Towards self-validating control loops

The possibility of developing self-validating control loops is examined for those loops that inherently eliminate steady state error, for instance by the incorporation of integral action. In these situations it should be possible to reason about abnormal, but steady state, deviations in measured, control loop related variables. The approach discussed here involves a search and test strategy which is based on a variant of the signed directed graph and on knowledge pertaining to control performance. The approach is applied to a simulated CSTR process successfully. Copyright (C) 1998 IFAC.