Machine learning device and associated methodology for adjusting parameters used to numerically control a machine tool

A numerical control device, comprising:

• a machine learning device, including:
• a processor;
• and
• a memory to store a program which, when executed by the processor, causes the processor to perform a process including:
• reading a plurality of machining programs for numerically controlling a machine tool;
• extracting, from parameters of the plurality of machining programs, a plurality of first parameters to be adjusted and
• extracting, from the parameters of the plurality of machining programs, a second parameter for the plurality of first parameters, the second parameter being a parameter used to adjust corresponding one of the plurality of first parameters, the second parameter being fixed and including at least a material property and shape of a material of a workpiece,
• generating a data set including the plurality of first parameters and the second parameter, and machine learning a value of one of the plurality of first parameters according to the data set to generate, for each of the plurality of first parameters, a learning model indicating a relationship between the one of the plurality of first parameters and the first parameter and the second parameter;
• reading a machining shape data of another machining program to obtain a present second parameter for inferring a present first parameter,
• inferring a present value of the present first parameter from a value of the present second parameter based on the learning model, and
• generating a new machining program based on the present value of the present first parameter inferred,
• wherein the numerical control device controls machining of the workpiece based on the new machining program, and
• the processor determines a type of the second parameter to be extracted for each type of first parameter, extracts the second parameter of the type determined, and inputs the second parameter to be subjected to the machine learning.

A machine learning method for a machine learning device, the machine learning method comprising:

• reading a plurality of machining programs for numerically controlling a machine tool;
• extracting, from parameters of the plurality of machining programs, a plurality of first parameters, and
• extracting, from the parameters of the plurality of machining programs, a second parameter for each of the plurality of first parameters, the second parameter being a parameter used to adjust corresponding one of the plurality of first parameters, the second parameter being fixed and including at least a material property and shape of a material of a workpiece;
• generating a data set including the plurality of first parameters and the second parameter,
• machine learning a value of one of the plurality of first parameters according to the data set to generate, for each of the plurality of first parameters, a learning model indicating a relationship between the one of the plurality of first parameters and the second parameter;
• providing at least the value learned through the machine learning to a machining program generation device to generate a machining program using the value learned through the machine learning,
• wherein the machining program is used to control machining of the workpiece, and
• a type of the second parameter to be extracted is determined for each type of the first parameter, the second parameter of the type determined is extracted, and the second parameter to be subjected to the machine learning is input.