To provide a learning control method capable of preventing learning accuracy from deteriorating if errors are greatly changed by factors that change over time or if learning has not been fully achieved as manufacturing conditions vary greatly, and also capable of preventing learning accuracy from deteriorating if model errors about the quantity to be controlled are large.

A manufacturing condition determining part 11 receives manufacturing conditions from a manufacturing condition setting device 20 and sends them to a learning calculating part 14. An after-boring length predicted value calculating part 12 receives the manufacturing conditions from the manufacturing condition setting device 20 and predicts the length of a steel pipe after boring according to the manufacturing conditions. The learning calculating part 14 calculates a new amount of learning for a controlled value according to a predicted value of length received from the after-boring length predicted value calculating part 12 and the amounts and times of learning of individual controlled values stored in a learning information storage part 16. The learning information storage part 16 stores the amounts and times of learning in association with the individual controlled values.