Molecular weight prediction apparatus, polymer compound production system, program, molecular weight prediction method, and polymer compound production method

Abstract

To sequentially predict a molecular weight distribution in a chemical treatment step for reducing a molecular weight of a polymer compound.SOLUTION: In a chemical treatment step in which a molecular weight of a polymer compound is reduced by a chemical reaction, a reaction rate calculation unit calculates a reaction rate of the chemical reaction by using a reaction rate prediction model on the basis of reaction condition variables at a time after a first time, and a molecular weight calculation unit simulates, as the chemical reaction, a process in which one polymer is used as a starting material and is decomposed into a plurality of polymers each having a molecular weight smaller than that of the starting material as target substances, on the basis of the reaction rate and a molecular weight distribution of the polymer compound actually measured at the first time, and estimates a molecular weight distribution of the polymer compound at a second time later than the first time.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present application relates to a molecular weight prediction device, a polymer compound production system, a program, a molecular weight prediction method, and a polymer compound production method.The present application relates to estimation or control of polymer compound production conditions in, for example, a chemical plant. [Background technology]

[0002] Conventionally, methods for predicting the molecular weight of a polymer compound have been proposed. For example, Patent Document 1 describes an information processing method in which molecular weight data indicating the measured value of the molecular weight in the molecular weight adjustment process of a polymer compound and production condition data indicating the production conditions of the polymer compound are acquired, a molecular weight prediction model is determined based on the measured value of the molecular weight and at least one of the production conditions, and a predicted value of the molecular weight is calculated using the determined prediction model. Patent Document 2 describes a polyester production system that acquires at least one piece of reaction condition information from a reactor in which a reaction accompanied by a change in molecular weight is taking place, acquires in-reaction molecular weight information, which is molecular weight information during the reaction, and predicts the molecular weight of the product during the reaction based on the reaction condition information, the in-reaction molecular weight information, and related information relating the reaction conditions and molecular weight conditions. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2022-182129 [Patent Document 2] International Publication No. 2022 / 163035 Summary of the Invention [Problem to be solved by the invention]

[0004] The techniques described in Patent Documents 1 and 2 predict a representative value of the molecular weight distribution, such as the weight-average molecular weight of a polymer compound. However, such a representative value of the molecular weight distribution is generally a statistical quantity that indicates the overall characteristics of the distribution of the content of each molecular weight in a collection of polymer compounds having different molecular weights, and information about the content of each molecular weight is discarded during the calculation of the statistical quantity. In other words, the distribution cannot be reconstructed from one or a finite number of representative values. On the other hand, even two polymer compounds of the same type that have the same representative value of the molecular weight distribution may have different molecular weight distributions, and these differences result in differences in the physical properties of the polymer compounds. Therefore, predicting the molecular weight distribution is important for accurately predicting the physical properties of polymer compounds. Despite this, the techniques described in Patent Documents 1 and 2 do not take into account the molecular weight distribution itself in their predictions.

[0005] The present application has been made in view of the above points, and one of the objects of the present application is to predict the molecular weight distribution in a chemical treatment process that reduces the molecular weight of a polymer compound. [Means for solving the problem]

[0006] (1) The present application has been made to solve the above-mentioned problems, and one aspect of the present application is a molecular weight prediction device that includes: a reaction rate calculation unit that calculates a reaction rate of a chemical reaction using a reaction rate prediction model based on reaction condition variables at a first time or later, in a chemical treatment process in which the molecular weight of a polymer compound is reduced by the chemical reaction; and a molecular weight calculation unit that simulates, as the chemical reaction, a process in which a single polymer is used as a starting material and is decomposed into multiple polymers as target materials each having a molecular weight smaller than that of the starting material, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimates the molecular weight distribution of the polymer compound at a second time after the first time.

[0007] (2) Another aspect of the present application is a molecular weight prediction method in a molecular weight prediction device, wherein the molecular weight prediction device executes a reaction rate calculation step of calculating a reaction rate of a chemical reaction using a reaction rate prediction model based on reaction condition variables at times after the first time in a chemical treatment process in which the molecular weight of a polymer compound is reduced by the chemical reaction; and a molecular weight calculation step of simulating, as the chemical reaction, a process in which one polymer is used as a starting material and is decomposed into multiple polymers as target materials each having a molecular weight smaller than that of the starting material, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimating the molecular weight distribution of the polymer compound at a second time after the first time. [Effects of the Invention]

[0008] According to an embodiment of the present application, it is possible to sequentially predict the molecular weight distribution in a chemical treatment process that reduces the molecular weight of a polymer compound. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic block diagram illustrating an example of the functional configuration of a production system according to an embodiment of the present invention. [Figure 2] FIG. 2 is an explanatory diagram showing learning and inference of a reaction rate prediction model according to the present embodiment. [Figure 3] FIG. 1 is an explanatory diagram for explaining a method for searching for a reaction rate based on an actual measurement value of molecular weight distribution. [Figure 4] FIG. 2 is a diagram showing an example of molecular weight distribution. [Figure 5] FIG. 10 is a diagram showing another example of molecular weight distribution. [Figure 6] FIG. 1 is a schematic block diagram showing an example of the hardware configuration of a molecular weight prediction device according to an embodiment of the present invention. [Figure 7] FIG. 1 is a diagram showing an example of molecular weight adjustment equipment for a polymer compound. [Figure 8] FIG. 2 is a diagram showing an example of the change in weight average molecular weight over time in the molecular weight adjustment step. [Figure 9]FIG. 2 is an explanatory diagram illustrating a reaction rate prediction model according to the present embodiment. [Figure 10] FIG. 10 is an explanatory diagram for explaining a decomposition process according to the present embodiment. [Figure 11] 1 is a flowchart showing an example of a simulation of molecular weight distribution according to the present embodiment. [Figure 12] FIG. 10 is a diagram illustrating a learning pattern of a reaction rate prediction model according to the present embodiment. [Figure 13] FIG. 1 is a diagram showing a comparison between estimated values ​​and actually measured values ​​for each pattern of number average molecular weight obtained by simulation using the reaction rate determined by inverse analysis. [Figure 14] FIG. 1 is a diagram showing a comparison between estimated values ​​and actually measured values ​​of weight-average molecular weights for each pattern obtained by simulation using reaction rates determined by inverse analysis. [Figure 15] FIG. 1 is a diagram showing a comparison between estimated and actually measured values ​​of Z-average molecular weights for each pattern obtained by simulation using the reaction rate determined by inverse analysis. [Figure 16] FIG. 10 is a diagram showing a comparison between estimated and actually measured values ​​of Z+1 average molecular weight for each pattern obtained by simulation using the reaction rate determined by inverse analysis. [Figure 17] FIG. 1 is a diagram illustrating the relative error of the estimated value of the weight average molecular weight by weighted averaging. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment of the present application will be described with reference to the drawings. First, an outline of the present embodiment will be described. Fig. 1 is a schematic block diagram showing an example of the functional configuration of a production system S1 according to the present embodiment. The production system S1 is a system for producing polymer compounds. The production process of polymer compounds includes a process for adjusting the molecular weight of the polymer compound that constitutes an intermediate product by reducing it through a chemical reaction (sometimes referred to herein as the "molecular weight adjustment process" or simply the "chemical reaction process"). The following explanation mainly focuses on the example of predicting the molecular weight in the molecular weight adjustment process. The production system S1 includes a molecular weight prediction device 10, a molecular weight measuring instrument 20, a process data detector 30, and a process control device 60.

[0011] The molecular weight prediction device 10 acquires molecular weight data from a molecular weight measuring instrument 20 and process data from a process data detector 30. The molecular weight data indicates the molecular weight of a polymer compound measured during a polymer compound production process. The process data indicates various variables (parameters, sometimes referred to herein as "process variables") that constitute the polymer compound production conditions. The polymer compound production process includes a chemical treatment process in which raw materials, intermediate products, etc. are subjected to chemical reactions to convert them into other substances. The chemical treatment process includes process data including variables indicating reaction conditions (sometimes referred to herein as "reaction condition variables"). The chemical treatment process may include a chemical reaction in which a polymer constituting a polymer compound is used as a starting material and is decomposed into multiple polymers with lower molecular weights as target materials. One such chemical reaction is hydrolysis. In hydrolysis, reaction condition variables include, for example, pH and temperature. This embodiment is suitable for a depolymerization reaction. This application mainly illustrates a case in which the chemical reaction in the chemical treatment process is depolymerization by hydrolysis.

[0012] The molecular weight prediction device 10 calculates the reaction rate of a chemical reaction using a predetermined reaction rate prediction model based on reaction condition variables from a certain reference time (sometimes referred to herein as the "first time") onward in a chemical treatment process for reducing the molecular weight of a polymer compound. The molecular weight prediction device 10 executes a simulation of the chemical reaction. The simulation includes a chemical reaction, i.e., a process in which a single polymer is used as a starting material and is decomposed into multiple polymers as target materials each having a molecular weight smaller than that of the starting material (sometimes referred to herein as the "decomposition process"). The actual measured value of the molecular weight distribution of the polymer compound at the first time is used as an initial value, and the reaction rate estimated by calculation is referenced as a parameter of the decomposition process. The simulation estimates the molecular weight distribution at a time later than the first time.

[0013] The molecular weight prediction device 10 generates control information for controlling a chemical treatment process based on the estimated molecular weight distribution. The molecular weight prediction device 10 outputs the generated control information to the process control device 60. In this application, the term "molecular weight distribution" refers to a set of contents of different molecular weights contained in a set of polymers in a chemical reaction system. In this application, the term "molecular weight distribution" refers to the number average molecular weight (M n ) to the weight average molecular weight (M w ) ratio (M w / M n This does not mean molecular weight dispersion, which is the molecular weight distribution of the polymer.

[0014] The molecular weight measuring instrument 20 is an instrument for measuring the molecular weight of a polymer compound. The molecular weight measuring instrument 20 collects a solution as a sample from an adjustment tank 72 (FIG. 7) where a chemical treatment process is carried out, and measures the molecular weight of the polymer compound dissolved or suspended (hereinafter referred to as "suspended, etc.") in the collected solution. The molecular weight measuring instrument 20 transmits molecular weight data indicating the measured molecular weight value to the molecular weight prediction device 10. The molecular weight measuring instrument 20 can measure the molecular weight distribution of the suspended, etc. solute using, for example, gel permeation chromatography (GPC). The GPC method is a type of liquid chromatography.

[0015] The process data detector 30 is installed in a polymer compound production facility and includes a detector (sensor) for detecting the production process, i.e., the state of the process. In a chemical treatment process, reaction condition variables are included as process state variables that indicate the state of a chemical reaction. The process data detector 30 outputs process data indicating the reaction condition variables that indicate the detected reaction conditions to the molecular weight prediction device 10. In the example of FIG. 7, pH and temperature are detected as reaction condition variables. The process data detector 30 includes a pH sensor 722 for measuring pH and a temperature sensor 724 for measuring temperature.

[0016] The process control device 60 controls the process state in the polymer compound production process based on the control information input from the molecular weight prediction device 10. The polymer compound production facility has an adjustment facility (not shown) for adjusting reaction condition variables in a chemical treatment process. The process control device 60 operates the adjustment facility so that the measured value of the reaction condition variable approaches its target value.

[0017] 7, the process control device 60 controls the hydrolysis of a polymer suspended in a polymer solution introduced into an adjustment tank 72. The process control device 60 adjusts the temperature of the solution by causing a temperature regulator (not shown) to cool or heat the solution contained in an adjustment tank 72 (described below) so that the measured value of the temperature in the chemical treatment process approaches its target value (sometimes referred to as a "temperature setpoint" in this application).

[0018] The process control device 60 may adjust the pH of the solution by instructing a pH adjuster to add a base (e.g., sodium hydroxide (NaOH)) or a solvent so that the measured pH value in the chemical treatment process approaches its target value (sometimes referred to as the "pH set value" in this application). The process control device 60 may use any of a number of techniques, such as PI control or PID control, to control the process state.

[0019] (Molecular weight prediction device) Next, a description will be given of an example of the functional configuration of the molecular weight prediction device 10. The molecular weight prediction device 10 includes an arithmetic control unit 120 and a storage unit 140. The arithmetic and control unit 120 is a computer system including, for example, a processor and a memory. The processor is a CPU (Central Processing Unit) or the like. The processor reads out a predetermined program stored in advance in the storage unit 140 and executes the processing indicated by the instructions written in the read program. This processing enables the molecular weight prediction device 10, including the arithmetic and control unit 120 described below, to perform its functions. In this application, executing the processing indicated by the instructions written in the program may be referred to as "executing a program" or "running a program." The arithmetic and control unit 120 may be realized using a dedicated component.

[0020] The storage unit 140 stores various types of data used in the processes executed by the arithmetic and control unit 120. The storage unit 140 stores various types of data acquired by the arithmetic and control unit 120. The storage unit 140 includes storage media such as a RAM (Random Access Memory) and a ROM (Read Only Memory).

[0021] Next, a description will be given of an example of the functional configuration of the arithmetic and control unit 120. The arithmetic and control unit 120 includes a reaction rate calculation unit 122, a molecular weight calculation unit 124, a control information output unit 128, and a model learning unit 132.

[0022] The reaction rate calculation unit 122 receives process data indicating reaction condition variables from the process data detector 30. Molecular weight data indicating molecular weight distribution may be input to the reaction rate calculation unit 122 from the molecular weight measurement instrument 20. The reaction rate calculation unit 122 uses a trained reaction rate prediction model to calculate the reaction rate, which serves as the objective variable (output value), from the reaction condition variables, which serve as explanatory variables (input values). The reaction rate calculation unit 122 may also use a representative value of the molecular weight distribution to calculate the reaction rate. The reaction rate calculation unit 122 outputs reaction rate data indicating the calculated reaction rate to the molecular weight calculation unit 124. A parameter set of the reaction rate prediction model is preset in the reaction rate calculation unit 122 before calculation is performed.

[0023] The molecular weight calculation unit 124 receives reaction rate data indicating the reaction rate from the reaction rate calculation unit 122 . The molecular weight calculation unit 124 performs a simulation of a chemical reaction performed in a chemical treatment process based on the reaction rate indicated by the reaction rate data. The chemical reaction uses one polymer forming a polymeric compound as a starting material and a decomposition process in which the target material is decomposed into two or more polymers having molecular weights smaller than that of the starting material. The molecular weight calculation unit 124 sets the molecular weight distribution indicated by the molecular weight data input from the molecular weight measuring instrument 20 as the initial value of the molecular weight distribution at a first time. The molecular weight calculation unit 124 performs a simulation to estimate the molecular weight distribution at each time after the first time. The first time is, for example, the reaction start time of the molecular weight adjustment process. The first time may be any time as long as it is the sampling time for a sample of molecular weight data already obtained through the molecular weight measuring instrument 20.

[0024] In the simulation, the reaction rate of a chemical reaction between a certain bond of monomers is defined as the probability that a bond between the monomers constituting the starting material will cleave per unit time. That is, the probability that a certain bond will cleave in a given time (the cleavage probability) is the product of the average reaction rate over that time and the given time. In other words, the reaction rate corresponds to the cleavage probability over that time normalized by unit time. The molecular weight calculation unit 124 randomly selects polymers and bonds to be subjected to the decomposition process from the set of polymers in the entire system involved in the simulation according to the bond cleavage probability at that time, and performs the cleavage. If the reaction rates of all bonds between all polymers in the system at a given time can be considered to be equal, the molecular weight calculation unit 124 may perform the simulation by treating the set of polymers in the entire system as the target for the decomposition process and using a uniform value within the system as the cleavage probability. The molecular weight calculation unit 124 outputs estimated molecular weight data indicating an estimated value of the molecular weight distribution at each time to the control information output unit 128. The molecular weight calculation unit 124 may output the process data input from the process data detector 30 to the control information output unit 128 in association with the estimated molecular weight data.

[0025] The control information output unit 128 generates control information for controlling the chemical reaction in the adjustment tank 72 based on the estimated value of molecular weight distribution indicated in the estimated molecular weight data input from the molecular weight calculation unit 124. For example, the control information output unit 128 calculates a predicted value of the weight average molecular weight from the estimated value of the molecular weight distribution. The control information output unit 128 generates control information indicating the start (trigger) of the reaction termination process of the molecular weight adjustment step at the time when the calculated predicted value of the weight average molecular weight reaches a predetermined target molecular weight.

[0026] More specifically, at the time when the predicted representative value of the molecular weight distribution reaches a predetermined target value, the control information output unit 128 generates a reaction stoppage treatment start command and outputs it to the process control device 60. In response to the input of the reaction stoppage treatment start command from the control information output unit 128, the process control device 60 starts the reaction stoppage treatment for the molecular weight adjustment step. At this time, as the reaction stoppage treatment, the process control device 60 starts cooling, for example. When starting cooling, the process control device 60 cools the temperature of the adjustment tank 72 from the predetermined reaction temperature for the molecular weight adjustment step to a low temperature that does not cause hydrolysis of the polymer compound, for example.

[0027] The model learning unit 132 uses training data to learn a reaction rate prediction model for predicting a reaction rate based on the molecular weight distribution and reaction condition variables. The model learning unit 132 acquires training data. The training data includes a large number of data pairs. As illustrated in FIG. 2 , each data pair includes, as input values ​​serving as explanatory variables, a representative value of the molecular weight distribution at a first time, reaction condition variables for each sample time from the first time to the second time, and the elapsed time from the first time. The output value serving as a response variable includes the average reaction rate from the first time to the second time. The reaction rate prediction model may be a statistical model, a machine learning model, a physical model, or a hybrid model of these. The model learning unit 132 learns a parameter set for the reaction rate prediction model so that the difference between the estimated reaction rate calculated from the input value using the reaction rate prediction model and the output value of the data pair related to the input value is minimized for the entire training data.

[0028] The loss function indicating the magnitude of the difference may be, for example, the sum of squared errors or the sum of absolute error values. For a reaction rate prediction model in which a parameter set that minimizes the loss function cannot be analytically determined, the model training unit 132 may use a method such as batch gradient descent or stochastic gradient descent as a parameter set training method. The model training unit 132 repeatedly updates the parameter set until the value of the parameter set converges. The model training unit 132 can determine whether convergence has occurred based on whether the amount of change in the parameter set or the loss function has reached a predetermined reference value or less. When determining that convergence has occurred, the model training unit 132 sets the obtained parameter set in the reaction rate calculation unit 122.

[0029] The model learning unit 132 calculates the reaction rates to be included in the training data before learning the reaction rate prediction model. This process is sometimes called "inverse analysis." As illustrated in FIG. 3, the model learning unit 132 performs a simulation using a method similar to that of the molecular weight calculation unit 124, using the measured value of the molecular weight distribution at the first time as the initial value. The model learning unit 132 searches for the fragmentation probability between the first time and the second time so that the estimated value of the molecular weight distribution at the second time obtained by the simulation is as close as possible to the measured value of the molecular weight distribution at the second time. Here, the model learning unit 132 excludes polymers having molecular weights outside the detection range of the molecular weight measuring device used for the actual measurement from the estimated value of the molecular weight distribution at the second time obtained by the simulation, and compares the estimated value with the measured value of the molecular weight distribution. As an index of the degree of approximation, for example, the sum of squares of the error between the measured value of the weight average molecular weight at the second time and the weight average molecular weight of the molecular weight distribution obtained by the simulation can be used. To search for a division probability that achieves an approximation, an optimization algorithm such as a genetic algorithm or the Nelder-Mead method can be used. The model learning unit 132 can calculate the average reaction rate within that period to be used as training data by dividing the determined disruption probability by the elapsed time from the first time to the second time.

[0030] As described above, in the simulation, the decomposition process is carried out at a frequency corresponding to the reaction rate for the molecular weight distribution at the first time point and at each time point after the first time point. Therefore, the transition of the molecular weight distribution over time is simulated. As illustrated in Figure 5, the molecular weight distribution at the second time point, which is later than the first time point, transitions to a region with lower molecular weight than the molecular weight distribution at the first time point (Figure 4).

[0031] Next, we will explain an example of the hardware configuration of the molecular weight prediction device 10. Fig. 6 is a schematic block diagram showing an example of the hardware configuration of the molecular weight prediction device 10 according to this embodiment. The molecular weight prediction device 10 is a computer that includes a processor 102 , an input unit 108 , an output unit 110 , a ROM 112 , a RAM 114 , an auxiliary storage unit 116 , and an interface unit 118 .

[0032] Processor 102 reads a predetermined program stored in advance in ROM 112 and executes processing instructed by various commands written in the read program. In this application, executing processing instructed by commands written in a program may be referred to as "executing a program" or "running a program." Processor 102 includes, for example, a CPU (Central Processing Unit).

[0033] The input unit 108 receives a user operation and generates an operation signal indicated by the received operation. The input unit 108 outputs the generated operation signal to the processor 102. The input unit 108 includes components such as a touch sensor, a keyboard, and a mouse. The output unit 110 outputs output information indicated by various output data input from the processor 102. The output unit 110 includes, for example, a display for presenting an image.

[0034] The ROM (Read Only Memory) 112 is a storage medium that permanently stores various data used by the processor 102 or programs to be executed. RAM (Random Access Memory) 114 is a storage medium that temporarily stores various data (parameters, etc.) used in the operation of processor 102 and various data acquired by processor 102. RAM 114 is used as a working area for processor 102. The auxiliary storage unit 116 is a storage medium that permanently stores various data acquired by the processor 102 and various data used by the processor 102. The auxiliary storage unit 116 may be configured to include, for example, one or a combination of an SSD (Solid State Drive) and an HDD (Hard Disk Drive).

[0035] The interface unit 118 is connected to other devices wirelessly or via a wire so as to be able to send and receive various types of data. The interface unit 118 may be connected to other devices using a communication network. The interface unit 118 may be configured to include, for example, an input / output interface, a communication interface, or a predetermined combination thereof. The functions of the arithmetic and control unit 120 and the storage unit 140 are realized in cooperation with the processor 102, RAM 114, auxiliary storage unit 116, interface unit 118, and other hardware. The functions of the memory unit 140 are realized by the ROM 112, RAM 114, and auxiliary memory unit 116 in cooperation with the processor 102.

[0036] (Production process) Next, the production process for polymer compounds will be described using an example in which P3HB3HH (Poly(3HB-co-3HHx), poly(3-hydroxybutyrate-co-hydroxyhexanoate)) is the product to be produced. P3HB3HH is a copolymer polyester obtained by polymerizing 3-hydroxybutyrate (3HB) and 3-hydroxyhexanoate (3HH). The production process for P3HB3HH comprises a production step, a purification step, and a post-treatment step. The production step comprises a culturing step and an inactivation step. The purification step comprises a molecular weight adjustment step, a cell membrane degradation step (Lys treatment), a proteolysis step (Alc treatment), a solubilization step (SDS treatment), and a separation step. The post-treatment step comprises a membrane concentration step, a pH adjustment step, and a drying step.

[0037] The cultivation process involves cultivating the bacterial strain and metabolizing organic matter. Polymeric compounds are synthesized during this process. The bacterial strain metabolizes natural organic materials such as palm oil and fructose to synthesize P3HB3HH. The cultivation process produces P3HB3HH with a weight-average molecular weight of approximately 1 to 2 million. The inactivation process is a process for inactivating the activity of the cultured bacterial strain. For example, the cultured bacterial strain is heated during the inactivation process. During the inactivation process, some of the P3HB3HH stored within the bacterial cells may undergo a chemical reaction due to heat, resulting in a decrease in molecular weight. This is followed by the molecular weight adjustment process. After the molecular weight adjustment process is completed, the process proceeds to the cell membrane degradation treatment process.

[0038] The cell membrane degradation process is a process for decomposing the cell membranes of bacterial strains contaminated in a solution. In the cell membrane degradation process, a phospholipid-degrading enzyme is added to a solution of inactivated microorganisms to decompose the cell membranes.

[0039] In the proteolysis process, a decomposing enzyme is added to the eluted protein to decompose it. During the cell membrane decomposition process and the proteolysis process, some of the P3HB3HH eluted in the solution may undergo a chemical reaction, resulting in a decrease in molecular weight. The solubilization step is a step for solubilizing other bacterial components and lipids contained in the P3HB3HH solution obtained through the molecular weight adjustment step. In the solubilization step, a surfactant with a denaturing effect is added to the molecular weight-adjusted solution.

[0040] The separation step is a step of centrifuging the dissolved P3HB3HH from the solution. The membrane concentration process is a process in which a concentrated solution of extracted P3HB3HH and a dilute solution are separated using a reverse osmosis membrane, and the concentrated solution is concentrated by applying a pressure higher than the osmotic pressure to the concentrated solution. The pH adjustment step is a step in which a pH adjuster is added to the concentrated solution to neutralize the pH of the solution. The drying step is a step in which the neutralized solution is dried to extract P3HB3HH. The order of the molecular weight adjustment step in the polyester production process does not necessarily have to be between the production step and the cell membrane degradation treatment step, and the molecular weight adjustment step may be any step after the inactivation treatment and before the proteolysis treatment step.

[0041] Next, the molecular weight adjustment process will be described. The molecular weight adjustment process is a process in which the molecular weight of P3HB3HH in a solution containing P3HB3HH as a solute is reduced to a level close to the molecular weight required for the product to be produced. As described above, the molecular weight is further reduced through subsequent processes, so the target molecular weight in this process is higher than the molecular weight required for the product. The target weight average molecular weight varies depending on the application, but is typically around 200,000 to 600,000. In the molecular weight adjustment process, a base such as sodium hydroxide (NaOH) is added to the solution to hydrolyze the P3HB3HH suspended in the solution. Hydrolysis occurs when hydroxide ions (OH) are introduced into the ester bonds between the monomers that make up one P3HB3HH. - This chemical reaction breaks the molecule down into a carboxylic acid R-COOH molecule with a terminal carboxyl group and a molecule with a hydroxyl group R-OH.

[0042] The molecular weight adjustment step is carried out in an adjustment tank 72, which serves as a reaction vessel constituting a P3HB3HH production facility 70. As illustrated in FIG. 7, a solution in which P3HB3HH is suspended and sodium hydroxide are charged into the adjustment tank 72. In the adjustment tank 72, the solution is stirred using an agitator 74 so that the pH and temperature of the solution become uniform. The adjustment tank 72 is also equipped with a pH sensor 722 and a temperature sensor 724. As described above, the pH sensor 722 and the temperature sensor 724 measure the pH and temperature of the solution, respectively. The adjustment tank 72 is equipped with a temperature regulator (not shown) for adjusting the temperature of the solution.

[0043] FIG. 8 shows an example of reaction control according to the change in weight average molecular weight (Mw) over time in the molecular weight adjustment process. At the beginning of the molecular weight adjustment process, the process control device 60 sets the pH and temperature of the solution to predetermined values. The process control device 60 also sets the target molecular weight of P3HB3HH to Mw. end In the example shown in FIG. 8, the weight average molecular weight at the beginning of the molecular weight adjustment process is Mw0. As hydrolysis progresses, the weight average molecular weight gradually decreases over time. Then, the process control device 60 determines whether the weight average molecular weight of P3HB3HH reaches the target molecular weight Mw end When the temperature reaches this value, the reaction termination process in the adjustment tank 72 is started, and the molecular weight adjustment process is completed.

[0044] The entire time required for one molecular weight adjustment process is approximately 6 to 14 hours. Because molecular weight measurement by GPC requires a considerable amount of time, it is usually possible to perform measurements only at one-hour intervals. In contrast, reaction condition variables are obtained at shorter time intervals (for example, one minute) and used to predict the hydrolysis rate. According to the simulation of this embodiment, the predicted hydrolysis rate is used to virtually obtain the molecular weight distribution at shorter time intervals (virtual monitoring), allowing for more precise adjustment of the molecular weight of the product.

[0045] Next, a specific example of a reaction rate prediction model used to predict the reaction rate will be described. FIG. 9 is an explanatory diagram illustrating a reaction rate prediction model according to this embodiment. The reaction rate calculation unit 122 receives as input the representative value of the measured molecular weight distribution, pH, and temperature as explanatory variables, and outputs the reaction rate as the response variable. As the representative value of the molecular weight distribution, the weight average molecular weight Mw is obtained at every measurement interval of the molecular weight distribution (for example, 1 hour). The pH and temperature are obtained at every measurement interval of the reaction condition variable (for example, 1 minute). The reaction rate is calculated at every measurement interval of the reaction condition variable (for example, 1 minute).

[0046] The reaction rate calculation unit 122 calculates the average pH value from the first time t1 to the latest time t as a first type intermediate value. The reaction rate calculation unit 122 divides the difference between the pH at the latest time t and the pH at the first time t1 by the elapsed time from the first time t - t1 to calculate the average rate of change of pH as a second type intermediate value. The reaction rate calculation unit 122 calculates the average reaction rate constant from the first time t1 to the latest time t as a third type intermediate value. The average reaction rate constant is given using equation (1).

[0047]

number

[0048] In equation (1), A and E a , R, T, [OH - ], Δt respectively represent the frequency factor, activation energy, gas constant, temperature, hydroxide ion concentration, pH, and the time interval between temperature measurement results. The hydroxide ion concentration is derived from pH using a predetermined calculation formula. The hydroxide ion concentration may or may not be temperature corrected. The denominator of equation (1) represents the elapsed time from the first time t1 to the latest time t. The numerator of equation (1) represents the sum of the average reaction rate constants at each time τ from the first time t1 to the latest time t. A, E a If these values ​​have not been experimentally determined, the reaction rate calculation unit 122 may perform a procedure such as cross-validation to search for optimal values ​​of these values ​​in advance.

[0049] The reaction rate calculation unit 122 calculates feature quantities (sometimes referred to herein as "polynomial terms") using a polynomial that includes the weight-average molecular weight at the first time t1, the first intermediate value, the second intermediate value, and the third intermediate value as elements. The calculated polynomial terms are used as explanatory variables that are directly input into the reaction rate prediction model in the model learning unit 132. The degree of the polynomial may be first order or may be second order or higher. By using a second order or higher order polynomial, interactions between different intermediate values ​​are taken into consideration.

[0050] The model learning unit 132 may use an adaptive (JIT: ​​Just in Time) model in learning the reaction rate prediction model. By using an adaptive model, learning is performed with emphasis on data pairs of explanatory variables and objective variables that are similar to the explanatory variables at that time. In learning, for example, Locally Weighted Partial Least Squares (LW-PLS) may be used. However, hyperparameters such as the number of components of LW-PLS and the similarity calculation method are determined in advance using a method such as cross-validation.

[0051] Next, an example of a simulation method according to this embodiment will be described. P3HB3HH is a polyester copolymer having 3HB and 3HH as structural units. Furthermore, the hydrolysis rate can be considered to be approximately equal regardless of the length of the polymer, the position of the bond, and whether the polymerized units in the carboxyl group at one end of the polymer and the hydroxyl group at the other end are 3HH or 3HB. Therefore, in this embodiment, P3HB3HH is approximated as a homopolymer, which is a chain of a single type of monomer. In this case, the molecular weight of each individual monomer is approximated as a weighted average of the mole fractions of the molecular weight units of 3HH and 3HB.

[0052] In the simulation according to this embodiment, it may be assumed that all bonds between monomers in a homopolymer are decomposed with equal probability. In this case, as illustrated in FIG. 10 , in the polymer population at the current step, bonds a1 and a2 in one polymer, bonds b1 to b4 in another polymer, and bonds c1 and c2 in yet another polymer are decomposed with equal probability. From another perspective, the expected number of decompositions that each polymer undergoes per unit time is proportional to the number of monomers constituting the polymer minus 1, i.e., the number of bonds between the monomers. Therefore, multiple bonds to be decomposed (sometimes referred to as "decomposition sites" in this application) may be set in one polymer within a given time. A polymer with n decomposition sites is decomposed into n+1 polymers. The maximum value of n is the number of bonds in the polymer. The decomposed polymers are then included in the polymer population in the next step, and the same process is repeated.

[0053] Next, the procedure of the simulation according to this embodiment will be described. FIG. 11 is a flowchart showing an example of a simulation of molecular weight distribution according to this embodiment. The molecular weight calculation unit 124 repeats the processing of steps S104 to S108, which are the decomposition process, from a first time to a second time. In the estimation of the molecular weight distribution, the first time corresponds to the start time of the simulation. The second time is the time at which the simulation is stopped, and in an actual production process, this may be the current time. The time interval between steps may be set arbitrarily. The time interval between steps may be, for example, the measurement time interval of the reaction condition variable or an integer multiple thereof. The time interval between steps is usually sufficiently shorter than the measurement time interval of the molecular weight distribution. (Step S102) The molecular weight calculation unit 124 generates an initial polymer population based on the actual measurement value of the molecular weight distribution at the first time. More specifically, the molecular weight calculation unit 124 includes, as polymers in the initial polymer population, chains made of monomers whose degree of polymerization is proportional to the number obtained by subtracting the molecular weight of the molecules located at both ends (i.e., the molecular weight of one water molecule) from the molecular weight, in a number proportional to the existence probability of the molecular weight that constitutes the actual measurement value. In the simulation program, the initial polymer population is represented, for example, as an array in which the number of elements for each polymer corresponds to the degree of polymerization, or as key-value format data in which the degree of polymerization is assigned to a key and the value for each key indicates the number of individuals with that degree of polymerization. The number of polymers in the initial polymer population is, for example, several thousand to tens of thousands or more.

[0054] (Step S104) The molecular weight calculation unit 124 sets the number of decomposition sites, which are bonds to be decomposed, as the total decomposition number so that the number is proportional to the product of the total number of bonds between monomers in all polymers in the polymer population and the reaction rate. (Step S106) The molecular weight calculation unit 124 randomly allocates decomposition sites, the number of which corresponds to the total number of decompositions, to the bonds of the polymers in the polymer population. (Step S108) The molecular weight calculation unit 124 decomposes the polymer into two or more new polymers at each of the assigned decomposition points to update the polymer population, and then returns to step S104.

[0055] Next, an example of a simulation according to this embodiment will be described. Using the molecular weight distributions actually measured in three patterns 1 to 3 shown in Fig. 12, an inverse analysis was performed to calculate the average reaction rate between each time point. Then, using the calculated reaction rate, a simulation was performed according to the method shown in Fig. 11. The validity of the molecular weight distribution calculated by the simulation was then verified.

[0056] Pattern 1 is at time t -1 The molecular weight distribution t measured at the first time -1The weight average molecular weight Mw of the molecular weight distribution estimated with time t0 as the second time is calculated by using the above formula, and the .... -1 This is a pattern in which reverse analysis is performed for the period from time t to time t0. -1 The period from time t0 to time t0 was set to 1 hour. Pattern 2 is at time t -2 The molecular weight distribution t measured at the first time -2 The weight average molecular weight Mw of the molecular weight distribution estimated with time t0 as the second time is calculated by using the above formula, and the .... -2 This is a pattern in which reverse analysis is performed for the period from time t to time t0. -2 The period from time t0 to time t0 was set to 2 hours. Pattern 3 is at time t -3 The molecular weight distribution t measured at the first time -3 The weight average molecular weight Mw of the molecular weight distribution estimated with time t0 as the second time is calculated by using the above formula, and the .... -3 This is a pattern in which reverse analysis is performed for the period from time t to time t0. -3 The period from time t0 to time t0 was set to 3 hours.

[0057] 13 to 16 show comparative examples of measured representative values ​​of molecular weight distribution and estimated representative values ​​of molecular weight distribution obtained by simulation using the reaction rate calculated by inverse analysis. In FIGS. 13 to 16, the horizontal axis indicates the second time t0 in FIG. 12, and the vertical axis indicates the representative value of molecular weight distribution. In the comparison between the measured values ​​and the estimated values ​​from each pattern, the number average molecular weight M n , weight average molecular weight M w , Z average molecular weight M z and Z+1 average molecular weight M z1 Generally, the number average molecular weight M n , weight average molecular weight M w , Z average molecular weight M z and Z+1 average molecular weight M z1The value increases in the order of . Also, the wider the molecular weight distribution, the higher the number average molecular weight M n , weight average molecular weight M w , Z average molecular weight M z and Z+1 average molecular weight M z1 The difference between them becomes larger.

[0058] 13, 14, 15, and 16 show the number average molecular weight M n , weight average molecular weight M w , Z average molecular weight M z , Z+1 average molecular weight M z1 The graph shows the time change of the number average molecular weight M. For each pattern, the representative values ​​of each type decrease with the passage of time in the same manner as the measured values, and no unusual change trend occurs. This indicates that the molecular weight distribution obtained by simulation can mimic the measured values. However, the degree of deviation between the measured values ​​and the estimated values ​​varies slightly depending on the type of representative value. For example, the number average molecular weight M n The estimated number average molecular weight M n The estimated value of number average molecular weight M tends to be smaller than the measured value. This indicates that the molecular weight distribution estimated by simulation contains a higher proportion of polymer compounds with lower molecular weights than the measured molecular weight distribution. n The estimated value of tends to differ more from the actually measured value in the order of patterns 1 to 3. Z+1 average molecular weight M z1 and Z-average molecular weight M z The estimated value of Z+1 average molecular weight M tends to be larger than the actual measured value. z1 and Z-average molecular weight M z The difference between the estimated patterns of Z+1 average molecular weight M z1 and Z-average molecular weight M z The estimated value of the number average molecular weight M n The difference between the estimated value and the actual measured value is not as significant as the estimated value.

[0059] The model learning unit 132 may perform inverse analysis based on the measured values ​​of molecular weight distribution obtained in different periods (patterns) for one second time t0 as shown in Figure 12, calculate the reaction rate at each time during that period, generate training data consisting of the reaction rate calculated for each time and reaction condition variables, and use these as individual training data to learn a reaction rate prediction model according to the above method.

[0060] In predicting the molecular weight distribution, the reaction rate calculation unit 122 uses a trained reaction rate prediction model to estimate the reaction rate for each different time period from the reaction condition variables and the representative value of the molecular weight at a first time for that time period. The molecular weight estimation unit 124 performs a simulation according to the estimated reaction rate using the actual measured value of the molecular weight distribution at the first time for each set period. The molecular weight estimation unit 124 calculates the average value of the molecular weight distribution for each set period obtained for the set period as the molecular weight distribution to be output. The average value may be a simple average or a weighted average. For example, the reaction rate calculation unit 122 sets a weighting coefficient so that the smaller the error between the actual measured value of the representative value of the molecular weight at the second time (e.g., number average molecular weight) for the set period is, the larger the average value may be. The reaction rate calculation unit 122 can calculate the weighted average value by summing the multiplication values ​​of the weighting coefficient set for each set period and the molecular weight distribution across the set period. By using the calculated weighted average, the resistance to measurement errors of the molecular weight distribution used as the initial value can be improved.

[0061] Here, the weighting coefficient used for the weighted average may be set to be larger as the period in which the distribution is reproduced by simulation becomes more accurate. For example, the weighting coefficient may be set to be larger as the period becomes shorter.

[0062] The applicant conducted cross-validation to evaluate the prediction performance of the molecular weight distribution obtained by weighted averaging. In the cross-validation, the molecular weight distribution estimated using the actual measured values ​​of the molecular weight distribution and reaction condition variables obtained from the molecular weight adjustment process of one lot was used as validation data. The actual measured values ​​of the molecular weight distribution and reaction condition variables obtained from the molecular weight adjustment process of another lot were used as training data. The average reaction rate was calculated by inverse analysis based on the actual measured values ​​of the molecular weight distribution of the other lot. Then, a reaction rate prediction model was trained using training data consisting of data pairs with the calculated average reaction rate as the objective variable and the molecular weight distribution and the actual measured values ​​of the reaction condition variables as the explanatory variables. In the validation, average reaction rates 1 to 3 for each of patterns 1 to 3 were estimated from the trained reaction prediction model based on the molecular weight distribution and reaction condition variables of the data of one validation lot. Then, a simulation was performed based on the estimated average reaction rates 1 to 3, and the weight-average molecular weight for the weighted average value of the obtained molecular weight distributions 1 to 3 was calculated as the estimated Mw value. Here, the weighting factors for molecular weight distributions 1, 2, and 3 were set to 20 / 26, 5 / 26, and 1 / 6, respectively. Then, the relative error between the estimated Mw value and the measured Mw value at each time point was calculated. The relative error corresponds to the ratio of the difference between the estimated Mw value and the measured Mw value to the measured Mw value (estimated Mw value - measured Mw value) / measured Mw value. The measured Mw value corresponds to the weight average molecular weight relative to the measured value of the molecular weight distribution.

[0063] Next, we explain the calculated relative error. Figure 17 illustrates the relative error of the Mw estimated value at the molecular weight measurement time closest to the reaction end time for each lot. Each lot is classified into one of three grades. Each grade corresponds to the difference between the reaction end point, i.e., the target Mw. Figure 17 shows that the relative error for each lot generally falls within the range of ±2.5%. This indicates that the weight-average molecular weight can be predicted with high accuracy by simulation based on the reaction rate estimated from the reaction rate prediction model trained over different time periods. For all lots, the standard deviation of the relative error was 1.7, and the percentage of lots with relative errors outside the ±2.5% range was 2 out of 28 lots. However, there are differences in estimation accuracy between grades. For example, the standard deviations for grades 1, 2, and 3 were 2.1, 0.95, and 0.26, respectively. The percentage of lots with relative errors outside the ±2.5% range was 2 out of 16 lots, 0 out of 9 lots, and 0 out of 3 lots, respectively.

[0064] As described above, the molecular weight prediction device 10 of this embodiment includes a reaction rate calculation unit 122 that calculates the reaction rate of a chemical reaction using a reaction rate prediction model based on reaction condition variables at a first time or later in a chemical treatment process (e.g., a molecular weight adjustment process) in which the molecular weight of a polymer compound is reduced by a chemical reaction, and a molecular weight calculation unit 124 that simulates a chemical reaction in which one polymer is used as a starting material and is decomposed into multiple polymers having molecular weights smaller than that of the starting material as target materials, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimates the molecular weight distribution of the polymer compound at a second time after the first time. The decomposition process for decomposing the starting material into a plurality of polymers with smaller molecular weights may include a process for breaking the bonds between the monomers constituting the starting material with a probability corresponding to the reaction rate. This configuration allows the decomposition process of polymer compounds with various molecular weights to be simulated at reaction rates corresponding to the reaction conditions, thereby successively estimating the molecular weight distribution, thereby obtaining a more accurate estimate of the molecular weight distribution than that calculated analytically.Furthermore, the estimated molecular weight distribution allows the derivation of a representative molecular weight value that is suitable for the purpose.

[0065] The probability of cleaving the bond between the monomers constituting the starting material may be constant regardless of the length and position of the bond of the polymer constituting the starting material. This method allows for the simulation of the fragmentation of a starting material into multiple polymers with a certain probability, regardless of the length of the polymer or the bond position. This allows for the estimation of molecular weight distribution under relatively simple conditions.

[0066] The reaction rate calculation unit 122 may further calculate the reaction rate using the molecular weight distribution of the polymer compound at the first time. According to this configuration, the reaction rate is calculated taking into consideration its dependency on the molecular weight distribution of the polymer compound at the first time, which makes it possible to estimate the reaction rate at each time after the first time more accurately than when the molecular weight distribution of the polymer compound at the first time is not taken into consideration.

[0067] The starting polymer may be a polymer compound containing 3-hydroxybutyrate (3HB) as a monomer. Furthermore, the polymer used as the starting material may be a polymer compound obtained by copolymerizing 3-hydroxybutyrate (3HB) and 3-hydroxyhexanoate (3HH) (ie, P3HB3HH). According to this configuration, by repeating the process of decomposing the starting material P3HB3HH into a plurality of P3HB3HHs with smaller molecular weights, a decrease in the molecular weight distribution of the entire P3HB3HH group can be simulated.

[0068] The process of breaking down a starting polymeric compound into multiple polymers is hydrolysis, and reaction variables may include pH or temperature. According to this configuration, the process of hydrolysis of a polymer compound as a starting material into a plurality of polymers at a reaction rate determined by the pH or temperature as a reaction condition is simulated.

[0069] The molecular weight prediction device 10 may include a model learning unit 132 that analyzes the reaction rate from the first time to the second time based on the actual measured value of the molecular weight distribution at the first time and the actual measured value of the molecular weight distribution at the second time of the training data, and learns a reaction rate prediction model that calculates the reaction rate based on the reaction condition variables based on the reaction condition variables at each time from the first time to the second time and the analyzed reaction rate. According to this configuration, the model learning unit 132 searches for a reaction rate at which an estimated value of molecular weight distribution at a second time obtained by performing a simulation based on the actual measured value of molecular weight distribution at a first time is as close as possible to the actual measured value of molecular weight distribution at the second time. The model learning unit 132 can learn a reaction rate prediction model so that the estimated value of reaction rate obtained from the reaction condition variables is as close as possible to the searched reaction rate. Therefore, based on the actual measured value of molecular weight distribution at the first time, a molecular weight distribution at the second time that is as close as possible to the actual measured value of molecular weight distribution at the second time is estimated by simulation.

[0070] The embodiments of the present invention have been described in detail above with reference to the drawings, but the specific configuration is not limited to that described above, and various design changes can be made within the scope of the gist of the present invention.

[0071] The production system S1 may be configured as a polymer compound production system including a reaction vessel (e.g., an adjustment tank 72) for carrying out a chemical treatment process, a molecular weight prediction device 10, and a control unit (e.g., a control information output unit 128, a process control device 60) for suppressing the above-mentioned chemical treatment process when a predetermined statistical quantity of molecular weight based on an estimated molecular weight distribution reaches a predetermined target value.

[0072] The production system S1 executes a polymer compound production method using a molecular weight prediction device 10, including the following molecular weight prediction method. The molecular weight prediction method includes a reaction rate calculation step of calculating a reaction rate of a chemical reaction using a reaction rate prediction model based on the molecular weight distribution of the polymer compound at a first time and reaction condition variables at times after the first time, in a chemical treatment process in which the molecular weight of the polymer compound is reduced by a chemical reaction; and a molecular weight calculation step of simulating the chemical reaction, i.e., the process in which a single polymer is used as a starting material and is decomposed into multiple polymers as target materials each having a molecular weight smaller than that of the starting material, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimating the molecular weight distribution of the polymer compound at a second time after the first time. The process control device 60 then executes a step of suppressing the chemical treatment process when a predetermined statistical quantity of molecular weight (e.g., weight average molecular weight) based on the molecular weight distribution estimated by the molecular weight prediction device 10 reaches a predetermined target value.

[0073] The above-described molecular weight prediction device 10 may operate in synchronization (online) with the polymer compound production process, or may operate asynchronously (offline) with the polymer compound production process. When operating online, the calculation and control unit 120 may monitor the actual measured value of molecular weight distribution indicated in the molecular weight data input from the molecular weight measuring instrument 20 and the estimated value of molecular weight distribution obtained by the molecular weight calculation unit 124. The calculation and control unit 120 may compare the actual measured value with the estimated value at the time of measurement of the actual measured value to determine the validity of the actual measured value. The calculation and control unit 120 can determine the presence or absence of an abnormality (abnormality detection) by, for example, whether the magnitude of the difference between the actual measured value and the estimated value is within a predetermined tolerance.

[0074] When an abnormality is detected, the arithmetic and control unit 120 may display notification information for notifying the abnormality on the display unit 50 or may transmit the notification information to another device (e.g., a mobile phone). By notifying the abnormality, various countermeasures can be promoted. For example, when an abnormality is detected, the arithmetic and control unit 120 may use the control information output unit 128 to instruct the process control device 60 to stop the molecular weight adjustment process. By enabling such control, it is possible to produce a polymer compound whose molecular weight has a small deviation from a predetermined target molecular weight. This in turn contributes to stabilizing the quality of the produced polymer compound.

[0075] When operating asynchronously with the polymer compound production process, the molecular weight prediction device 10 may use, for example, pre-collected process data and molecular weight data. The production system S1 may be equipped with a database system that associates the collected process data and molecular weight data and temporarily or permanently stores them. In this case, the molecular weight prediction device 10 accesses the database system to acquire the process data and molecular weight data.

[0076] In the above description, the reaction condition variables are pH and temperature, but are not limited to these. The reaction condition variables may further include some or all of the stirring speed, the compound addition rate, and the volume of the solution held in the adjustment tank 72. Although the reaction stopping measure has been described as being performed by the process control device 60 to lower the temperature in the adjustment tank 72, the present invention is not limited to this. Instead of or in addition to lowering the temperature, the process control device 60 may lower the pH in the adjustment tank 72 from a predetermined pH in the molecular weight adjustment step to a level at which hydrolysis does not occur.

[0077] Some parts of the molecular weight prediction apparatus 10, such as the control information output unit 128 and the model learning unit 132, may be omitted or may be configured as separate devices from the molecular weight prediction apparatus 10. The control information output unit 128 and the model learning unit 132, configured as separate devices, may be connected to the molecular weight prediction apparatus 10 wirelessly or via a wire so as to be able to send and receive various types of data.

[0078] The model learning unit 132 may estimate the reaction rate prediction model in synchronization with the reaction rate estimation by the reaction rate calculation unit 122, or may estimate the reaction rate prediction model asynchronously with the reaction rate estimation. The reaction rate prediction model may be trained in a device separate from the molecular weight prediction device 10 , and a parameter set of the reaction rate prediction model obtained by training may be set in the reaction rate calculation unit 122 .

[0079] The above-described simulation was performed using an example in which the polymer compound P3HB3HH was applied to ester hydrolysis in an alkaline solution in which it was suspended, but the present invention is not limited to this. The above-described simulation may also be applied to other polymer compounds, such as poly(3-hydroxyalkanoates) (P3HA: poly(3-hydroxyalcanoate)) other than P3HB3HH, such as poly(3-hydroxybutyrate) (P3HB: poly(3-hydroxybutyrate)) and poly(3-hydroxybutyrate-co-4-hydroxyhexanoate) (P3HB4HB). Furthermore, the above-described simulation is not limited to polyesters containing 3HB as a monomer, but may also be applied to other polyesters, such as polylactic acid, polybutylene succinate, polybutylene succinate adipate, polybutylene adipate terephthalate, polybutylene succinate terephthalate, polycaprolactone, and polyethylene terephthalate (PET). The simulation described above can also be applied to polymers other than polyester, such as polystyrene and polyvinyl chloride. The chemical reaction to which the simulation is applied is not limited to hydrolysis, but depends on the polymer to which the simulation is applied. For example, the simulation can be applied to thermal decomposition, oxidation, and the like.

[0080] Furthermore, part or all of the molecular weight prediction device 10 in the above-described embodiment may be realized as an integrated circuit such as an LSI (Large Scale Integration), or may be configured as dedicated hardware. Each functional block of the molecular weight prediction device 10 may be individually implemented as a processor, or part or all of the functional blocks may be integrated into a processor. Furthermore, the integrated circuit implementation method is not limited to LSI, and may be implemented using a dedicated circuit or a general-purpose processor. Furthermore, if an integrated circuit implementation technology that can replace LSI emerges due to advances in semiconductor technology, an integrated circuit based on that technology may be used. [Explanation of symbols]

[0081] S1...production system, 10...molecular weight prediction device, 20...molecular weight measurement device, 30...process data detector, 50...display unit, 60...process control device, 120...arithmetic and control unit, 122...reaction rate calculation unit, 124...molecular weight calculation unit, 128...control information output unit, 132...model learning unit, 140...storage unit

Claims

1. In chemical processing processes where the molecular weight of polymer compounds is reduced by chemical reactions, a reaction rate calculation unit that calculates a reaction rate of the chemical reaction using a reaction rate prediction model based on reaction condition variables at a first time point or later; and a molecular weight calculation unit that simulates, as the chemical reaction, a process in which one polymer is used as a starting material and is decomposed into a plurality of polymers as target materials each having a molecular weight smaller than that of the starting material, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimates the molecular weight distribution of the polymer compound at a second time that is later than the first time. Molecular weight prediction device.

2. The process includes a treatment of breaking the bond between the monomers constituting the starting material with a probability corresponding to the reaction rate. The molecular weight prediction device according to claim 1 .

3. The probability is constant regardless of the length of the starting polymer and the position of the bond. The molecular weight prediction device according to claim 2 .

4. The reaction rate calculation unit further calculates the reaction rate using the molecular weight distribution of the polymer compound at the first time. The molecular weight prediction device according to claim 1 .

5. The polymer is a polymer compound containing 3-hydroxybutyrate as a monomer. The molecular weight prediction device according to claim 1 .

6. The polymer is a polymer compound further containing 3-hydroxyhexanoate as a monomer. The molecular weight prediction device according to claim 5 .

7. the chemical reaction is hydrolysis, The reaction condition variables include pH or temperature. The molecular weight prediction device according to claim 5 .

8. analyzing the reaction rate from the first time to the second time based on the measured value of the molecular weight distribution at the first time and the measured value of the molecular weight distribution at the second time; a model learning unit that learns the reaction rate prediction model based on the reaction condition variables and the reaction rates at each time from the first time to the second time. The molecular weight prediction device according to claim 1 .

9. a reaction vessel in which the chemical treatment process is carried out; The molecular weight prediction device according to claim 1; When a predetermined statistical quantity of molecular weight based on the estimated molecular weight distribution reaches a predetermined target value, a control unit that suppresses the chemical treatment process Polymer compound production system.

10. A program for causing a computer to function as the molecular weight prediction device according to claim 1.

11. A molecular weight prediction method in a molecular weight prediction device, The molecular weight prediction device In chemical processing processes where the molecular weight of polymer compounds is reduced by chemical reactions, a reaction rate calculation step of calculating a reaction rate of the chemical reaction using a reaction rate prediction model based on reaction condition variables at a first time or later; a molecular weight calculation step of simulating, as the chemical reaction, a process in which one polymer is used as a starting material and is decomposed into a plurality of polymers as target materials each having a molecular weight smaller than that of the starting material, based on the reaction rate and the molecular weight distribution of the polymer compound actually measured at the first time, and estimating the molecular weight distribution of the polymer compound at a second time that is later than the first time. Molecular weight prediction methods.

12. The reaction rate calculation step further calculates the reaction rate using a molecular weight distribution of the polymer compound at the first time. The molecular weight prediction method according to claim 11.

13. Execute the molecular weight prediction method according to claim 11 or 12, When a predetermined statistical quantity of molecular weight based on the estimated molecular weight distribution reaches a predetermined target value, Inhibit the chemical treatment process A method for producing polymeric compounds.

Images