ABSTRACT
Identification of both steady-state (SS) and transient state in noisy time- or sequence-based signals is important. A conceptually simple approach to identify SS would be to look at data values in a recent time-window and if the range between high and low is acceptably small, declare SS. Cao and Rhinehart and Shrowti, et al. investigated the robustness of the R-statistic distributions to a variety of noise distributions in the process variable. The basis for the R-statistic method presumes that there is no autocorrelation in the time-series process data when at SS. If the autocorrelation in the PV is due to sustained oscillations from a feedback mechanism, then this method of delaying the sample will only work if the delay (lag) matches the cycle time of the oscillation. In a multivariable analysis, there is a greater chance that data vagaries will cause a false TS claim in any one monitored variable.
