Diagnosis of Fuel Cells for stationary and automotive applications (2nd phase) – DIAPASON 2

Fuel cell is a promising technology for both automotive and stationary applications. However, its reliability and its lifetime remain major hurdles for its wide access to these two markets. It is therefore necessary to develop reliable diagnosis tools with minimal instrumentation and even without any additional instrumentation to the one included in the systems. For that purpose, the DIAPASON project (ANR PAN-H 006-04) aimed at developing different diagnosis methods for fuel cell, using only the stack as sensor and by comparing diagnosis methods based on physical or black box models or based on pattern recognition algorithms. DIAPASON project has already resulted in developing:
* an impedance spectrometer able to perform impedance spectras on stacks with voltages up to 500 V.
* diagnosis algorithms for detection of fuel cells' failures based on dynamic neural networks and a patented diagnosis method based on pattern recognition..
* a 24 channels cardboard measuring voltages through GMR (Giant Magneto-Resistance) that was interfaced with a multiplexing cardboard, that was itself controlled by a digitalization cardboard connected to a PC working under Labview and Matlab. This set-up was experimentally successfully validated.
* a Labview and Matlab interface between diagnosis algorithms with acquisition cardboard, able to diagnose fuel cell's state of health.
This project is a continuation of DIAPASON project and aims, in the same time, at bringing several major breakthroughs:
* new algorithms développement bbased on new innovative approaches (statistic methods or operating point follow-up).
* extension of the validity range of both the abovementioned approaches and of those already studied during the first phase (black box models and pattern recognition based algorithms) to other defaults that were not considered in phase 1.
* extension of present measurements possibility (now, only voltage with cycle time of 1.6 ms) to: shorter cycle times (~100 µHz for impedance measurements), other variables (current, pressure drops,…), higher power ranges (100 to 200 cells stacks).
* integration, with the help of the skills and know-how of a SME specialized in electronic components integration and project partner, all the modules developed in DIAPASON and improved in this project in a single hybride on-board module gathering GMR cells, instrumentation, calculation core and extern memory.
* integration of well known algorithms (FFT, …) or developped during this project in the shape of IP software components, easily integrable in a chip.
These diagnosis techniques could then be used either in real time (by coupling with fuel cell's regulation system: "On-board diagnostic") or during regular and planned maintenance operations. Such techniques will then allow improving fuel cell system's reliability and increasing its lifetime by anticipating degradation phenomenon.
In fine, this project will pay a special attention to the validation of the performances of the developed integrated (hardware + algorithm) diagnosis tool on two real systems, dedicated to automotive and microcogeneration applications. This tool will therefore represent a project's major expected breakthrough, that will in fine allow increasing the system's reliability and increasing its lifetime. It will also set and strengthen a solid know-how at national level, guaranteeing a rapid valorisation of project's results by fuel cell based systems users or manufacturers.