Quality prediction method, device and equipment of chemical product and storage medium

Abstract

The invention discloses a chemical product quality prediction method and device, equipment and a storage medium, and relates to the field of industrial big data. The method comprises the following steps: acquiring process operation parameters and actually measured quality parameters of a chemical product in a production process; based on the process operation parameters and the actually measured quality parameters, a training sample of the product quality prediction model is generated, and sample data of the training sample comprises the process operation parameters; wherein the product quality prediction model is an artificial intelligence model constructed based on a limit gradient lifting tree algorithm; training the product quality prediction model by adopting the training sample to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used for predicting the quality of the chemical product. According to the method, the product quality prediction model based on the limit gradient boosting tree algorithm is trained, so that the quality of the chemical product is predicted, and the labor cost and the medicament consumption are reduced.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of industrial big data, and in particular to a quality prediction method, device, equipment and storage medium for chemical products. Background Art

[0002] With the development of the digital age, the digitization and intelligentization of the petrochemical industry are ushering in a major transformation of the era. Innovative organizational and operating methods that deeply integrate industrial technology and information technology are constantly emerging in the industry.

[0003] Traditional product quality control methods require manual sampling and testing, which have the problems of slow detection speed and high latency. At present, related technologies obtain the data required for modeling by matching the similarity between the current sample and the historical data in the database, and then perform modeling based on the partial least squares method to obtain a product quality prediction model, which can predict product quality parameters through the model.

[0004] However, the model building process in the related technology requires matching processing of a large amount of data, which makes the model building process more complicated. Summary of the invention

[0005] The present application provides a method, device, equipment and storage medium for predicting the quality of chemical products. The technical solution is as follows:

[0006] According to one aspect of an embodiment of the present application, a method for predicting the quality of a chemical product is provided, the method comprising:

[0007] Obtaining process operation parameters and measured quality parameters of the chemical product during the production process, wherein the process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product;

[0008] Based on the process operation parameters and the measured quality parameters, a training sample of a product quality prediction model is generated, wherein the sample data of the training sample includes the process operation parameters, and the label data of the training sample includes the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an artificial intelligence (AI) model constructed based on an extreme gradient boosting tree algorithm;

[0009] The product quality prediction model is trained using the training samples to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of the chemical product.

[0010] According to one aspect of an embodiment of the present application, a device for predicting the quality of a chemical product is provided, the device comprising:

[0011] An acquisition module, used to acquire process operation parameters and measured quality parameters of the chemical product during the production process, wherein the process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product;

[0012] A generation module, used to generate training samples of a product quality prediction model based on the process operation parameters and the measured quality parameters, wherein the sample data of the training samples include the process operation parameters, and the label data of the training samples include the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an AI model constructed based on the extreme gradient boosting tree algorithm;

[0013] A training module uses the training samples to train the product quality prediction model to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of the chemical product.

[0014] According to one aspect of an embodiment of the present application, a computer device is provided, the computer device comprising a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the above-mentioned chemical product quality prediction method.

[0015] According to one aspect of an embodiment of the present application, a computer-readable storage medium is provided, in which a computer program is stored. The computer program is used to be executed by a processor to implement the above-mentioned chemical product quality prediction method.

[0016] According to one aspect of an embodiment of the present application, a computer program product is provided, which includes a computer program stored in a computer-readable storage medium, and a processor reads and executes the computer program from the computer-readable storage medium to implement the above-mentioned chemical product quality prediction method.

[0017] The technical solution provided by the embodiments of the present application includes at least the following beneficial effects:

[0018] A large number of high-quality training samples are generated by using the process operating parameters and measured quality parameters in the production process of chemical products. The process operating parameters in the training samples are used as sample data, and the measured quality parameters are used as label data. The sample data is used to train the product quality prediction model based on the extreme boosting tree algorithm, and the label data is continuously fitted to finally obtain a product quality prediction model that can accurately predict the quality parameters of the chemical product. The present application trains a product quality prediction model based on the extreme boosting tree algorithm, and the product quality prediction model can be used to predict the quality of chemical products. The prediction results are highly accurate, which fully reduces labor costs and drug consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of an implementation environment of a solution provided by an embodiment of the present application;

[0020] Figure 2 is a flow chart of a chemical product quality prediction method provided by an embodiment of the present application;

[0021] Figure 3 is a schematic diagram of a polyethylene production process provided by an embodiment of the present application;

[0022] Figure 4 is a training flow chart of a polyethylene product quality prediction model provided by an embodiment of the present application;

[0023] Figure 5 is a block diagram of a chemical product quality prediction device provided by an embodiment of the present application;

[0024] Figure 6 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0026] The Laboratory Information Management System (LIMS) is composed of computer hardware and application software, which can complete the collection, analysis, reporting and management of laboratory data and information. LIMS is based on a computer local area network and is specially designed for the overall environment of a laboratory. It is an efficient integrated system including signal acquisition equipment, data communication software, and database management software.

[0027] Distributed Control System (DCS) is a computerized control system used in factories or processes. Generally, there are several control loops, and autonomous controllers are scattered in the system without central operator supervision. This is the opposite of the concept of traditional centralized controllers (which may be many discrete controllers placed in a central control room, or a centrally controlled computer). Distributed control systems can increase the reliability of the system, place the controller close to the process, and then monitor and control it remotely, which can save the cost of setting up.

[0028] Please refer to Figure 1 , which shows a schematic diagram of a solution implementation environment provided by an embodiment of the present application. The solution implementation environment can be implemented as a training and use system for a product quality prediction model. The solution implementation environment may include: a model training device 10 and a model use device 20.

[0029] The model training device 10 may be an electronic device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a server, or other electronic devices with strong computing capabilities. The model training device 10 is used to train product quality prediction.

[0030] In the embodiment of the present application, the product quality prediction model is a machine learning model obtained through training, which is used to predict the quality parameters of chemical products in real time. For example, according to the real-time acquisition of the production data of the chemical product, such as reaction temperature, reaction pressure, hydrogen feed and other data, the quality parameters of the chemical product at this time are obtained after calculation by the product quality prediction model. The model training device 10 can train the product quality prediction model in a machine learning manner so that it has the ability to predict the quality parameters of chemical products in real time. The specific model training method can refer to the following embodiment.

[0031] The trained product quality prediction model can be deployed in the model using device 20 for use. The model using device 20 can be a terminal device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, or a server. When product quality prediction is required, the model using device 20 can implement the above product quality prediction function through the trained product quality prediction model.

[0032] The model training device 10 and the model using device 20 can be two independent devices or the same device.

[0033] In the method provided in the embodiment of the present application, the execution subject of each step may be a computer device, which refers to an electronic device with data calculation, processing and storage functions. The computer device may be a terminal device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, or a server. The server may be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. For example, the electronic device may be a Figure 1 The model training device 10 in.

[0034] Please refer to Figure 2 , which shows a flow chart of a chemical product quality prediction method provided by an embodiment of the present application. The method may include at least one of the following steps 210 to 230:

[0035] Step 210, obtaining the process operation parameters and measured quality parameters of the chemical product during the production process, wherein the process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product.

[0036] Chemical products are various chemical substances produced by the chemical industry, such as but not limited to polyethylene, polypropylene, polyvinyl chloride, polystyrene, pesticides, energy chemicals, etc. Chemical products can be any of the above chemical substances. For example, the chemical product is polyethylene. Based on the process operation parameters and measured quality parameters of polyethylene in the production process, a product quality prediction model for predicting the quality parameters of polyethylene products can be trained.

[0037] Process operating parameters, also known as process parameters, are controllable variables used in the production process of chemical products. They record the process of adjustment in the production process to ensure product quality, reduce costs, improve efficiency, etc. For example, the process operating parameters involved in the polyethylene production process may include ethylene feed, reaction temperature, reaction pressure, circulating gas volume, hydrogen feed, etc. Process operating parameters are usually monitored by industrial control systems and can be collected through different sensors. For example, the reaction temperature and reaction pressure of polyethylene in the production process are monitored by the DCS system, the reaction temperature is collected by the temperature sensor in the production device, and the reaction pressure is collected by the pressure sensor in the production device, and the collected data is transmitted to the DCS system. Since the process operating parameters involved in the production of different chemical products are different, the types and collection methods of specific process operating parameters are not limited in this application.

[0038] Measured quality parameters refer to indicators obtained through manual experiments and tests that can be used to evaluate the quality of chemical products. For example, the measured quality parameters of polyethylene are polymer density and melt index. By testing the polyethylene product to obtain the polymer density and melt index, it can be determined whether the polyethylene meets the quality requirements. Measured quality parameters are usually saved in an information management system, in which data management, data collation, data analysis, etc. can be performed on the measured quality parameters. For example, if the information system is a LIMS system, the measured quality parameters of chemical products are saved in the LIMS system and called when needed.

[0039] In some embodiments, data features are selected from process operation parameters. Data features are some variables with high correlation with product quality, which are a subset of the features of process operation parameters. There are many ways to select data features, including experience-based selection, filtering methods, wrapping methods, embedded methods, etc. Exemplarily, during the training process of the product quality prediction model, the importance of all features in the training data is calculated, sorted, and the data features with the strongest correlation with the target label are selected.

[0040] In this embodiment, a product quality prediction model for predicting the product quality index of a chemical product is trained based on the process operation parameters, data features and measured quality parameters in the production process of a chemical product, which can accurately predict the product quality index of the chemical product in real time. Different process operation parameters are collected for different chemical products, and product quality prediction models for different chemical products are trained in a targeted manner.

[0041] By collecting process operation parameters and measured quality parameters, a large amount of data can be obtained for training the product quality prediction model, which makes the product quality prediction model have better generalization ability and helps to improve the prediction effect of the product quality prediction model.

[0042] Step 220, based on the process operation parameters and the measured quality parameters, generate training samples of the product quality prediction model, the sample data of the training samples include the process operation parameters, and the label data of the training samples include the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an AI model constructed based on the extreme gradient boosting tree algorithm.

[0043] The Extreme Gradient Boosting (XGBoost) algorithm is a machine learning algorithm for solving classification and regression problems. It is an algorithm that optimizes and extends the Gradient Boosting Decision Trees (GBDT) algorithm. The XGBoost algorithm is specifically described in the following embodiments.

[0044] In some embodiments, the process operating parameters and the measured quality parameters are preprocessed to obtain preprocessed process operating parameters and preprocessed measured quality parameters; wherein the preprocessing includes at least one of the following: eliminating missing values, eliminating outliers, and eliminating noise; based on the preprocessed process operating parameters and the preprocessed measured quality parameters, training samples for a product quality prediction model are generated.

[0045] The method of preprocessing the process operation parameters and the measured operation quality parameters includes but is not limited to using specific tools, functions of existing models, etc. For example, using the scikit-learn library in Python for preprocessing.

[0046] Through the method of this embodiment, the process of preprocessing the process operation parameters and the measured operation quality parameters is realized, and high-quality training samples are obtained, which will have a better effect on the subsequent training of the product quality prediction model.

[0047] Step 230, using training samples to train the product quality prediction model to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of chemical products.

[0048] In some embodiments, a product quality prediction model is obtained by constructing a weak learner based on the process operation parameters included in the sample data of the training sample, where the weak learner refers to a machine learning model used for prediction; the parameters of the product quality prediction model are adjusted based on the training sample to obtain a trained product quality prediction model.

[0049] In the embodiment of the present application, the product quality prediction model is an additive model, which is composed of multiple weak learners to form a strong learner, wherein the weak learner adopts a decision tree model.

[0050] In some embodiments, a prediction value is determined based on process operation data; the residual between the prediction value and the measured quality parameter is calculated, wherein the residual is the difference between the initial prediction value and the measured quality parameter; the residual is used as label data to construct and train a weak learner, wherein the weak learner is used to indicate the prediction content corresponding to the process operation parameter; wherein the prediction content is used to fit the residual; the prediction value is updated, and new weak learners are continuously constructed in the same way until the residual is minimized; and a product prediction model is obtained based on the constructed weak learner.

[0051] The prediction ability of a single weak learner is relatively weak, but each weak learner has its own characteristics that it is good at predicting. By integrating multiple weak learners through the product quality prediction model, a model with stronger prediction ability can be constructed, and its output is the sum of the outputs of each weak learner. Exemplarily, the weak learner is a decision tree, and a process operation parameter is input into the current product quality prediction model. The process operation parameter will obtain the prediction content of each decision tree along the tree structure from the first decision tree, and the final product quality prediction value is the sum of all the prediction contents obtained.

[0052] In some embodiments, a first loss value is calculated based on the difference between the initial prediction value and the measured quality parameter, and the first loss value is used to measure the error between the initial prediction value and the measured quality parameter; a second loss value is calculated based on the regularization term included in the product quality prediction model, and the second loss value is used to measure the complexity of the product quality prediction model; a loss function value of the product quality prediction model is calculated based on the first loss value and the second loss value; a weak learner is constructed based on the loss function value and the residual. Among them, the second loss value is used to prevent the model from overfitting. The first loss value is usually calculated by the loss function of the XGBoost algorithm; the second loss value is usually the total complexity of the product quality prediction model, which can be calculated by the regularization term of XGBoost. The calculation method of the first loss value and the second loss value is introduced below and will not be repeated here.

[0053] In some embodiments, when the chemical product is a polyethylene product, the process operating parameters include at least one of the following: ethylene feed, reaction temperature, reaction pressure, circulating gas volume, hydrogen feed, reaction wall temperature, and catalyst feed amount; the quality evaluation results include at least one of the following: melt index and polymer density.

[0054] In this embodiment, the trained product quality prediction model is deployed on the server, and the process operation parameters one hour before and one hour after the time point at which the product quality needs to be predicted are obtained. The process operation parameters are processed to eliminate missing values, noise, outliers, etc., and the average value of the process operation data is taken and input into the product quality prediction model to obtain the final predicted quality parameters.

[0055] A large number of high-quality training samples are generated by using the process operating parameters and measured quality parameters in the production process of chemical products. The process operating parameters in the training samples are used as sample data, and the measured quality parameters are used as label data. The sample data is used to train the product quality prediction model based on the extreme boosting tree algorithm, and the label data is continuously fitted to finally obtain a product quality prediction model that can accurately predict the quality parameters of the chemical product. The present application trains a product quality prediction model based on the extreme boosting tree algorithm, and the product quality prediction model can be used to predict the quality of chemical products. The prediction results are highly accurate, which fully reduces labor costs and drug consumption.

[0056] Next, the process of obtaining the process operating parameters and measured quality parameters of chemical products during the production process is introduced and explained.

[0057] In some embodiments, the measured quality parameters corresponding to the first timestamp of a chemical product in the production process are obtained; the process operation parameters of the chemical product in a first time period in the production process are obtained; wherein the first time period refers to a time period obtained by combining a time period before the first timestamp and a time period after the first timestamp, and the process operation parameters in the first time period include: process operation parameters corresponding to multiple timestamps in the first time period; based on the process operation parameters corresponding to multiple timestamps in the first time period, the process operation parameters corresponding to the first timestamp are calculated; the process operation parameters corresponding to the first timestamp and the measured quality parameters corresponding to the first timestamp are taken as a set of corresponding process operation parameters and measured quality parameters.

[0058] The above-mentioned process operation parameters are collected at a frequency of seconds through the industrial control system, while the above-mentioned measured quality parameters are obtained by the operator at intervals of time. There is a significant difference in the amount of data between the two types of data. In this embodiment, according to the time of the measured quality parameter test, that is, the first timestamp, the process operation parameters for a period of time before the first timestamp and a period of time after the first timestamp are collected, and a process operation parameter corresponding to the first timestamp is determined in a certain way. At this time, the above-mentioned measured quality parameters and the obtained process operation parameters are regarded as a group of training samples. Among them, the method for determining the process operation parameter corresponding to the first timestamp can be a method of taking the median of the process operation parameter in the first time period, or a method of taking the average of the process operation parameter in the first time period, or other methods, which are not limited in this application.

[0059] The technical solution provided in this embodiment solves the problem of data volume mismatch between the measured quality parameters and the process operation parameters by determining the most representative process operation parameters in the first time period and associating them with the corresponding measured quality parameters.

[0060] Below, the principle of the XGBoost algorithm is introduced.

[0061] The XGBoost algorithm is an algorithm based on the GBDT algorithm. The GBDT algorithm is an additive model based on the idea of ​​boosting ensembles. During training, the forward distribution algorithm is used for greedy learning. Each iteration learns a decision tree to fit the residual between the prediction result of the previous decision tree and the true value of the training sample.

[0062] The basic idea of ​​the XGBoost algorithm is the same as that of the GBDT algorithm. The difference is that the objective function of the XGBoost algorithm consists of two parts: the loss function and the regularization term, as shown in the following formula:

[0063]

[0064] Where i is a positive integer; n is the total number of samples; K is the total number of decision trees; k is a positive integer; y i is the training data set T = {(x1, y1), (x2, y2), ..., (x n ,y n )}, the true value of the i-th sample, x i represents the i-th sample; is the predicted value of the i-th sample; Represents the complexity of k trees.

[0065] Next, the XGBoost objective function is derived:

[0066] The predicted value of the i-th sample can be expressed as the following formula, which means the predicted value of the i-th sample in the t-th round is the predicted value of the previous t-1 round and the sample x in the t-th decision tree i The predicted content of t (x i ) and:

[0067]

[0068] Then the objective function L of the tth round is (t) It can be expressed as:

[0069]

[0070] Assume l(x)=l(y i , x), for l(y i , x) in Performing a second-order Taylor expansion yields:

[0071]

[0072] make And the first-order derivative is The second-order derivative is You can get The second-order Taylor expansion of :

[0073]

[0074] Substituting it back into the objective function, we get:

[0075]

[0076] Since the structure of the first t-1 decision trees has been determined, is a constant term, and the objective function can be simplified to:

[0077]

[0078] because because is a constant, so the objective function can be simplified to:

[0079]

[0080] In this algorithm, the regularization term uses l2 regularization, that is:

[0081]

[0082] Where T is the number of leaf nodes of the t-th decision tree; w j is the predicted content of the jth leaf node.

[0083] All samples x belonging to the jth leaf node i , is classified into a leaf node sample set, that is, where j is a positive integer:

[0084] I j ={i|q(x i )=j}

[0085] At this time, the objective function is:

[0086]

[0087]

[0088] Assumptions That is, the final objective function is:

[0089]

[0090] Then when When the loss function can reach the minimum value At this point, the best decision tree is obtained.

[0091] Optionally, in the construction process of each weak learner, that is, a single decision tree, an exact greedy strategy is adopted at each step to select the feature that can reduce the loss function the most for splitting, until the loss function cannot be reduced or the reduction slows down.

[0092] Please refer to Figure 3 , which shows a schematic diagram of a polyethylene production device process provided by an embodiment of the present application. Figure 3 The polyethylene device shown has the following relationships:

[0093] Nitrogen S101, ethylene S102, and carbon monoxide S103 are input into the gas phase reactor R101;

[0094] The outlets of butene-1 tank V101, hexene-1 tank V102, and isopentane tank V103 are connected to the input end of gas phase reactor R101;

[0095] The output end of the nitrogen compressor C101 is connected to the input end of the gas phase reactor R101;

[0096] The top output end of the gas phase reactor R101 is connected to the input end of the circulating gas compressor C102;

[0097] The output end of the circulating gas compressor C102 is connected to the input end of the condensate collection tank V104;

[0098] The output end of the condensate collection tank V104 is connected to the input end of the purge gas compressor C103;

[0099] The purge gas compressor C103 is connected to the input end of the gas phase reactor R101;

[0100] The lower output end of the gas phase reactor R101 is connected to the input end of the intermediate tank V105;

[0101] The output end of the material processing intermediate tank V105 is connected to the input end of the polymer separation tank V106;

[0102] The bottom output end of the polymer separation tank V106 is connected to the input end of the devolatilization tank V107;

[0103] The top output end of the devolatilization tank V107 is connected to the top input end of the polymer separation tank V106;

[0104] The bottom output end of the devolatilization tank V107 is connected to the input end of the boundary area outer package S104.

[0105] By collecting data from sensors in the upper component, the process operating parameters in the polyethylene production process can be obtained.

[0106] The training process of the polyethylene product quality prediction model is as follows: Figure 4 shown.

[0107] The obtained process operating parameters and measured quality parameters are preprocessed to eliminate missing values, noise and outliers, and the process operating parameters are sorted according to the test time of the measured quality parameters to generate training samples; a product quality prediction model based on the extreme gradient boosting tree algorithm is constructed and trained, and the product quality prediction model includes multiple decision trees; a grid search method is used to perform hyperparameter tuning on the trained product quality prediction model according to the training samples to obtain the final product quality prediction model.

[0108] The technical solution of the present application can also be applied to the production scenarios of chemical products such as polyvinyl chloride, polystyrene, and plastics. The specific implementation method can refer to the above embodiments and will not be described in detail.

[0109] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0110] Please refer to Figure 5 , which shows a block diagram of a chemical product quality prediction device provided by an embodiment of the present application. The device has the function of implementing the above-mentioned chemical product quality prediction method, and the function can be implemented by hardware or by hardware executing corresponding software. The device can be the computer device described above, or it can be set in a computer device. Figure 5 As shown, the device 500 may include: an acquisition module 510 , a generation module 520 and a training module 530 .

[0111] The acquisition module 510 is used to obtain the process operation parameters and measured quality parameters of the chemical product during the production process. The process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product.

[0112] The generation module 520 is used to generate training samples of a product quality prediction model based on the process operation parameters and the measured quality parameters, wherein the sample data of the training samples include the process operation parameters, and the label data of the training samples include the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an AI model constructed based on the extreme gradient boosting tree algorithm.

[0113] The training module 530 uses the training samples to train the product quality prediction model to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of the chemical product.

[0114] In some embodiments, the acquisition module 510 is used to: obtain the measured quality parameters corresponding to the first timestamp of the chemical product in the production process; obtain the process operation parameters of the chemical product in a first time period in the production process; wherein the first time period refers to a time period obtained by combining a time period before the first timestamp and a time period after the first timestamp, and the process operation parameters in the first time period include: process operation parameters corresponding to multiple timestamps in the first time period; according to the process operation parameters corresponding to multiple timestamps in the first time period, the process operation parameters corresponding to the first timestamp are calculated; the process operation parameters corresponding to the first timestamp and the measured quality parameters corresponding to the first timestamp are taken as a set of corresponding process operation parameters and measured quality parameters.

[0115] In some embodiments, the generation module 520 is used to: preprocess the process operation parameters and the measured quality parameters to obtain preprocessed process operation parameters and preprocessed measured quality parameters; wherein the preprocessing includes at least one of the following: eliminating missing values, eliminating outliers, and eliminating noise; based on the preprocessed process operation parameters and the preprocessed measured quality parameters, generate training samples for the product quality prediction model.

[0116] In some embodiments, the training module 530 includes: a construction unit 531 and an adjustment unit 532 .

[0117] The construction unit 531 is used to obtain the product quality prediction model by constructing a weak learner according to the process operation parameters included in the sample data of the training sample, and the weak learner refers to a machine learning model used for prediction.

[0118] The adjusting unit 532 is used to adjust the parameters of the product quality prediction model according to the training samples to obtain the trained product quality prediction model.

[0119] In some embodiments, the construction unit 531 is used to: determine an initial prediction value according to the process operation data through the product quality prediction model; calculate the residual between the initial prediction value and the measured quality parameter, the residual being the difference between the initial prediction value and the measured quality parameter; construct and train a weak learner based on the residual as label data, the weak learner being used to indicate the prediction content corresponding to the process operation parameter; wherein the prediction content is used to fit the residual; update the initial prediction value, and continuously construct new weak learners in the same way until the residual is minimized; and obtain the product prediction model based on the constructed weak learner.

[0120] In some embodiments, the construction unit 531 includes: a determination subunit, a residual subunit, a construction subunit, an update subunit and an acquisition subunit.

[0121] The determination subunit is used to determine the predicted value according to the process operation data.

[0122] The residual subunit is used to calculate the residual between the predicted value and the measured quality parameter, where the residual is the difference between the predicted value and the measured quality parameter.

[0123] A subunit is constructed, which is used to use the residual as label data to construct and train a weak learner, wherein the weak learner is used to indicate the prediction content corresponding to the process operation parameter; wherein the prediction content is used to fit the residual.

[0124] An updating subunit, used for updating the predicted value, and continuously constructing new weak learners in the same way until the residual is minimized;

[0125] A subunit is obtained, which is used to obtain the product prediction model based on the constructed weak learner.

[0126] In some embodiments, the construction subunit is used to: calculate a first loss value based on the difference between the initial prediction value and the measured quality parameter, and the first loss value is used to measure the error between the initial prediction value and the measured quality parameter; calculate a second loss value based on the regularization term included in the product quality prediction model, and the second loss value is used to measure the complexity of the product quality prediction model; calculate the loss function value of the product quality prediction model based on the first loss value and the second loss value; and construct the weak learner based on the loss function value and the residual.

[0127] The technical solution provided in the embodiment of the present application generates a large number of high-quality training samples by using the process operating parameters and measured quality parameters in the production process of chemical products. The process operating parameters in the training samples are used as sample data, and the measured quality parameters are used as label data. The sample data is used to train the product quality prediction model based on the extreme boosting tree algorithm, and the label data is continuously fitted to finally obtain a product quality prediction model that can accurately predict the quality parameters of the chemical product. The present application trains a product quality prediction model based on the extreme boosting tree algorithm, and the product quality prediction model can be used to predict the quality of chemical products. The prediction results are highly accurate, which fully reduces labor costs and drug consumption.

[0128] One point that needs to be explained is that the device provided in the above embodiment only uses the division of the above-mentioned functional modules as an example to implement its functions. In actual applications, the above-mentioned functions can be assigned to different functional modules according to actual needs, that is, the content structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0129] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0130] Please refer to Figure 6 , which shows a schematic diagram of the structure of a computer device 600 provided in one embodiment of the present application. The computer device 600 can be used to execute the method steps in the above embodiment. The computer device 600 may include a bus 601, a user interface 602, a processor 603, a network interface 604 and a memory 605 to implement the functions of each module in the above device embodiment.

[0131] The bus 601 is used to connect the processor 603 and the memory 605; the processor 603 is used to execute the software program in the memory 605; the network interface 604 is used to communicate with the real-time database system and LIMS; the user interface 602 is used to display the results on the PC.

[0132] The memory 605 is a non-transitory computer-readable storage medium that can be used to store non-transitory software programs, non-transitory computer executable programs and modules. The processor 603 executes various functional applications and data processing of the electronic device by running the non-transitory software programs, instructions and modules stored in the memory 605, that is, the processing method of the above method embodiment is implemented.

[0133] The memory 605 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data, etc. In addition, the memory 605 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 605 may optionally include a memory remotely arranged relative to the processor 603, and these remote memories may be connected to the processing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0134] For details not described in detail in the above embodiments, please refer to the introduction in the above method embodiments, which will not be repeated here.

[0135] The embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored, and the computer program is used to be executed by a processor to implement the quality prediction method of the above-mentioned chemical product. Optionally, the computer-readable storage medium may include: a read-only memory (ROM), a random access memory (RAM), a solid-state drive (SSD) or an optical disk, etc. Among them, the random access memory may include a resistance random access memory (ReRAM) and a dynamic random access memory (DRAM).

[0136] An embodiment of the present application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. A processor reads and executes the computer program from the computer-readable storage medium to implement the above-mentioned chemical product quality prediction method.

[0137] The term "multiple" as used herein refers to two or more than two. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the related objects are in an "or" relationship.

[0138] The term “greater than or equal to” mentioned herein may mean greater than or equal to, or greater than, and the term “less than or equal to” may mean less than or equal to, or less than.

[0139] In addition, the step numbers described in this document only illustrate a possible execution order between the steps. In some other embodiments, the above steps may not be executed in the order of the numbers, such as two steps with different numbers are executed at the same time, or two steps with different numbers are executed in the opposite order to that shown in the figure. The embodiments of the present application are not limited to this.

[0140] The above description is only an exemplary embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the quality of a chemical product, characterized in that: The method comprises: Obtaining process operation parameters and measured quality parameters of the chemical product during the production process, wherein the process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product; Based on the process operation parameters and the measured quality parameters, a training sample of a product quality prediction model is generated, wherein the sample data of the training sample includes the process operation parameters, and the label data of the training sample includes the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an artificial intelligence AI model constructed based on an extreme gradient boosting tree algorithm; The product quality prediction model is trained using the training samples to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of the chemical product.

2. The method according to claim 1, characterized in that The adopting the training samples to train the product quality prediction model to obtain the trained product quality prediction model includes: According to the process operation parameters included in the sample data of the training samples, the product quality prediction model is obtained by constructing a weak learner, where the weak learner refers to a machine learning model used for prediction; The parameters of the product quality prediction model are adjusted according to the training samples to obtain the trained product quality prediction model.

3. The method according to claim 2, characterized in that The process of obtaining the product quality prediction model by constructing a weak learner based on the process operation parameters included in the sample data of the training sample includes: Determining a predicted value based on the process operation data; Calculating a residual between the predicted value and the measured quality parameter, the residual being the difference between the predicted value and the measured quality parameter; Using the residual as label data, constructing and training a weak learner, wherein the weak learner is used to indicate the prediction content corresponding to the process operation parameter; wherein the prediction content is used to fit the residual; Update the predicted value, and continuously construct new weak learners in the same way until the residual is minimized; Based on the constructed weak learner, the product prediction model is obtained.

4. The method according to claim 3, characterized in that The constructing a weak learner based on the residual includes: Calculate a first loss value according to the difference between the initial prediction value and the measured quality parameter, where the first loss value is used to measure the error between the initial prediction value and the measured quality parameter; According to the regularization term included in the product quality prediction model, a second loss value is calculated, where the second loss value is used to measure the complexity of the product quality prediction model; Calculating a loss function value of the product quality prediction model according to the first loss value and the second loss value; The weak learner is constructed according to the loss function value and the residual.

5. The method according to claim 1, characterized in that The generating of training samples of a product quality prediction model based on the process operation parameters and the measured quality parameters comprises: Preprocessing the process operation parameters and the measured quality parameters to obtain preprocessed process operation parameters and preprocessed measured quality parameters; wherein the preprocessing includes at least one of the following: eliminating missing values, eliminating abnormal values, and eliminating noise; Based on the preprocessed process operating parameters and the preprocessed measured quality parameters, training samples of the product quality prediction model are generated.

6. The method according to claim 1, characterized in that The obtaining of the process operation parameters and measured quality parameters of the chemical product during the production process includes: Obtaining a measured quality parameter corresponding to a first timestamp of the chemical product during the production process; Acquire process operation parameters of the chemical product in a first time period during the production process; wherein the first time period includes a time period before the first timestamp and a time period after the first timestamp, and the process operation parameters in the first time period include: process operation parameters corresponding to multiple timestamps in the first time period; Calculate the process operation parameter corresponding to the first timestamp according to the process operation parameters respectively corresponding to the multiple timestamps in the first time period; The process operation parameters corresponding to the first timestamp and the actually measured quality parameters corresponding to the first timestamp are taken as a set of corresponding process operation parameters and actually measured quality parameters.

7. The method according to any one of claims 1 to 6, characterized in that: The chemical product is a polyethylene product; The process operation parameters include at least one of the following: ethylene feed, reaction temperature, reaction pressure, circulating gas volume, hydrogen feed, reaction wall temperature, catalyst feed volume; The quality evaluation result includes at least one of the following: melt index and polymer density.

8. A quality prediction device for a chemical product, characterized in that: The device comprises: An acquisition module, used to acquire process operation parameters and measured quality parameters of the chemical product during the production process, wherein the process operation parameters are relevant parameters measured during the production process of the chemical product, and the measured quality parameters are quality evaluation results obtained by experimental testing of the chemical product; A generation module, used to generate training samples of a product quality prediction model based on the process operation parameters and the measured quality parameters, wherein the sample data of the training samples include the process operation parameters, and the label data of the training samples include the measured quality parameters corresponding to the process operation parameters; wherein the product quality prediction model is an artificial intelligence AI model constructed based on an extreme gradient boosting tree algorithm; A training module uses the training samples to train the product quality prediction model to obtain a trained product quality prediction model; wherein the trained product quality prediction model is used to predict the quality of the chemical product.

9. A computer device, characterized in that: The computer device comprises a processor and a memory, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and the computer program is used to be executed by a processor to implement the method according to any one of claims 1 to 7.

Images