This paper presents an analysis of processes with regard to their circulating flows and mainly focuses on the characterization of variables depending on their redundancy level. Previous studies in this field were performed on processes described by linear equations (i.e. when considering a model of total conservation of material or energy). Here, the authors propose a novel methodology, permitting extension of observability analysis to processes involving linear and bilinear equations (i.e. partial and total balance equations). The aim here is the classification of process variables according to their observability level and, on a more practical side, the extraction of the subset of unmeasured variables deducible from those measured and from balance equations. The proposed technique is based on the analysis of the circuit matrix associated to the graph of the process under consideration. This analysis is performed depending on the number and position of the measurement devices.