Granted Patent B2

US 12,682,986 · App. 18/556,092 · Granted Jul 14, 2026

Inventor: Kyohei Hanaoka (Tokyo, JP)

G16C20/30G16C20/50G16C20/70

Abstract

Claims (49)

1 . A property prediction system predicting properties of a material based on a plurality of raw materials having a known structure, the system comprising

at least one processor,

wherein the at least one processor is configured to:

acquire partial structure data indicating a partial structure from a database;

receive input comprising:

raw material structure data for specifying the structure of each of the raw materials; and

blending ratio data indicating a ratio of blending of each of the raw materials;

generate partial structure input data indicating the partial structure existing in the structure of each of the raw materials, on the basis of the partial structure data and the raw material structure data of each of the raw materials, by converting the partial structure data to a vector using molecular description, wherein the vector corresponds to the partial structure input data of each of the raw materials;

generate input data by:

multiplying each of elements of the vector by the blending ratio data corresponding to the partial structure input data, and

adding or averaging elements of each of the vectors for each of the raw materials;

input the input data to a machine learning model; and

cause the machine learning model to output predicting result of properties of the material.

2 . The property prediction system according to claim 1 ,

wherein the at least one processor is configured to:

specify a number of the partial structure existing in the structure of the raw material; and

generate the input data by reflecting a value obtained by multiplying the blending ratio data and the number of the partial structure together on the partial structure input data of the raw material.

3 . The property prediction system according to claim 1 ,

wherein the partial structure input data is molecular structure information indicating a structure of the partial structure.

4 . The property prediction system according to claim 1 ,

wherein the at least one processor

generates the input data by further reflecting a value indicating a difference in the raw materials on a plurality of data pieces, which are partial structure input data for each of the plurality of raw materials, and compiling the data pieces on one data piece.

5 . The property prediction system according to claim 1 ,

wherein the at least one processor

acquires a plurality of types of the partial structure data pieces from the database;

generates a plurality of types of the input data pieces by using the plurality of types of partial structure data pieces; and

inputs the plurality of types of input data pieces to a plurality of machine learning models to build an ensemble learning machine.

6 . A property prediction method for predicting properties of a material based on a plurality of raw materials having a known structure, the method being executed by a computer including at least one processor, the method comprising:

acquiring partial structure data indicating a partial structure from a database;

receiving input comprising:

raw material structure data for specifying the structure of each of the raw materials; and

blending ratio data indicating a ratio of blending of each of the raw materials;

generating partial structure input data indicating the partial structure existing in the structure of each of the raw materials, on the basis of the partial structure data and the raw material structure data of each of the raw materials, by converting the partial structure data to a vector using molecular description, wherein the vector corresponds to the partial structure input data of each of the raw materials;

generating input data by:

multiplying each of elements of the vector by the blending ratio data corresponding to the partial structure input data, and

adding or averaging elements of each of the vectors for each of the raw materials;

inputting the input data to a machine learning model; and

causing the machine learning model to output predicting result of properties of the material.

7 . A non-transitory computer-readable storage medium storing a property prediction program for predicting properties of a material based on a plurality of raw materials having a known structure, the program allowing a computer to execute:

acquiring partial structure data indicating a partial structure from a database;

receiving input comprising:

raw material structure data for specifying the structure of each of the raw materials; and

blending ratio data indicating a ratio of blending of each of the raw materials;

generating partial structure input data indicating the partial structure existing in the structure of each of the raw materials, on the basis of the partial structure data and the raw material structure data of each of the raw materials, by converting the partial structure data to a vector using molecular description, wherein the vector corresponds to the partial structure input data of each of the raw materials;

generating input data by:

multiplying each of elements of the vector by the blending ratio data corresponding to the partial structure input data, and

adding or averaging elements of each of the vectors for each of the raw materials;

inputting the input data to a machine learning model; and

causing the machine learning model to output predicting result of properties of the material.

Assignments (2)

Priority Claims (1)

Continuity (1)

References Cited (18)