Fuel cells: Neural networks provide new insights into Nafion

Not even the powerful beamlines at ESRF and ILL can accurately ascertain the multi-scale (nanometer to centimeter) structure of Nafion as a function of water content, one of the keys to PEMFC performance. Researchers at Irig found a workaround in the form of a convolutional neural network.

They used nanostructure/water content data on ionic surfactants, whose behavior is well-known, to teach the algorithm. The network then expresses Nafion’s nanostructure for different water contents as a distribution of ionic surfactant behaviors with the corresponding probabilities. Nafion’s self-organization behavior is described by analogy, without a model or initial hypothesis, and the results are much more accurate.

Contact: stefano.mossa@cea.fr

Facebook

Twitter