Automatic control system failure detection via parameter identification techniques

The authors examine a building block of automatic supervision, the detection of system failures. Included is a comparison of two methods of system parameter identification for the purpose of detecting system failures. The identification methods utilized are the final value theorem applied to a step response and the modified least squares technique. Both methods were implemented using a personal computer with a 12-b A/D and D/A board. The application of failure detection in armature winding resistance of a DC servomotor is presented for the two methods. Also included is an example of the detection of a failure in an analog second-order system. It is concluded that system failures can be detected using parameter identification. The modified least squares technique, because it is an optimal identification technique, gives better results in most cases than the final value technique