Machine learning device, control system, and machine learning method

A machine learning device for calculating an optimized virtual temperature calculation equation including a first coefficient for determining a heat generation amount and a second coefficient for determining a heat dissipation amount in order to estimate a temperature of a specific portion of a machine based on heat generation factor data of the machine having a thermally expandable machine element, the machine learning device comprising:

• a memory configured to store a program;
• and
• a processor configured to execute the program and control the machine learning device to:
• acquire actual temperature data within a predetermined period from at least one temperature sensor mounted to the specific portion of the machine;
• acquire, from the machine, heat generation factor data including any of a driving current of a motor, a rotation speed of a spindle of the machine, load of the motor, temperature of the motor, and an atmospheric temperature within the predetermined period;
• randomly set the first coefficient for determining the heat generation amount and the second coefficient for determining the heat dissipation amount of the specific portion of the machine and
• based on an error between the virtual temperature data of the specific portion of the machine obtained by substituting the heat generation factor in the virtual temperature calculation equation and the actual temperature data acquired from the at least one temperature sensor mounted to the specific portion,
• calculate a virtual temperature calculation equation optimized by machine learning to search for the first coefficient and the second coefficient.

A machine learning method executed by a machine learning device including a memory configured to store a program and a processor for calculating an optimized virtual temperature calculation equation including a first coefficient for determining a heat generation amount and a second coefficient for determining a heat dissipation amount in order to estimate a temperature of a specific portion of a machine based on heat generation factor data of the machine having a thermally expandable machine element, the processor executing the program to perform operations of the machine learning method comprising:

• acquiring actual temperature data within a predetermined period from at least one temperature sensor mounted to a portion other than the specific portion of the machine;
• acquiring, from the machine, heat generation factor data including any of a driving current of a motor, a rotation speed of a spindle of the machine, load of the motor, temperature of the motor, and an atmospheric temperature within the predetermined period;
• randomly setting the first coefficient for determining the heat generation amount and the second coefficient for determining the heat dissipation amount of the specific portion of the machine, and using the heat generation factor data to acquire virtual temperature data within the predetermined period of the specific portion of the machine calculated by the virtual temperature calculation equation including the first coefficient and the second coefficient;
• acquiring an actually measured value of a thermal displacement amount within the predetermined period of the specific portion of the machine;
• defining the actual temperature data acquired, the heat generation factor data acquired, and the virtual temperature data acquired as input data, defining the actually measured value of the thermal displacement amount acquired as a label, and storing the input data and the label as training data in association with each other;
• calculating a thermal displacement estimation amount calculation equation optimized by performing machine learning and including a third coefficient based on an error between an estimated value of a thermal displacement amount of the specific portion of the machine calculated by substituting the input data stored as the training data in the thermal displacement estimation amount calculation equation including the third coefficient, and the actually measured value of the thermal displacement amount of the specific portion of the machine stored as the label;
• calculating a virtual temperature calculation equation optimized by performing machine learning and including the first coefficient and the second coefficient to search for the first coefficient and the second coefficient based on an error between the thermal displacement estimation value of the specific portion calculated by the actual temperature data and the virtual temperature data obtained by the virtual temperature calculation equation, and the actually measured thermal displacement of the specific portion;
• and
• controlling the machine learning method based on the calculated virtual temperature calculation equation.

A machine learning method executed by a machine learning device including a memory configured to store a program and a processor for calculating an optimized virtual temperature calculation equation including a first coefficient for determining a heat generation amount and a second coefficient for determining a heat dissipation amount in order to estimate a temperature of a specific portion of a machine based on heat generation factor data of the machine having a thermally expandable machine element, the processor executing the program to perform operations of the machine learning method comprising:

• acquiring actual temperature data within a predetermined period from at least one temperature sensor mounted to the specific portion of the machine;
• acquiring, from the machine, heat generation factor data including any of a driving current of a motor, a rotation speed of a spindle of the machine, load of the motor, temperature of the motor, and an atmospheric temperature within the predetermined period;
• randomly setting the first coefficient for determining the heat generation amount and the second coefficient for determining the heat dissipation amount of the specific portion of the machine;
• calculating a virtual temperature calculation equation optimized by machine learning to search for the first coefficient and the second coefficient based on an error between the virtual temperature data of the specific portion of the machine obtained by substituting the heat generation factor in the virtual temperature calculation equation and the actual temperature data acquired from the at least one temperature sensor mounted to the specific portion;
• and
• controlling the machine learning method based on the calculated virtual temperature calculation equation.