ABSTRACT Clamping pressure applied during proton exchange membrane fuel cell (PEMFC) assembly reduces interfacial contact resistance but impedes mass transport, yielding coupled beneficial and adverse effects on performance, which makes its optimal specification nontrivial. This study establishes an integrated framework that couples three‐dimensional multiphase non‐isothermal computational fluid dynamics (CFD) with two‐dimensional finite element analysis (FEA) and data‐driven surrogate modeling to quantify these interactions and to enable rapid optimization. Deformation induced changes in geometry, gas diffusion layer porosity and permeability, and pressure dependent interfacial contact resistance are propagated into the transport and electrochemical model. Three surrogate models, namely the radial basis function neural network (RBFNN), support vector regression (SVR), and Gaussian process regression (GPR) are trained on the coupled CFD and FEA dataset to predict power density over the design space of operating voltage and clamping pressure. Results reveal region‐specific effects of clamping pressure: In the activation‐loss‐dominated regime, performance is low and insensitive to pressure; in the ohmic‐loss‐dominated regime, power density peaks at moderate pressure; in the concentration‐loss‐dominated regime, power density decreases monotonically with increasing pressure. All surrogate models achieved R 2 values exceeding 0.995, with fast predictions. Coupled CFD‐FEA simulation yielded a maximum power density of 0.777 W·cm −2 , with a recommended range of 0.75–1.25 MPa clamping pressure and 0.60–0.65 V voltage. The artificial intelligence‐genetic algorithm framework refined it to 0.8–1.2 MPa and 0.61–0.63 V, where the predicted power density consistently exceeds 0.78 W·cm −2 . These findings provide quantitative insights for the optimal assembly and operation of PEMFCs.
Building similarity graph...
Analyzing shared references across papers
Loading...
Lei Chen
Jianmin Wu
Guoqiu Liu
Fuel Cells
Xi'an Jiaotong University
Building similarity graph...
Analyzing shared references across papers
Loading...
Chen et al. (Tue,) studied this question.
www.synapsesocial.com/papers/69af95ee70916d39fea4e02b — DOI: https://doi.org/10.1002/fuce.70062