| 摘要 |
Method of monitoring and/or diagnosing the operational state of a complicated, dynamic system constituted by a machine or an industrial process installation, a plurality (M) of state variables being measured at a plurality (N) of instants for the assembly of a corresponding measurement data (Xik), which are processed statistically to form at least one reference class corresponding to a specific operational state. There is used here a group of measurement data (Xik) serving as a learning group, which is adapted by linear projection to a new coordinate system with fewer coordinate axes. In a corresponding way, current measurement or calculation data are adapted to the new coordinate system, whereupon the distance between the current measurement or calculation data, on the one hand, and the measurement data corresponding to the learning group, on the other hand, is determined. If this distance exceeds a predetermined value, it is established which state variables that cause the deviation, by calculating the effect of the respective state variable on the variance along the respective coordinate axis in the new coordinate system. |
