JP2022146429A - Prediction device, control device, manufacturing method, and program - Google Patents

Abstract

To provide prediction devices or the like capable of predicting the concentration of an oil as a substrate.SOLUTION: A prediction device according to an aspect of the present invention is a prediction device in a culture system for producing a biodegradable polymer by culturing microorganisms using an oil as a substrate, comprising: a near-infrared spectroscopic analysis unit that analyzes components contained in a culture solution; an acquisition unit that acquires variables related to a culture by the culture system; and a prediction unit that predicts the concentration of the substrate based on the analysis result by the near-infrared spectroscopic analysis unit and the variables acquired by the acquisition unit.SELECTED DRAWING: Figure 1

Description

The present invention relates to a prediction device, control device, manufacturing method, and program.

BACKGROUND ART Conventionally, biodegradable polymers, which are polymers produced by microorganisms using vegetable oils as raw materials and are biodegradable even in seawater, have been known. This biodegradable polymer is, for example, PHBH (Kaneka Biodegradable Polymer, registered trademark). A biodegradable polymer production process includes, for example, a culturing step, a purification step, and a post-treatment step. The culture step includes, for example, a seed culture step and a main culture step, and the main culture step includes a medium preparation step, an inoculation step, a carbon source (oil) feeding step, and a culture step.

For example, Patent Document 1 is known as a technique for culturing cells. The cell culture control system described in Patent Document 1 measures metabolite concentration values in the culture solution, generates time-series data, and based on the feature points of the time-series data, physical and chemical variables of the culture solution. is controlling Furthermore, the cell culture control system measures the nutrient concentration value, metabolite concentration value, pH value, temperature value, dissolved oxygen concentration value, and osmotic pressure value of the culture solution, and calculates the predicted value of the time series data. 7, and extracts the feature points of the predicted values.

A state detection technique using a near-infrared spectrophotometer in the technique of culturing microorganisms is described in Non-Patent Document 1, for example.

Furthermore, Patent Documents 2 and 3, for example, are known as techniques using a near-infrared spectroscopic analyzer. In Patent Document 2, using a near-infrared spectrometer, the absorption spectrum of the sample in the reactor in the polymerization process is measured at a wavelength of 800 to 2500 nm, and based on the measured value, the aromatic petroleum obtained in the final process A method for producing an aromatic petroleum resin is described in which the physical properties of the resin are predicted and the production process is controlled based on the predicted values.

In Patent Document 3, a measurement device including a near-infrared spectrometer having an analysis device capable of performing multivariate analysis by chemometrics is used to measure the near-infrared absorption spectrum of the object to be measured during the dehydration condensation reaction. Subsequently, based on the measurement data, at least one physical property value selected from terminal group number (or terminal group concentration), water content, concentration, and molecular weight is calculated, and from these calculated values, using an arithmetic device, A process for producing aliphatic polyesters is described in which instrument control information is obtained and this information is used to remotely operate reactor instrumentation to control specific reaction conditions and reaction times.

JP 2015-216886 A JP-A-2002-145966 JP-A-11-60711

M.V. Cruz et al., "Online monitoring of P(3HB) produced from used cooking oil with near-infrared spectroscopy", Journal of Biotechnology 194 (2015) 1-9

In the above-described fed-batch culture process using oil as a substrate, the inventors have found that it is important to appropriately control the substrate concentration in the culture system in consideration of productivity and the like. If the amount of oil is small, the rate of assimilation will decrease. Conversely, if the amount of oil supplied is excessive, foaming will make stable cultivation difficult. However, in the techniques described in Non-Patent Document 1 and Patent Documents 2-3 described above, the absorption position of the functional group (e.g., methyl group, carbonyl group) of the skeleton of PHBH in the near-infrared spectrophotometer and the oil Since the absorption positions of functional groups (eg, methyl group, carbonyl group) appear close to each other, it may be difficult to obtain a good calibration curve. In this case, for example, a large amount of data is required in order to obtain a good calibration curve in the near-infrared spectrophotometer.

Furthermore, the degree of polymerization of the polymer, its concentration in the culture system, the degree of influence of the immiscibility of the oil, and the degree of influence of the state of foaming may cause an error in the measurement result of the oil concentration with the near-infrared spectrophotometer. Furthermore, the error in the measurement results of the oil concentration in the near-infrared spectrophotometer is due to the physical and chemical effects of NIR such as the temperature in the culture system, the particle size and pH of the fungus, general molecular vibration and It affects not only the absorption of near-infrared light originating from rotation, but also the electronic transition caused by near-infrared light and the scattering of near-infrared light. FIG. 13 is a diagram showing an example of the relationship between the measured NIR intensity and the wave number.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a prediction device, a control device, a manufacturing method, and a program capable of predicting the concentration of oil as a substrate.

A prediction device according to an aspect of the present invention is a prediction device in a culture system for producing a biodegradable polymer by culturing microorganisms using oil as a substrate, and includes near-infrared spectroscopy for analyzing components contained in the culture solution. an analysis unit, an acquisition unit that acquires variables related to culture by the culture system, and a prediction unit that predicts the concentration of the substrate based on the analysis result by the near-infrared spectroscopic analysis unit and the variables acquired by the acquisition unit. And prepare.

A control device according to an aspect of the present invention is a control device in a culture system for producing a biodegradable polymer by culturing microorganisms using oil as a substrate, and includes near-infrared spectroscopy for analyzing components contained in the culture solution. an analysis unit, an acquisition unit that acquires variables related to culture by the culture system, and a prediction unit that predicts the concentration of the substrate based on the analysis result by the near-infrared spectroscopic analysis unit and the variables acquired by the acquisition unit. and a control unit that controls the culture system based on the prediction result of the prediction unit.

A production method according to an aspect of the present invention is a production method in a culture system for producing a biodegradable polymer by culturing microorganisms using oil as a substrate, comprising the steps of: analyzing components contained in the culture solution; Obtaining variables related to culturing by the culturing system and predicting concentrations for the substrate based on the analytical results of the components and the variables.

A program according to an aspect of the present invention is provided in a computer in a culture system for producing biodegradable polymers by culturing microorganisms using oil as a substrate, analyzing components contained in a culture solution, and culturing by the culture system. obtaining a variable for and predicting a concentration for the substrate based on the analytical results of the component and the variable.

According to the invention, concentrations can be predicted for oils as substrates.

It is a figure showing an example of culture system 1 of an embodiment. 4 is a block diagram showing an example of a learning device 400 in an embodiment; FIG. It is a figure which shows the explanatory variable in embodiment, a prediction model, and the relationship of an objective variable. It is a figure which shows the result of having evaluated the prediction model by the learning process in embodiment. It is a figure which shows evaluation of the result of having predicted the oil concentration by the prediction model in embodiment. FIG. 5 is a diagram showing the relationship between the oil concentration predicted by the embodiment and the oil concentration measured as a comparative example; (A) is a diagram showing the relationship between the oil concentration measured by the gas chromatography device (GC) and the oil concentration measured by NIR, and (B) is the oil concentration measured by the gas chromatography device (GC). and oil concentration predicted by the prediction model of the embodiment. It is a figure which shows the result of having evaluated the prediction model by the learning process in embodiment. It is a figure which shows evaluation of the result of having predicted the oil concentration by the prediction model in embodiment. FIG. 5 is a diagram showing the relationship between the oil concentration predicted by the embodiment and the oil concentration measured as a comparative example; (A) is a diagram showing the relationship between the oil concentration measured by the gas chromatography device (GC) and the oil concentration measured by NIR, and (B) is the oil concentration measured by the gas chromatography device (GC). and oil concentration predicted by the prediction model of the embodiment. (A) is a histogram showing the residual difference between the oil concentration measured by the gas chromatograph (GC) and the oil concentration measured by the NIR, and (B) is the oil measured by the gas chromatograph (GC). 4 is a histogram showing residuals between concentrations and oil concentrations predicted by the prediction model of the embodiment. It is a figure which shows an example of the relationship between the measured intensity of NIR, and a wavenumber.

A prediction device, a control device, a manufacturing method, and a program to which the present invention is applied will be described below with reference to the drawings.

[Overview of culture system 1]
FIG. 1 is a diagram showing an example of a culture system 1 of an embodiment. The culture system 1 of the embodiment is a system for culturing microorganisms in a process of producing a biodegradable polymer by culturing microorganisms using oil as a substrate. The biodegradable polymer is, for example, PHBH (Kaneka Biodegradable Polymer) (registered trademark), PHBH is a type of PHA (polyhydroxyalkanoate), and 3HB (3-hydroxybutyrate) and 3HH (3 -hydroxyhexanoate). PHBH is a raw material for biodegradable plastics. However, the production object of the present embodiment is not limited to PHBH, and may be other copolymers such as PHA. PHA in the embodiment is P (3HB) which is a homopolymer of 3-hydroxybutanoic acid (3HB), P (3HB-co-3HV) which is a copolymer of 3HB and 3-hydroxyvaleryl acid (3HV), 3HB and Examples include a copolymer P (3HB-co-3HH) (abbreviation: PHBH) of 3-hydroxyhexanoic acid (3HH) and a copolymer of 3HB and 4-hydroxybutanoic acid (4HB).

The production process of PHBH is carried out, for example, in the order of culture process, purification process, and post-treatment process. The culture process includes, for example, a seed culture process and a process called a main culture process, and the main culture process includes a medium preparation process, an inoculation process, a feeding process of oil as a substrate, and a culture process. include. In the oil fed-batch process, phosphoric acid and aqueous ammonia are fed as other substrates. The culture system 1 of the embodiment predicts the oil concentration in the culture solution in order to manage the oil flow rate in the oil feeding process. As a result, the culture system 1 can make the administrator recognize the oil concentration by, for example, displaying the predicted oil concentration in real time. Furthermore, the culture system 1 of the embodiment may automatically control the oil flow rate based on the predicted oil concentration in the culture solution.

A PHA microbial culture substrate in embodiments is, for example, an oil. Oils include fats, fatty acids, and glycerin. Fats and oils are esters of fatty acids and glycerin. However, the substrate for PHA microbial culture is not limited to oils, but may be fatty acids, carbon sources such as glucose, or combinations thereof. Fats and oils used in culture in the embodiments are, for example, triacylglycerol, diacylglycerol, monoacylglycerol, and the like. Fatty acids used in the culture in the embodiment are, for example, palmitic acid, oleic acid, linoleic acid, stearic acid and the like.

[Configuration of culture system 1]
The culture system 1 includes a culture device 100 as shown in FIG. The culture apparatus 100 includes, for example, a culture tank 110, an agitator 120, a pump system 130, and sensors (140, 142, 144, 146, and 148). A culture solution is stored in the culture tank 110 . The stirring device 120 includes, for example, a motor, a motor driving device (not shown), a stirring rod 112 and the like, and stirs the culture solution stored in the culture tank 110 . The pump system 130 includes a pump, a pump driving device (not shown), a flow path connected to the pump, and the like, and supplies materials necessary for culture such as oil, NH4OH, and H3PO4 into the culture tank 110. The exhaust gas sensor 140 outputs a signal based on the oxygen concentration and a signal based on the CO2 concentration in the gas discharged from the culture tank 110 . A dissolved oxygen (DO) sensor 142 outputs a signal based on the concentration of oxygen dissolved in the culture medium. A temperature sensor 144 outputs a signal based on the temperature of the culture solution. The pH sensor 146 outputs a signal regarding the pH of the culture solution. The NIR (near-infrared) sensor 148 includes, for example, a light source that emits near-infrared rays and a photodetector that detects reflected light or transmitted light from the culture solution. NIR sensor 148 outputs a signal based on the reflected light.

Furthermore, the culture apparatus 100 includes an O2/CO2 acquisition section 200 , a DO acquisition section 202 , a temperature acquisition section 204 , a pH acquisition section 206 and an NIR analysis section 208 . The O2/CO2 acquisition unit 200 acquires a signal based on the oxygen concentration in the gas and a signal based on the CO2 concentration from the exhaust gas sensor 140 and supplies O2 concentration information and CO2 concentration information of the exhaust gas to the prediction device 300 . The DO acquisition unit 202 acquires a signal based on the concentration of oxygen dissolved in the culture solution from the dissolved oxygen sensor 142 and supplies dissolved oxygen concentration information to the prediction device 300 . The temperature acquisition unit 204 acquires a signal based on the temperature of the culture fluid from the temperature sensor 144 and supplies culture fluid temperature information to the prediction device 300 . The pH acquisition unit 206 acquires a signal based on the pH of the culture solution from the pH sensor 146 and supplies pH information to the prediction device 300 . The NIR analyzer 208 is a near-infrared spectrophotometer that acquires a signal based on reflected light or transmitted light from the NIR sensor 148 , analyzes the acquired signal, and supplies the analysis results to the prediction device 300 . The NIR analysis unit 208 is an example of a near-infrared spectroscopic analysis unit that analyzes components contained in the culture solution. The O2/CO2 acquisition unit 200 , the DO acquisition unit 202 , the temperature acquisition unit 204 , and the pH acquisition unit 206 are examples of acquisition units that acquire variables related to culture by the culture system 1 .

Furthermore, culture device 100 includes prediction device 300 , learning device 400 , display device 500 , and control device 600 . The prediction device 300 executes processing based on the prediction model 304 by the prediction processing unit 302 to predict the oil concentration in the culture fluid based on the analysis result by the NIR analysis unit 208 and variables related to culture. The prediction processing unit 302 is implemented by a processor such as a CPU (Central Processing Unit) executing a program stored in a program memory. The predictive model 304 is, for example, a regression model that receives an explanatory variable as a variable related to culture and outputs an objective variable as the oil concentration in the culture solution. The explanatory variables are, for example, physically acquired information and information stored in the explanatory variable storage unit 310 .

The display device 500 is, for example, a display device that can be viewed by an administrator of the culture system 1 . The display device 500 displays information about the oil concentration predicted by the prediction device 300, for example.

The control device 600 controls each part of the culture system 1 based on the oil concentration predicted by the prediction device 300 . The control device 600 includes a processor such as a CPU, and is implemented by the processor executing a program stored in a program memory. The control device 600, for example, calculates a target oil flow rate based on the oil concentration predicted at a predetermined timing in the culture process, and controls the pump system 130 so as to approach the target oil flow rate. The predetermined timing is desirably set to occur continuously, for example, every few minutes, so that the control device 600 continuously acquires the oil concentration and performs control based on the oil concentration. is desirable.

Further, the controller 600 desirably controls the pump system 130 to achieve the optimum oil flow rate based on the O2 concentration and CO2 concentration in the exhaust gas, and the optimum oil flow rate based on the bubbling in the culture tank 110. It is desirable to control the flow rate. In this case, it is desirable that the control device 600 controls the oil flow rate based on the oil concentration predicted by the prediction device 300 so that the oil concentration in the culture solution falls within a predetermined range.

Learning device 400 learns predictive model 402 based on teacher data stored in teacher data storage unit 410 . Prediction model 402 learned by learning device 400 is introduced into prediction device 300 as prediction model 304 .

FIG. 2 is a block diagram showing an example of the learning device 400 according to the embodiment. The learning device 400 includes, for example, a teacher data acquisition unit 420, a model construction unit 430, a prediction model 402, and a learning result storage unit 440.

The teacher data acquisition unit 420 acquires the explanatory variable x and the objective variable y as teacher data from the teacher data storage unit 410 . The explanatory variable x and objective variable y in one teacher data have a corresponding relationship. That is, when the culture process is performed under the calculated NIR value and/or the culture conditions specified by the explanatory variable x in certain teacher data, the oil concentration is determined by the objective variable y in the teacher data.

The model construction unit 430 learns a function f or a processing parameter that specifies the prediction model 402 so that the explanatory variable x is input to the prediction model 402 and the objective variable y corresponding to the explanatory variable x is output from the prediction model 402. . The model construction unit 430 recursively calculates (updates) the function f or the processing parameter so that, for example, the difference between the objective variable y corresponding to the explanatory variable x and the output of the prediction model 402 becomes smaller. The prediction model 402 is, for example, a regression model before being learned. The function f in the prediction model 402 is, for example, an arithmetic expression for calculating the explanatory variable x, and the processing parameter is, for example, a weighting factor in the arithmetic expression. The model construction unit 430 saves the calculated (updated) function f and the processing parameter in the learning result storage unit 440 .

FIG. 3 is a diagram showing the relationship among explanatory variables, prediction models, and objective variables in the embodiment. The explanatory variable storage unit 310 is a storage device including, for example, a hard disk device and a database management device. The explanatory variable storage unit 310 stores explanatory variables including NIR calculated values (analysis results), culture conditions, polymer information, substrate information, process parameters, and microorganism information. The information stored in the explanatory variable storage unit 310 includes information acquired by the acquiring units (200, 202, 204, 206, and 208), information based on signals detected by sensors (not shown), and information received by the explanatory variable setting unit 312. information. The explanatory variable setting unit 312 is, for example, a personal computer or the like used by the administrator of the culture system 1, and stores explanatory variables in the explanatory variable storage unit 310 based on the operation of the administrator.

The prediction model in the embodiment is a model using a regression formula, but the regression formula for realizing the prediction model is multiple regression in linear regression, sparse regression, PLS (partial least squares regression) regression, etc. Good, but not limited to, known regression equations that can input explanatory variables and output objective variables for predicting oil concentrations, and various types that can be applied to predict oil concentrations in the future A regression equation may be used. The model construction unit 430 selects, for example, one of multiple regression, sparse regression, and PLS regression, and selects a combination of explanatory variables that can predict the oil concentration with high accuracy using the selected regression. A prediction model can be constructed by Furthermore, the model building unit 430 may predict the oil concentration using the optimum combination of explanatory variables for each regression equation, and the optimum regression equation based on the combination of explanatory variables that can be acquired (measured) in the culture system 1 may be adopted.

NIR calculated value (analysis result) information includes, for example, the concentration of oil in the culture medium, information on the copolymer in the culture medium, and weight average molecular weight (Mw) information for each component contained in the culture medium. , at least one of the information on the dry cell weight (DCW) in the culture medium. The calculated value (analysis result) information of the NIR as the explanatory variable may be information that can be output by the NIR analysis unit 208 and is information that correlates with the oil concentration as the prediction target. The NIR calculation value information is the analysis result of the NIR analysis unit 208, but is not limited to this. Just do it. The value obtained by processing the analysis result of the NIR analysis unit 208 is, for example, an average value of the analysis result, a normalized value, a data group from which peculiar data are excluded, and the like.

The culture condition information includes, for example, the feed amount of oil (substrate) per unit time, the integrated feed amount of oil (substrate), the feed amount of alkaline solution (eg, NH4OH) per unit time, and the feed amount of alkaline solution (eg, NH4OH). integrated feed amount, acid solution (e.g., H3PO4) feed amount per unit time, acid solution (e.g., H3PO4) feed amount integrated value, culture time information, agitation number information, agitation power information, culture tank 110 It includes at least one of internal pressure information, aeration amount information in the culture tank 110, culture solution temperature information, culture solution pH information, culture solution volume information, and inoculation volume information. Information on the internal pressure of the culture tank 110 and information on the amount of aeration in the culture tank 110 may be detected by sensors (not shown), and may be predetermined values such as target values and empirical values. It should be noted that the culture condition information as an explanatory variable may be any information that correlates with the oil concentration as a prediction target among the culture conditions. The culture condition information may be values obtained by processing the culture conditions. Values obtained by processing culture conditions are, for example, average values of culture conditions, normalized values, data groups from which peculiar data are excluded, and the like.

Polymer information is an example of information about biodegradable polymers. The polymer information includes at least one of the concentration of the polymer (biodegradable polymer) in the culture medium, the ratio information of the copolymer as the primary structure, and the inter- and intramolecular composition distribution information of the polymer. The primary structure is, for example, a structure showing the state before the polymer (biodegradable polymer) is formed (e.g., copolymer, comonomer (e.g., 3HH)), and the ratio of the copolymer is the copolymer and polymer in the culture medium. and the comonomer ratio indicates the ratio of comonomer and polymer in the culture medium. Note that the copolymer ratio and comonomer ratio are examples of information about the primary structure of the biodegradable polymer.

The polymer information may be a measured value obtained by sampling and analyzing the culture fluid, or may be a predetermined value such as a target value or an empirical value. The polymer information as an explanatory variable may also include other information about the polymer to which the embodiment can be applied, as long as it is information correlated with the oil concentration as a prediction target among the information about the polymer. The polymer information may be a value obtained by processing the polymer information. The value obtained by processing the polymer information is, for example, an average value of values related to the polymer, a normalized value, a data group from which peculiar data are excluded, and the like.

The substrate information includes, for example, at least one of the decomposition state of the substrate, the type and composition ratio of oils and fats contained in the culture medium, the type and composition ratio of fatty acids contained in the culture medium, and the consumption rate of the substrate. Fats and oils used in culture in the embodiments are, for example, triacylglycerol, diacylglycerol, monoacylglycerol, and the like. Fatty acids used in the culture in the embodiment are, for example, palmitic acid, oleic acid, linoleic acid, stearic acid and the like. The substrate information may be a measured value obtained by sampling and analyzing the culture solution, or may be a predetermined value such as a target value or an empirical value. Substrate information as an explanatory variable may also include other substrate-related information to which the embodiments can be applied, as long as it is information correlated with the oil concentration as a prediction target among the substrate-related information. Substrate information may be a value obtained by processing a value relating to a substrate. The value obtained by processing the substrate-related value is, for example, the average value of the substrate-related value, the normalized value, or a data group from which peculiar data are excluded.

A process parameter is an example of information regarding the state of the culture system 1 . The process parameters are, for example, the amount of oxygen in the culture solution (dissolved oxygen concentration; DO), the O2 concentration and CO2 concentration of the exhaust gas, the oxygen uptake rate (OUR), the respiratory quotient during the culture (respiratory quotient; RQ ), including at least one of the parameters relating to the state of foaming. The parameters relating to the foaming state are, for example, information such as the position of the foaming liquid surface, changes in the position, and generation period. The process parameter may be a measured value, a target value, or a predetermined value such as an empirical value. The process parameter as the explanatory variable may include other information related to the process to which the embodiment can be applied, as long as it is information correlated with the oil concentration as the target of prediction among the information related to the process parameter. The process parameter may be a processed value of the process parameter. The value obtained by processing the value of the process parameter is, for example, an average value of the values of the process parameter, a normalized value, or a data group from which peculiar data are excluded.

OUR is the amount of oxygen consumed by the cells, and is calculated based on the detected values of exhaust gas O2 and CO2. A high OUR indicates a high oil utilization rate (productivity). An increase in OUR indicates an increase in cell activity, and a decrease in OUR indicates at least one of a decrease in cell activity, a decrease in oxygen supply rate, and an oil shortage. represents that A high OUR is desirable for improving productivity.

DO is the concentration of oxygen remaining dissolved in the culture medium, and a large DO indicates a state in which there is excess oxygen in the culture medium and the rate of oil assimilation is low, resulting in a shortage of oil. A state in which DO is small (≈0) represents a state in which all the oxygen supplied to the culture medium is consumed. An increase in DO indicates that there is excess oxygen in the culture solution, and at least one of a decrease in bacterial cell activity or a shortage of oil has occurred, and a decrease in DO indicates that It indicates that there is enough oxygen in the culture medium. A low DO is desirable for improving productivity.

RQ is the ratio of carbon dioxide (CO2) to consumed oxygen (O2), and a large RQ indicates a state in which oil is wasted. The state of wastefully consuming oil is, for example, a state in which the consumed C is converted into CO2 at a high rate. A state in which RQ is small represents a state in which oil is efficiently converted into polymer. A state in which the oil is efficiently converted to polymer is a state in which the rate of conversion of the consumed C to CO2 is low. A low RQ is desirable for improving productivity.

Microorganism information is an example of information on microorganisms such as bacteria, yeast, fungi, and archaea that produce biodegradable polymers. The microbial information includes, for example, at least one of the strain type, the particle size of the biodegradable polymer-containing microbial cell, and the microbial cell weight in the dry microbial cell weight of the biodegradable polymer-containing microbial cell. The microbial information may be a measured value, or a predetermined value such as a target value or an empirical value. Microorganism information as an explanatory variable may be any information related to microorganisms that is correlated with the oil concentration to be predicted. Microorganism information may be a value obtained by processing a value related to microorganisms. Values obtained by processing microorganism-related values are, for example, average values of microorganism-related values, normalized values, data groups from which peculiar data are excluded, and the like.

[Example]
FIG. 4 is a diagram showing the results of evaluating the prediction model by the learning process in the above-described embodiment, and the MSE ( mean squared error), RMSE (root mean squared error), MAE (mean absolute error), R2 (coefficient of determination). FIG. 5 is a diagram showing the evaluation of the results of predicting the oil concentration by the prediction model in the above-described embodiment, and the MSE, RMSE, MAE, when using the PLS regression model and the linear regression model as the prediction models. Shows R2.

FIG. 6 is a diagram showing the relationship between the oil concentration predicted by the embodiment and the oil concentration measured as a comparative example. In FIG. 4, the plotted points indicate the predicted oil concentration. The measured oil concentration is the result of sampling the oil concentration in the culture solution and measuring it using a gas chromatograph (Gas Chromatograph). The prediction model was partial least squares regression. The explanatory variables in the examples are the oil concentration and the copolymer ratio (ratio of PHBH and 3HH) as calculated values of NIR, the oil feed amount, the ammonia water feed amount, and the culture time as culture conditions, strain information, OUR (oxygen consumption rate) as a process parameter. The example yielded a root mean square error (RMSE) of 0.149 and a coefficient of determination (R2) of 0.723 for the predictive model. Also, it can be seen that the predicted oil concentration has a high correlation with the measured oil concentration.

FIG. 7A is a diagram showing the relationship between the oil concentration measured by the gas chromatography device (GC) and the oil concentration measured by NIR, and FIG. It is a figure which shows the relationship between the measured oil concentration and the oil concentration predicted by the prediction model of embodiment. According to FIG. 7, the error in the oil concentration measured by NIR with respect to the oil concentration measured by GC is large, and the error in the oil concentration predicted by the prediction model of the embodiment with respect to the oil concentration measured by GC is small.

FIG. 8 is a diagram showing the results of evaluating the prediction model by the learning process in the above-described embodiment. show. FIG. 9 is a diagram showing the evaluation of the result of oil concentration prediction by the prediction model in the embodiment described above, showing MSE, RMSE, MAE, and R2 when the PLS regression model is used as the prediction model. The explanatory variables in the examples were the oil concentration and dry cell weight as calculated values of NIR, the integrated ammonia water feed amount and oil feed amount as culture conditions, and strain information (dummy variables).

FIG. 10 is a diagram showing the relationship between the oil concentration predicted by the embodiment and the oil concentration measured as a comparative example. In FIG. 10, the plotted points indicate the predicted oil concentration. The measured oil concentration is the result of sampling the oil concentration in the culture solution and measuring it using a gas chromatograph (Gas Chromatograph). The prediction model was partial least squares regression. Explanatory variables in the example are the oil concentration and the dry cell weight as calculated values of NIR, the integrated value of ammonia water feed amount and oil feed amount as culture conditions, and strain information. The example yielded a root mean square error (RMSE) of 0.094 and a coefficient of determination (R2) of 0.924 for the predictive model. Also, it can be seen that the predicted oil concentration has a high correlation with the measured oil concentration.

FIG. 11(A) is a diagram showing the relationship between the oil concentration measured by the gas chromatography device (GC) and the oil concentration measured by NIR, and FIG. 11(B) is the gas chromatography device (GC) It is a figure which shows the relationship between the measured oil concentration and the oil concentration predicted by the prediction model of embodiment. The lower part of FIG. 11(A) shows MSE, RMSE, MAE, and R2 as accuracy indicators for the oil concentration measured by NIR, and the lower part of FIG. 11(B) shows the accuracy of the oil concentration predicted by the prediction model in the embodiment. MSE, RMSE, MAE, and R2 are shown as indices. According to FIG. 11 (A), the error in the oil concentration measured by NIR with respect to the oil concentration measured by GC is large, and according to FIG. 11 (B), the prediction model of the embodiment for the oil concentration measured by GC It can be seen that the error in the oil concentration calculated is small.

FIG. 12(A) is a histogram showing the residual difference between the oil concentration measured by the gas chromatography device (GC) and the oil concentration measured by NIR, and FIG. 12(B) is the gas chromatography device (GC) 3 is a histogram showing the residual difference between the oil concentration measured in , and the oil concentration predicted by the prediction model of the embodiment. According to FIG. 12, it can be seen that the error in the oil concentration predicted by the prediction model of the embodiment is smaller than the error in the oil concentration measured by NIR.

[Effects of Embodiment]
According to the culture system 1 of the embodiment, the prediction device 300 in the culture system 1 that manufactures biodegradable polymers by culturing microorganisms using oil as a substrate, and the NIR analysis unit that analyzes the components contained in the culture solution 208, an acquisition unit (such as 200) that acquires variables related to culture by the culture system 1, and a prediction processing unit 302 that predicts the concentration of the substrate based on the analysis results of the NIR analysis unit 208 and the variables acquired by the acquisition unit. and the prediction device 300 can be realized. According to the culture system 1 of the embodiment, the oil concentration can be predicted. As a result, according to the culture system 1, it is possible to eliminate the work of extracting the culture medium and measuring the oil concentration using GC, which is required in the existing method, thereby reducing the work load. In addition, according to the culture system 1, by continuously predicting the oil concentration, the oil concentration in the culture solution can be controlled with high accuracy, and the productivity of the biodegradable polymer can be improved.

Further, according to the culture system 1 of the embodiment, the prediction model 304 is learned by using the analysis result by the NIR analysis unit 208, other explanatory variables, and the oil concentration as teacher data, and the NIR analysis unit 208 and the output of the predictive model 304 in response to inputting the variables obtained by the obtaining unit (such as 200 ) into the predictive model 304 . As a result, according to the culture system 1, the prediction accuracy can be increased by repeatedly learning the prediction model 304 using the analysis results of the NIR analysis unit 208 that are correlated with the oil concentration and other explanatory variables. Furthermore, productivity can be improved.

Furthermore, according to the culture system 1 of the embodiment, the control device in the culture system 1 that manufactures biodegradable polymers by culturing microorganisms using oil as a substrate, and the NIR analysis that analyzes the components contained in the culture solution A unit 208, an acquisition unit (such as 200) that acquires variables related to culture by the culture system 1, and a prediction device 300 that predicts the concentration of the substrate based on the analysis results of the NIR analysis unit 208 and the variables acquired by the acquisition unit. and a control unit ( 600 ) that controls the culture system 1 based on the prediction result of the prediction device 300 . As a result, according to the culture system 1, the oil concentration can be continuously predicted, the oil concentration in the culture solution can be controlled with high accuracy, and the productivity of the biodegradable polymer can be improved.

Furthermore, according to the culture system 1 of the embodiment, a production method in the culture system 1 for producing a biodegradable polymer by culturing microorganisms using oil as a substrate, comprising the step of analyzing the components contained in the culture solution; , acquiring variables related to culture by the culture system 1, and predicting concentrations related to the substrate based on the analysis results of the components and the variables. According to the culture system 1 of the embodiment, the computer in the culture system 1 that manufactures biodegradable polymers by culturing microorganisms using oil as a substrate analyzes the components contained in the culture solution; A program can be implemented that causes the steps of obtaining variables for the culture and predicting concentrations for the substrate based on the analytical results of the components and the variables. As a result, according to the culture system 1, it is possible to eliminate the work of extracting the culture medium and measuring the oil concentration using GC, which is required in the existing method, thereby reducing the work load. In addition, according to the culture system 1, by continuously predicting the oil concentration, the oil concentration in the culture solution can be controlled with high accuracy, and the productivity of the biodegradable polymer can be improved.

As described above, the mode for carrying out the present invention has been described using the embodiments, but the present invention is not limited to such embodiments at all, and various modifications and replacements can be made without departing from the scope of the present invention. can be added.

1 Culture System 100 Culture Apparatus 120 Stirrer 130 Pump System 140 Exhaust Gas Sensor 142 Dissolved Oxygen Sensor 144 Temperature Sensor 146 pH Sensor 148 NIR Sensor 200 O2/CO2 Acquisition Part 202 DO Acquisition Part 204 Temperature Acquisition Part 206 pH Acquisition Part 208 NIR Analysis Part 300 prediction device 302 prediction processing unit 304 prediction model 310 explanatory variable storage unit 312 explanatory variable setting unit 400 learning device 410 teacher data storage unit 430 model construction unit 440 learning result storage unit 500 display device 600 control device

Claims (13)

A prediction device in a culture system for producing biodegradable polymers by culturing microorganisms using oil as a substrate,
a near-infrared spectroscopic analysis unit that analyzes components contained in the culture solution;
an acquisition unit that acquires variables related to culture by the culture system;
a prediction unit that predicts the concentration of the substrate based on the analysis result by the near-infrared spectroscopic analysis unit and the variables acquired by the acquisition unit;
A prediction device comprising a The prediction unit includes a prediction model trained by using the analysis result by the near-infrared spectroscopic analysis unit, the variable, and the concentration of the substrate as teacher data, and the analysis result by the near-infrared spectroscopic analysis unit and the acquisition unit. 2. The prediction device according to claim 1, which obtains an output of concentration for said substrate of said prediction model in response to inputting a variable obtained by into said prediction model. The variables related to the culture include information on culture conditions, information on biodegradable polymers, information on microorganisms such as bacteria, yeasts, molds, and archaea that produce biodegradable polymers, information on the state of the substrate, and the culture system. 3. A prediction device according to claim 1 or 2, comprising at least one of information relating to the state of the . The culture condition information includes the substrate feeding amount, the alkaline solution feeding amount, the acid solution feeding amount, the culture time, the number of stirring, the stirring power, the internal pressure, the aeration rate, the pH of the culture solution, and the culture solution. 4. The prediction device according to claim 3, comprising at least one of the temperature of the culture solution, the amount of the culture solution, and the amount of inoculation. The prediction device according to claim 3, wherein the information on the biodegradable polymer includes at least one of information on the concentration of the biodegradable polymer and information on the primary structure of the biodegradable polymer. Information on microorganisms such as bacteria, yeast, fungi, and archaea that produce the biodegradable polymer includes strain type, particle size of the biodegradable polymer-containing bacterial cell, and biodegradable polymer-containing bacterial cell size. 4. The prediction device according to claim 3, comprising at least one of the bacterial cell weight in the dry bacterial cell weight. The information on the state of the substrate includes at least the decomposition state of the substrate, the type and composition ratio of oils and fats contained in the culture medium, the type and composition ratio of fatty acids contained in the culture medium, and the consumption rate of the substrate. 4. The prediction device of claim 3, comprising one. Information on the state of the culture system includes the amount of oxygen in the culture solution (dissolved oxygen concentration), the amount of oxygen in the exhaust gas, the amount of carbon dioxide in the exhaust gas, the oxygen consumption rate, the respiratory quotient (RQ), and the state of foaming. 4. Prediction device according to claim 3, comprising at least one of the parameters relating to The analysis results by the near-infrared spectroscopic analysis unit are the concentration analysis result of the substrate, the analysis result of the comonomer ratio of the biodegradable polymer, the average weight molecular weight of the biodegradable polymer, the biodegradable polymer 4. The prediction device according to claim 3, comprising at least one of the cell weight of microorganisms such as bacteria, yeast, fungi, and archaea that produce . A control device in a culture system for producing a biodegradable polymer by culturing microorganisms using oil as a substrate,
a near-infrared spectroscopic analysis unit that analyzes components contained in the culture solution;
an acquisition unit that acquires variables related to culture by the culture system;
a prediction unit that predicts the concentration of the substrate based on the analysis result by the near-infrared spectroscopic analysis unit and the variables acquired by the acquisition unit;
a control unit that controls the culture system based on the prediction result of the prediction unit;
A control device comprising: A production method in a culture system for producing a biodegradable polymer by culturing microorganisms using oil as a substrate,
A step of analyzing components contained in the culture medium;
obtaining variables related to culturing by the culturing system;
predicting a concentration for the substrate based on the analytical results of the component and the variables;
Manufacturing method including. In the step of supplying the substrate to the culture medium by the culture system, controlling the supply amount of the substrate based on the predicted concentration of the substrate;
The manufacturing method according to claim 11. In the computer in the culture system that manufactures biodegradable polymers by culturing microorganisms using oil as a substrate,
A step of analyzing components contained in the culture medium;
obtaining variables related to culturing by the culturing system;
predicting a concentration for the substrate based on the analytical results of the component and the variables;
The program that causes the to run.

Images