Most of car manufacturers around the world have launched important research programs on the integration of fuel cell (FC) power generators into cars. Despite the first achievements, fuel cell systems are still badly known, particularly when talking about fault diagnosis and predictive maintenance. This paper proposes a first step in this way by introducing a simple but also efficient diagnosis-oriented model of a proton exchange membrane fuel cell (PEMFC). The considered diagnosis model is here a fuzzy one and is tuned thanks to genetic algorithms. (C) 2003 Elsevier B.V. All rights reserved.