Estimation model generation device, estimation device, estimation model generation method, estimation method, and program

Abstract

To estimate a status of a disperse system.SOLUTION: An estimation model generation method generates an estimation model which outputs status data of a disperse system to be estimated when process data of a disperse system manufacturing device which manufactures the disperse system to be estimated, raw material data of the disperse system to be estimated, and device data of the disperse system manufacturing device which manufactures the disperse system to be estimated are input, using the process data which is a feature amount in a manufacturing process in the disperse system manufacturing device, the raw material data which is a feature amount of raw materials of the disperse system manufactured by the disperse system manufacturing device, the device data which is a feature amount indicating the disperse system manufacturing device, and status data which is a feature amount of the status of the disperse system and a feature amount not including the process data, the raw material data and the device data as teacher data.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an estimation model generation device, an estimation device, an estimation model generation method, an estimation method, and a program.

Background Art

[0002] There is a method for manufacturing a dispersion system such as an emulsion. In the production of a dispersion system, it is required to acquire quality information of the dispersion system in order to improve the quality of the produced dispersion system. For example, Patent Document 1 discloses performing sequential sampling from a stirring tank to evaluate the physical properties of an emulsion. For example, Patent Document 2 discloses inserting probes into a stirring tank and measuring the physical properties of an emulsion in real time.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the methods disclosed in Patent Document 1 and Patent Document 2, it is necessary to insert sensors and probes for measuring the state and various physical properties of the emulsion into the emulsion in the stirring tank. These methods cannot be used when sensors and probes cannot be inserted into the inside of a dispersion system such as an emulsion for reasons such as contamination prevention. An object of the present invention is to provide an estimation model generation device, an estimation device, an estimation model generation method, an estimation method, and a program for estimating the state of a dispersion system.

Means for Solving the Problems

[0005] One aspect of the present invention is to generate an estimation model that outputs state data of a dispersion system to be estimated, using, as teacher data, process data that is a characteristic quantity in a manufacturing process in a dispersion system manufacturing apparatus, raw material data that is a characteristic quantity of raw materials of a dispersion system manufactured by the dispersion system manufacturing apparatus, apparatus data that is a characteristic quantity indicating the dispersion system manufacturing apparatus, and state data that is a characteristic quantity of the state of the dispersion system and does not include the process data, the raw material data, and the apparatus data. The process data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the apparatus data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated are used as input data.

[0006] One aspect of the present invention is an estimation method of estimating the state of a dispersion system to be estimated. When process data that is a characteristic quantity in a manufacturing process in a dispersion system manufacturing apparatus, raw material data that is a characteristic quantity of raw materials of a dispersion system manufactured by the dispersion system manufacturing apparatus, and apparatus data that is a characteristic quantity indicating the dispersion system manufacturing apparatus are input, state data that is a characteristic quantity of the state of the dispersion system and does not include the process data, the raw material data, and the apparatus data is output by an estimation model learned by a machine learning method. Then, the process data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the apparatus data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated are input to the estimation model.

[0007] One aspect of the present invention uses, as teacher data, analysis condition data included in data related to a dispersion system and state data that is a characteristic quantity of the state of the dispersion system and does not include data related to the dispersion system. When the first data included in the analysis condition data of the dispersion system to be estimated and the state data of the dispersion system to be estimated are used as input data and input, A method for generating an estimation model that generates second data included in the analysis condition data of the dispersion system to be estimated and not included in the first data. The data related to the dispersion system is process data that is a feature quantity in the manufacturing process in a dispersion system manufacturing apparatus, raw material data that is a feature quantity of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, and apparatus data that is a feature quantity indicating the dispersion system manufacturing apparatus. It is a method for generating an estimation model.

[0008] One aspect of the present invention is that when first data included in data related to a dispersion system and state data that is a feature quantity of the state of the dispersion system and does not include data related to the dispersion system are input, the data related to the dispersion system is included, and the second data not included in the first data is output. By inputting the first data of the dispersion system to be estimated and the state data of the dispersion system to be estimated into the learned estimation model, the second data of the dispersion system to be estimated is estimated. The data related to the dispersion system is process data that is a feature quantity in the manufacturing process in a dispersion system manufacturing apparatus, raw material data that is a feature quantity of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, and apparatus data that is a feature quantity indicating the dispersion system manufacturing apparatus. It is an estimation method.

Advantages of the Invention

[0009] According to the present invention, the state of the dispersion system can be estimated.

Brief Description of the Drawings

[0010]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Embodiments for Carrying Out the Invention

[0011] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. (First Embodiment) 《Configuration of Estimation Model Generation Device》 FIG. 1 is a diagram showing the configuration of an estimation model generation device 1 according to the first embodiment. The estimation model generation device 1 includes a teacher data acquisition unit 10, an estimation model generation unit 12, and an estimation model output unit 14.

[0012] The teacher data acquisition unit 10 acquires teacher data. The teacher data is data in which process data, raw material data, device data of a dispersion manufacturing apparatus, and state data of a dispersion manufactured by the dispersion manufacturing apparatus are associated.

[0013] The dispersion system manufacturing apparatus is an apparatus for manufacturing a substance of a dispersion system. The substance of the dispersion system is not particularly limited, and examples thereof include substances of gas-liquid, gas-solid, solid-liquid, liquid-liquid, and gas-liquid-solid three-phase dispersion systems. The substance of the dispersion system is, for example, an emulsion, a suspension, an aerosol, a bubble, and more specifically, dairy products such as whipped cream, ice cream, condensed milk, and milk coffee, oil and fat compositions such as margarine, butter, and shortening, bakery and confectionery foods such as sponge cake, biscuit, chocolate, and chewing gum, beverages such as coffee, fruit juice beverage, lactic acid bacteria beverage, soy milk beverage, and vegetable juice, seasonings such as mayonnaise, ketchup, soy sauce, sauce, and emulsified dressing, cosmetics such as facial cleanser, lotion, emulsion / cream, foundation, lipstick, eyeshadow, eyeliner, mascara, eyebrow pencil, nail enamel, shampoo, conditioner, hair wax (clay / cream type), sunscreen, and others, pharmaceuticals, electronic materials, agricultural chemicals, soap, natural rubber latex, synthetic resin, carbon material, ceramics, cement, lubricating oil, grease, adhesive, sealing material, printing ink, pigment, coloring material, fragrance, and the like.

[0014] Process data is a characteristic quantity in the manufacturing process in the dispersion system manufacturing apparatus, and includes, for example, the temperature, pressure, pH, viscosity, liquid density, liquid volume, electrical conductivity, flow rate, stirring number of the stirrer, shear rate, stirring current value, voltage value, power, power value, current value of the stirring motor, voltage value, power, power consumption, aeration rate, gas-liquid ratio, process time, and the like of the product manufactured by the manufacturing apparatus. Raw material data is a characteristic quantity of the raw materials of the dispersion system manufactured by the dispersion system manufacturing apparatus, and includes, for example, the type and amount of the raw materials used in the dispersion system, composition, as well as the temperature, pressure, pH, viscosity, liquid density, liquid volume, electrical conductivity, process time, and the like of the raw materials. Apparatus data is a characteristic quantity indicating the dispersion system manufacturing apparatus, and includes, for example, the tank diameter, tank height, type / shape of the tank bottom of the stirring tank for manufacturing the dispersion system, blade diameter of the stirring blade, type / shape of the stirring blade, 3D CAD data of the apparatus, and the like.

[0015] The state data of the dispersion system are characteristic quantities of the state of the dispersion system manufactured by the dispersion system manufacturing apparatus. The characteristic quantities of the state of the dispersion system do not include the characteristic quantities included in the process data, raw material data, and apparatus data. The state data of the dispersion system are, for example, the distribution of the size of the dispersed phase, the value indicating the degree of sedimentation of the dispersed phase, the value indicating the degree of dispersion of the dispersed phase, the value indicating the dispersion stability of the dispersed phase, the value indicating the texture of the dispersion system, the value indicating the feel and spreadability of the dispersion system on the skin, etc. When the dispersion system is an oil-in-water (O / W type) emulsion, the distribution of the size of the dispersed phase is the distribution of the size of the oil droplets. Further, the state data of the dispersion system may be data indicating the state of unit operations such as production (reaction operation, culture operation, etc.), purification and separation (filtration, membrane separation, floatation separation, etc.) in systems other than the homogeneous phase, storage stability in tanks, storage tanks, or product packages.

[0016] The estimation model generation unit 12 generates an estimation model by a machine learning method based on the teacher data. The estimation model is a model that takes the process data, raw material data, and apparatus data of the dispersion system manufacturing apparatus as input data and outputs the state of the dispersion system manufactured by the dispersion system manufacturing apparatus, which is different from the process data, raw material data, and apparatus data. Here, there are mainly two patterns in the configuration of the estimation model. First, all are estimated by a black box model. Second, it is estimated by a gray box model that combines a black box and a white box, which is a physical model based on phenomenology.

[0017] Examples of white box models include reaction rate equations and equations of motion, and equations derived from general physical equations are also included in the white box. Next, as examples of black box models, in the case of a linear regression method, there are multiple regression, logistic regression, Ridge regression, Elastic Net regression, Lasso regression, PLS regression, PCR, GPR, etc. In the case of a non-linear regression method, there are SVR, GPR and PCR using kernel tricks, ensemble learning models, neural networks, deep learning models, and other population balance equations related to splitting and coalescence aggregation, Bayesian estimation methods. The gray box model includes a model that combines the white box model and the black box model shown above.

[0018] Examples of ensemble learning models include RandomForest, XGBoost, LightGBM, etc. Instead of using a single model presented by the above black box model, multiple models may be stacked to improve the prediction accuracy.

[0019] The estimated model output unit 14 outputs the estimated model generated by the estimated model generation unit 12. The estimated model output unit 14 outputs the estimated model to, for example, the estimation device 2 described later.

[0020] FIG. 2 is a flowchart showing the operation of the estimated model generation device 1 according to the first embodiment. The teacher data acquisition unit 10 acquires teacher data (step S101). The estimated model generation unit 12 generates an estimated model (step S102). The estimated model output unit 14 outputs the estimated model (step S103).

[0021] 《Configuration of the Estimation Device》 FIG. 3 is a diagram showing the configuration of the estimation device 2 according to the first embodiment. The estimation device 2 includes a data acquisition unit 20, a storage unit 22, an estimation unit 24, and an estimation result output unit 26. The estimation device 2 estimates a state of the dispersion system to be estimated that is not included in the process data, raw material data, and device data, based on the process data of the dispersion system manufacturing device that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the device data of the dispersion system manufacturing device that manufactures the dispersion system to be estimated.

[0022] The data acquisition unit 20 acquires input data. The input data is the process data of the dispersion system manufacturing device that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the device data of the dispersion system manufacturing device that manufactures the dispersion system to be estimated. The storage unit 22 stores the estimated model. The storage unit 22 stores, for example, the estimated model output by the estimated model generation device 1.

[0023] Based on the input data, the estimation unit 24 estimates the state of the dispersion system to be estimated that is not included in the process data, raw material data, and equipment data. The estimation unit 24 inputs the input data into the estimation model stored in the storage unit 22 to output the state data of the dispersion system to be estimated, which is different from the process data, raw material data, and equipment data.

[0024] The estimation result output unit 26 outputs the estimation result by the estimation unit 24. The estimation result output unit 26 outputs the estimation result to, for example, an external storage device and accumulates the estimation result. The estimation result output unit 26 outputs the estimation result to, for example, an external display device and displays the estimation result.

[0025] FIG. 4 is a flowchart showing the operation of the estimation device 2 according to the first embodiment. The data acquisition unit 20 acquires input data (step S201). The estimation unit 24 estimates the state of the dispersion system to be estimated based on the input data (step S202). The estimation result output unit 26 outputs the estimation result by the estimation unit 24 (step S203).

[0026] The estimation model generated by the estimation model generation device 1 according to the first embodiment includes the process data, raw material data, and equipment data of the dispersion system manufacturing apparatus to be estimated as input data. Among the states of the dispersion system, the states not included in the process data, raw material data, and equipment data change depending on the process, raw materials, and equipment. Therefore, the estimation model generated by the estimation model generation device 1 according to the first embodiment can estimate the state of the dispersion system.

[0027] (Second Embodiment) In the estimation model generation device 1 according to the second embodiment, the teacher data acquired by the teacher data acquisition unit 10 includes, in addition to the process data, raw material data, equipment data, and state data of the dispersion system manufacturing apparatus, the feature amount of the non-uniformity inside the dispersion system manufacturing apparatus (hereinafter referred to as non-uniformity data). The non-uniformity data includes, for example, the flow velocity distribution of the dispersion medium of the dispersion system in the dispersion system manufacturing apparatus, the shear rate distribution in the tank, the collision frequency of the dispersed phase, the velocity difference, and the like.

[0028] The non-uniformity data may be estimated based on process data, raw material data, and equipment data. The non-uniformity data is estimated, for example, by CFD (computational fluid dynamics) simulation, DEM (Discrete Element Method) simulation, or CFD-DEM combining CFD and DEM based on process data, raw material data, and equipment data. The CFD simulation is a simulation method for the continuous phase. Based on the Euler equations or the Navier-Stokes equations, there are a lattice method for Eulerian simulation of the flow of the continuous phase and a particle method for simulating the continuous phase as virtual particles based on the Lagrangian description of the equation of motion with continuum approximation. The DEM simulation is a method for simulating a discrete body that is a physical aggregate of particles.

[0029] The CFD simulation method can use different models depending on the flow state. In the case of laminar flow, a laminar flow model is used, and in the case of turbulent flow, RANS (Reynolds-Averaged Navier-Stokes), LES (Large Eddy Simulation), DNS (Direct Numerical Simulation), etc. are used. In the case of turbulent flow, vortices are generated, and the accuracy and calculation time are determined by how far the vortices are approximated without calculation. RANS takes a time average of the vortices, LES approximates only the vortices smaller than the mesh with an interpolation model, and DNS is a method of directly calculating without any approximation. The less the approximation, the greater the computational load, and the large-scale calculation by DNS has an unrealistic computational load, so usually RANS or LES is used. In a dispersion manufacturing apparatus, the method can be selectively used depending on the state in the tank where the dispersion is manufactured and the physical quantity that is desired to be accurately determined. For example, in the case of a multiphase flow, models such as the Euler-Euler multiphase flow model, the Euler-Lagrange multiphase flow model, and VOF (Volume of Fluid) are used. Here, if the interface state is to be tracked, VOF is used; if changes such as breakup, coalescence, growth, and dissolution of the dispersed phase are to be tracked, a combination of the Euler-Euler multiphase flow model and the population balance model or the Euler-Lagrange model is used. Also, if it is desired to consider the elasticity and adhesion force of the powder during fluidization, CFD-DEM is used.

[0030] As shown in FIG. 5, the estimation model generation apparatus 1 according to the second embodiment includes a non-uniformity estimation unit 16, and the non-uniformity estimation unit 16 may estimate non-uniformity based on the process data, raw material data, and apparatus data of the dispersion manufacturing apparatus.

[0031] The estimation model generation unit 12 according to the second embodiment generates an estimation model that estimates a state not included in the process data, raw material data, apparatus data, and non-uniformity data among the states of the dispersion, using the process data, raw material data, apparatus data, and non-uniformity data of the dispersion manufacturing apparatus as input data based on the teacher data.

[0032] In the estimation apparatus 2 according to the second embodiment, the data acquisition unit 20 acquires the process data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated, the raw material data of the dispersion to be estimated, the apparatus data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated, and the non-uniformity data of the dispersion to be estimated. The estimation unit 24 according to the second embodiment inputs the process data, raw material data, apparatus data, and non-uniformity data related to the dispersion to be estimated into the estimation model stored in the storage unit 22, thereby outputting state data not included in the process data, raw material data, apparatus data, and non-uniformity data related to the dispersion to be estimated.

[0033] The non-uniformity data may be information estimated based on process data, raw material data, and equipment data related to the dispersion system to be estimated, similar to the non-uniformity data included in the teacher data. Note that the non-uniformity data acquired by the data acquisition unit 20 may be data estimated by a surrogate model. The surrogate model is a model created by replacing, by machine learning, the simulation used to estimate the non-uniformity data included in the teacher data.

[0034] As shown in FIG. 6, the estimation device 2 according to the second embodiment may include a non-uniformity estimation unit 28, and may estimate non-uniformity based on process data, raw material data, and equipment data related to the dispersion system to be estimated. The non-uniformity estimation unit 28 may estimate non-uniformity using a surrogate model.

[0035] The non-uniformity data is difficult to acquire when the dispersion system is not directly sensed. Since the estimation model generated by the estimation model generation device 1 according to the second embodiment includes non-uniformity data as an explanatory variable, it is possible to estimate the state of the dispersion system more accurately while reflecting non-uniformity.

[0036] Also, the simulation for estimating non-uniformity requires a large amount of computing resources and time. In the estimation device 2 according to the second embodiment, by using the non-uniformity data estimated using a surrogate model, non-uniformity can be estimated in less time. Thereby, the estimation device 2 can estimate the state of the dispersion system to be estimated in a form close to real time.

[0037] <Other Embodiments> As described above, one embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to the above, and various design changes and the like can be made without departing from the gist of the present invention.

[0038] The estimation model generation device 1 and the estimation device 2 may perform inverse analysis. The estimation model generation unit 12 of the estimation model generation device 1 uses the analysis condition data and the state data of the distributed system included in the data related to the distributed system (process data, raw material data, and device data) as teacher data. The estimation model generation unit 12 uses the teacher data to generate an estimation model that outputs second data when the first data of the distributed system to be estimated and the state data of the distributed system to be estimated are input as input data. The first data is the data included in the analysis condition data of the distributed system to be estimated. Here, the analysis condition data of the distributed system to be estimated is the same type of data as the analysis condition data of the distributed system in the teacher data. For example, when the analysis condition data of the distributed system in the teacher data is A and B classified as process data, C and D classified as raw material data, and E and F classified as device data, the analysis condition data of the distributed system to be estimated is A, B, C, D, E, and F of the distributed system to be estimated.

[0039] The second data is the data included in the analysis condition data and not included in the first data. For example, when the first data is A, B, C, D, and E, the second data is F. The input data and output data of the model may include different data classified into the same category. For example, the input data may be the composition information of the raw material classified as raw material data and the state data of the distributed system, and the output data may be the amount data of the raw material classified as raw material data. The output data of the model may be two or more pieces of data. For example, the first data may be A, B, C, D, and the second data may be E, F. As methods for optimizing objective variables in multiple numbers, methods such as Bayesian optimization and multi-objective optimization methods for selecting Pareto optimal solutions are used. Specifically, there are NSGA-II, NSGA-III, MOEA / D, IBEA, Epsilon-MOEA, SPEA2, GDE3, OMOPSO, SMPSO, Epsilon-NSGA-II, R-NSGA-II, U-NSGA-III, R-NSGA-III, MOEA / D, C-TAEA, etc. For the optimization of a single objective variable, there are GA, DE, PSO, Nelder-Mead method, Hooke-Jeeves pattern search method, BRKGA, etc.

[0040] The estimation device 2 acquires first data of the dispersion system to be estimated, inputs it into the estimation model, outputs second data of the dispersion system to be estimated, and estimates the feature amount indicated by the second data.

[0041] The estimation model generation device 1 and the estimation device 2 can estimate process data, raw material data, or device data for obtaining desired state data by performing inverse analysis. After acquiring the estimation result, by changing the manufacturing process of the dispersion system, the raw material of the dispersion system, or the dispersion system manufacturing device according to the estimation result, a product in a desired state can be obtained.

Explanation of Signs

[0042] 1 Estimation model generation device, 10 Teacher data acquisition unit, 12 Estimation model generation unit, 14 Estimation model output unit, 16 Non-uniformity estimation unit, 2 Estimation device, 20 Data acquisition unit, 22 Storage unit, 24 Estimation unit, 26 Estimation result output unit, 28 Non-uniformity estimation unit

Claims

1. Using, as teacher data, process data that is a characteristic quantity in the manufacturing process of a dispersion manufacturing apparatus, raw material data that is a characteristic quantity of the raw material of the dispersion manufactured by the dispersion manufacturing apparatus, apparatus data that is a characteristic quantity indicating the dispersion manufacturing apparatus, and state data that is a characteristic quantity of the state of the dispersion and does not include the process data, the raw material data, and the apparatus data, generating an estimation model that outputs the state data of the dispersion to be estimated when inputting, as input data, the process data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated, the raw material data of the dispersion to be estimated, and the apparatus data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated. Estimation model generation method.

2. The teacher data includes non-uniformity data that is a characteristic quantity of non-uniformity in the dispersion manufacturing apparatus related to the teacher data, and the input data includes non-uniformity data that is a characteristic quantity of non-uniformity in the dispersion manufacturing apparatus that manufactures the dispersion to be estimated. The estimation model generation method according to Claim 1.

3. The non-uniformity data included in the teacher data is data generated by simulating the process data, raw material data, and apparatus data included in the teacher data. The estimation model generation method according to Claim 2.

4. Inputting the process data that is a characteristic quantity in the manufacturing process of the dispersion manufacturing apparatus, the raw material data that is a characteristic quantity of the raw material of the dispersion manufactured by the dispersion manufacturing apparatus, and the apparatus data that is a characteristic quantity indicating the dispersion manufacturing apparatus into an estimation model learned to output the state data of the dispersion, which is a characteristic quantity of the state of the dispersion and does not include the process data, the raw material data, and the apparatus data, to estimate the state of the dispersion to be estimated by inputting the process data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated, the raw material data of the dispersion to be estimated, and the apparatus data of the dispersion manufacturing apparatus that manufactures the dispersion to be estimated. Estimation method.

5. The estimation model is learned to output state data when inputting process data, raw material data, apparatus data, and non-uniformity data that is a characteristic quantity of non-uniformity in the dispersion manufacturing apparatus. By inputting process data, raw material data, equipment data related to the dispersion system to be estimated, and non-uniformity data of the dispersion system to be estimated into the estimation model, the state of the dispersion system to be estimated is estimated. The estimation method according to claim 4.

6. The non-uniformity data of the dispersion system to be estimated is calculated by simulation based on the process data, raw material data, and equipment data related to the dispersion system to be estimated. The estimation method according to claim 5.

7. Using process data, which is a characteristic quantity in the manufacturing process of a dispersion system manufacturing apparatus, raw material data, which is a characteristic quantity of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, equipment data, which is a characteristic quantity indicating the dispersion system manufacturing apparatus, and state data, which is a characteristic quantity of the state of the dispersion system and does not include the process data, the raw material data, and the equipment data, as teacher data, when inputting the process data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the equipment data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, generating an estimation model that outputs the state data of the dispersion system to be estimated. Estimation model generation device.

8. When inputting process data, which is a characteristic quantity in the manufacturing process of a dispersion system manufacturing apparatus, raw material data, which is a characteristic quantity of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, and equipment data, which is a characteristic quantity indicating the dispersion system manufacturing apparatus, into an estimation model learned by a machine learning method so as to output state data, which is a characteristic quantity of the state of the dispersion system and does not include the process data, the raw material data, and the equipment data, by inputting the process data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, the raw material data of the dispersion system to be estimated, and the equipment data of the dispersion system manufacturing apparatus that manufactures the dispersion system to be estimated, the state of the dispersion system to be estimated is estimated. Estimation device.

9. A program for causing a computer to execute the estimation model generation method according to claim 1.

10. A program for causing a computer to execute the estimation method according to claim 4.

11. Using the analysis condition data included in the data related to the dispersion system and the state data, which is a characteristic quantity of the state of the dispersion system and does not include the data related to the dispersion system, as teacher data. When the first data included in the analysis condition data of the dispersion system to be estimated and the state data of the dispersion system to be estimated are input as input data, Generate an estimation model that outputs second data included in the analysis condition data of the dispersion system to be estimated and not included in the first data. An estimation model generation method, The data related to the dispersion system is process data that is a feature amount in the manufacturing process in the dispersion system manufacturing apparatus, raw material data that is a feature amount of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, and apparatus data that is a feature amount indicating the dispersion system manufacturing apparatus. Estimation model generation method.

12. When the first data included in the data related to the dispersion system and the state data that is a feature amount of the state of the dispersion system and does not include the data related to the dispersion system are input, To an estimation model learned to output second data included in the data related to the dispersion system and not included in the first data, By inputting the first data of the dispersion system to be estimated and the state data of the dispersion system to be estimated, estimate the second data of the dispersion system to be estimated. An estimation method, The data related to the dispersion system is process data that is a feature amount in the manufacturing process in the dispersion system manufacturing apparatus, raw material data that is a feature amount of the raw material of the dispersion system manufactured by the dispersion system manufacturing apparatus, and apparatus data that is a feature amount indicating the dispersion system manufacturing apparatus. Estimation method.

Images