Search device, operation method of search device, operation program of search device, and flow reaction equipment

1 . Flow reaction equipment comprising:

a search device;

a reaction section inside which is a flow path, configured so that through the flow path a mixture of a first raw material and a second raw material flows, and a flow reaction is performed;

a recovery section that is connected to the reaction section and is configured to recover a product from the reaction section;

a system controller that is configured to control the reaction section under a reaction condition as a condition parameter that is set, the condition parameter including items of concentration and flow speed of the first raw material, concentration and flow speed of the second raw material, shape of a merging portion, reaction path diameter, reaction path length, and reaction temperature; and

a detection unit that is configured to detect a reaction result as a result parameter in the reaction section, and is configured to output the detected reaction result to the search device via the system controller, the result parameter including yield, property of a product, molecular weight dispersion, purity of the product, molecular weight, concentration of the product in a solution, or yield or purity of a by-product,

wherein

the search device comprises:

a prediction data set generation unit comprising a learning unit, a trained model storage unit, and a search unit and configured to generate a trained model by performing machine learning using a known data set as supervised data based on a learning model by the learning unit, wherein the trained model is a function representing a relationship between the condition parameter as an explanatory variable and the result parameter as a response variable, wherein the trained model storage unit is configured to store the learning model generated by the learning unit, and wherein the search unit, by executing a search algorithm, is configured to: generate, based on the known data set, a prediction data set composed of a plurality of prediction data in which for each explanatory variable a value of the response variable is unknown, and associate each explanatory variable in the prediction data with a respective prediction value of the response variable based on the trained model stored in the trained model storage unit;

a first actual measurement value acquisition unit configured to acquire a first actual measurement value, DE, of the response variable for a value of the explanatory variable by registering the value of the explanatory variable included in the prediction data closest to a target value, DT, into the system controller, and operating the flow reaction device with said value of the explanatory variable;

an improvement rate calculation unit configured to calculate an improvement rate using a mathematical expression R=(DK−DE)×100/DK, R=(DE−DK)×100/DK or R=|DK−DE|×100/DK, wherein R represents the improvement rate, DK represents a reference value which is a known value of the response variable within the known data set and is closest to the target value, DT, and DE represents the first actual measurement value, and wherein the mathematical expression R=(DK−DE)×100/DK is used when the actual measurement value DE approaches the target value DT as the value DE becomes smaller, the mathematical expression R=(DE−DK)×100/DK is used when the response variable is a parameter that approaches the target value DT as the value becomes larger, such as the yield in flow synthesis, and the mathematical expression R=|DK−DE|×100/DK is used when a parameter, such as the molecular weight, which preferably converges to a specific value is used as the response variable;

a known data set update unit configured to add the first actual measurement value and a value of the explanatory variable corresponding to the first actual measurement value to the known data set as known data in a case in which the improvement rate is equal to or higher than a target improvement rate; and

a second actual measurement value acquisition unit configured to acquire, in a case where the improvement rate is lower than the target improvement rate, a second actual measurement value, DE, of the response variable for a value of the explanatory variable included in the prediction data, wherein said value of the explanatory variable is not used for acquiring the first actual measurement value by the first actual measurement value acquisition unit and deviates from the value of explanatory variable used for acquiring the first actual measurement value, DE, by a value equal to or more than a threshold value, and wherein the second actual measurement value is used for recalculating the improvement rate,

wherein the system controller is configured to operate the search device using the reaction condition as the explanatory variable and the reaction result as the response variable, and is configured to set the value of the explanatory variable corresponding to the first actual measurement value or the second actual measurement value in the reaction section as the reaction condition in a case in which said actual measurement value achieves the target value, and

wherein the flow reaction equipment obtains a product by reaction processing.

2 . The flow reaction equipment according to claim 1 , further comprising:

a prediction data set specifying unit configured to specify the prediction data set generated by the prediction data set generation unit with reference to data of a past improvement rate accumulated in a storage unit, and based on the known data set updated by the known data set update unit immediately after obtaining a highest improvement rate in the past, in a case in which the improvement rate is lower than the target improvement rate,

wherein the second actual measurement value acquisition unit is configured to acquire the second actual measurement value of the response variable for the value of the explanatory variable included in the prediction data, which is not used for acquiring the first actual measurement value by the first actual measurement value acquisition unit, in the prediction data set specified by the prediction data set specifying unit,

the improvement rate calculation unit is configured to calculate the improvement rate based on the known data set updated by the known data set update unit, and the second actual measurement value acquired by the second actual measurement value acquisition unit, and

the known data set update unit is configured to add, as the known data, the second actual measurement value acquired by the second actual measurement value acquisition unit, and the value of the explanatory variable corresponding to the second actual measurement value to the known data set in a case in which the improvement rate is equal to or higher than the target improvement rate.

3 . The flow reaction equipment according to claim 2 ,

wherein the second actual measurement value acquisition unit is configured to acquire the second actual measurement value based on the prediction data including a value of the explanatory variable deviating from the value of the explanatory variable included in the prediction data used by the first actual measurement value acquisition unit by a value equal to or more than a threshold value, in the prediction data set specified by the prediction data set specifying unit.

4 . The flow reaction equipment according to claim 3 ,

wherein the second actual measurement value acquisition unit is configured to acquire the second actual measurement value by selecting the prediction data in which the prediction value is closest to the target value in a case in which a plurality of the prediction data including the value of the explanatory variable deviating from the value of the explanatory variable included in the prediction data used by the first actual measurement value acquisition unit by the value equal to or more than the threshold value are present.

5 . The flow reaction equipment according to claim 1 , further comprising:

a target improvement rate change unit that is configured to change the target improvement rate in a case in which the known data set update unit updates the known data set.

6 . The flow reaction equipment according to claim 2 ,

wherein the search algorithm is a genetic algorithm.

7 . An operation method of flow reaction equipment, the method comprising:

a prediction data set generation step, by executing a search algorithm, of: generating, based on a known data set, a prediction data set composed of a plurality of prediction data in which for each explanatory variable a value of a response variable is unknown and associating each explanatory variable in the prediction data with a respective prediction value of the response variable based on a trained model stored in a trained model storage unit;

a first actual measurement value acquisition step of acquiring a first actual measurement value, DE, of the response variable for a value of the explanatory variable by registering the value of the explanatory variable included in the prediction data closest to a target value, DT, into a system controller, and operating a flow reaction device with said value of the explanatory variable;

an improvement rate calculation step of calculating an improvement rate using a mathematical expression R=(DK−DE)×100/DK, R=(DE−DK)×100/DK or R=|DK−DE|×100/DK, wherein R represents the improvement rate, DK represents a reference value which is a known value of the response variable within the known data set and is closest to the target value, DT, and DE represents the first actual measurement value, and wherein the mathematical expression R=(DK−DE)×100/DK is used when the actual measurement value DE approaches the target value DT as the value DE becomes smaller, the mathematical expression R=(DE−DK)×100/DK is used when the response variable is a parameter that approaches the target value DT as the value becomes larger, such as yield in flow synthesis, and the mathematical expression R=|DK−DE|×100/DK is used when a parameter, such as molecular weight, which preferably converges to a specific value is used as the response variable;

a known data set update step of adding the first actual measurement value and a value of the explanatory variable corresponding to the first actual measurement value to the known data set as the known data in a case in which the improvement rate is equal to or higher than a target improvement rate; and

a second actual measurement value acquisition step of acquiring, in a case where the improvement rate is lower than the target improvement rate, a second actual measurement value, DE, of the response variable for a value of the explanatory variable included in the prediction data, wherein said value of the explanatory variable is not used for acquiring the first actual measurement value in the first actual measurement value acquisition step and deviates from the value of explanatory variable used for acquiring the first actual measurement value, DE, by a value equal to or more than a threshold value, and wherein the second actual measurement value is used for recalculating the improvement rate, and

operating the search device using a reaction condition as the explanatory variable and a reaction result as the response variable, and sets the value of the explanatory variable corresponding to the first actual measurement value or the second actual measurement value in a reaction section as the reaction condition in a case in which said actual measurement value achieves the target value.

8 . A non-transitory computer readable recording medium storing instructions to cause a computer to perform the steps of the method according to claim 7 .

9 . An operation method of the flow reaction equipment according to claim 1 .