Examining apparatus, examining method and recording medium
To easily perform examination of at least one facility based on detection signals of a plurality of sensors installed in the facility. Provided are an examining apparatus, an examining method and a recording medium, including: a group designation acquiring unit to acquire designation of a targeted group including a plurality of targeted sensors to be analyzed among a plurality of sensors installed in at least one facility; a sensor data acquiring unit to acquire sensor data from each targeted sensor included in the targeted group; a learning unit to learn an analysis model by using the sensor data from each targeted sensor included in the targeted group; and an examining unit to examine the facility by using the learned analysis model.
1. An examining apparatus comprising:
at least one processor;
a group designation acquiring section of the at least one processor, when executed by the at least one processor, configured to acquire designation of a targeted group including a plurality of targeted sensors to be analyzed among a plurality of sensors installed in at least one facility;
a sensor data acquiring section of the at least one processor, when executed by the at least one processor, configured to acquire sensor data from each targeted sensor included in the targeted group;
a machine learning section of the at least one processor, when executed by the at least one processor, configured to learn an analysis model by using the sensor data from each targeted sensor included in the targeted group; and
an examining section of the at least one processor, when executed by the at least one processor, configured to examine the at least one facility by using the learned analysis model, wherein
the machine learning section of the at least one processor is further configured to learn a plurality of types of analysis models by using the sensor data from each targeted sensor included in the targeted group,
the examining apparatus further comprises a model selecting section of the at least one processor, when executed by the at least one processor, configured to select at least one analysis model used for examining the at least one facility from among the learned plurality of types of analysis models,
the machine learning section of the at least one processor is further configured to learn a third analysis model among the plurality of types of analysis models by using the sensor data from all the targeted sensors included in the targeted group,
the machine learning section of the at least one processor is further configured to learn a fourth analysis model among the plurality of types of analysis models by using the sensor data from each targeted sensor, except for some targeted sensors which remain operational during the machine learning of the fourth analysis model, included in the targeted group,
the model selecting section of the at least one processor generates an evaluation result for the third analysis model and an evaluation result for the fourth analysis model using a formulated generalization coefficient for the third analysis model and for the fourth analysis model,
when the generalization coefficient for the third analysis model is higher than the generalization coefficient for the fourth analysis model, the model selecting section of the at least one processor selects the third analysis model over the fourth analysis model, and
when the generalization coefficient for the fourth analysis model is higher than the generalization coefficient for the third analysis model, the model selecting section of the at least one processor selects the fourth analysis model over the third analysis model.
2. The examining apparatus according to claim 1 , further comprising a user interface section of the at least one processor, when executed by the at least one processor, configured to input feedback on an examination result obtained by using the analysis model, wherein
the machine learning section of the at least one processor learns the analysis model by further using the feedback.
3. The examining apparatus according to claim 2 , further comprising an alarm output section of the at least one processor, when executed by the at least one processor, configured to output an alarm if the examination result obtained by using the analysis model indicates an abnormality in the at least one facility, wherein
the user interface section inputs the feedback indicating whether the alarm is correct.
4. The examining apparatus according to claim 3 , further comprising a storage section of the at least one processor, when executed by the at least one processor, configured to store labeled data obtained by labeling, with the feedback on the alarm output by the alarm output section, a portion of the sensor data from each targeted sensor included in the targeted group, on which portion the feedback is made, wherein
The machine learning section learns the analysis model by using the labeled data stored in the storage section.
5. The Examining apparatus according to claim 1 , further comprising:
an operational record acquiring section of the at least one processor, when executed by the at least one processor, configured to acquire an operational record of the at least one facility;
an identification section of the at least one processor, when executed by the at least one processor, configured to identify, based on the operational record, a period in which the machine learning of the analysis model is not to be performed; and
an exclusion processing section of the at least one processor, when executed by the at least one processor, configured to exclude, from targets for the machine learning, a piece of the sensor data from each targeted sensor included in the targeted group, which piece corresponds to the period in which the machine learning of the analysis model is not to be performed.
6. The examining apparatus according to claim 5 , wherein the identification section of the at least one processor, when executed by the at least one processor, configured to identify, from the operational record, at least one of a maintenance period of the at least one facility, a stop period of the at least one facility, a startup period of the at least one facility and a test running period of the at least one facility as the period in which the machine learning of the analysis model is not to be performed.
7. The examining apparatus according to claim 1 , wherein
the machine learning section of the at least one processor is further configured to learn a first analysis model among the plurality of types of analysis models by means of supervised machine learning using the sensor data from each targeted sensor included in the targeted group and an examination result that should be output as an examination result of the at least one facility, and
the machine learning section of the at least one processor is further configured to learn a second analysis model among the plurality of types of analysis models by the means of unsupervised machine learning using the sensor data from each targeted sensor included in the targeted group.
8. The examining apparatus according to claim 1 , wherein the model selecting section of the at least one processor is further configured to select the analysis model used for examining the at least one facility from among the learned plurality of types of analysis models according to examination accuracy.
9. The examining apparatus according to claim 1 , wherein
the group designation acquiring section of the at least one processor is further configured to acquire designation of a plurality of targeted groups,
the sensor data acquiring section of the at least one processor is further configured to acquire sensor data from each targeted sensor included in each of the plurality of targeted groups,
the machine learning section of the at least one processor is further configured to learn analysis models respectively associated with the plurality of targeted groups by using the sensor data from each targeted sensor included in each of the plurality of targeted groups, and
the examining section of the at least one processor is further configured to perform examination of the at least one facility by using each of the learned analysis models.
10. An examining apparatus comprising:
at least one processor;
a sensor data acquiring section of the at least one processor, when executed by the at least one processor, configured to acquire sensor data from each targeted sensor included in a plurality of targeted sensors to be analyzed among a plurality of sensors installed in at least one facility;
a machine learning section of the at least one processor, when executed by the at least one processor, configured to learn an analysis model by using the sensor data from each targeted sensor;
an examining section of the at least one processor, when executed by the at least one processor, configured to examine the at least one facility by using the learned analysis model; and
a user interface section of the at least one processor, when executed by the at least one processor, configured to input feedback on an examination result obtained by using the analysis model, wherein
the machine learning section of the at least one processor, when executed by the at least one processor, is further configured to learn the analysis model by further using the feedback,
the machine learning section of the at least one processor is further configured to learn a plurality of types of analysis models by using the sensor data from each targeted sensor included in the targeted group,
the examining apparatus further comprises a model selecting section of the at least one processor, when executed by the at least one processor, configured to select at least one analysis model used for examining the at least one facility from among the learned plurality of types of analysis models,
the machine learning section of the at least one processor is further configured to learn a third analysis model among the plurality of types of analysis models by using the sensor data from all the targeted sensors included in the targeted group,
the machine learning section of the at least one processor is further configured to learn a fourth analysis model among the plurality of types of analysis models by using the sensor data from each targeted sensor, except for some targeted sensors which remain operational during the machine learning of the fourth analysis model, included in the targeted group,
the model selecting section of the at least one processor generates an evaluation result for the third analysis model and an evaluation result for the fourth analysis model using a formulated generalization coefficient for the third analysis model and for the fourth analysis model,
when the generalization coefficient for the third analysis model is higher than the generalization coefficient for the fourth analysis model, the model selecting section of the at least one processor selects the third analysis model over the fourth analysis model, and
when the generalization coefficient for the fourth analysis model is higher than the generalization coefficient for the third analysis model, the model selecting section of the at least one processor selects the fourth analysis model over the third analysis model.
11. A non-transitory recording medium having recorded thereon a program that causes a computer to function as an examining apparatus comprising:
a group designation acquiring section of the computer, when executed by the computer, to acquire designation of a targeted group including a plurality of targeted sensors to be analyzed among a plurality of sensors installed in at least one facility;
a sensor data acquiring section of the computer, when executed by the computer, to acquire sensor data from each targeted sensor included in the targeted group;
a machine learning section of the computer, when executed by the computer, to learn an analysis model by using the sensor data from each targeted sensor included in the targeted group; and
an examining section of the computer, when executed by the computer, to examine the at least one facility by using the learned analysis model, wherein
the machine learning section of the computer learns a plurality of types of analysis models by using the sensor data from each targeted sensor included in the targeted group,
the examining apparatus further comprises a model selecting section of the computer, when executed by the computer, to select at least one analysis model used for examining the at least one facility from among the learned plurality of types of analysis models,
the machine learning section of the computer learns a third analysis model among the plurality of types of analysis models by using the sensor data from all the targeted sensors included in the targeted group,
the machine learning section of the computer learns a fourth analysis model among the plurality of types of analysis models by using the sensor data from each targeted sensor, except for some targeted sensors which remain operational during the machine learning of the fourth analysis model, included in the targeted group,
the model selecting section of the at least one processor generates an evaluation result for the third analysis model and an evaluation result for the fourth analysis model using a formulated generalization coefficient for the third analysis model and for the fourth analysis model,
when the generalization coefficient for the third analysis model is higher than the generalization coefficient for the fourth analysis model, the model selecting section of the at least one processor selects the third analysis model over the fourth analysis model, and
when the generalization coefficient for the fourth analysis model is higher than the generalization coefficient for the third analysis model, the model selecting section of the at least one processor selects the fourth analysis model over the third analysis model.
12. A facility monitoring method comprising:
acquiring designation of a targeted group including a plurality of targeted sensors to be analyzed among a plurality of sensors installed in at least one facility;
acquiring sensor data from each targeted sensor included in the targeted group;
machine learning at least four types of analysis models by using the sensor data from each targeted sensor included in the targeted group; and
examining the at least one facility by using the learned analysis model, wherein
the machine learning comprises learning a plurality of types of analysis models by using the sensor data from each targeted sensor included in the targeted group,
the examining comprises selecting at least one analysis model used for the monitoring of the at least one facility from among the learned plurality of types of analysis models,
the machine learning further comprises,
learning a third analysis model among the plurality of types of analysis models by using the sensor data from all the targeted sensors included in the targeted group, and
learning a fourth analysis model among the plurality of types of analysis models by using the sensor data from each targeted sensor, except for some targeted sensors which remain operational during the machine learning of the fourth analysis model, included in the targeted group,
the examining further comprises,
generating an evaluation result for the third analysis model and an evaluation result for the fourth analysis model using a formulated generalization coefficient for the third analysis model and for the fourth analysis model,
selecting the third analysis model over the fourth analysis model when the generalization coefficient for the third analysis model is higher than the generalization coefficient for the fourth analysis model, and
selecting the fourth analysis model over the third analysis model when the generalization coefficient for the fourth analysis model is higher than the generalization coefficient for the third analysis model.