MACHINE LEARNING APPARATUS AND METHOD LEARNING PREDICTED LIFE OF POWER DEVICE, AND LIFE PREDICTION APPARATUS AND MOTOR DRIVING APPARATUS INCLUDING MACHINE LEARNING APPARATUS

• US 20170117841A1
• Filed: 10/14/2016
• Published: 04/27/2017
• Est. Priority Date: 10/26/2015
• Status: Active Grant

• Current Assignee: Fanuc Corporation
• Original Assignee: Fanuc Corporation
• Inventors: Matsumoto, Yasuyuki, Watanabe, Masato
• Granted Patent: US 10,090,798 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 5/045 Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
• CPC Class Codes: H02P 27/08 with pulse width modulation
• CPC Class Codes: H02P 29/68 based on the temperature of a drive component or a semiconductor component
• CPC Class Codes: H02P 6/14 Electronic commutators

Abstract

A machine learning apparatus that learns predicted life of a power device of a motor driving apparatus converting DC power into AC power by a switching operation of the power device to supply the AC power to an AC motor includes: a state observation unit that observes a state variable constituted by data regarding the number of times of switching of the power device, data regarding junction temperature of the power device, and data regarding presence or absence of a failure of the power device; and a learning unit that learns the predicted life of the power device in accordance with a training data set defined by the state variable.

Claims

1. A machine learning apparatus that learns predicted life of a power device of a motor driving apparatus converting DC power into AC power by a switching operation of the power device to supply the AC power to an AC motor, the machine learning apparatus comprising:

a state observation unit that observes a state variable constituted by data regarding the number of times of switching of the power device, data regarding junction temperature of the power device, and data regarding presence or absence of a failure of the power device; and

a learning unit that learns the predicted life of the power device in accordance with a training data set defined by the state variable.

2. The machine learning apparatus according to claim 1, wherein the junction temperature of the power device is calculated on the basis of temperature of a fin cooling the power device, ambient air temperature of the motor driving apparatus, a value of current supplied to the AC motor from the motor driving apparatus, and a value of voltage applied to the AC motor from the motor driving apparatus.

3. The machine learning apparatus according to claim 1, wherein the learning unit comprises:

a reward computation unit that computes a reward on the basis of the number of times of switching of the power device, and the presence or absence of the failure of the power device; and

a function update unit that updates on the basis of the state variable and the reward, functions for computing the predicted life of the power device.

4. The machine learning apparatus according to claim 3, wherein the reward computation unit decreases the reward when the number of times of switching of the power device exceeds a specified number of times.

5. The machine learning apparatus according to claim 3, wherein the reward computation unit decreases the reward when a failure of the power device occurs under a state in which the number of times of switching of the power device does not exceed the specified number of times.

6. The machine learning apparatus according to claim 3, wherein the function update unit updates the functions for computing the predicted life of the power device on the basis of the state variable and the reward in accordance with a neural network model.

7. The machine learning apparatus according to claim 1, wherein the learning unit comprises:

an error computation unit that computes an error on the basis of the state variable; and

a learning model update unit that updates a learning model for computing the predicted life of the power device, on the basis of the state variable and the error.

8. The machine learning apparatus according to claim 1, wherein the learning unit is configured to learn the predicted life of the power device in accordance with the training data set acquired for a plurality of motor driving apparatuses.

9. A life prediction apparatus for a power device in a motor driving apparatus, the life prediction apparatus comprising the machine learning apparatus according to claim 1 wherein the life prediction apparatus further comprises

a decision-making unit that computes the predicted life of the power device, on the basis of a result of learning by the learning unit in accordance with the training data set, in response to an input of the present state variable.

10. The life prediction apparatus according to claim 9, further comprising

a notification unit that notifies an operator of the predicted life computed by the decision-making unit.

11. The life prediction apparatus according to claim 9, further comprising

a notification part that notifies an operator of the information for prompting replacement of the power device or replacement of the motor driving apparatus on the basis of the predicted life computed by the decision-making unit.

12. The life prediction apparatus according to claim 9, wherein

the learning unit is configured to re-learn and update the predicted life of the power device in accordance with an additional training data set defined by the present state variable.

13. A motor driving apparatus comprising:

the life prediction apparatus according to claim 9;

a power supply unit that converts DC power into AC power by a switching operation of the power device to supply the AC power to an AC motor;

a number of times of switching acquisition unit that acquires data regarding the number of times of switching on the basis of a switching command to the power device; and

a failure determination unit that acquires data regarding the presence or absence of the failure of the power device on the basis of the switching command to the power device and the AC power that is output from the power supply unit.

14. The motor driving apparatus according to claim 13, further comprising

a temperature calculation unit that calculates junction temperature of the power device on the basis of the temperature of the fin cooling the power device, the ambient air temperature of the motor driving apparatus, the value of current supplied to the AC motor from the motor driving apparatus, and the value of voltage applied to the AC motor from the motor driving apparatus.

15. The motor driving apparatus according to claim 13, further comprising

a temperature sensor measuring junction temperature of the power device.

16. A machine learning method that learns predicted life of a power device of a motor driving apparatus and converts DC power into AC power by a switching operation of the power device to supply the AC power to an AC motor, the machine learning method comprising:

a state observation step observing a state variable constituted by data regarding the number of times of switching of the power device, data regarding junction temperature of the power device, and data regarding presence or absence of a failure of the power device; and

a learning step learning the predicted life of the power device in accordance with a training data set defined by the state variable.