Process data play a vital role in monitoring plant performance and yield accounting. The importance of timely detection of gross errors in process data is widely recognized. Most detection methods assume that the data are serially uncorrelated. The assumption is mathematically convenient but is often contradicted by experimental evidence. Correlated observations may be caused by process dead time, process dynamics, process control, and other physical factors. This paper examines the effect of serial correlations and evaluates two methods which allow previously developed gross error detection methods to be adapted to serially correlated data. The measurement test (MT) is used as the framework of reference in the evaluation.
ASJC Scopus subject areas
- Chemical Engineering(all)
- Industrial and Manufacturing Engineering