Analysis of Autocorrelations in Dynamic Processes

Abstract
Data collected by process information and control systems are almost always correlated due to process dynamics combined with short sampling times. The traditional time series approach for dealing with autocorrelated data has been to model the autocorrelation. This modeling effort is substantially more difficult than simply treating the time series as a sequence of independent data. In this article, we develop and demonstrate a methodology that yields insight into the error introduced by not modeling the autocorrelation of the process data when performing material balances around process equipment. This work shows that in many cases data analysis (e.g., gross-error detection) can be done using the simplified models.