Information processing device, information processing method, computer-readable recording medium, and model generation method

An information processing method comprising:

• inputting, to a physical model, raw material information based on a raw material of a product produced at a plant;
• acquiring a first output result output from the physical model in response to the inputting of the raw material information;
• inputting, to a machine learning model, the first output result and component information based on optical spectrum data obtained by spectroscopic sensing for the raw material;
• acquiring a second output result output from the machine learning model in response to the inputting of first output result and the component information;
• outputting plant state information concerning a state of the plant based on the first output result and the second output result;
• acquiring the first output result by inputting, to the physical model, a first component of the optical spectrum data as the raw material information, and
• acquiring the second output result by inputting, to the machine learning model, a second component of the optical spectrum data as the component information, wherein the second component is provided by eliminating the first component from the optical spectrum data.

A non-transitory computer-readable recording medium having stored therein information processing instructions that cause a computer to perform a process comprising:

• inputting, to a physical model, raw material information based on a raw material of a product produced at a plant;
• acquiring a first output result output from the physical model in response to the inputting of the raw material information;
• inputting, to a machine learning model, the first output result and component information based on optical spectrum data obtained by spectroscopic sensing for the raw material;
• acquiring a second output result output from the machine learning model in response to the inputting of the first output result and the component information;
• outputting plant state information concerning a state of the plant based on the first output result and the second output result;
• acquiring the first output result by inputting, to the physical model, a first component of the optical spectrum data as the raw material information, and
• acquiring the second output result by inputting, to the machine learning model, a second component of the optical spectrum data as the component information, wherein the second component is provided by eliminating the first component from the optical spectrum data.

An information processing device comprising:

• a processor configured to generate:
• a physical model that predicts a state of a plant by a simulation, based on raw material information based on a raw material of a product produced at the plant, and
• a machine learning model that predicts an error of an output of the physical model from component information based on optical spectrum data obtained by spectroscopic sensing for the raw material and the output of the physical model, by machine learning that uses the component information and the output of the physical model as training data, wherein
• the raw material information is a first component of the optical spectrum data, and
• the component information is a second component of the optical spectrum data, wherein the second component is provided by eliminating the first component from the optical spectrum data.

A model generation method comprising:

• generating a physical model that predicts a state of a plant by a simulation, based on raw material information based on a raw material of a product produced at the plant;
• and
• generating a machine learning model that predicts an error of an output of the physical model from component information based on optical spectrum data obtained by spectroscopic sensing for the raw material and the output of the physical model, by machine learning that uses the component information and the output of the physical model as training data, wherein
• the raw material information is a first component of the optical spectrum data, and
• the component information is a second component of the optical spectrum data, wherein the second component is provided by eliminating the first component from the optical spectrum data.

A non-transitory computer-readable recording medium having stored therein model generation instructions that cause a computer to perform a process comprising:

• generating a physical model that predicts a state of a plant by a simulation, based on raw material information based on a raw material of a product produced at the plant;
• and
• generating a machine learning model that predicts an error of an output of the physical model from component information based on optical spectrum data obtained by spectroscopic sensing for the raw material and the output of the physical model, by machine learning that uses the component information and the output of the physical model as training data, wherein
• the raw material information is a first component of the optical spectrum data, and
• the component information is a second component of the optical spectrum data, wherein the second component is provided by eliminating the first component from the optical spectrum data.