Granted Patent B2

US 12,277,143 · App. 17/693,401 · Granted Apr 15, 2025

Inventor: Ryohei Fujii (Tokyo, JP)

Assignee: Yokogawa Electric Corporation

G06F16/285G06F16/258

Abstract

Claims (98)

1. An apparatus comprising:

at least one processor;

an acquisition unit that uses the at least one processor for acquiring target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added;

a conversion generation unit that uses the at least one processor for generating a conversion rule of the target state data and the source state data into data of a common space based on a distribution of the target state data and a distribution of the source state data;

a model generation unit that uses the at least one processor for generating classification model for classifying a quality of a state indicated by the target state data in the common space using data of the common space to which the label is added;

an evaluation unit that uses the at least one processor for evaluating the classification model using the source state data converted into data of the common space;

a setting unit that uses the at least one processor for setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

a determination unit that uses the at least one processor for determining a quality of a state based on a classification result that is output from the classification model in response to an input of the target state data that are newly acquired and converted into data of the common space; and

a display control unit that uses the at least one processor for displaying a determination result;

wherein the conversion generation unit uses the at least one processor to generate the conversion rule using first source state data and the target state data,

the evaluation unit uses the at least one processor to evaluate the classification model using second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein the conversion generation unit uses the at least one processor to generate the conversion rule so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting unit uses the at least one processor for setting, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the display control unit alternatively uses the at least one processor to switch and display, together with the determination result, the measurement value to be displayed among measurement values of the target state da ta newly acquired by the acquisition unit and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired by the acquisition unit is converted.

2. The apparatus according to claim 1 , wherein the acquisition unit uses the at least one processor to acquire the source state data corresponding to a good state and the source state data corresponding to a poor state in a pre-specified ratio based on a label added to the source state data.

3. The apparatus according to claim 1 , wherein the acquisition unit uses the at least one processor to further acquire state data outside a distribution range of the target state data that have been already acquired, as the target state data for generating the conversion rule.

4. The apparatus according to claim 1 , comprising a label addition unit that uses the at least one processor for adding the label to the target state data measured at a moment specified by an operator.

5. The apparatus according to claim 4 , comprising a detection unit that uses the at least one processor for detecting that the target state data to which a label is added by the label addition unit reaches a reference amount.

6. The apparatus according to claim 1 , wherein the determination unit uses the at least one processor to determine a quality of a state based on classification results that are output from a plurality of classification models that includes the classification model and is different from each other.

7. A method comprising:

acquiring target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added;

generating a conversion rule for converting the target state data and the source state data into data of a common space based on a distribution of the target state data and a distribution of the source state data;

generating a classification model for classifying a quality of a state indicated by the target state data in the common space using data of the common space to which the label is added;

evaluating the classification model using the source state data converted into data of the common space;

setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

determining a quality of a state based on a classification result that is output from the classification model in response to an input of the target state data newly acquired and converted into data of the common space; and

displaying a determination result;

wherein

generating the conversion rule includes using first source state data and the target state data,

evaluating the classification model uses second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein

the conversion rule is generated so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting includes setting, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the displaying includes alternatively switching and displaying, together with the determination result, the measurement value to be displayed among measurement values of the target state data newly acquired and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired is converted.

8. A method comprising:

converting, using a conversion rule based on a distribution of target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added, the target state data into data of a common space shared with the source state data;

newly acquiring the target state data;

determining a quality of a state based on a classification result that is output from a classification model in response to the target state data that are newly acquired and converted into data of the common space being input into a classification model for classifying a quality of a state indicated by the target state data in the common space based on data of a common space to which the label is added;

setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

controlling a display of a determination result; and

evaluating the classification model using the source state data converted into data of the common space;

wherein

the conversion rule is generated using first source state data and the target state data,

the evaluating the classification model uses second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein

the conversion rule is generated so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting includes setting, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the controlling the displaying includes alternatively controlling switching and displaying, together with the determination result, the measurement value to be displayed among measurement values of the target state data newly acquired and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired is converted.

9. A non-transitory computer readable medium having a program recorded thereon, wherein the program causes a computer to function as:

an acquisition unit for acquiring target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added;

a conversion generation unit for generating a conversion rule of the target state data and the source state data into data of a common space based on a distribution of the target state data and a distribution of the source state data;

a model generation unit for generating classification model for classifying a quality of a state indicated by the target state data in the common space using data of the common space to which the label is added;

an evaluation unit for evaluating the classification model using the source state data converted into data of the common space;

a setting unit for setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

a determination unit for determining a quality of a state based on a classification result that is output from the classification model in response to an input of the target state data that are newly acquired and converted into data of the common space; and

a display control unit for displaying a determination result;

wherein the conversion generation unit generates the conversion rule using first source state data and the target state data,

the evaluation unit evaluates the classification model using second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein

the conversion generation unit generates the conversion rule so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting unit sets, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the display control unit alternatively switches and displays, together with the determination result, the measurement value to be displayed among measurement values of the target state data newly acquired by the acquisition unit and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired by the acquisition unit is converted.

10. A non-transitory computer readable medium having a program recorded thereon, wherein the program causes a computer to function as:

a conversion execution unit for converting, using a conversion rule based on a distribution of target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added, the target state data into data of a common space shared with the source state data;

a classification model for classifying a quality of a state indicated in the target state data in the common space based on data of a common space to which the label is added;

an acquisition unit for newly acquiring the target state data;

a determination unit for determining a quality of a state based on a classification result that is output from the classification model in response to an input of the target state data that are newly acquired by the acquisition unit and is converted into data of the common space by the conversion execution unit;

a setting unit for setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

a display control unit for displaying a determination result; and

an evaluation unit for evaluating the classification model using the source state data converted into data of the common space;

wherein

the conversion rule is generated using first source state data and the target state data,

the evaluation unit uses second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein

the conversion rule is generated so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting unit sets, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the display control unit alternatively switches and displays, together with the determination result, the measurement value to be displayed among measurement values of the target state data newly acquired by the acquisition unit and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired by the acquisition unit is converted.

11. An apparatus comprising:

at least one processor;

a conversion execution unit that uses the at least one processor for converting, using a conversion rule based on a distribution of target state data that are state data in a target domain and source state data that are state data in a source domain and to which a label indicating a quality of a state is added, the target state data into data of a common space shared with the source state data;

a classification model that uses the at least one processor for classifying a quality of a state indicated in the target state data in a common space based on data of the common space to which the label is added;

an acquisition unit that uses the at least one processor for newly acquiring the target state data;

a determination unit that uses the at least one processor for determining a quality of a state based on a classification result that is output from the classification model in response to an input of the target state data that are newly acquired by the acquisition unit and is converted into data of the common space by the conversion execution unit;

a setting unit that uses the at least one processor for setting the classification model as a classification model used for classifying the target state data, in response to an evaluation of the classification model being higher than a reference;

a display control unit that uses the at least one processor for displaying a determination result; and

an evaluation unit for evaluating the classification model using the source state data converted into data of the common space;

wherein

the conversion rule is generated using first source state data and the target state data,

the evaluation unit uses second source state data that is different from the first source state data, and

the first source state data and the second source state data are included in the source state data;

wherein the conversion generation unit uses the at least one processor to generate the conversion rule so that a distribution of the target state data that have been converted is approximately equal to a distribution of the source state data that have been converted and a variance of each distribution is maximized;

wherein the setting unit uses the at least one process or for setting, as a measurement value to be displayed, at least one measurement value whose influence on a classification result of the classification model is greater than a reference among several types of measurement values included in the target state data, and setting, as a parameter to be displayed, at least one parameter whose influence on a classification result of the classification model is greater than a reference among a plurality of parameters included in data of the common space; and

wherein in response to an operation by an operator, the display control unit alternatively uses the at least one processor to switch and display, together with the determination result, the measurement value to be displayed among measurement values of the target state data newly acquired by the acquisition unit and a value of the parameter to be displayed among parameters of the common space into which the target state data newly acquired by the acquisition unit is converted.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (18)