Visualization of damage progress on solid oxide fuel cell

Fuel cell is regarded as a highly efficient power generation system as well as low-pollution. In particular, SOFC (Solid Oxide Fuel Cell) has high generation efficiency. However, a crucial issue in putting SOFC into practical use is the establishment of a technique for evaluating the deterioration. We have previously developed a technique to measure the mechanical damage of SOFC using Acoustic Emission (AE) method. This paper applied the kernel Self-Organizing Map (SOM), which is an extended neural network model, to AE data observed from damage progress on SOFC to produce a cluster map reflecting similarity of AE waves. The obtained map visualized the change of occurrence patterns of similar AE waves showing four phases of damage progress. The interpretation of the result as physical phenomenon is limited at this stage, though our methodology provides a common foundation for comprehensive damage evaluation system as well as monitoring.