In this study, a methodology for parametric identification of phenomenological based semiphysical models (PBSMs) is presented. The proposed methodology relies on a hierarchy of the relevance of parameters with respect to the model outputs. This hierarchy is accomplished by means of the Hankel matrix of the process model and its singular value decomposition (SVD). In this way, parameters having a major impact on the process output are prioritized. Two concepts, parameter interpretability and sacrifice parameter, are coined to be used in such a methodology. The proposed scheme is tested in simulation by using two realistic examples, both selected as batch processes due to their inherent difficulty:
δ‐endotoxins production by means of Bacillus thuringiensis and the production of polyhydroxyalkanoates (PHA). Our results show a reduction of 92 % and 39 % in the integral of time‐weighted absolute error (ITAE), in the first and second examples, respectively, with respect to a conventional identification procedure.