Using MATLAB Image Processing to Optimize Injection Molding Process for Bipolar Plates in Hydrogen Fuel Cells

Abstract:

THydrogen fuel cells use a chemical reaction to produce energy, yielding only water and heat as byproducts. One fuel cell yields a small amount of energy, but a stack of fuel cells can power cars, trains, or any multitude of other devices. Bipolar plates are the parts of the fuel cell that provide structure to the cell and stacks of fuel cells. They are commonly machined from steel or graphite. These materials are rigid to support stacks of cells yet also conductive to route energy for use in applications. Hydrogen fuel cells are underused in commercial settings in part due to the high cost of these materials and machining practices used to manufacture them. Reducing the cost of manufacturing would make hydrogen fuel cells more accessible. Previous research has shown that an injection-moldable blend of nylon and nickel-coated carbon fibers meets Department of Energy standards for bipolar plate conductivity. However, the manufacturing process largely shapes the material properties of any part. Refining the injection molding process could increase the conductivity of future plates.
One factor that contributed to the conductivity of the plates is the average angle of nickel-coated carbon fibers throughout the plate. Analyzing the relationship between different fiber angles, molding practices, and conductivity would allow for more conductive plates to be made. However, the average angle is painstakingly difficult to obtain and verify experimentally. Thousands of fibers through hundreds of images must be analyzed to determine the angle of the fibers, taking weeks of research time. This research focuses on developing an image processing program to reduce the analysis time to mere minutes with minimal penalty to accuracy. The program uses MATLAB Image Processing toolkit to distinguish fibers from the surrounding polymer and image artifacts to create a measure of average fiber angle accurate to previous research.