A controller for a machine tool includes a machine learning apparatus configured to learn an Nth-order time-derivative component of a speed of each axis of the machine tool. The machine learning apparatus includes: a state observation section configured to observe first state data representing the Nth-order time-derivative component of the speed of each axis as a state variable representing a current state of an environment; a determination data acquisition section configured to acquire determination data representing a properness determination result of at least any one of machining accuracy, surface quality, and machining time of the machined workpiece; and a learning section configured to learn the Nth-order time-derivative component of the speed of each axis in relation to at least any one of the machining accuracy, the surface quality, and the machining time of the machined workpiece using the state variable and the determination data.

A machine tool includes an operation evaluation section that evaluates an operation thereof and a machine learning device that performs the machine learning of a movement amount of an axis thereof. The machine learning device calculates a reward based on state data including the output of the operation evaluation section, performs the machine learning of the determination of the movement amount of the axis, and determines the movement amount of the axis based on a machine learning result and outputs the determined movement amount. The machine learning device performs the machine learning of the determination of the movement amount of the axis based on the determined movement amount of the axis, the acquired state data, and the calculated reward.

A positioning control system is provided with a state-feedback control system with a state observer as a full-closed control system for driving and controlling a motor so that a load shaft, which is an output shaft of a strain wave gearing, is positioned at a target position on the basis of a load shaft position actually detected. The state observer estimates a motor shaft position and a motor velocity based on a control input for the motor and the load shaft position. The state-feedback control system feeds back the state quantity of the object of control using the load shaft position as well as estimated motor shaft position and estimated motor velocity obtained by the state observer. It is possible to suppress resonant vibration caused by angular transmission error in the strain wave gearing and perform highly accurate positioning.

A controller for a machine tool includes a machine learning apparatus configured to learn an Nth-order time-derivative component of a speed of each axis of the machine tool. The machine learning apparatus includes: a state observation section configured to observe first state data representing the Nth-order time-derivative component of the speed of each axis as a state variable representing a current state of an environment; a determination data acquisition section configured to acquire determination data representing a properness determination result of at least any one of machining accuracy, surface quality, and machining time of the machined workpiece; and a learning section configured to learn the Nth-order time-derivative component of the speed of each axis in relation to at least any one of the machining accuracy, the surface quality, and the machining time of the machined workpiece using the state variable and the determination data.

A positioning control system is provided with a state-feedback control system with a state observer as a full-closed control system for driving and controlling a motor so that a load shaft, which is an output shaft of a strain wave gearing, is positioned at a target position on the basis of a load shaft position actually detected. The state observer estimates a motor shaft position and a motor velocity based on a control input for the motor and the load shaft position. The state-feedback control system feeds back the state quantity of the object of control using the load shaft position as well as estimated motor shaft position and estimated motor velocity obtained by the state observer. It is possible to suppress resonant vibration caused by angular transmission error in the strain wave gearing and perform highly accurate positioning.

A machine tool includes an operation evaluation section that evaluates an operation thereof and a machine learning device that performs the machine learning of a movement amount of an axis thereof. The machine learning device calculates a reward based on state data including the output of the operation evaluation section, performs the machine learning of the determination of the movement amount of the axis, and determines the movement amount of the axis based on a machine learning result and outputs the determined movement amount. The machine learning device performs the machine learning of the determination of the movement amount of the axis based on the determined movement amount of the axis, the acquired state data, and the calculated reward.

A machine tool includes an operation evaluation section that evaluates an operation thereof and a machine learning device that performs the machine learning of a movement amount of an axis thereof. The machine learning device calculates a reward based on state data of the machine tool including output data from the operation evaluation section, performs the machine learning of the determination of the movement amount of the axis, and determines the movement amount of the axis based on a machine learning result and outputs the determined movement amount. The machine learning device performs the machine learning of the determination of the movement amount of the axis based on the determined movement amount of the axis, the acquired state data, and the calculated reward.