Working machine learning diagnostic system, the state diagnostic apparatus and state learning device

A learning diagnostic system for a working machine is capable of making a fault diagnosis of the working machine universally relative to various types of sensor information, and preventing a failure in the working machine by enabling a fault diagnosis even in a transient operating state that represents a transitional state between operating states. A state learning device 201 sorts inputted sensor data 101a into one under a steady operating state and one under a transient operating state, and generates, through learning, steady state data 102a and intermediate state data 103a, each including a permissible error. A state diagnostic device 202 uses the steady state data 102a to determine whether an operating state of the working machine related to the inputted sensor data is the steady operating state or the transient operating state, and make a fault determination in the steady operating state. The state diagnostic device 202 also uses the intermediate state data 103a to make a fault determination in the determined transient operating state. The intermediate state data 103a also includes intermediate point information.