set;

[0055] In addition, due to the limitation of experimental conditions, the number of samples in the original data set of polymer polymerization reaction product evaluation obtained in step 1 is small. Small sample data sets often have defects such as loose data structure, discrete distribution, information gaps between sample points, and inability to obtain effective information. Therefore, the prediction model built based on it often fails to meet the accuracy requirements or has poor learning generalization ability, and is prone to "overfitting" phenomenon;

[0056] Therefore, after obtaining the pre-amplification data set, it is necessary to use the quantum genetic optimization algorithm to effectively amplify the data within the set conditions, and finally output the optimal individual and the best fitness value, where the optimal individual includes the optimal monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size required by the model;

[0057] Repeat the above steps multiple times to generate multiple sets of different optimal artificial data sets, merge them with the pre-amplification data set, generate new raw data sets for polymer polymerization reaction product evaluation, achieve effective amplification of the data set, and thus meet the training and testing requirements of subsequent evaluation models;

[0058] The original data set for evaluating new polymer polymerization products is divided into a training sample set and a test sample set at a ratio of 80 / 20%;

[0059] Step 3, constructing a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtaining a polymer polymerization reaction product evaluation model according to the constructed training sample set;

[0060] Step 4, verifying the correctness of the polymer polymerization reaction product evaluation model based on the polymer polymerization reaction product evaluation model trained in step 3 and the test sample set in step 2;

[0061] Specifically, the static adsorption value predicted by the polymer polymerization product evaluation model is calculated with the static adsorption value measured by the original test to solve the root mean square error evaluation index, and the calculation formula is: Where X t and X o are respectively the model prediction value and the actual test value of the static adsorption amount of the polymerization product, and N is the number of test samples, in order to verify the accuracy of the evaluation model of the polymer polymerization product after training; if the root mean square error between the predicted value and the actual value of the trained model is extremely small, it means that the difference between the predicted value and the actual value is small, and the evaluation model has good stability;

[0062] Step 5, in a real polymer polymerization reaction experimental environment, use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables, calibrate the evaluation model according to the feedback deviation, explore the impact of different influencing factors on the evaluation results, and form a mature and stable polymer polymerization reaction product evaluation model.

[0063] In view of the problem of less data in the original data set for evaluating polymer polymerization products, this application further introduces a quantum genetic optimization algorithm to generate new polymer polymerization test data within a reasonable range based on the original data set for evaluating polymer polymerization products, thereby achieving effective expansion of the data set to avoid defects such as poor prediction accuracy of the constructed model or insufficient learning generalization ability due to too few data set samples. Through the above data processing operation process, while retaining the authenticity of the data to the greatest extent, the influence of variables of different dimensions and small sample data is taken into account, and the original data set is transformed into a new polymer polymerization reaction data set.

[0064] Specifically, since the ultimate goal is to minimize the difference between the artificially generated data set and the real test data set, the root mean square error is used as the fitness value function of the quantum genetic optimization algorithm to generate the optimal artificial data set, and a new artificial sample data set is obtained within the set condition framework. The specific steps are as follows:

[0065] Set parameters such as population size, maximum number of genetic iterations, chromosome length, etc.;

[0066] Initialize the population and randomly generate chromosomes encoded with a certain number of quantum bits;

[0067] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0068] The binary array obtained by measurement is converted into decimal according to the set range of each reaction variable, and is substituted into the support vector machine model to realize fitness evaluation for the measurement results; wherein the fitness value function for fitness evaluation is the root mean square error;

[0069] Record the original optimal individual and the corresponding optimal fitness value;

[0070] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0071] Conduct fitness assessment based on the measurement results;

[0072] Update the best fitness and related index information;

[0073] Use quantum rotating gates to update the probability amplitude of quantum bits to achieve genetic variation of population individuals and obtain new population individuals;

[0074] The number of iterations is increased by 1, and the original optimal individual and the corresponding optimal fitness value of the execution record are returned until all iterations are completed;

[0075] Finally, all the recorded optimal individuals and optimal fitness values ​​are output as the optimal artificial data set.

[0076] Since the quantum genetic optimization algorithm can find the optimal solution in a shorter time based on a small sample data set, the small amount of sample data does not affect the algorithm performance, and the individual diversity in the population can be maintained. In addition, it has the advantages of high search efficiency, good global search capability and strong adaptability, and is very suitable for the amplification processing of small sample data sets related to the polymerization reaction test in the present invention.

[0077] Furthermore, the specific steps of step 3 are as follows:

[0078] Set the training sample set to in is the jth model input vector composed of the jth training sample; y j is the j-th model output vector composed of the j-th training sample; N is the dimension of the model input vector and M is the number of model input vectors, corresponding to the total number of training samples;

[0079] The goal is to find the optimal classification line Ensure that the above sample points are separated to the greatest extent possible, where is the weight vector, b is the bias constant, and the maximum classification interval is expressed as Right now Therefore, the classification problem will be transformed into the following optimization problem:

[0080]

[0081] Introducing error variable e j , used to indicate the allowable deviation of data points, then the objective function is transformed into

[0082] Among them, γ is the penalty coefficient;

[0083] Convert the constraints from inequality constraints to equality constraints:

[0084] Introducing the Lagrange function multiplier Construct the Lagrangian function:

[0085]

[0086] Among them, α j ≥0 corresponds to The Lagrange multiplier of is [α1,α2,…,α j ,…,α M ], is the error vector;

[0087] Based on the KKT condition, we can find the partial derivative of the Lagrangian function and obtain the linear equation system.

[0088] Where I = [1, 1, …, 1], and are the j-th and k-th model input vectors, j, k = 1, 2, …, M;

[0089] Based on the training sample set, the HHL quantum algorithm is used to solve the linear equations Solve and get the parameters of the optimal hyperplane in is α j The optimal solution of , and finally obtain the maximum interval hyperplane, to achieve the construction of the evaluation model of polymer polymerization reaction products, the mathematical expression of the function is:

[0090] In the specific implementation, SWAP-test is used to calculate the kernel function k(x j ,x k ), such as Figure 2 As shown:

[0091]

[0092] The above formula is expressed using Dirac notation, and They are and The corresponding quantum states, |0> and |1> are both auxiliary qubit states; for |0>,|1> corresponding tensor product; <0|1>,<0|0>, They are |0>,|1> corresponding inner product, where For the ultimate goal;

[0093] H1 refers to the Hadamard gate transformation of the first quantum state, Swap 2，3 It means exchanging the second quantum state and the third quantum state using the Swap gate; measure1 means measuring the first quantum state.

[0094] In the specific implementation, the specific steps of the HHL quantum algorithm circuit are as follows Figure 3 As shown, including:

[0095] Prepare input quantum states in is the input quantum state, and set u j is an auxiliary variable.

[0096] Convert the matrix F to a unitary operation The converted unitary matrix should meet the unitary property requirements of quantum computing for quantum gates. If the matrix is an s-sparse matrix, then the unitary matrix The simulation time is O(log(N));

[0097] Apply quantum phase estimation to the clock register and input register, decomposing on the eigenvector basis The phase estimation module includes Hadamard gate, unitary operation and inverse quantum Fourier transform; after this operation, the clock register and input register will obtain matrices respectively. (equivalent to e iFt )'s eigenvalue |λ j > and the eigenvector When the phase estimation is accurate, entanglement will occur between the two registers, and their values ​​will become

[0098] The clock register is used as a control qubit to rotate the auxiliary qubit and convert it into a superposition state of |0> and |1>; the controlled rotation operation is from the ground state |λ j > respectively extract λ j To probability amplitude and in The three register values ​​will become In actual operation, the auxiliary qubit extraction ratio directly clocks the eigenvalues ​​|λ in the register and the input register. j > and the eigenvector

[0099] Recover |λ using inverse phase estimation i >(i.e. |λ i >→|0>), this inverse phase estimation includes quantum Fourier transform, unitary operation, Hadamard gate, and then the amplitude amplification module is used to increase the amplitude to |1>;

[0100] Measure the auxiliary register. If the result is |1>, the result of the input register will be the same as Proportional calculation results;

[0101] After all the above processes are completed, the input register will be become in This allows the linear equations to be solved. The computational complexity of this process is reduced from O(M 2 ) becomes O(log(M)), based on the optimal Can be obtained respectively Finally, the maximum margin hyperplane is constructed.

[0102] In this step, the quantum support vector machine QSVM algorithm is introduced to train the polymer polymerization reaction product evaluation model, focusing on the inner product kernel function calculation and the solution of the linear equation system, and the computational complexity is reduced from O(log(∈ -1 )poly(N,M)) is reduced to O(log(N,M)), where M and N are the number of samples in the training set and the dimension of the feature vector respectively. At the same time, the influence of various complex variable factors such as the ratio of reaction monomers, the ratio of initiators, the reaction temperature, the reaction time, etc. on the performance of the experimental product can be fully considered, and the advantages of quantum algorithms in complex environments can be brought into play, the computational complexity can be reduced, and the prediction efficiency can be improved.

[0103] The second aspect of the present invention provides a polymer polymerization reaction product evaluation system based on quantum support vector machine, comprising:

[0104] The data acquisition module is configured to acquire raw data sets related to the polymer polymerization process

[0105] Different reaction conditions were set, different reaction products were prepared by polymer polymerization reaction and classified and numbered, and they were in white powder after treatment. Under the same solubility, temperature and time, static adsorption capacity test was carried out with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers were taken, and absorbance was tested by ultraviolet spectrophotometer, and then the concentration value was calculated according to the standard curve, and finally the static adsorption capacity of the product was output;

[0106] Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size;

[0107] The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for evaluating polymer polymerization reaction products;

[0108] The original data set processing and classification module is configured to perform standardization processing on the original data set for evaluating the polymer polymerization reaction product in sequence to obtain a pre-amplification data set; based on the pre-amplification data set, a quantum genetic optimization algorithm is used to generate an optimal artificial data set in the sample space; the above steps are repeated multiple times to generate multiple groups of different optimal artificial data sets, which are merged with the pre-amplification data set to generate a new original data set for evaluating the polymer polymerization reaction product; the new artificial sample data set is divided into a training sample set and a test sample set at a ratio of 80 / 20%;

[0109] A polymer polymerization reaction product evaluation model training module is configured to construct a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtain a polymer polymerization reaction product evaluation model based on a constructed training sample set;

[0110] A model testing module is configured to verify the correctness of the polymer polymerization reaction product evaluation model based on the trained polymer polymerization reaction product evaluation model and the test sample set;

[0111] The prediction and correction module is configured to use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables in a real polymer polymerization reaction experimental environment, and to correct the evaluation model according to the feedback deviation.

[0112] In specific implementation, based on the pre-amplified data set, the quantum genetic optimization algorithm is used to generate the optimal artificial data set in the sample space, including:

[0113] Set the population size, maximum number of genetic iterations, and chromosome length. Each individual in the population corresponds to a model input vector, and the chromosome length is the total length of the binary strings of all elements in the model input vector.

[0114] Initialize the population and randomly generate chromosomes encoded with a certain number of quantum bits;

[0115] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0116] The binary array obtained by measurement is converted into decimal according to the set range of each reaction variable, and is substituted into the support vector machine model to realize fitness evaluation for the measurement results, wherein the root mean square error is used as the fitness value function;

[0117] Record the original optimal individual and the corresponding optimal fitness value;

[0118] The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code;

[0119] Conduct fitness assessment based on the measurement results;

[0120] Update the best fitness and related index information;

[0121] Use quantum rotating gates to update the probability amplitude of quantum bits to achieve genetic variation of population individuals and obtain new population individuals;

[0122] The number of iterations is increased by 1, and the original optimal individual and the corresponding optimal fitness value of the execution record are returned until all iterations are completed;

[0123] Finally, all the recorded optimal individuals and optimal fitness values ​​are output as the optimal artificial data set.

[0124] Furthermore, based on the quantum support vector machine, a polymer polymerization reaction product evaluation model is constructed, and according to the constructed training sample set, the specific steps of obtaining the polymer polymerization reaction product evaluation model are as follows:

[0125] Set the training sample set to in is the jth model input vector composed of the jth training sample; y j is the j-th model output vector composed of the j-th training sample; N is the dimension of the model input vector and M is the number of model input vectors, corresponding to the total number of training samples;

[0126] The goal is to find the optimal classification line Ensure that the above sample points are separated to the greatest extent possible, where is the weight vector, b is the bias constant, and the maximum classification interval is expressed as Right now Therefore, the classification problem will be transformed into the following optimization problem:

[0127]

[0128] Introducing error variable e j , used to indicate the allowable deviation of data points, then the objective function is transformed into

[0129] Among them, γ is the penalty coefficient;

[0130] Convert the constraints from inequality constraints to equality constraints:

[0131] Introducing the Lagrange function multiplier Construct the Lagrangian function:

[0132]

[0133] Among them, α j ≥0 corresponds to The Lagrange multiplier of is [α1,α2,…,α j ,…,α M ], is the error vector;

[0134] Based on the KKT condition, we can find the partial derivative of the Lagrangian function and obtain the linear equation system.

[0135] Where I = [1, 1, …, 1], and are the j-th and k-th model input vectors, j, k = 1, 2, …, M;

[0136] Based on the training sample set, the HHL quantum algorithm is used to solve the linear equations Solve and get the parameters of the optimal hyperplane in is α j The optimal solution of , and finally obtain the maximum interval hyperplane, to achieve the construction of the evaluation model of polymer polymerization reaction products, the mathematical expression of the function is:

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. A method for evaluating polymer polymerization products based on quantum support vector machine, characterized in that: Step 1, obtaining the original data set related to the polymer polymerization reaction process; Different reaction conditions were set, different reaction products were prepared by polymer polymerization reaction and classified and numbered, and they were in white powder after treatment. Under the same solubility, temperature and time, static adsorption capacity test was carried out with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers were taken, and absorbance was tested by ultraviolet spectrophotometer, and then the concentration value was calculated according to the standard curve, and finally the static adsorption capacity of the product was output; Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size; The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for polymer polymerization reaction product evaluation; Step 2, preprocessing and amplification of the raw data set for evaluation of polymer polymerization reaction products; The raw data set for evaluating the polymer polymerization reaction product is sequentially standardized to obtain a pre-amplification data set; Based on the pre-amplified dataset, a quantum genetic optimization algorithm is used to generate the optimal artificial dataset in the sample space; Repeat the above steps multiple times to generate multiple sets of different optimal artificial data sets, merge them with the pre-amplification data set, and generate a new raw data set for evaluating polymer polymerization reaction products; The new artificial sample data set is divided into a training sample set and a test sample set in a ratio of 80 / 20%; Step 3, constructing a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtaining a polymer polymerization reaction product evaluation model according to the constructed training sample set; The specific steps of step 3 include: Set the training sample set to in is the jth model input vector composed of the jth training sample; y j is the j-th model output vector composed of the j-th training sample; N is the dimension of the model input vector and M is the number of model input vectors, corresponding to the total number of training samples; The goal is to find the optimal classification line Ensure that the above sample points are separated to the greatest extent possible, where is the weight vector, b is the bias constant, and the maximum classification interval is expressed as Right now Therefore, the classification problem will be transformed into the following optimization problem: Introducing error variable e j , used to indicate the allowable deviation of data points, then the objective function is transformed into Among them, γ is the penalty coefficient; Convert the constraints from inequality constraints to equality constraints: Introducing the Lagrange function multiplier Construct the Lagrangian function: Among them, α j ≥0 corresponds to The Lagrange multiplier of is [α1,α2,…,α j ,…,α M ], is the error vector; Based on the KKT condition, we can find the partial derivative of the Lagrangian function and obtain the linear equation system. Where I = [1, 1, …, 1], and are the j-th and k-th model input vectors, j, k = 1, 2, …, M; Based on the training sample set, the HHL quantum algorithm is used to solve the linear equations Solve and get the parameters of the optimal hyperplane in is α j The optimal solution of , and finally obtain the maximum interval hyperplane, to achieve the construction of the evaluation model of polymer polymerization reaction products, the mathematical expression of the function is: Among them, SWAP-test is used to calculate the kernel function k(x j ,x k ): The above formula is expressed using Dirac notation, and They are and The corresponding quantum states, |0> and |1> are both auxiliary qubit states; for The corresponding tensor product; They are The corresponding inner product is For the ultimate goal; H1 refers to the Hadamard gate transformation of the first quantum state, Swap 2，3 It means that the second quantum state and the third quantum state are exchanged using the Swap gate; measure1 means measuring the first quantum state; Step 4, verifying the correctness of the polymer polymerization reaction product evaluation model based on the polymer polymerization reaction product evaluation model trained in step 3 and the test sample set in step 2; Step 5, in a real polymer polymerization reaction experimental environment, use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables, and calibrate the evaluation model according to the feedback deviation.

2. The method for evaluating polymer polymerization products based on quantum support vector machine according to claim 1, characterized in that: Based on the pre-amplified dataset, the quantum genetic optimization algorithm is used to generate the optimal artificial dataset in the sample space, including: Set the population size, maximum number of genetic iterations, and chromosome length. Each individual in the population corresponds to a model input vector, and the chromosome length is the total length of the binary strings of all elements in the model input vector. Initialize the population and randomly generate chromosomes encoded with a certain number of quantum bits; The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code; The binary array obtained by measurement is converted into decimal according to the set range of each reaction variable, and is substituted into the support vector machine model to realize fitness evaluation for the measurement results, wherein the root mean square error is used as the fitness value function; Record the original optimal individual and the corresponding optimal fitness value; The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code; Conduct fitness assessment based on the measurement results; Update the best fitness and related index information; Use quantum rotating gates to update the probability amplitude of quantum bits to achieve genetic variation of population individuals and obtain new population individuals; The number of iterations is increased by 1, and the original optimal individual and the corresponding optimal fitness value of the execution record are returned until all iterations are completed; Finally, all the recorded optimal individuals and optimal fitness values ​​are output as the optimal artificial data set.

3. A polymer polymerization reaction product evaluation system based on quantum support vector machine, characterized in that: include: The data acquisition module is configured to acquire raw data sets related to the polymer polymerization process Different reaction conditions were set, different reaction products were prepared by polymer polymerization reaction and classified and numbered, and they were in white powder after treatment. Under the same solubility, temperature and time, static adsorption capacity test was carried out with simulated salt water as solvent and quartz sand as adsorption medium. Samples of solutions before and after adsorption corresponding to different numbers were taken, and absorbance was tested by ultraviolet spectrophotometer, and then the concentration value was calculated according to the standard curve, and finally the static adsorption capacity of the product was output; Among them, different reaction conditions include monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size; The monomer dosage, initiator ratio, reaction temperature, reaction time, drying temperature, drying time and fine particle size are used as model input vectors, and the static adsorption amount is used as the model output vector to form the original data set for polymer polymerization reaction product evaluation; The original data set processing and classification module is configured to perform standardization processing on the original data set for evaluating the polymer polymerization reaction product in sequence to obtain a pre-amplification data set; based on the pre-amplification data set, a quantum genetic optimization algorithm is used to generate an optimal artificial data set in the sample space; the above steps are repeated multiple times to generate multiple groups of different optimal artificial data sets, which are merged with the pre-amplification data set to generate a new original data set for evaluating the polymer polymerization reaction product; the new artificial sample data set is divided into a training sample set and a test sample set at a ratio of 80 / 20%; A polymer polymerization reaction product evaluation model training module is configured to construct a polymer polymerization reaction product evaluation model based on a quantum support vector machine, and obtain a polymer polymerization reaction product evaluation model based on a constructed training sample set; The polymer polymerization product evaluation model training module constructs a polymer polymerization product evaluation model based on a quantum support vector machine, and obtains the polymer polymerization product evaluation model according to the constructed training sample set. The specific steps include: Set the training sample set to in is the jth model input vector composed of the jth training sample; y j is the j-th model output vector composed of the j-th training sample; N is the dimension of the model input vector and M is the number of model input vectors, corresponding to the total number of training samples; The goal is to find the optimal classification line Ensure that the above sample points are separated to the greatest extent possible, where is the weight vector, b is the bias constant, and the maximum classification interval is expressed as Right now Therefore, the classification problem will be transformed into the following optimization problem: Introducing error variable e j , used to indicate the allowable deviation of data points, then the objective function is transformed into Among them, γ is the penalty coefficient; Convert the constraints from inequality constraints to equality constraints: Introducing the Lagrange function multiplier Construct the Lagrangian function: Among them, α j ≥0 corresponds to The Lagrange multiplier of is [α1,α2,…,α j ,…,α M ], is the error vector; Based on the KKT condition, we can find the partial derivative of the Lagrangian function and obtain the linear equation system. Where I = [1, 1, …, 1], and are the j-th and k-th model input vectors, j, k = 1, 2, …, M; Based on the training sample set, the HHL quantum algorithm is used to solve the linear equations Solve and get the parameters of the optimal hyperplane in is α j The optimal solution of , and finally obtain the maximum interval hyperplane, to achieve the construction of the evaluation model of polymer polymerization reaction products, the mathematical expression of the function is: Among them, SWAP-test is used to calculate the kernel function k(x i ,x j ): The above formula is expressed using Dirac notation, and They are and The corresponding quantum states, |0> and |1> are both auxiliary qubit states; for The corresponding tensor product; They are The corresponding inner product is For the ultimate goal; H1 refers to the Hadamard gate transformation of the first quantum state, Swap 2，3 It means that the second quantum state and the third quantum state are exchanged using the Swap gate; measure1 means measuring the first quantum state; A model testing module is configured to verify the correctness of the polymer polymerization reaction product evaluation model based on the trained polymer polymerization reaction product evaluation model and the test sample set; The prediction and correction module is configured to use the tested polymer polymerization reaction product evaluation model to predict and comprehensively analyze the experimental results under more different types of reaction process variables in a real polymer polymerization reaction experimental environment, and to correct the evaluation model according to the feedback deviation.

4. The polymer polymerization product evaluation system based on quantum support vector machine according to claim 3 is characterized in that: Based on the pre-amplified dataset, the quantum genetic optimization algorithm is used to generate the optimal artificial dataset in the sample space, including: Set the population size, maximum number of genetic iterations, and chromosome length. Each individual in the population corresponds to a model input vector, and the chromosome length is the total length of the binary strings of all elements in the model input vector. Initialize the population and randomly generate chromosomes encoded with a certain number of quantum bits; The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code; The binary array obtained by measurement is converted into decimal according to the set range of each reaction variable, and is substituted into the support vector machine model to realize fitness evaluation for the measurement results, wherein the root mean square error is used as the fitness value function; Record the original optimal individual and the corresponding optimal fitness value; The individuals in the population are measured sequentially to obtain a certain state, i.e., a binary code; Conduct fitness assessment based on the measurement results; Update the best fitness and related index information; Use quantum rotating gates to update the probability amplitude of quantum bits to achieve genetic variation of population individuals and obtain new population individuals; The number of iterations is increased by 1, and the original optimal individual and the corresponding optimal fitness value of the execution record are returned until all iterations are completed; Finally, all the recorded optimal individuals and optimal fitness values ​​are output as the optimal artificial data set.

Images