MOTOR CONTROL APPARATUS HAVING PROTECTION OPERATION UNIT, AND MACHINE LEARNING APPARATUS AND METHOD THEREOF

• US 20170033726A1
• Filed: 02/12/2016
• Published: 02/02/2017
• Est. Priority Date: 07/31/2015
• Status: Active Grant

• Current Assignee: Fanuc Corporation
• Original Assignee: Fanuc Corporation
• Inventors: YAMAMOTO, Kenta
• Granted Patent: US 9,787,245 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/006 based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
• CPC Class Codes: H02H 1/0092 concerning the data processing means, e.g. expert systems, neural networks
• CPC Class Codes: H02H 3/00 Emergency protective circuit arrangements for automatic disconnection directly responsive to an undesired change from normal electric working condition with or without subsequent reconnection specially adapted for specific types of electric machines or apparatus or for sectionalised protection of cable or line systems H02H7/00; systems for change-over to standby supply H02J9/00 ; integrated protection for motors H02H7/0822
• CPC Class Codes: H02H 7/122 for inverters, i.e. dc/ac converters
• CPC Class Codes: H02P 27/06 using dc to ac converters or inverters H02P27/05 takes precedence
• CPC Class Codes: H02P 29/028 the motor continuing operation despite the fault condition, e.g. eliminating, compensating for or remedying the fault
• CPC Class Codes: H02P 29/032 Preventing damage to the motor, e.g. setting individual current limits for different drive conditions

Abstract

A machine learning apparatus learns conditions associated with power failure on the side of an AC power supply in a motor control apparatus which converts AC power into DC power, outputs the DC power to a DC link, further converts the DC power into AC power for driving a motor, and supplies the AC power to the motor, includes a state observation unit which observes a state variable including at least one of data associated with the value of a power supply voltage on the AC power supply side, data associated with the amount of energy stored in a DC link capacitor provided in the DC link, and data indicating whether a protective operation for the motor control apparatus is successful, and a learning unit which learns conditions associated with power failure on the AC power supply side in accordance with a training data set defined by the state variable.

Claims

1-8. -8. (canceled)

9. A machine learning apparatus for learning (i) a power failure detection level, which is a value of a power supply voltage defined as a criterion for determination as to whether power failure has occurred on an AC power supply side of a motor control apparatus, and (ii) a power failure detection time, which is a time for which the power supply voltage on the AC power supply side of the motor control apparatus has been lower than the power failure detection level, in the motor control apparatus which converts AC power supplied from the AC power supply side into DC power, outputs the DC power to a DC link, further converts the DC power into AC power for driving a motor, and supplies the AC power to the motor,

the machine learning apparatus comprising:

a state observation unit which observes a state variable including data associated with a value of a power supply voltage on the side of the AC power supply, data associated with an amount of energy stored in a DC link capacitor provided in the DC link, data indicating whether a protective operation for the motor control apparatus is successful, and data associated with motor output; and

a learning unit which learns the power failure detection level and the power failure detection time, by repeating update of a function for changing the power failure detection level and the power failure detection time, based on the state variable.

10. The machine learning apparatus according to claim 9, wherein the learning unit comprises:

a reward computation unit which computes a reward for the result of changing the power failure detection level and the power failure detection time, based on the state variable; and

a function update unit which updates, based on the reward, the function, wherein

the learning unit learns the power failure detection level and the power failure detection time which achieve the greatest reward, by repeating update of the function by the function update unit.

11. The machine learning apparatus according to claim 10, wherein the reward computation unit increases the reward when the amount of energy stored in the DC link capacitor and observed by the state observation unit is smaller than a predetermined value, and the reward computation unit reduces the reward when the amount of energy stored in the DC link capacitor and observed by the state observation unit is not less than the predetermined value.

12. The machine learning apparatus according to claim 10, wherein the reward computation unit reduces the reward when failure of the protective operation for the motor control apparatus is observed by the state observation unit.

13. The machine learning apparatus according to claim 9, wherein the learning unit is configured to learn the power failure detection level and the power failure detection time, based on state variables obtained for a plurality of motor control apparatuses.

14. A motor control apparatus comprising the machine learning apparatus according to claim 9, the motor control apparatus further comprising:

a decision-making unit which decides the power failure detection level and the power failure detection time in response to input of a current state variable of the state variable, based on a result of learning by the learning unit;

a rectifier which converts AC power supplied from the AC power supply into DC power and outputs the DC power to the DC link;

an inverter which is connected to the DC link, converts the DC power in the DC link into AC power, and supplies the AC power to the motor;

a power failure detection unit which determines that power failure has occurred on the side of the AC power supply when the power supply voltage on the side of the AC power supply has been lower than the power failure detection level for the power failure detection time; and

a protective operation control unit which outputs a protective operation command to the inverter to output power for performing a predetermined protective operation by the motor, when the power failure detection unit detects power failure on the side of the AC power supply.

15. The motor control apparatus according to claim 14, wherein the learning unit is configured to re-learn and update the power failure detection level and the power failure detection time, based on the state variable.

16. A machine learning method for learning (i) a power failure detection level, which is a value of a power supply voltage defined as a criterion for determination as to whether power failure has occurred on an AC power supply side of a motor control apparatus, and (ii) a power failure detection time, which is a time for which the power supply voltage on the AC power supply side of the motor control apparatus has been lower than the power failure detection level, in the motor control apparatus which converts AC power supplied from the AC power supply side into DC power, outputs the DC power to a DC link, further converts the DC power into AC power for driving a motor, and supplies the AC power to the motor,

the method comprising:

a state observation step of observing a state variable including data associated with a value of a power supply voltage on the side of the AC power supply, data associated with an amount of energy stored in a DC link capacitor provided in the DC link, data indicating whether a protective operation for the motor control apparatus is successful, and data associated with motor output; and

a learning step of learning the power failure detection level and the power failure detection time, by repeating update of a function for changing the power failure detection level and the power failure detection time, based on the state variable.