| 摘要 |
PURPOSE:To allow efficient and automatic generation of diagnosis algorism free from subjectiveness of human being by sorting data for fault diagnosis basing on attribute values at normal time and attribute values in which accuracy of sensor at abnormal time is taken into account. CONSTITUTION:Electrical amounts, e.g. residual current, residual voltage, load current, and ground voltage, are calculated or measured at normal time thus obtaining attribute values at normal time. Line constants and resistance at ground fault point are then calculated or measured at abnormal time using the attribute values at the normal time as parameters thus obtaining attribute values at the abnormal time. Attribute values at the normal time and attribute values at the abnormal time, where accuracy of sensor is taken into account, are then calculated and the data for fault diagnosis are sorted basing on the attribute values thus calculated. This method allows automatic generation of algorism through mechanical (computor) learning based on recursive learning. Furthermore, algorism can be generated efficiently by introducing an evaluation function for the attributes. |
