Scientific research on cell proton exchange fuel cells (PEMFC) have, until recently, almost exclusively concerned fundamental aspects of electrochemistry, particularly the design, sizing, the electrochemical performance and diagnostics. Recently, the objectives of life cycle have opened a new direction of research on the mechanical behavior of the PEMFC leading to its static and dynamic optimization. At the same time new environmental facilities of the test platform "Fuel Cell Systems" at Belfort are developed. Vicky ROUSS thesis sustained in 2008 shows the importance and the potential of the black box modeling to simulate the mechanical behavior of the PEMFC and experimental mechanical signatures to highlight the presence of physical phenomena inside PEMFC. In this context the work of this thesis concerned the monitoring of durability tests by simulation and real-time black-box operation to explore the physical phenomena inside the PEMFC. Modeling neural networks simple systems such as harmonic oscillator represented the first step towards the definition of a neural control model of real time environmental tests. Then, it was considered the case of the harmonic oscillator excited by the base, which corresponds to the fuel cell mounted on the vibration platform. The optimal neural architecture has been defined in several stages using different algorithms. This architecture uses as input the control signal of the system and the measured signal on the fuel cell at the time t and as output the predicted response behavior of the fuel cell at time t+1. This architecture has been developed and validated by tests on the platform. Other tests have allowed demonstrating the different behavior of the fuel cell in accordance with the amplitude of solicitation, the pressure and temperature of the fuel cell. Mechanical signatures made from tests complete the existing library of signatures and demonstrate new behaviors of the fuel cell.
