Analysis method, analysis device, and analysis program for analyzing chemical reaction

Abstract

To improve cost performance for analyzing a chemical reaction.SOLUTION: An analysis method for analyzing a chemical reaction includes: condition setting steps S1, S2 of setting a reaction condition for the chemical reaction; a model setting step S3 of setting a prediction model M for outputting a result of the chemical reaction according to input of reaction information on the chemical reaction; an execution step S4-1 of executing the chemical reaction according to the reaction condition; an information acquisition step S5-1 of acquiring the reaction information measured in the chemical reaction in the execution step S4-1; a condition update step S8 of updating the reaction condition on the basis of a result of execution of the chemical reaction; and a training step S9 of training the prediction model M on the basis of the measured reaction information and the result of execution of the chemical reaction.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a technique for analyzing chemical reactions. [Background technology]

[0002] Conventionally, a technique for analyzing a chemical reaction using an artificial intelligence model has been proposed. For example, Patent Document 1 discloses a technique for predicting the physical property value of a product using a trained model that has been trained on synthesis information of a chemical reaction and synthesis result information of the product. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7219374 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in order to create data for learning a model, it is necessary to carry out a large number of chemical reactions, which is time-consuming. In addition, analytical methods for the reaction process must be introduced and verified individually, and it takes time to establish the reaction process. Therefore, there is a problem in that the cost performance of analyzing chemical reactions is low.

[0005] The present invention has been made in consideration of the above problems, and an object of the present invention is to improve the cost performance for analyzing chemical reactions. [Means for solving the problem]

[0006] In order to solve the above problems, the present invention includes the following aspects. Section 1. An analysis method for analyzing a chemical reaction, comprising: A condition setting step of setting reaction conditions for the chemical reaction; a model setting step of setting a prediction model for outputting a result of the chemical reaction in response to input of reaction information of the chemical reaction; carrying out said chemical reaction according to said reaction conditions; an information acquiring step of acquiring reaction information measured in the chemical reaction in the carrying out step; A condition update step of updating the reaction conditions based on the results of the chemical reaction; a training step of training the prediction model based on the measured reaction information and the results of the chemical reaction; The analysis method comprises: Section 2. The analysis method described in item 1, wherein the implementation step, the condition update step, and the training step are repeated while updating the reaction conditions. Section 3. 3. The analysis method according to item 1 or 2, wherein the reaction information includes optical information measured during the chemical reaction. Section 4. An analysis device for analyzing a chemical reaction, comprising: A condition setting unit for setting reaction conditions for the chemical reaction; a model setting unit that sets a prediction model that outputs a result of the chemical reaction in response to input of reaction information of the chemical reaction; an information acquiring unit that acquires reaction information measured in the chemical reaction carried out under the reaction conditions; a condition update unit that updates the reaction conditions based on a result of the chemical reaction; A training unit that trains the prediction model based on the measured reaction information and the results of the chemical reaction; An analysis device comprising: Section 5. An analysis program for analyzing a chemical reaction, a condition setting unit for setting reaction conditions for the chemical reaction; a model setting unit that sets a prediction model that outputs a result of the chemical reaction in response to input of reaction information of the chemical reaction; an information acquiring unit for acquiring reaction information measured in the chemical reaction carried out under the reaction conditions; a condition update unit that updates the reaction conditions based on the results of the chemical reaction; a training unit that trains the prediction model based on the measured reaction information and the results of the chemical reaction; An analysis program that runs a computer as a Effect of the Invention

[0007] According to the present invention, it is possible to improve the cost performance for analyzing chemical reactions. [Brief description of the drawings]

[0008] [Figure 1] 1 is a block diagram showing a configuration of an analysis system according to an embodiment of the present invention. [Diagram 2] 4 is a flowchart showing a processing procedure of an analysis method according to the present embodiment. [Diagram 3] 13 is a sub-flowchart showing a detailed processing procedure of a condition updating step. [Figure 4] 13 is a sub-flowchart showing detailed processing procedures of a training step. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. Note that the present invention is not limited to the following embodiment, and various modifications are possible without departing from the spirit of the present invention.

[0010] (System Configuration) 1 is a block diagram showing a configuration of an analysis system 100 according to one embodiment of the present invention. The analysis system 100 includes an analysis device 1, a reaction device 2, an optical sensor 3, and an analysis device 4.

[0011] The analysis device 1 is a device for analyzing chemical reactions, and in particular has a function of simultaneously optimizing reaction conditions of the chemical reactions and creating a prediction model for predicting the results of the chemical reactions from optical information of the substances. In this embodiment, the chemical reactions are synthetic reactions, but are not limited thereto.

[0012] The analysis device 1 can be configured as a general-purpose computer, and includes, as its hardware configuration, input devices such as a mouse and a keyboard, an output device such as a display, a processor such as a CPU or a GPU, a main storage device such as a DRAM or an SRAM (not shown), and an auxiliary storage device 10 such as an HDD or an SSD. The auxiliary storage device 10 stores an analysis program P and the like.

[0013] The analysis device 1 includes, as functional blocks, a condition setting unit 11, a model setting unit 12, an information acquisition unit 13, a training unit 14, a data recording unit 15, a comparison unit 16, and a condition update unit 17. These functional blocks are realized by the processor of the analysis device 1 reading out an analysis program P into a main storage device and executing it. The analysis program P may be downloaded to the analysis device 1 via a communication network such as the Internet, or may be recorded in a computer-readable non-transitory recording medium such as a CD-ROM or an SD card and installed in the analysis device 1 via the recording medium. The functions of each functional block will be described later.

[0014] The reactor 2 is a device for carrying out a chemical reaction such as a synthesis reaction, etc. The reactor 2 carries out a chemical reaction according to set reaction conditions (solvent, raw material, temperature, flow rate, pressure, reaction time, etc.).

[0015] The optical sensor 3 is a device that acquires optical information of a substance undergoing a chemical reaction, and in this embodiment, is configured as a spectroscopic sensor. The optical sensor 3 irradiates light onto the substance undergoing a chemical reaction, detects the intensity of each wavelength or wave number of the light guided by the spectroscopic means, and measures spectroscopic data.

[0016] The analysis device 4 is a device that analyzes products produced by chemical reactions, and is configured by, for example, gas chromatography or liquid chromatography. In this embodiment, the analysis device 4 samples the target product and impurities produced by the synthesis reaction, and analyzes the yield and purity of the target product.

[0017] (Processing Procedure) FIG. 2 is a flowchart showing the processing procedure of the analysis method according to this embodiment.

[0018] In steps S1 and S2 (condition setting steps), the condition setting unit 11 of the analysis device 1 sets reaction conditions for the chemical reaction. In this embodiment, the condition setting unit 11 sets an objective variable for the synthesis reaction in step S1, and sets an explanatory variable for the chemical reaction in step S2. The objective variable is a target characteristic value and its tolerance. In this embodiment, the target characteristic value is the yield of the target product by the chemical reaction, but is not limited to this, and may be the purity of the target product or the concentration of impurities. The explanatory variable is a control factor for the chemical reaction and its tolerance, and the control factor is a solvent, raw material, temperature, flow rate, pressure, reaction time, etc. The control factor can be designed by, for example, experimental design.

[0019] In step S3 (model setting step), the model setting unit 12 sets a prediction model (quantitative evaluation model, measurement model) M. The prediction model M is a model that outputs the result of a chemical reaction in response to the input of reaction information of the chemical reaction. In this embodiment, the prediction model M is a multiple regression model, but is not limited to this, and may be, for example, a least squares regression model or a principal component regression model. The reaction information includes the temperature, pressure, and flow rate of the reactants of the chemical reaction, as well as optical information measured by the optical sensor 3. The result of the chemical reaction is the yield of the target product, but may also be, for example, the concentration of the target product or the concentration of impurities. The model setting unit 12 acquires initial data by performing an offline simulation or offline experiment according to the target variable set by the user, and sets the prediction model M using the initial data.

[0020] Next, steps S4-1 to S4-3 and steps S5-1 to S5-3 are carried out in parallel.

[0021] In steps S4-1 to S4-3, a chemical reaction is carried out according to the set reaction conditions, and the results of the chemical reaction are acquired and recorded. In this embodiment, in step S4-1 (carrying out step), the reaction device 2 carries out a synthesis reaction experiment according to the explanatory variables set in step S2.

[0022] In parallel with step S4-1, in step S5-1 (information acquisition step), the information acquisition unit 13 acquires reaction information measured in the chemical reaction in step S4-1. In this embodiment, various sensors provided in the reaction device 2 measure the temperature, pressure, flow rate, and implementation time (explanatory variables) in the synthesis reaction, and the optical sensor 3 measures optical information (spectroscopic data). The measured reaction information is transmitted to the analysis device 1 as measurement data D2 and acquired by the information acquisition unit 13.

[0023] In step S4-2, the analysis device 4 analyzes the product produced by the chemical reaction. In this embodiment, the analysis device 4 analyzes the target product and impurities at the end point of the synthesis reaction, and calculates the yield of the target product as analysis data D1. The analysis data D1 is transmitted to the analysis device 1 and acquired by the information acquisition unit 13.

[0024] In step S5-2, the training unit 14 applies the data to the prediction model M. In this embodiment, the training unit 14 inputs the measurement data D2 (temperature, pressure, flow rate, and spectroscopic data) measured in step S5-1 to the prediction model M. The prediction model M outputs a response variable (yield of the target product) as prediction data D3.

[0025] In step S4-3, the data recording unit 15 records in the auxiliary storage device 10 the analysis data D1 acquired in step S4-2.

[0026] In step S5-3, the data recording unit 15 records in the auxiliary storage device 10 the measurement data D2 acquired in step S5-1 and the prediction data D3 output from the prediction model M in step S5-2.

[0027] In step S6, the comparison unit 16 compares the analysis value of the analysis data D1 with the objective variable set in step S1. In this embodiment, the comparison unit 16 compares the yield of the objective. Feedback is applied to optimize the explanatory variables and the measurement model until the analysis value of the analysis data D1 becomes greater than the objective variable (YES in step S6).

[0028] If the analysis value of the analysis data D1 is less than or equal to the objective variable (NO in step S6), the process proceeds to step S7. In step S7, the data recording unit 15 reads out the analysis data D1, the measurement data D2, and the prediction data D3 recorded in steps S4-3 and S5-3 from the auxiliary storage device 10.

[0029] In step S8 (condition update step), the condition setting unit 11 updates the reaction conditions for the chemical reaction. The details of step S8 will be described later.

[0030] In step S9 (training step), the training unit 14 trains the prediction model M based on the reaction information measured in the chemical reaction and the results of the chemical reaction. In this embodiment, the training unit 14 trains the prediction model M based on the measurement data D2 acquired in step S5-1 and the analysis data D1 acquired in step S4-2. The details of step S9 will be described later.

[0031] 3 is a sub-flowchart showing the detailed processing procedure of step S8. Step S8 is made up of steps S8-1 to S8-3, and these steps are carried out by the condition setting unit 11.

[0032] In step S8-1, a residual analysis with respect to the objective variable is performed. Specifically, the difference between the target characteristic value (yield) set in step S2 and the analysis data D1 acquired in step S4-2 is analyzed.

[0033] In step S8-2, the contribution of the explanatory variables is analyzed. Specifically, the contribution of the explanatory variables (control factors) set in step S2 to the multiple regression model (prediction model M) is analyzed by, for example, distance extraction by principal component analysis using the analysis values ​​of the analysis data D1, the measurement values ​​of the measurement data D2, statistics converted from the measurement values ​​according to the purpose, and the prediction values ​​of the prediction data D3.

[0034] In step S8-3, important explanatory variables are estimated. Specifically, in step S8-2, explanatory variables that are critical to the reaction conditions (for example, temperature and its tolerance) are estimated from explanatory variables with high contributions.

[0035] As a result of the above, in step S8, the explanatory variables for the synthesis reaction are updated.

[0036] 4 is a sub-flowchart showing the detailed processing procedure of step S9. Step S9 is made up of steps S9-1 to S9-4, and these steps are carried out by the training unit .

[0037] In step S9-1, a residual analysis with respect to the objective variable is performed. Specifically, the difference between the target characteristic value (yield) set in step S2 and the prediction data D3 output from the prediction model M in step S5-2 is analyzed.

[0038] In step S9-2, explanatory variables of the prediction model M are selected. For example, important analysis regions (for example, important wavenumber ranges of the spectrum) are extracted from the optical information (in-line data) input to the prediction model M (multiple regression model).

[0039] In step S9-3, pre-processing parameters are selected, for example, a signal pre-processing method for the in-line data (for example, a filtering method for the spectral data) and parameters (parameters for the filtering method).

[0040] In step S9-4, a multiple regression model is reconstructed based on the difference analyzed in step S9-1, the explanatory variables selected in step S9-2, and the pre-processing parameters selected in step S9-3.

[0041] As a result, in step S9, the prediction model M is trained.

[0042] Then, returning to step S2, the explanatory variables updated in step S8 among the explanatory variables previously set are weighted while considering the predetermined constraint conditions, and the explanatory variables are reset (the experimental plan is updated). In step S3, the prediction model M trained in step S9 is reset. Then, steps S4-1 to S4-3 and steps S5-1 to S5-3 are performed. Then, feedback (steps S7 to S9, S2, S3, steps S4-1 to S4-3, and steps S5-1 to S5-3) is repeated until the analysis value (yield) of the analysis data D1 becomes larger than the objective variable set in step S1 (YES in step S6). As a result, the reaction conditions and the prediction model M are optimized in parallel.

[0043] In step S10, the optimized reaction conditions (important explanatory variables and their ranges), objective variables (characteristic data (evaluation data)), and prediction model M (optimized analysis range, preprocessing parameters, model formula, evaluation results) are output (displayed).

[0044] (Summary) As described above, in this embodiment, the optimization of reaction conditions can be performed in parallel with the optimization (training) of the prediction model M. That is, the added value of the optimization of reaction conditions is generated in the training process of the prediction model M. Therefore, the development cost of the prediction model M can be reduced, and the cost performance for analyzing chemical reactions can be improved.

[0045] In addition, in the process of optimizing the reaction conditions, various reaction conditions and results of chemical reactions are obtained, and by using these to train the prediction model M, a robust prediction model M can be constructed.

[0046] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Forms obtained by appropriately combining the technical means disclosed in the embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0047] 1 Analysis device 10 Auxiliary storage 11 Condition setting section 12 Model setting section 13 Information acquisition department 14 Training Department 15 Data recording section 16 Comparison section 17 Condition update section 2. Reactor 3. Optical Sensor 4 Analyzer 100 Analysis System D1 Analysis Data D2 Measurement data D3 Prediction Data M Prediction Model P Analysis Program

Claims

1. An analysis method for analyzing a chemical reaction, comprising: A condition setting step of setting reaction conditions for the chemical reaction; a model setting step of setting a prediction model for outputting a result of the chemical reaction in response to input of reaction information of the chemical reaction; carrying out said chemical reaction according to said reaction conditions; an information acquiring step of acquiring reaction information measured in the chemical reaction in the carrying out step; A condition update step of updating the reaction conditions based on the results of the chemical reaction; a training step of training the prediction model based on the measured reaction information and the results of the chemical reaction; The analysis method comprises:

2. The analysis method according to claim 1 , wherein the implementing step, the condition updating step, and the training step are repeated while updating the reaction conditions.

3. The analysis method according to claim 1 , wherein the reaction information includes optical information measured during the chemical reaction.

4. An analysis device for analyzing a chemical reaction, comprising: A condition setting unit for setting reaction conditions for the chemical reaction; a model setting unit that sets a prediction model that outputs a result of the chemical reaction in response to input of reaction information of the chemical reaction; an information acquiring unit that acquires reaction information measured in the chemical reaction carried out under the reaction conditions; a condition update unit that updates the reaction conditions based on a result of the chemical reaction; A training unit that trains the prediction model based on the measured reaction information and the results of the chemical reaction; An analysis device comprising:

5. An analysis program for analyzing a chemical reaction, a condition setting unit for setting reaction conditions for the chemical reaction; a model setting unit that sets a prediction model that outputs a result of the chemical reaction in response to input of reaction information of the chemical reaction; an information acquiring unit for acquiring reaction information measured in the chemical reaction carried out under the reaction conditions; a condition update unit that updates the reaction conditions based on the results of the chemical reaction; a training unit that trains the prediction model based on the measured reaction information and the results of the chemical reaction; An analysis program that runs a computer as a

Images