Machine learning method and machine learning apparatus learning operating command to electric motor and machine tool including machine learning apparatus

1. A machine learning apparatus for learning an operating command to an electric motor in a machine tool,

the machine learning apparatus comprising a processor configured to:

observe an amount of power fed to the electric motor, a temperature of the electric motor, a load of the electric motor, and a vibration of the electric motor as status variables, and

perform learning whether or not it is necessary to correct the operating command to the electric motor according to

the status variables, and

a result of judging whether an abnormality has occurred in the machine tool based on the status variables,

wherein, in said learning, the processor is further configured to

calculate a reward based on

a difference of the amount of power fed and a predetermined amount of power fed,

a difference of the temperature and a predetermined temperature,

a difference of the load and a predetermined load, and

a difference of the vibration and a predetermined vibration,

wherein the processor is configured to calculate the reward by

increasing the reward when said result of judging indicates that no abnormality has occurred in the machine tool based on the status variables, and

decreasing the reward when said result of judging indicates that an abnormality has occurred in the machine tool based on the status variables, and

update a function for determining the operating command based on the reward, and

wherein the electric motor is controlled by the operating command determined by the updated function.

2. The machine learning apparatus according to claim 1 , wherein the processor is configured to hold values of all of pairs of the status variables and the operating command as an action value table and update the action value table based on the reward.

3. The machine learning apparatus according to claim 1 , wherein the processor is configured to learn the operating command in accordance with a set of training data prepared for a plurality of electric motors.

4. A machine tool, comprising:

a drive shaft or rotating shaft driven by at least one first electric motor;

a spindle driven by at least one second electric motor; and

a machine learning apparatus comprising a processor configured to:

observe an amount of power fed to the electric motors, a temperature of the electric motors, a load of the electric motors, and a vibration of the electric motors as status variables,

perform learning whether or not it is necessary to correct an operating command to the electric motors according to

the status variables, and

a result of judging whether an abnormality has occurred in the machine tool based on the status variables, and

perform at least one of feed speed correction, spindle speed correction, acceleration/deceleration time constant correction, and position correction feed based on results of said learning,

wherein, in said learning, the processor is further configured to

calculate a reward based on

a difference of the amount of power fed and a predetermined amount of power fed,

a difference of the temperature and a predetermined temperature,

a difference of the load and a predetermined load, and

a difference of the vibration and a predetermined vibration,

wherein the processor is configured to calculate the reward by

increasing the reward when said result of judging indicates that no abnormality has occurred in the machine tool based on the status variables, and

decreasing the reward when said result of judging indicates that an abnormality has occurred in the machine tool based on the status variables, and

update a function for determining the operating command based on the reward, and

wherein the machine tool is configured to control the first electric motor and the second electric motor in accordance with the operating command determined by the updated function.

5. The machine tool according to claim 4 , wherein

the processor is further configured to

observe a temperature and a vibration of a spindle head as additional status variables, and

calculate the reward based on

a difference of the temperature of the spindle head and a predetermined temperature, and

a difference of the vibration of the spindle head and a predetermined vibration.

6. A machine learning method learning an operating command to an electric motor in a machine tool, the machine learning method comprising

observing an amount of power fed to the electric motor, a temperature of the electric motor, a load of the electric motor, and a vibration of the electric motor as status variables; and

learning whether or not it is necessary to correct the operating command to the electric motor according to

the status variables, and

a result of judging whether an abnormality has occurred in the machine tool based on the status variables,

wherein said learning comprises:

calculating a reward based on

a difference of the amount of power fed and a predetermined amount of power fed,

a difference of the temperature and a predetermined temperature,

a difference of the load and a predetermined load, and

a difference of the vibration and a predetermined vibration,

wherein said calculating the reward comprises

increasing the reward when said result of judging indicates that no abnormality has occurred in the machine tool based on the status variables, and

decreasing the reward when said result of judging indicates that an abnormality has occurred in the machine tool based on the status variables, and

updating a function for determining the operating command based on the reward, and

wherein the method further comprises controlling the electric motor by the operating command determined by the updated function.