Simplified conical tube network modeling for the characterization of gas diffusion layer

SCHMIDHALTER I.; AIMO C.E.; AGUIRRE P.A.

Buenos Aires

Congreso; WCCE11 - 11th WORLD CONGRESS OF CHEMICAL ENGINEERING; 2023

Asociación Argentina de Ingenieros Químicos

Liquid water distribution within Gas Diffusion Layer (GDL) of a Polymer Electrolyte Membrane Fuel Cells (PEMFC) have a significant impact on the fluid-dynamic models of fuel cells or stacks and a simplified model to deal with it is still unknown. Other authors tackled this issue from different point of view: simulating the two-phase flow in porous media as a square lattice of tubes connected together at the nodes; dealing with the distribution of liquid water in the GDL of a PEM-FC through a Nano-scale Monte Carlo study; modeling a fibrous GDL as a regular cubic network of pore bodies and pore throats taking into account the size distributions. A novel and simplified model for the characterization of the porous GDL in PEMFC is proposed. The porous structure is represented as a network of a set of interconnected truncated right circular cone tubes so-called as Conical Tube Network (CTN). Each component of the network is characterized by the small tube radius, the large tube radius, and the length and number of tubes. All these variables are obtained from the solution of a optimization problem minimizing the approximation error between Mercury Intrusion Porosimetry (MIP) experimental data and the model prediction. The main idea behind the model is the assumption that the CTN is filled during MIP experiments beginning from the large tube radius. In filling CTN, cumulative mercury volume intruded in the sample changes as a function of the tube radius. The geometry of this filling process is mathematically modeled. Experimental MIP results are represented as cumulative mercury volume as a function of the pore radius just intruded. For a given value of this experimental pore radius, tubes showing radius greater than the actual pore radius are completely filled, tubes with radius lower than de actual value are totally free, whereas tubes showing one radius greater than the experimental pore radius and the other lower than the experimental pore radius are partially filled in a fraction that can be computed. Optimization variables are length and number of tubes. Tubes radius are selected covering the range of experimental pore radius. However, resorting to discrete optimization, the set of tube radius is optimally selected. The maximum number of tubes radius is controlled with an upper bound constraint. The objective function to be minimized corresponds to the linear error between the cumulative tubes volumes computed by the model and the experimental cumulative volumes obtained from MIP data. The equations of the model define a Mixed Integer Linear Programming (MILP) model. The characterization model is applied to a GDL 25BC from Sigracet. The model is implemented in General Algebraic Modeling System (GAMS) and it is solved with Gurobi solver.A simplified set of 14 characteristic tubes radius is obtained resorting to optimization to describe accurately the cumulative volume distribution for the complete GDL. The cumulative volume distribution of the MIP experimental data is reproduced by mean of the CTN with a maximum relative error lower than 6.5% for the complete range of tube radius.