Multiway principal component analysis and case base reasoning methodology for abnormal situation detection in a nutrient removing SBR

Abstract

Multiway Principal Component Analysis (MPCA) and Case-Based Reasoning (CBR) approaches are applied in a biological nutrient removal process. The goal is monitoring of normal and abnormal operation conditions in this process. MPCA is used as a compression tool where with few variables the process can be described, as well as, to detect batches with abnormal conditions. However, some abnormal conditions (alarms) are omitted or, otherwise some false alarms are produced. Then, CBR is proposed. It assumes that similar problems should have similar solutions. In biological process, these problems generally are operational situations under normal or abnormal situations: low ORP, high pH, sensors fault, among others. These symptoms can be stored in a Case Base (CB) in order to diagnose future situations. Several proof are made in order to find the better methodology.