Controller and machine learning device

A numerical controller which, during machining performed by controlling, on the basis of a machining program, a machine having a spindle including a tool and an axis that drives the spindle, executes PID control for controlling a movement rate of the axis so that a spindle load of the spindle becomes constant, the numerical controller comprising

• a machine learning device that has learned a parameter of the PID control in relation to a machining condition and a machining environment of the machining for suppressing depletion of the life of the tool within a range where the cycle time of the machining performed does not greatly increase, wherein
• the machine learning device includes:
• a state observation unit that observes, as state variables representing a current state of an environment, PID control parameter data indicating the parameter of the PID control during the machining, machining condition data indicating the machining condition of the machining, and machining environment data relating to the machining environment of the machining;
• a learning unit that has learned the machining condition and the machining environment of the machining, and correlation of the parameter of the PID control for suppressing depletion of the life of the tool within a range where the cycle time of the machining performed does not greatly increase in association with each other;
• and
• an estimation result output unit that outputs the parameter of the PID control on the basis of the state variables observed by the state observation unit and a learning result generated by the learning unit,
• wherein the learning unit includes:
• a reward calculation unit that determines a reward relating to the appropriateness determination result;
• and
• a value function updating unit that uses the reward to update a function representing a value of the parameter of the PID control in relation to the machining condition and the machining environment of the machining, and the reward calculation unit issues a steadily higher award as depletion of the life of the tool decreases and the cycle time of the machining shortens.

A machine learning device which, when PID control is executed during machining performed by controlling, on the basis of a machining program, a machine having a spindle including a tool and an axis that drives the spindle, the PID control being executed to control a movement rate of the axis so that a spindle load of the spindle becomes constant, learns a parameter of the PID control in relation to a machining condition and a machining environment of the machining, the machine learning device comprising:

• a state observation unit that observes, as state variables representing a current state of an environment, PID control parameter data indicating the parameter of the PID control during the machining, machining condition data indicating the machining condition of the machining, and machining environment data relating to the machining environment of the machining;
• a determination data acquisition unit that acquires, as determination data, tool life determination data indicating an appropriateness determination result relating to depletion of a life of the tool during the machining, and cycle time determination data indicating an appropriateness determination result relating to a cycle time of the machining;
• and
• a learning unit that uses the state variables and the determination data to learn the machining condition and the machining environment of the machining, and the parameter of the PID control in association with each other,
• wherein the learning unit includes:
• a reward calculation unit that determines a reward relating to the appropriateness determination result;
• and
• a value function updating unit that uses the reward to update a function representing a value of the parameter of the PID control in relation to the machining condition and the machining environment of the machining, and the reward calculation unit issues a steadily higher award as depletion of the life of the tool decreases and the cycle time of the machining shortens.

A machine learning device which, when PID control is executed during machining performed by controlling, on the basis of a machining program, a machine having a spindle including a tool and an axis that drives the spindle, the PID control being executed to control a movement rate of the axis so that a spindle load of the spindle becomes constant, has learned a parameter of the PID control in relation to a machining condition and a machining environment of the machining for suppressing depletion of the life of the tool within a range where the cycle time of the machining performed does not greatly increase, the machine learning device comprising:

• a state observation unit that observes, as state variables representing a current state of an environment, PID control parameter data indicating the parameter of the PID control during the machining, machining condition data indicating the machining condition of the machining, and machining environment data relating to the machining environment of the machining;
• a learning unit that has learned the machining condition and the machining environment of the machining, and correlation of the parameter of the PID control for suppressing depletion of the life of the tool within a range where the cycle time of the machining performed does not greatly increase in association with each other;
• and
• an estimation result output unit that outputs the parameter of the PID control on the basis of the state variables observed by the state observation unit and a learning result generated by the learning unit,
• wherein the learning unit includes:
• a reward calculation unit that determines a reward relating to the appropriateness determination result;
• and
• a value function updating unit that uses the reward to update a function representing a value of the parameter of the PID control in relation to the machining condition and the machining environment of the machining, and the reward calculation unit issues a steadily higher award as depletion of the life of the tool decreases and the cycle time of the machining shortens.