Logo image
Back
An Artificial Neural Network for Capillary Transport Characterization of Fuel Cell Diffusion Media
Journal article

An Artificial Neural Network for Capillary Transport Characterization of Fuel Cell Diffusion Media

Emin C. Kumbur, Kendra V. Sharp and Matthew M. Mench
ECS transactions, v 11(1), pp 675-681
28 Sep 2007
DOI: https://doi.org/10.1149/1.2780980

Abstract

This study addresses the development of a design algorithm based on artificial neural network (ANN) that can precisely predict the capillary transport characteristics of fuel cell diffusion media (DM). A three-layered ANN architecture processing the feed-forward error back propagation methodology has been constructed. The designed neural network was systematically trained with the novel benchmark data generated from direct measurements of capillary pressure-saturation of differently engineered DMs under a wide range of conditions [8-10]. Once the trained network learned the complex non-linear relationship between the transport properties and measured parameters of the tested DM samples, it was utilized to predict the capillary pressure of the DM as a function of the hydrophobic additive content and assembly compression pressure at the intermediate conditions, in which the experimental data are not available.

Metrics

4 Record Views

Details

Logo image