Finish-machining amount prediction apparatus and machine learning device

A finish-machining amount prediction apparatus for predicting, when performing finish machining on respective parts of a component to be attached to a machine, finish-machining amounts of the respective parts of the component, the finish-machining amount prediction apparatus comprising:

• a machine learning device for learning the finish-machining amounts of the respective parts of the component in the finish machining, wherein
• the machine learning device has a processor configured to:
• observe, as state variables expressing a current state of an environment, (a) finish-machining amount data indicating the finish-machining amounts of the respective parts of the component in the finish machining and (b) accuracy data indicating accuracy of respective parts of the machine, to which the component measured before a start of the finish machining is attached, wherein
• the finish machining includes grinding and scraping,
• the finish-machining amount data include at least one of grinding frequencies, grinding amounts, and scraping frequencies of the respective parts of the component,
• acquire determination data indicating propriety determination results of the accuracy of the respective parts of the machine, to which the component after being subjected to the finish machining is attached,
• learn the finish-machining amounts of the respective parts of the component in the finish machining in association with the accuracy data by using the state variables and the determination data,
• calculate an error between (i) a correlation model that derives the finish-machining amounts of the respective parts of the component in the finish machining from the state variables and the determination data and (ii) a correlation feature identified from teacher data prepared in advance, and
• update the correlation model to reduce the error, and
• the teacher data is configured by experimental values accumulated when a worker records determined finish-machining amounts of the respective parts of the component in past finish machining for the component.