Fuel cells are a promising technology that has the potential to significantly revolutionize the energy sector by providing clean and sustainable sources of power for transportation, aerospace, and electricity applications. However, optimizing the performance of these devices remains a significant challenge as it largely depends on the catalyst layer in the electrodes of these devices, composed of carbon-supported Pt nanoparticle catalyst (Pt/C), perflurosulfonic acid (PFSA), ionomer binder, and a solvent. The catalyst layer design and operation require extensive knowledge of complex physical and chemical processes especially when achieving a low (<0.1 mgcm^-2) Pt loading.

Furthermore, the degradation of fuel cells is usually associated with either mechanical and/or electrochemical processes of the catalyst layer during operations, ultimately resulting in performance loss. While combinations of factors are responsible for the degradation of the catalyst layer in the fuel cell electrodes, it is still challenging to evaluate which factor contributes the most to the degradation of the catalyst layer.

Machine learning algorithms, being powerful tools for predicting and optimizing the behavior of systems can be used to model complex processes, identify optimal operating compositions, and conditions, and reduce the laborious time and cost required for experimentation. They can be used to analyze data obtained from various experiments to identify patterns, trends, and anomalies. In this study, machine learning algorithms are employed for the prediction of catalyst layer degradation. Furthermore, an assessment of feature importance is conducted, elucidating the relative contributions of individual features to the degradation of catalyst layers. Leveraging on machine learning, we can develop an efficient, reliable, and cost-effective fuel cell system.