CN117500845A

CN117500845A - Prediction apparatus, manufacturing apparatus, and prediction method

Info

Publication number: CN117500845A
Application number: CN202280042560.6A
Authority: CN
Inventors: 伊与田淳平
Original assignee: Daikin Industries Ltd
Current assignee: Daikin Industries Ltd
Priority date: 2021-06-17
Filing date: 2022-06-07
Publication date: 2024-02-02
Also published as: JP2023000306A; WO2022264885A1

Abstract

In the production of a fluororesin, the MFR value of a mixture containing the fluororesin in the polymerization step is predicted. A prediction device (10) predicts the MFR value of a fluororesin in a polymerization vessel at a predetermined time after the raw material is fed and before the completion of the production in a fluororesin polymerization step, comprising: an acquisition unit (112) that acquires, as a sensor value for prediction, a pressure value in a polymerization tank measured in a current polymerization step of a fluororesin and a stirring current value of a stirrer that stirs the inside of the polymerization tank; and a prediction unit (113) for predicting the MFR value at a predetermined time of the mixture containing the fluororesin in the current polymerization step by using the prediction sensor value acquired by the acquisition unit (112) and using a learned model obtained by learning by machine learning the relationship between the learning sensor value measured in the past polymerization step of the fluororesin of the same type as the fluororesin and the MFR value measured in the fluororesin at the predetermined time of the past polymerization step.

Description

Prediction apparatus, manufacturing apparatus, and prediction method

Technical Field

The present invention relates to a prediction apparatus, a prediction method, and a production apparatus for producing a fluororesin, which predict an MFR value in a polymerization step of a fluororesin.

Background

Patent document 1 describes a method for producing a thermoplastic resin composition. Patent document 1 describes a preferable range of Melt Flow Rate (MFR) of a fluororesin and a measurement method thereof. In this way, MFR was measured and performance was evaluated at the time of manufacturing a fluororesin or the like.

Prior art literature

Patent literature

Patent document 1: japanese patent laid-open publication No. 2019-151833

Disclosure of Invention

Technical problem to be solved by the invention

The present invention provides a prediction device, a prediction method and a manufacturing device capable of predicting an MFR value of a mixture containing a fluororesin without performing MFR measurement in a fluororesin polymerization process.

Technical scheme for solving technical problems

The prediction device of the present invention predicts an MFR value of a mixture containing a fluororesin in a polymerization vessel at a time point after a raw material is fed and before completion of production in a polymerization step of the fluororesin, and comprises: an acquisition unit that acquires information including a pressure value in the polymerization tank measured in the current polymerization step of the fluororesin and a stirring current value of a stirrer stirring the inside of the polymerization tank, as sensor values for prediction; and a prediction unit for predicting the MFR value at the predetermined time of the mixture containing the fluororesin in the current polymerization step by using a learned model obtained by learning a relation between a learning sensor value measured in the past polymerization step of the fluororesin of the same type and the MFR value measured for the fluororesin at the predetermined time of the past polymerization step by machine learning, and using the prediction sensor value obtained by the obtaining unit.

The prediction device further includes a timer unit for measuring a current elapsed time from a start of the current polymerization step as a polymerization time, and the prediction unit predicts the MFR value using the sensor value for prediction and the polymerization time measured by the timer unit.

The prediction apparatus further includes an output processing unit that outputs information about the end time of the current polymerization step obtained by using the MFR value predicted by the prediction unit.

In the above-described prediction apparatus, in the polymerization step of the fluororesin, the material to be charged may be added to the polymerization vessel, the acquisition unit may acquire the pressure value and the stirring current value, and an integrated amount of the material to be charged into the polymerization vessel as the prediction sensor value, and the prediction unit may predict the MFR value using the prediction sensor value by using a learned model obtained by learning using a learning sensor value including the integrated amount obtained in the past polymerization step.

In the above-described prediction apparatus, the prediction unit predicts the MFR value using the pressure value measured during a part of the polymerization process of the fluororesin.

In the above-described prediction apparatus, the prediction unit predicts the MFR value using at least one of the average value, the maximum value, the minimum value, and the variance of the pressure values during the partial period.

In the above-described prediction apparatus, the prediction unit predicts the MFR value using the stirring current value measured during a part of the polymerization process of the fluororesin.

In the above-described prediction apparatus, the prediction unit predicts the MFR value using an average value, a maximum value, a minimum value, or a variance of the stirring current value during the partial period.

In the above-described prediction apparatus, the partial period is selected from a plurality of periods obtained by dividing the time from the start to the end of the polymerization step of the fluororesin.

In the above-mentioned prediction apparatus, the above-mentioned polymerization step of the fluororesin is carried out at least 2 times, the above-mentioned obtaining part obtains the MFR value of the fluororesin obtained in the previous polymerization step, and the above-mentioned prediction part predicts the MFR value of the above-mentioned fluororesin using the sensor value for prediction including the MFR value of the fluororesin obtained in the previous polymerization step obtained by the above-mentioned obtaining part.

The production apparatus of the present invention is an apparatus for producing a fluororesin, comprising: a polymerization tank into which the fluororesin material can be introduced; a stirrer for stirring the polymerization tank; an acquisition unit that acquires, as a sensor value for prediction, a pressure value in the polymerization tank and a stirring current value of the stirrer, which are measured in a current polymerization step of the fluororesin; a prediction unit for predicting the MFR value of the fluororesin-containing mixture in the current polymerization step by using a learned model obtained by learning a relation between a learning sensor value measured in a past polymerization step of a fluororesin of the same type as the fluororesin, a time taken for the past polymerization step, and an MFR value measured for the fluororesin obtained in the past polymerization step by machine learning, and using the prediction sensor value obtained by the obtaining unit; and an output processing unit that outputs information on the end time of the current polymerization step, using the MFR value predicted by the prediction unit.

The method for producing a fluororesin of the present invention comprises: a step of charging a fluororesin material into a polymerization vessel; stirring the inside of the polymerization tank; a step of acquiring, as a sensor value for prediction, a pressure value in the polymerization vessel and a stirring current value of the stirrer, which are measured in a current polymerization step of the fluororesin; a step of predicting the MFR value of the fluororesin-containing mixture in the current polymerization step by using the obtained prediction sensor value by means of a learned model obtained by learning the relation between the learning sensor value measured in the past polymerization step of the fluororesin of the same type as the fluororesin, the time taken in the past polymerization step, and the MFR value measured for the fluororesin obtained in the past polymerization step; and a step of ending the current production using information on the end time of the current polymerization step obtained by using the predicted MFR value.

The prediction method of the present invention is a prediction method for predicting an MFR value of a mixture containing a fluororesin in a polymerization vessel in a polymerization step of the fluororesin, and comprises: an acquisition step of acquiring, as a sensor value for prediction, a pressure value in the polymerization vessel and a stirring current value of a stirrer stirring the inside of the polymerization vessel, which are measured in a current polymerization step of the fluororesin; and a prediction step of predicting the MFR value of the mixture containing the fluororesin in the current polymerization step using a learned model obtained by learning a relation between the learning sensor value measured in the past polymerization step of the fluororesin of the same type as the fluororesin, the time taken for the past polymerization step, and the MFR value measured for the fluororesin obtained in the past polymerization step by machine learning, and using the prediction sensor value obtained by the obtaining unit.

The above-described production method is a production method for producing a learned model for predicting the MFR value of a fluororesin in a polymerization vessel in a polymerization step of the fluororesin, and comprises: a step of acquiring a plurality of sets of data including training data (training data) including a sensor value measured in a polymerization step of a fluororesin of the same kind as the fluororesin and a polymerization time used in the polymerization step, and correct result data (correct answer date) for the training data, the correct result data being an MFR value measured for the fluororesin obtained in the polymerization step; and generating a learned model from the sensor values obtained in the polymerization step of the fluororesin based on the plurality of training data and the accurate result data by using a learner, the learned model having the MFR value of the fluororesin in the polymerization step as an output.

These general and specific ways may be implemented by systems, methods, and computer programs, as well as combinations thereof.

Effects of the invention

According to the prediction apparatus, the production apparatus, and the prediction method of the present invention, the MFR value of the mixture containing the fluororesin in the polymerization step can be predicted when the fluororesin is produced.

Drawings

Fig. 1 is a block diagram showing a configuration of a manufacturing apparatus.

Fig. 2A is a conceptual diagram showing generation of a learned model.

Fig. 2B is a conceptual diagram showing the use of a learned model.

Fig. 3 is a conceptual diagram illustrating learning data for generating a learned model.

Fig. 4 is a diagram showing an example of a display screen for outputting a prediction result obtained by the prediction device.

Fig. 5 is a conceptual diagram illustrating learning data.

Fig. 6 is a diagram illustrating an example of data items of learning data.

Fig. 7 is a conceptual diagram illustrating an example of acquiring data.

Fig. 8 is a flowchart illustrating a process of the manufacturing apparatus.

Detailed Description

The prediction apparatus, the manufacturing apparatus, and the prediction method according to the embodiments are described below with reference to the drawings. The prediction device of the present invention does not need to actually measure the MFR value of the fluororesin-containing mixture in the polymerization step, but performs prediction using various measured sensor values. Thus, the prediction means is a so-called "soft sensor". The production apparatus of the present invention predicts the MFR value of the fluororesin. In the following description, the prediction apparatus is described as being included in a manufacturing apparatus for manufacturing a fluororesin. In the following description, the same components are denoted by the same reference numerals, and description thereof is omitted.

In the following description, the "polymerization step" refers to a polymerization step of a fluororesin. Specifically, the "polymerization step" will be described as a process of obtaining a fluororesin from a material for producing a fluororesin by polymerization reaction.

Fluorine resin

The "fluororesin" is a resin containing fluorine atoms, and is a resin obtained by polymerizing monomers containing at least 1 kind of fluorine-containing monomer.

Examples of the fluorine-containing monomer include tetrafluoroethylene, vinylidene fluoride, hexafluoropropylene, chlorotrifluoroethylene, trifluoroethylene, trifluoropropene, tetrafluoropropene, pentafluoropropene, trifluorobutene, tetrafluoroisobutene, hexafluoroisobutene,Fluoroethylene, CH ₂ ＝CFRf ^a Indicated (perfluoroalkyl) ethylene, CH ₂ ＝CF(ORf ^a ) The (perfluoroalkyl) vinyl ether, CH ₂ ＝CF(Rf ^b ORf ^a ) Perfluoro (alkoxyalkyl vinyl ether) of the formula [ wherein Rf ^a Is a linear or branched perfluoroalkyl group having 1 to 6 carbon atoms, rf ^b Is a linear or branched perfluoroalkylene group having 1 to 6 carbon atoms, preferably 1 to 3 carbon atoms]Etc. The Rf described above ^a The linear or branched perfluoroalkyl group is preferably a linear or branched perfluoroalkyl group having 1 to 3 carbon atoms, more preferably a linear or branched perfluoroalkyl group having 1 or 2 carbon atoms. In one embodiment, the perfluoroalkyl group is linear. In another embodiment, the perfluoroalkyl group is branched. The Rf described above ^b The linear or branched perfluoroalkylene group is preferably one having 1 to 3 carbon atoms, and more preferably one having 1 or 2 carbon atoms. In one embodiment, the Quan Fuya alkyl group is a linear chain. In another embodiment, the Quan Fuya alkyl group is branched.

Examples of the other monomer include ethylene, propylene and CH ₂ In the formula =cf (OR) [ wherein R is a linear OR branched alkyl group having 1 to 6 carbon atoms]Alkyl vinyl ethers as shown, and the like. The above R is preferably a linear or branched alkyl group having 1 to 3 carbon atoms, more preferably a linear or branched alkyl group having 1 or 2 carbon atoms. In one embodiment, the alkyl group is a straight chain. In another embodiment, the alkyl group is branched.

The fluororesin may be a homopolymer of the above-mentioned fluoromonomer or a copolymer of 2 or more monomers including at least 1 of the above-mentioned fluoromonomer.

Examples of the fluororesin include Polytetrafluoroethylene (PTFE), polychlorotrifluoroethylene (PCTFE), tetrafluoroethylene (TFE)/Hexafluoropropylene (HFP) copolymers, TFE/HFP/perfluoro (alkyl vinyl ether) (PAVE) copolymers, TFE/PAVE copolymers, (tetrafluoroethylene-perfluoroalkyl vinyl ether copolymers (PFA) and tetrafluoroethylene-perfluoromethyl vinyl ether copolymers (MFA)), ethylene (Et)/TFE copolymers, et/TFE/HFP copolymers, chlorotrifluoroethylene (CTFE)/TFE copolymers, et/CTFE copolymers, TFE/vinylidene fluoride (VDF) copolymers, VDF/HFP/TFE copolymers, and VDF/HFP copolymers.

In a preferred embodiment, the fluororesin is a fluororesin containing a tetrafluoroethylene unit. Examples of the fluororesin containing a tetrafluoroethylene unit include a homopolymer of TFE, a copolymer of TFE/HFP, TFE/PAVE, TFE/ethylene, TFE/vinyl ether, TFE/vinyl ester/vinyl ether, and a copolymer of TFE/vinyl ether/allyl ether. In a preferred manner, the fluororesin may be TFE/ethylene.

In the case where the fluororesin is a fluororesin containing tetrafluoroethylene units, the content of the tetrafluoroethylene units is preferably 20 to 80 mol%, more preferably 30 to 70 mol%, and even more preferably 35 to 60 mol%. In particular, in the case where the fluororesin is an ethylene/tetrafluoroethylene copolymer, the molar ratio of ethylene/tetrafluoroethylene is preferably 20/80 to 80/20, more preferably 25/75 to 70/30, still more preferably 30/70 to 60/40.

The melting point of the fluororesin is preferably 200℃or higher, more preferably 230℃or higher.

The melt flow rate of the fluororesin is preferably 0.1 to 60, more preferably 1 to 50. The melt flow rate is a value obtained by heating the resin to be measured in a cylindrical barrel, applying a constant load to the heated resin, extruding the heated resin, and measuring the amount of the extruded resin within 10 minutes. The higher the MFR value, the better the flow under the same temperature, load conditions. For example, the above value is an example of a value obtained when a cylinder made of hastelloy is used and the temperature is 300 ℃ and the load is 5.00 kg.

Machine learning

The "machine learning" is a technique of learning features included in input data and generating a "model" for estimating a result corresponding to newly input data. Specifically, machine learning uses a "learner" for learning. The model thus generated is referred to as a "learned model".

As shown in fig. 2A, a plurality of sets of learning data are input to the "learner". Thus, the learner learns the relationship of the learning data, and generates a learned model that parametrically represents the relationship of the learning data. Further, by using the generated learned model, as shown in fig. 2B, an output to be obtained for a new input can be obtained. Machine learning includes supervised learning, unsupervised learning, reinforcement learning, and the like, and examples of using supervised learning are described in the following embodiments. Therefore, fig. 2A also shows an example of supervised learning of the data "objective function" associated with the correct result corresponding to each "explanatory function". As the learner, for example, a Neural Network (NN), a Support Vector Machine (SVM), a decision tree, gradient boosting, random forest, XGBoost, linear regression, ridge regression, lasso regression, elastic net (elastic network regression), partial least squares regression, gaussian process regression, principal Component Analysis (PCA), and the like can be cited as the machine learning technique. The learner may use these techniques in combination.

Here, "learning data" is data used when generating a learning model in machine learning. In the case of supervised learning, an "explanatory function" and a "objective function" as a correct result corresponding to the explanatory function are input to the learner as a set of learning data. More specifically, in machine learning, as shown in fig. 3, in order to generate a learned model, a "training data set" formed of a plurality of sets of "description function" and "objective function" as a set is used. It is also possible to use a part of the prepared data set as a "learning data set" to generate a learning model, and then use another part of the data set as a "test data set" to test and improve the prediction accuracy of the generated learning model. The other part of the data set can be further used as a test data set, and finally the prediction precision of the checked and improved learning model is confirmed. The "explanation function" may be a pressure value in the polymerization vessel, a stirring current value of the stirrer, or the like. The "objective function" is the MFR value of the fluororesin-containing mixture.

Manufacturing device

As shown in fig. 1, the manufacturing apparatus 1 includes a prediction apparatus 10, a polymerization tank 2, a stirrer 3, and a control apparatus 4. The fluororesin material is charged into the polymerization vessel 2. For example, the material is fed at the beginning of the polymerization process. In addition, a part of the material may be added in the polymerization step. The polymerization tank 2 is hermetically formed to block outside air and pressure and temperature. The polymerization tank 2 is connected to a pressure sensor 21 for measuring an internal pressure value and/or a temperature sensor 22 for measuring an internal temperature value. The sensor values to which the respective sensors 21, 22 are connected are acquired by the prediction apparatus 10.

The stirrer 3 stirs the mixture containing the material charged into the polymerization vessel 2 from the start to the end of the polymerization step in the production of the fluororesin. The agitator 3 includes a motor 31, a rotation shaft 32 connected to the motor 31, and an agitating blade 33 supported by the rotation shaft 32. The stirrer 3 can stir the mixture containing the material in the polymerization vessel 2 by rotating the stirring blade 33 located in the polymerization vessel 2 using the motor 31. The polymerization reaction proceeds by stirring with the stirrer 3, and the fluororesin is produced from the mixture in the polymerization vessel 2. Here, the power and the rotation speed of the stirrer 3 depend on the stirring current as the current supplied to the motor 31. Therefore, the greater the stirring current, the greater the power and the higher the rotational speed. If the same rotation speed is maintained, the stirring current supplied to the motor 31 needs to be increased as the polymerization in the polymerization tank 2 progresses and the viscosity of the mixture increases.

The control device 4 controls the mixer 3. The control device 4 controls, for example, the stirring current supplied to the motor 31 so that the rotation speed of the stirrer 3 is constant during the polymerization step. Here, the control device 4 may feedback-control the stirring current value measured in the motor 31 or the stirring current supplied to the motor 31 in accordance with the rotation speed of the stirrer 3. The stirring current may be feedforward-controlled by using data from the prediction device 10 described later.

The control device 4 controls the environment in the polymerization vessel 2. The control device 4 controls, for example, the compressor so that the pressure value measured by the pressure sensor 21 in the polymerization step reaches a target value. Further, for example, the control device 4 controls the thermostat such that the temperature value measured by the temperature sensor 22 reaches a target value. Here, the control device 4 may feedback-control the pressure or temperature in the polymerization vessel 2 based on the sensor value measured by the pressure sensor 21 or the temperature sensor 22. The pressure or temperature may be feedforward-controlled by using data from the prediction device 10 described later. Wherein the control device 4 does not have to perform both pressure and temperature measurement and control. The control device 4 may, for example, only measure and control the pressure.

Prediction device

The prediction apparatus 10 is an information processing apparatus having a control unit 11, a storage unit 12, and a communication unit 13. The control unit 11 is a controller that performs overall control of the prediction apparatus 10. The control section 11 reads and executes the prediction program 123 stored in the storage section 12, thereby realizing processing as the timer section 111, the acquisition section 112, the prediction section 113, and the output processing section 114. The control unit 11 is not limited to a portion that realizes a predetermined function by cooperation of hardware and software, and may be a hardware circuit specifically designed to realize the predetermined function. The control unit 11 may be implemented by various processors such as CPU, MPU, GPU, FPGA, DSP, ASIC.

The storage unit 12 is a recording medium that records various information. The storage unit 12 may be implemented by, for example, RAM, ROM, flash memory, SSD (Solid State Drive, solid state disk), hard disk, other storage devices, or an appropriate combination thereof. The storage unit 12 stores various data for learning and prediction, and the like, in addition to the prediction program 123 executed by the control unit 11. The storage section 12 stores acquired data 121, a learned model 122, and a prediction program 123.

The communication unit 13 is an interface circuit (module) for realizing data communication with an external device (not shown).

In addition, the prediction apparatus 10 may have an input unit 14 and an output unit 15. The input unit 14 is an input device such as an operation button, a mouse, and a keyboard for inputting operation signals and data. The output unit 15 is an output device such as a display for outputting the processing result and data.

Here, the prediction apparatus 10 may be executed by 1 computer or by a combination of a plurality of computers connected via a network. Although not shown, for example, all or part of the data stored in the storage unit 12 may be stored in an external recording medium connected via a network, and the prediction apparatus 10 may use the data stored in the external recording medium.

The timer 111 measures the time elapsed from the start of the polymerization process of the fluororesin in the manufacturing apparatus 1 as the polymerization time. In this case, the production apparatus 1 may perform the polymerization process of the fluororesin a plurality of times. In this case, the timer unit 111 measures the polymerization time in each polymerization step. Therefore, the timer 111 measures the polymerization time in the current polymerization step.

The acquisition unit 112 acquires, as the sensor values for prediction, a pressure value in the polymerization tank measured in the polymerization step of the fluororesin and a stirring current value of a stirrer stirring the material in the polymerization tank or the mixture containing the fluororesin. The acquisition unit 112 also associates the acquired sensor value for prediction with time information, and stores the associated sensor value in the storage unit 12 as acquired data 121. For example, the time information is an aggregation time measured by the timer unit 111. The time information may be time, but is preferably information relating data to an aggregation time at the time of acquisition.

The timing at which the acquisition unit 112 acquires each sensor value may be determined according to the usage pattern. The time at which the sensor value is acquired is a regular time such as every 1 second, every 1 minute, every 10 minutes, or the like.

The prediction unit 113 predicts the MFR value at a predetermined time of the fluororesin in the current polymerization step by using the learned model learned by machine learning and the prediction sensor value acquired by the acquisition unit 112. For example, the predetermined time is a time every 10 hours or the like. Here, the respective times are not necessarily equally spaced as long as they are predetermined times. The prediction unit 113 can predict the MFR value using the prediction sensor value and the polymerization time measured by the timer unit 111.

The learned model is generated by learning the relationship between the learning sensor value measured in the past polymerization step of the same type of fluororesin as the fluororesin produced in the current polymerization step, the time required for the past polymerization step, and the MFR value measured for the fluororesin obtained in the past polymerization step. The generation of the learned model is described in detail later.

At this time, the prediction unit 113 can predict the MFR value using the pressure value measured during the portion of the polymerization process of the fluororesin. The prediction unit 113 can predict the MFR value using the stirring current value measured during the portion of the polymerization process of the fluororesin. The partial period is selected from a plurality of periods of time from the start to the end of the polymerization step of the fluororesin. Here, the prediction unit 113 can predict the MFR value using the average value of the pressure values measured during the partial period. The prediction unit 113 can predict the MFR value using the average value of the stirring current values measured during a part of the period. At this time, the period for the portion for the average pressure value may be the same as or different from the period for the average stirring current value. The prediction in the prediction unit 113 will be described in detail later. In this way, the prediction apparatus 10 can efficiently improve the accuracy of predicting the MFR value by using the sensor value during the portion having a large degree of influence on the MFR value. In particular, the MFR value is also liable to vary depending on the person to be measured, and it is difficult to realize high-precision measurement. In contrast, by using the sensor value, the MFR value is predicted by the predicting unit 113, and thus a highly reliable MFR value can be obtained.

The prediction unit 113 obtains information on the current polymerization completion time by using the predicted MFR value. For example, "information about the end time" is "at the time of the expected end" when the current aggregation is expected to end. Alternatively, the "information about the end time" may be "remaining time" required from the present time to the end of the expected polymerization step. Further, "information about the end time" may be a time when a specific operation for ending the manufacture is performed. Specific operations for ending the manufacture also include, for example, stopping raw material input, stopping stirring, reducing tank pressure, starting cooling, and the like. For example, "information about the end time" may be found from the sensor values obtained when the polymer is manufactured. Specifically, the prediction unit 113 can calculate information about the end time using a predetermined calculation formula or a correspondence formula using the calculated MFR value. In this way, by obtaining information about the end time by using the predicting unit 113, information about the end time is determined without actually taking out the polymer from the polymerization tank and measuring the performance thereof in the polymerization step of the fluororesin.

The output processing unit 114 outputs the information on the current end time of aggregation obtained by the prediction unit 113. By using the predicted information on the end time in this way, it is possible to determine the end of the polymerization step, and thereby to improve the accuracy of the fluororesin obtained by the manufacturing apparatus 1. The output processing unit 114 may output the MFR value itself of the desired current mixture.

As shown in fig. 4, for example, the output processing unit 114 outputs a display screen W including a display unit W1 and a display unit W2 to the output unit 15, wherein the display unit W1 displays a curve indicating a change in the MFR value predicted from the beginning, and the display unit W2 indicates the "remaining time" required until the end. The display unit w1 may include an index t1 indicating the current time point and an index t2 indicating the end time point predicted by the prediction unit 113.

Study in prediction unit 113

The prediction unit 113 generates a learned model using a learner. Here, the description is given of the case where the learner is included in the prediction unit 113, but the present invention is not limited to this. For example, the learner may also be external to the prediction apparatus 10. In the case where the learner exists outside the prediction apparatus 10, the prediction unit 113 uses the learned model generated and stored in the storage unit 12 by the external learner.

Specifically, as described above with reference to fig. 2A, the "description function" used by the prediction unit 113 in the machine learning is data including the "pressure value" in the polymerization tank 2 obtained in the past polymerization step of the fluororesin, the "stirring current value" of the stirrer 3, and the time "polymerization time" required for the past polymerization step. In addition, in this machine learning, the "objective function" as a correct result is the "MFR value" measured for the fluororesin polymerized in the corresponding past polymerization step. The learner then uses a plurality of sets of data including the description function and the destination function. Therefore, the learned model used by the prediction unit 113 is a parameter representing a plurality of relations indicating "pressure value", "stirring current value", "polymerization time" and "MFR value" obtained in the past polymerization step. Here, the MFR value measured for the fluororesin polymerized in the past polymerization step includes, in addition to the MFR value measured for the fluororesin obtained as the final product, the MFR value of the mixture extracted by the sampling operation in the middle of the polymerization step. The prediction unit 113 predicts the MFR value by using the learned model generated by machine learning in this way and using the pressure value and the stirring current value.

In this case, since the change in the polymerization step and the like are different depending on the types of the fluorine resins, the "description function" uses a different function for each fluorine resin. Therefore, a learning model generated using data obtained in the polymerization of a specific fluororesin (for example, ETFE) as an explanatory function is used for predicting the MFR value of the specific fluorine. Here, a description of the evaluation of the model obtained by learning is omitted.

Pretreatment of learning data

Here, the "polymerization time" and "MFR value" may be obtained only 1 time or 2 to 3 times or more in the 1-time polymerization step, but the "pressure value" and "stirring current value" may be obtained multiple times. Therefore, as for the learner, as the learning data, the learning data obtained by the preprocessing may be used instead of directly using the plurality of pressure values and the current values as the learning data. Specifically, the learner may use an average value of the plurality of pressure values and an average value of the plurality of agitation current values for a specific period as learning data included in the explanatory function. For example, as shown in fig. 5, the total time is divided into groups for each polymerization step, and the average value of the pressure value and the stirring current value is obtained for each group, and the learner uses these average values as learning data as an explanatory function. Here, the polymerization steps a to C are polymerization steps of the same fluororesin to be performed at different times.

The example of fig. 5 is an example in which the polymerization time is divided into a plurality of groups every 10 hours for each of the polymerization steps a to C, and the values used for learning data are obtained for each group. For each polymerization step, for example, the pressure value and the stirring current value measured from the beginning to 10 hours become the 1 st group data, and the pressure value and the stirring current value measured from 10 hours to 20 hours after stirring become the 2 nd group data. Specifically, in the polymerization step a, the average value of the pressure values (pressure average value A1) obtained from 10 hours after the start and the average value of the stirring current values (stirring current average value A1) obtained at the same time become the learning data for group 1. The pressure value and the stirring current value measured from 10 hours to 20 hours after the start were set as the data of group 2. Therefore, the average value of the pressure values (pressure average value A2) obtained from 10 hours after the start to 20 hours after the start and the average value of the stirring current values (stirring current average value b 2) obtained at the same time become the group 2 learning data. Similarly, the pressure average value and the stirring current average value were obtained for groups 3 to 20, and these were the learning data for each group in the polymerization step A. In addition, in the polymerization steps B and C, the pressure average value and the stirring current average value are also obtained for each group as learning data by defining each group corresponding to the polymerization time. The learner may use the learning data thus obtained by the preprocessing. In the example shown in fig. 5, the time is described as 10 hours, but the time is not limited to this, and may be arbitrarily set according to circumstances.

In addition, the learning is not performed using all the data of the group obtained in this way, and the values used in the learning are different depending on which data is used to obtain which time MFR value is required, such as the learning of only group 1, the learning of only group 2, the learning of the group 1 and the group 2.

Here, the learner does not need to take the pressure average value and the agitation current average value of all the groups as learning data, and may take the pressure average value and the agitation current average value of a specific group as learning data. Fig. 6 is an example of a data item of a learning sensor value used in a learner. In the example of fig. 6, as the sensor values for learning, the average values of the pressures of the 20 th and 19 th groups corresponding to the final stage of the polymerization process and the average value of the pressure of the 1 st group corresponding to the initial stage of the polymerization process are used. As the sensor values for learning, the average value of the stirring currents of the 20 th and 18 th groups with respect to the initial stage of the polymerization step and the average value of the stirring currents of the 8 th group corresponding to the middle stage of the polymerization step were used. As described above, the learner uses the sensor value obtained during the portion selected in the aggregation step as the learning sensor value. For example, the sensor values used herein are determined by selecting from among the values considered valid based on the sensor values of each group. In this case, the pressure average value and the stirring current average value do not have to be obtained for all the groups, and the pressure average value and the stirring current average value may be obtained for the selected group.

Therefore, the learner learns the relation between the "pressure average value" of each group, the "stirring current average value" and the "polymerization time" of each group, and the "MFR value" of the "objective function" as the "explanation function", and obtains the relation as a learned model for predicting the MFR value from the inputs of the pressure value, the stirring current value, and the like.

The preprocessing of the learning data is performed by the control unit 11, but is not limited to this. For example, the prediction unit 113 may be executed outside the prediction apparatus 10, and the learning data which is input from the outside and stored in the storage unit 12 and which has been subjected to the preprocessing may be used.

Prediction by the use prediction unit 113

The prediction unit 113 predicts the MFR value of the mixture in the polymerization vessel in the current polymerization step using the learned model generated as described above and the sensor value acquired by the acquisition unit 112. Here, the learner learns using learning data including an average value of the pressure values of each of the plurality of groups set during the part of the polymerization process and an average value of the agitation current value as described above. Therefore, the prediction unit 113 also performs preprocessing on the acquired data using the average value obtained for each of the same number of groups as the learning data. Then, using the obtained data after the preprocessing, the MFR value is predicted by the learned model.

As shown in fig. 7, when 150 hours have elapsed after the start of the current polymerization step, the prediction unit 113 groups the acquired data 121 at predetermined times, calculates the average value of the pressure value and the stirring current value for each group, and predicts the MFR value using the learned model. Here, for example, when a learner that learns using fig. 5 is used, if MFR values are predicted after 10 hours, a learned model that learns using only the 1 st group data is used. In addition, when the MFR value is predicted after 130 hours, for example, a learned model learned by data of the 1 st to 13 th groups may be used.

The time at which the MFR value is predicted by the prediction unit 113 is the time at which the predetermined set of prediction sensor values required for all prediction are obtained. Therefore, the prediction unit 113 does not need to predict the MFR value at the end of each group. For example, in the example shown in fig. 7, it is considered that the polymerization time is sufficiently shorter than the general end time of the polymerization step, such as 10 hours later, 20 hours later, and 30 hours later. Therefore, in this case, the MFR value may not be predicted, for example, each time the group determined from the shortest time of the past polymerization process is ended, after the lapse of a predetermined time period in which the polymerization process is unlikely to be ended.

Treatment using manufacturing apparatus

The flow of the process of the polymerization step of the fluororesin to be executed in the production apparatus 1 will be described with reference to the flowchart shown in fig. 8. First, in the manufacturing apparatus 1, a fluororesin material is charged into the polymerization vessel 2 (S1).

Further, the control of the environment in the polymerization vessel 2 and the control of the stirrer 3 are started by the control of the control device 4. Then, the timer 111 starts counting the aggregation time (S2). The polymerization reaction of the fluororesin is started in the polymerization vessel 2 under the control of the control device.

When control is started by the control device 4 in step S2, the acquisition unit 112 starts acquiring the input sensor value (S3). The acquisition unit 112 acquires, for example, a pressure value measured by the pressure sensor 21 and a stirring current value of the current supplied to the motor 31.

Then, the prediction unit 113 predicts the MFR value by the learned model 122 (S4). As described above, the prediction sensor values used in the learned model 122 may be preprocessed.

The prediction unit 113 predicts the end time using the MFR value predicted in step S4 (S5). Here, the end time is a time for ending a specific operation of the manufacturing, and the like, and the end may be determined in the subsequent processing corresponding to the time.

The output processing unit 114 outputs the end time predicted in step S5 (S6). The output processing unit 114 can output the end time by displaying the display screen W on the output unit 15 using, for example, fig. 4.

When the polymerization process is not completed (no in S7), the prediction apparatus 10 returns to step S3, and repeats the processing in steps S3 to S7. For example, the end of the aggregation process may be determined corresponding to the end time predicted in step S5. For example, when the time required for the prediction unit 113 to finish is predicted, if the time predicted in step S5 is 0, the polymerization process is finished.

On the other hand, when the polymerization step is completed (yes in S7), the control device 4 ends the control. The timer unit 111 then ends the timer (S8). This completes the polymerization step 1 in the production apparatus 1.

As described above, in the prediction apparatus 10, the MFR value of the mixture can be predicted not actually but using the sensor value. The prediction device 10 can improve the accuracy of predicting the MFR value by using the polymerization time. Thus, the manufacturing apparatus 1 can predict the timing of the end of the polymerization of the fluororesin in the polymerization step of the fluororesin using the MFR value predicted in the prediction apparatus 10. Thus, by using the MFR value predicted by the prediction device 10, it is not necessary to take out a part of the mixture for MFR value measurement from the polymerization tank in the middle of the polymerization step to actually measure the MFR value.

Modification 1

In the above description, the case where the prediction apparatus 10 generates the learned model 122 using the average value of the pressure values and the average value of the stirring current values as the sensor values as the learning data is described. However, the present invention is not limited to this, and a maximum value, a minimum value, or a variance of the sensor value may be used. Also, the average value, the maximum value, the minimum value, and the variance may be used in combination.

In this case, the prediction unit 113 predicts the MFR value by using at least any one of the average value, the maximum value, the minimum value, and the variance in combination. The prediction apparatus 10 can effectively use various values obtained from the measured values to improve the accuracy of predicting the MFR value.

Modification 2

In the above description, an example in which the manufacturing apparatus 1 inputs a material of a fluororesin in step S1 is described. On the other hand, the case where a material or the like is additionally charged into the polymerization vessel 2 in the polymerization step of the fluororesin is also included. In this case, the prediction apparatus 10 uses the cumulative amount of the additional input material for the prediction of the MFR value.

Specifically, the acquisition unit 112 acquires the pressure value, the stirring current value, and the cumulative amount of the material charged into the polymerization tank 2 as the sensor value for prediction.

At this time, the prediction unit 113 uses the learned model 122 learned by learning data including the pressure value and the stirring current value, and the cumulative amount of the material used in the past polymerization. The prediction unit 113 predicts the MFR value using a sensor value for prediction including the cumulative amount of the material acquired by the acquisition unit 112.

The prediction apparatus 10 can improve the prediction accuracy by predicting the MFR value using the cumulative amount of the material. This is because the amount of the material to be charged affects the MFR value of the fluororesin in the polymerization step. In this way, the prediction apparatus 10 can improve the accuracy of predicting the MFR value by using the cumulative amount of the materials thus charged.

Modification 3

In the above description, an example in which the MFR value is predicted using the sensor values obtained in the polymerization step of each fluororesin is described. On the other hand, when the polymerization step is repeatedly performed a plurality of times, the data obtained in the previous polymerization step is used for the prediction of the MFR value.

Specifically, the obtaining unit 112 may obtain the MFR value of the fluororesin obtained in the previous polymerization step. In other words, the obtaining unit 112 may obtain the MFR value of the fluororesin obtained in the previous batch.

In this case, the prediction unit 113 uses the learned model 122 learned by the learning data including the sensor values and the MFR value of the fluororesin obtained in the previous polymerization step. The prediction unit 113 predicts the MFR value using the sensor value for prediction and the MFR value of the fluororesin obtained in the previous polymerization step acquired by the acquisition unit 112. Here, when the MFR value obtained in the previous polymerization step is used, the prediction accuracy can be further improved when the same kind of fluororesin as that in the previous polymerization step is produced.

The prediction apparatus 10 can improve the prediction accuracy by predicting the MFR value of the fluororesin in the previous polymerization step by using the MFR value obtained in the previous polymerization step. This is because the MFR values of the produced fluororesin are similar in the continuous polymerization process because the various environments are similar. The prediction apparatus 10 can improve the accuracy of predicting the MFR value by using the MFR value obtained in the previous batch.

Other modifications

As the learning data, various data used in the aggregation step or various data acquired in the aggregation step may be used. In this case, as the data used in the learned model, the same kind of data may be used. For example, the temperature value measured by the temperature sensor 22 may be used as the learning sensor value and the prediction sensor value.

In the above description, the number of learned models 122 used by the prediction unit 113 is not increased, but the number of learned models used by the prediction unit 113 is not limited. For example, the prediction unit 113 may use 2 or more learned models 122 in combination.

Effects and supplements

As described above, the above embodiments are described as an example of the technology of the invention in the present application. However, the technique of the present invention is not limited to this, and can be applied to embodiments in which modifications, substitutions, additions, omissions, and the like are appropriately made.

The prediction apparatus, the manufacturing apparatus, and the prediction method described in the claims of the present invention can be realized by hardware resources such as cooperation of a processor, a memory, and a program.

Industrial applicability

The prediction apparatus, the production apparatus, and the prediction method of the present invention can be effectively used, for example, to determine the time point when polymerization ends in the polymerization step of the fluororesin.

The present application claims priority based on japanese patent application 2021-101048, filed on japanese 6-17 of 2021, the entire contents of which are incorporated herein by reference.

Description of the reference numerals

1 manufacturing apparatus

10 prediction device

11 control part

111 timing part

112 acquisition unit

113 prediction unit

114 output processing unit

12 storage part

121 obtain data

122 learned model

123 prediction program

13 communication unit

14 input part

15 output part

2 polymerization tank

21 pressure sensor

22 temperature sensor

3 stirrer

31 motor

32 rotation axis

33 stirring vane

4 control means.

Claims

1. A prediction apparatus for predicting an MFR value of a mixture containing a fluororesin in a polymerization vessel at a predetermined time after a raw material is charged and before completion of production in a polymerization process of the fluororesin, the prediction apparatus comprising:

an acquisition unit that acquires information including a pressure value in the polymerization tank measured in a current polymerization step of the fluororesin and a stirring current value of a stirrer stirring the inside of the polymerization tank, as a sensor value for prediction; and

a prediction unit that predicts an MFR value at a predetermined time of a mixture containing a fluororesin in a current polymerization step by using a learned model obtained by learning a relation between a learning sensor value measured in a past polymerization step of a fluororesin of the same type as the fluororesin and an MFR value measured for the fluororesin at the predetermined time in the past polymerization step by machine learning, and using the prediction sensor value obtained by the obtaining unit.

2. The predictive device of claim 1 wherein,

further comprising a timer for measuring the current time elapsed from the start of the current polymerization step as the polymerization time,

the prediction unit predicts the MFR value using the sensor value for prediction and the polymerization time measured by the timer unit.

3. The prediction apparatus according to claim 1 or 2, wherein,

the polymerization apparatus further includes an output processing unit that outputs information on the end time of the current polymerization step obtained by using the MFR value predicted by the prediction unit.

4. The prediction apparatus according to claim 1 to 3,

in the step of polymerizing the fluororesin, a material may be additionally charged into the polymerization vessel,

the acquisition unit acquires the pressure value, the stirring current value, and the cumulative amount of the material charged into the polymerization vessel as sensor values for prediction,

the prediction unit predicts the MFR value using the prediction sensor value by using a learned model obtained by learning using a learning sensor value including the integrated amount obtained in the past polymerization step.

5. The prediction apparatus according to claim 1 to 4,

The prediction unit predicts the MFR value using the pressure value measured during a part of the polymerization process of the fluororesin.

6. The predictive device of claim 5 wherein,

the prediction unit predicts the MFR value using at least any one of the average value, the maximum value, the minimum value, and the variance of the pressure values during the partial period.

7. The prediction apparatus according to claim 1 to 4,

the prediction unit predicts the MFR value using the stirring current value measured during a part of the polymerization process of the fluororesin.

8. The predictive device of claim 7 wherein,

the prediction unit predicts an MFR value using an average value, a maximum value, a minimum value, or a variance of the stirring current values during the partial period.

9. The prediction apparatus according to claim 5 to 8,

the partial period is selected from a plurality of periods in which the time from the start to the end of the polymerization step of the fluororesin is divided.

10. The prediction apparatus according to any one of claim 1 to 9,

the fluororesin polymerization step is carried out at least 2 times,

The obtaining unit obtains the MFR value of the fluororesin obtained in the previous polymerization step,

the prediction unit predicts the MFR value of the fluororesin using the prediction sensor value including the MFR value of the fluororesin obtained in the last polymerization step obtained by the obtaining unit.

11. A computing device for obtaining a target time for performing a specific operation for ending the production of a fluororesin in a polymerization vessel in a process for polymerizing the fluororesin, the computing device comprising:

a receiving unit that receives a target MFR value indicating a performance target of the fluororesin in the current production and an MFR value at a predetermined time after the raw material of the fluororesin is fed and before the production is completed; and

a calculation unit that obtains a target time for determining the end of the production of the fluororesin using a relationship between an MFR value of the fluororesin at the predetermined time of the past production of the fluororesin of the same type, an MFR value of the fluororesin obtained in the past production, and a target time of the past production, and an MFR value of the target MFR value and the predetermined time received by the reception unit.

12. An arithmetic device for obtaining a target time for judging the end of the production of a fluororesin in a polymerization step of the fluororesin, the arithmetic device comprising:

A receiving unit that receives a sensor value observed as a value related to the production of the fluororesin during the current production, and that receives a target MFR value representing a performance target of the fluororesin during the current production as a target value, with the sensor value being used as a sensor value for calculation; and

and a calculation unit that obtains the current manufacturing target time based on the calculation sensor value received by the reception unit, using a relationship between a sensor value obtained in a past manufacturing process of a fluororesin of the same type as the fluororesin, an MFR value of the fluororesin obtained in the past manufacturing process, and a target time of the past manufacturing process.

13. A production apparatus for producing a fluororesin, comprising:

a polymerization tank into which the fluororesin material can be introduced;

a stirrer for stirring the inside of the polymerization tank;

an acquisition unit that acquires, as sensor values for prediction, a pressure value in the polymerization tank and a stirring current value of the stirrer, which are measured in a current polymerization step of the fluororesin;

a prediction unit that predicts an MFR value of a mixture containing a fluororesin in a current polymerization process by using a learned model obtained by learning a relation between a learning sensor value measured in a past polymerization process of a fluororesin of the same type as the fluororesin, a time taken for the past polymerization process, and an MFR value measured for the fluororesin obtained in the past polymerization process by machine learning, and using the prediction sensor value obtained by the obtaining unit; and

And an output processing unit that outputs information on the end time of the current polymerization step, the information being obtained by using the MFR value predicted by the prediction unit.

14. A method for producing a fluororesin, comprising:

a step of charging a fluororesin material into a polymerization vessel;

stirring the inside of the polymerization tank;

a step of acquiring, as sensor values for prediction, a pressure value in the polymerization vessel and a stirring current value of a stirrer stirring the inside of the polymerization vessel, which are measured in a current polymerization step of the fluororesin;

a step of predicting the MFR value of a fluororesin-containing mixture in a current polymerization step using the obtained prediction sensor value by using a learned model obtained by learning a relation between a learning sensor value measured in a past polymerization step of a fluororesin of the same type as the fluororesin, a time taken in the past polymerization step, and an MFR value measured for the fluororesin obtained in the past polymerization step; and

and a step of ending the current production using information on the end time of the current polymerization step obtained by using the predicted MFR value.

15. A method for producing a fluororesin, comprising:

a step of charging a raw material of a fluororesin into a polymerization vessel;

a step of acquiring an MFR value observed as a value related to the production of the fluororesin in the current production of the fluororesin, and using the MFR value as a sensor value for prediction;

a step of predicting the MFR value at a predetermined time of the fluororesin currently produced based on the obtained sensor value for prediction, using a relation between a sensor value obtained in a past production of a fluororesin of the same kind as the fluororesin and the MFR value of the fluororesin at the predetermined time of the past production;

a step of receiving a target MFR value representing a performance target of the fluororesin and a predicted MFR value at the predetermined time in the current production;

a step of obtaining a target time for determining the end of the production of the fluororesin using a relationship between an MFR value of the fluororesin at the predetermined time of the past production of the fluororesin of the same kind, an MFR value of the fluororesin obtained at the past production, and a target time of the past production, and an MFR value of the received target MFR value and the MFR value at the predetermined time; and

And ending the current manufacture by using the obtained target time.

16. A prediction method for predicting an MFR value of a mixture containing a fluororesin in a polymerization vessel in a polymerization process of the fluororesin, the prediction method comprising:

an acquisition step of acquiring, as a sensor value for prediction, a pressure value in the polymerization tank and a stirring current value of a stirrer stirring the inside of the polymerization tank, which are measured in a current polymerization step of the fluororesin; and

a prediction step of predicting the MFR value of the fluororesin-containing mixture in the current polymerization step using the prediction sensor value obtained in the obtaining step, by using a learned model obtained by learning a relation between the learning sensor value measured in the past polymerization step of the fluororesin of the same type, the time taken in the past polymerization step, and the MFR value measured for the fluororesin obtained in the past polymerization step.

17. A production method for producing a learned model for predicting an MFR value of a fluororesin in a polymerization vessel in a polymerization process of the fluororesin, the production method comprising:

A step of acquiring a plurality of sets of data including training data including a sensor value measured in a polymerization step of a fluororesin of the same kind as the fluororesin and a polymerization time used in the polymerization step, and correct result data for the training data, the correct result data being an MFR value measured for the fluororesin obtained in the polymerization step; and

and generating a learned model by using a learner, wherein the learned model is generated by using a sensor value obtained in the polymerization step of the fluororesin based on a plurality of training data and the correct result data, and the MFR value of the fluororesin in the polymerization step is output.