This paper presents a novel method for assessing multiple fault diagnosability and detectability of nonlinear parametrized dynamical models. This method is based on computer algebra algorithms which return precomputed values of algebraic expressions characterizing the presence of some constant multiple fault(s). Estimations of these expressions, obtained from input and output measurements, permit then the detection and the isolation of multiple faults acting on the system. The application of this method on a coupled water-tank model attests the relevance of the suggested approach. (C) 2018 Elsevier Ltd. All rights reserved.