US-12626153-B2 - Learning apparatus, learning method, and failure prediction system

US12626153B2US 12626153 B2US12626153 B2US 12626153B2US-12626153-B2

Abstract

The learning apparatus according to one exemplary embodiment includes: a pattern extractor that extracts a time fluctuation pattern of an amplitude of a feature frequency from state observation signal data up to a first time point, the state observation signal data indicating an operation state of equipment, the feature frequency being associated with a part of the equipment; a training data generator that generates, based on the time fluctuation pattern of the amplitude of the feature frequency, simulated state observation signal data representing the time fluctuation pattern of the amplitude of the feature frequency at and after the first time point, and generates training data including the simulated state observation signal data; and a learner that generates a classification model for determination of a failure state of the part of the equipment using the training data.

Inventors

Naganori Shirakata
Zhiqi LIU
Tenta KOMATSU
Takayuki Tsukizawa

Assignees

PANASONIC INTELLECTUAL PROPERTY MANAGEMENT CO., LTD.

Dates

Publication Date: 20260512
Application Date: 20230525
Priority Date: 20201130

Claims (10)

1 . A learning apparatus, comprising: a pattern extractor that extracts a time fluctuation pattern of an amplitude of a feature frequency from state observation signal data up to a first time point, the state observation signal data indicating an operation state of equipment, the feature frequency being associated with a part of the equipment; a training data generator that generates, based on the time fluctuation pattern of the amplitude of the feature frequency, simulated state observation signal data representing the time fluctuation pattern of the amplitude of the feature frequency at and after the first time point, and generates training data including the simulated state observation signal data; and a learner that generates a classification model for determination of a failure state of the part of the equipment using the training data.
2 . The learning apparatus according to claim 1 , wherein the training data generator generates the training data in which the simulated state observation signal data, a frequency label representing the feature frequency, and a time point label representing an elapsed time from the first time point are combined as a set.
3 . The learning apparatus according to claim 1 , wherein the training data generator generates the simulated state observation signal data by performing extrapolation of the time fluctuation pattern of the amplitude extracted from the state observation signal data up to the first time point.
4 . The learning apparatus according to claim 1 , further comprising: a simulator that estimates the feature frequency by performing a simulation using an equipment model modeling the equipment.
5 . The learning apparatus according to claim 4 , wherein when the simulator simulates the failure state of the part of the equipment using the equipment model modeling the equipment, the training data generator generates the training data such that the training data converges to the simulated failure state of the part of the equipment.
6 . A learning method performed by a learning apparatus, the learning method comprising: extracting a time fluctuation pattern of an amplitude of a feature frequency from state observation signal data up to a first time point, the state observation signal data indicating an operation state of equipment, the feature frequency being associated with a part of the equipment; generating, based on the time fluctuation pattern of the amplitude of the feature frequency, simulated state observation signal data representing the time fluctuation pattern of the amplitude of the feature frequency at and after the first time point; generating training data including the simulated state observation signal data; and generating a classification model for determination of a failure state of the part of the equipment using the training data.
7 . A failure prediction system, comprising: a learning apparatus according to claim 1 ; and a state determiner that determines the failure state of the part of the equipment using current state observation signal data indicating a current operation state of the equipment and the classification model.
8 . The failure prediction system according to claim 7 , further comprising: a display that displays a determination result of the failure state of the part of the equipment.
9 . The failure prediction system according to claim 8 , wherein the display displays a time fluctuation of the amplitude of the feature frequency.
10 . The failure prediction system according to claim 8 , wherein: the part of the equipment comprises a plurality of the parts of the equipment and the feature frequency comprises a plurality of the feature frequencies, and the display displays determination results of failure states of the plurality of parts of the equipment.

Description

TECHNICAL FIELD The present disclosure relates to a learning apparatus, a learning method, and a failure prediction system. BACKGROUND ART A large number of motors, gears, and the like are used in industrial equipment, industrial machinery, industrial robots, and the like for performing production in factories and the like. Abnormalities in apparatuses due to aging degradation and wear degradation, as well as sudden device troubles, lead to line stoppage, and there is a concern that productivity may decrease and/or accidents may occur. Therefore, there is an increasing demand for a failure prediction system that monitors the state of equipment including these apparatuses and devices and supports efficient planned maintenance according to the state of the equipment. In connection with such a failure prediction system, Patent Literature (hereinafter, referred to as “PTL”) 1 discloses a technique of learning conditions associated with a failure of an industrial machine according to a training data set created based on a combination of a state variable including sensor data reflecting a state of the industrial machine and determination data resulting from determination of a degree of the failure of the industrial machine. By learning by supervised learning using training data (learning data) as described above, the accuracy of prediction of the failure of the equipment is improved as compared with learning by unsupervised learning that does not use training data. CITATION LIST Patent Literature PTL 1 Japanese Patent Application Laid-Open No. 2017-033526 SUMMARY OF INVENTION However, for individual motors, gears, and the like in apparatuses having entirely different operating conditions and/or entirely different configurations, it is difficult to collect, from operating equipment, training data consisting of a combination of state variables and a plurality of failure states. For example, it is practically difficult to collect a large amount of data indicating a state in which a motor, a gear, or the like is actually in an abnormal state or a failure state (hereinafter, the abnormal and failure states are collectively referred to as “failure state”). For this reason, a method for predicting the failure state by performing learning of the normal state by unsupervised learning and detecting a deviation from the normal state is often used in practice. As described above, for the learning for predicting or determining the failure state of the equipment by the supervised learning, there is room for examination in terms of the accuracy of the prediction or determination. One non-limiting and exemplary embodiment of the present disclosure facilitates providing a learning apparatus, a learning method, and a failure prediction system that easily acquire data clearly indicating a failure state used as training data in the supervised learning, and perform learning for accurately determining the failure state of equipment. A learning apparatus according to one exemplary embodiment of the present disclosure includes: a pattern extractor that extracts a time fluctuation pattern of an amplitude of a feature frequency from state observation signal data up to a first time point, the state observation signal data indicating an operation state of equipment, the feature frequency being associated with a part of the equipment; a training data generator that generates, based on the time fluctuation pattern of the amplitude of the feature frequency, simulated state observation signal data representing the time fluctuation pattern of the amplitude of the feature frequency at and after the first time point, and generates training data including the simulated state observation signal data; and a learner that generates a classification model for determination of a failure state of the part of the equipment using the training data. A learning method according to one exemplary embodiment of the present disclosure is performed by a learning apparatus and includes steps performed by the learning apparatus of: extracting a time fluctuation pattern of an amplitude of a feature frequency from state observation signal data up to a first time point, the state observation signal data indicating an operation state of equipment, the feature frequency being associated with a part of the equipment; generating, based on the time fluctuation pattern of the amplitude of the feature frequency, simulated state observation signal data representing the time fluctuation pattern of the amplitude of the feature frequency at and after the first time point; generating training data including the simulated state observation signal data; and generating a classification model for determination of a failure state of the part of the equipment using the training data. A failure prediction system according to one exemplary embodiment of the present disclosure includes: the above-described learning apparatus; and a state determiner that determines the failure state of the