Type-II critical values for a steady-state identifier

Identification of steady and transient states in process signals is used to select data segments for modeling, model adaptation, and historian archives; and to trigger on-line interventions and stages. Cao and Rhinehart [20] published an easily implemented statistic to evaluate the null hypothesis (signal is at steady-state) and published critical values (1997) to reject the steady-state (SS) hypothesis (accept the transient state condition). This work provides the complementary critical values, those needed to reject the transient state hypothesis (accept the steady-state condition). Critical values for Type-II error are reported for a variety of signal-to-noise conditions, and for a variety of non-steady-state behaviors. A critical value of 0.8 for the ratio-statistic permits acceptance of SS for processes visually judged to be at SS, and rejects processes that are visually judged not-at-SS with greater than 99.99% confidence. The result is substantially independent of the process conditions and noise distribution. (C) 2010 Elsevier Ltd. All rights reserved.