Fuel cells are currently experiencing an invigorating resurgence, both the industrial and research levels. Diagnosis of stack performance is of importance to proton exchange membrane fuel cell (PEMFC) research. In this paper, a bond graph (BG) approach is used for modelling, simulation and robust diagnosis of a PEMFC. In literature, several PEMFC diagnosis methodologies are outlined in terms of efficiency and applicability. This paper describes the linear fractional transformations (LFT) method to make it capable of handling the PEMFC diagnostics; an approach based on LFT-BG is developed to diagnose hydration and cells deterioration faults that may occur within a fuel cell. Simulation and experimental diagnostic testing results of a 1.2 kW Nexa fuel cell are presented and used to show the dynamic behaviour of the system variables and assessing the performance of the observer.