Conditional base maintenance has been gaining widespread acceptance, with numerous sensors attached to equipment for constant monitoring of its operational state, the resulting sensor data being compared with those about the equipment in the normal state for a diagnosis to determine whether the equipment is currently operating normally, the result of the diagnosis being used to conduct maintenance. Conditional base maintenance can rapidly detect aging deterioration of the equipment, so that abnormal sates that were not detected before in time base maintenance can now be detected. However, although conventional diagnosis technology can distinguish between the normal state and anomaly, it has been difficult with such technology to identify causes including abnormal phenomena and parts. If it is determined that the target equipment to be diagnosed deviates from the normal state, the degree of deviation of each parameter from the normal state as the reference is calculated as an abnormal contribution ratio. A failure cause is estimated from a similarity ratio between the calculated abnormal contribution ratio and the abnormal contribution ratio of each of the failure causes collected in the past and including failure phenomena and failure parts.