Information processing device, control device, and optimization method

An information processing device configured to, when generating a thermal displacement amount prediction formula by performing machine learning for, based on a measurement data group including temperature data regarding a mechanical element that undergoes thermal expansion and is in a machine tool and regarding a periphery of the mechanical element and/or operating state data regarding the mechanical element, estimating an amount of thermal displacement of the mechanical element, optimize values for a plurality of hyperparameters included in the thermal displacement amount prediction formula, the information processing device comprising:

• a measurement data acquisition unit configured to acquire the measurement data group;
• a thermal displacement amount acquisition unit configured to acquire an actual measurement value for an amount of thermal displacement of the mechanical element;
• a storage unit configured to set the measurement data group acquired by the measurement data acquisition unit as input data, set the actual measurement value for the amount of thermal displacement of the mechanical element acquired by the thermal displacement amount acquisition unit as a label, and store the input data and the label associated with each other as teacher data;
• a parameter selection method determination unit configured to select, as a first hyperparameter, at least one hyperparameter to be optimized from among the plurality of hyperparameters;
• a parameter selection unit configured to set/change a value for the first hyperparameter, set a hyperparameter that was not selected by the parameter selection method determination unit as a second hyperparameter, and fix a value for the second hyperparameter;
• a machine learning unit configured to, by performing machine learning based on the measurement data group and the actual measurement value for the amount of thermal displacement of the mechanical element for each combination of the value for the first hyperparameter and the value for the second hyperparameter, generate the thermal displacement amount prediction formula for each set/changed value for the first hyperparameter while setting the value for the second hyperparameter to a fixed value;
• and
• a model evaluation unit configured to calculate, as an evaluation value, an error between the actual measurement value for the amount of thermal displacement of the mechanical element and an amount of thermal displacement estimated by inputting the measurement data group to the thermal displacement amount prediction formula for each value for the first hyperparameter while setting the value for the second hyperparameter to the fixed value, and store the calculated evaluation value in the storage unit in association with the value for the first hyperparameter while setting the value for the second hyperparameter to the fixed value,
• wherein the parameter selection method determination unit, based on a history for the value for the first hyperparameter and the evaluation value that are stored in the storage unit, sets the value for the first hyperparameter at a time when the evaluation value is a minimum to an optimal value while setting the value for the second hyperparameter to the fixed value.