Machine learning device, motor control system, and machine learning method for learning cleaning interval of fan motor

1. A machine learning device which learns a cleaning frequency of a fan motor which cools an electric motor mounted in a machine when the fan motor is cleaned, the cleaning frequency of the fan motor being related to a life of the electric motor and an operating ratio of the machine, comprising:

a state observation unit which observes a state variable representing a current, a voltage, a temperature and a slip of the electric motor, a cleaning interval of the fan motor, and the operating ratio of the machine; and

a learning unit which updates an action value table for cleaning the fan motor, on the basis of the state variable observed by the state observation unit, wherein

the cleaning interval of the fan motor is determined on the basis of the action value table.

2. The machine learning device according to claim 1 , wherein the state observation unit

receives a first state variable representing the current, the voltage, the temperature, and the slip of the electric motor from the electric motor as a feedback signal to an electric motor control unit which controls the electric motor, and

receives a second state variable representing the cleaning interval of the fan motor and the operating ratio of the machine from the electric motor control unit.

3. The machine learning device according to claim 1 , wherein

the state observation unit further observes a state variable representing an ambient temperature of the electric motor, and

the learning unit updates the action value table for cleaning the fan motor, in consideration of the state variable representing the ambient temperature of the electric motor observed by the state observation unit, as well.

4. The machine learning device according to claim 3 , wherein the state observation unit receives a third state variable representing an ambient temperature of the electric motor from a temperature measuring unit located near the electric motor.

5. The machine learning device according to claim 1 , wherein the learning unit comprises:

a reward computation unit which computes a reward on the basis of the state variable observed by the state observation unit; and

a value function update unit which updates a function for determining an amount of change in the cleaning interval of the fan motor from a current state variable, on the basis of the reward computed by the reward computation unit.

6. The machine learning device according to claim 5 , further comprising: a decision unit which decides whether the fan motor is to be cleaned from the current state variable, on the basis of a result of learning by the learning unit.

7. The machine learning device according to claim 6 , wherein the reward computation unit

provides a smaller reward on the basis of prolongation of the cleaning interval of the fan motor, a reduction in a difference between the observed temperature of the electric motor and a predicted temperature of the electric motor, or a reduction in the operating ratio of the machine, and

provides a larger reward on the basis of shortening of the cleaning interval of the fan motor, an increase in the difference between the observed temperature of the electric motor and the predicted temperature of the electric motor, or an increase in the operating ratio of the machine.

8. The machine learning device according to claim 6 , further comprising: an alarm output unit which outputs an alarm on the basis of output from the decision unit when the fan motor is cleaned.

9. The machine learning device according to claim 1 , wherein the machine learning device further comprises a neural network.

10. A motor control system comprising a plurality of machine learning devices according to claim 5 , the system further comprising:

a data exchange unit which exchanges data between the plurality of machine learning devices, wherein a function updated by a value function update unit of one machine learning device among the plurality of machine learning devices is used to update functions of value function update units of the remaining machine learning devices.

11. The motor control system according to claim 10 , wherein the motor control system comprises a plurality of electric motors and a plurality of fan motors, and the machine learning device is provided in correspondence with each of the fan motors to share or exchange the state variables learned by the machine learning devices, via a communication medium.

12. A machine learning method for learning a cleaning frequency of a fan motor which cools an electric motor mounted in a machine when the fan motor is cleaned, the cleaning frequency of the fan motor being related to a life of the electric motor and an operating ratio of the machine, the method comprising:

observing a state variable representing a current, a voltage, a temperature and a slip of the electric motor, a cleaning interval of the fan motor, and the operating ratio of the machine;

updating an action value table for cleaning the fan motor, on the basis of an observed state variable; and

determining the cleaning interval of the fan motor on the basis of the action value table.

13. The machine learning method according to claim 12 , wherein

observing the state of the machine further comprises observing a state variable representing an ambient temperature of the electric motor, and

updating the action value table for cleaning the fan motor comprises updating the action value table for cleaning the fan motor, in consideration of the observed state variable representing the ambient temperature of the electric motor, as well.