A training model creating method includes creating a database by breaking a component in a laser device on a Digital Twin constructed by modeling electrical hardware and software of the laser device, and accumulating data in which the broken component and a failure phenomenon output from the Digital Twin are associated with each other; and training a training model using the data in the database as training data for machine learning so that, upon receiving an input of information on a failure phenomenon, the training model outputs information on a malfunction location corresponding to the input.