Development of a process monitoring strategy for facilities producing multiple portfolio of products

Abstract

The process monitoring strategies are the methodologies devised and applied for newlineconstant surveillance of the processes facilitating the detection of abnormal newlinebehaviour or upset conditions normally designated as faults. The proposed work aims newlineto develop a stratified process monitoring strategy for a multistage production facility newlineengaged in the production of multiple portfolio of products. In the pursuit of the newlineabove mentioned goal the addresses several objectives viz. a) Devising a unified newlinemathematical process model or representation to facilitate monitoring of multiple newlineproducts or multiple grades of the same product. b) Development of a fault detection newlinestatistic to enable early detection of the developing fault c) Development of fault newlinediagnostic statistic to determine the contribution of the main or primary newlinecharacteristics and the associate characteristics for a particular fault or upset newlinecondition. d) Validation of the devised monitoring strategy via a case study from an newlineintegrated steel plant (ISP). newlineIn the present study two stratified process monitoring strategies were suggested and newlineapplied to a case study. The first strategy was employed for the monitoring of newlineindividual product portfolios whereas the second monitoring strategy intended to newlinedevelop a single unified monitoring strategy for multiple portfolio of products newlineproduced in the facility. The devised strategies were based on the multi-block variant newlineof PCA. The strategies hence devised were employed to detect fault(s) in the process newlineusing Hotelling T2 newlinecontrol charts. A novel fault detection statistic was suggested for newlineearly detection of the developing fault. In the study a strategy for stratified diagnosis newlineof fault has also been proposed. The stratified diagnosis of the faults encompasses newlinethe estimation of the relative contributions of the stages, sub-stages and the process newlineand feedstock characteristics towards the detected fault or out of control newlineobservations.