Investigation of a novel compression concept for fuel cell stacks

The usage of hydrogen in fuel cells gains more and more importance due to the need for emission-low alternatives in various technical areas. The focus of this work lies on the polymer electrolyte membrane fuel cell. One parameter to reduce performance losses is the contact pressure between the bipolar plate and the gas diffusion layer. It is influenced by first the assembly force and second the swelling of the membrane during operation which is caused by changes in temperature and humidity. To improve the stress distribution and to keep the stress increase during operation low, a novel compression concept is proposed and investigated in an FEM analysis. Different variants of springs are compared to each other and to the reference, a disc spring stack on tie rods. The results show that an even stress distribution is achievable through a concept with 15 small springs where the force flow enters the end plate at various points over the stack. High contact pressure at the edges of the active area can be reduced by reducing the end plate bending. This is achieved by moving the springs closer together so that the inner springs touch each other. A difference between minimum and maximum stress along one path in the cross-section of 0.2-0.4 MPa is achievable and of 0.9-1.3 MPa along the longitudinal section. It is found that the stress increase from assembly to operation depends on the spring characteristic. The maximum increase at the edge of the active area can be reduced to 1.7 MPa, and the smallest stress increase in the centre of the active area is around 0.5 MPa. The proposed concepts show better results than the disc spring reference with a good stress distribution. Although the stress increase in the proposed concepts is also better than in the disc spring stack the value is not yet satisfactory because it strongly influences the porosity of the GDL during operation. The most promising spring concept is tested experimentally. The experimental results show a qualitative accordance with the simulation.